The Oxford Handbook of Ergativity 9780198739371, 0198739370

Citation preview

 

T h e Ox f o r d H a n d b o o k o f

E RG AT I V I T Y

 

OXFOR D HANDBOOKS IN LINGUISTICS Recently Published

THE OXFORD HANDBOOK OF CORPUS PHONOLOGY Edited by Jacques Durand, Ulrike Gut, and Gjert Kristoffersen

THE OXFORD HANDBOOK OF LINGUISTIC FIELDWORK Edited by Nicholas Thieberger

THE OXFORD HANDBOOK OF DERIVATIONAL MORPHOLOGY Edited by Rochelle Lieber and Pavol Štekauer

THE OXFORD HANDBOOK OF HISTORICAL PHONOLOGY Edited by Patrick Honeybone and Joseph Salmons

THE OXFORD HANDBOOK OF LINGUISTIC ANALYSIS Second Edition Edited by Bernd Heine and Heiko Narrog

THE OXFORD HANDBOOK OF THE WORD Edited by John R. Taylor

THE OXFORD HANDBOOK OF INFLECTION Edited by Matthew Baerman

THE OXFORD HANDBOOK OF DEVELOPMENTAL LINGUISTICS Edited by Jeffrey Lidz, William Snyder, and Joe Pater

THE OXFORD HANDBOOK OF LEXICOGRAPHY Edited by Philip Durkin

THE OXFORD HANDBOOK OF NAMES AND NAMING Edited by Carole Hough

THE OXFORD HANDBOOK OF INFORMATION STRUCTURE Edited by Caroline Féry and Shinichiro Ishihara

THE OXFORD HANDBOOK OF MODALITY AND MOOD Edited by Jan Nuyts and Johan van der Auwera

THE OXFORD HANDBOOK OF LANGUAGE AND LAW Edited by Peter M. Tiersma and Lawrence M. Solan

THE OXFORD HANDBOOK OF PRAGMATICS Edited by Yan Huang

For a complete list of Oxford Handbooks in Linguistics, please see pp 1263–1264

 

The Oxford Handbook of

ERGATIVITY Edited by

JESSICA COON, DIANE MASSAM, and

LISA DEMENA TRAVIS

1

 

3 Great Clarendon Street, Oxford, ox2 6dp, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Jessica Coon, Diane Massam, and Lisa deMena Travis 2017 © the chapters their several authors 2017 The moral rights of the authors have been asserted First Edition published in 2017 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2016944782 ISBN 978–​0–​19–​873937–​1 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

 

Contents

Notes on Contributors 

xi

1. Introduction  Jessica Coon, Diane Massam, and Lisa demena Travis

1

PA RT I   AC C OU N T I N G F OR E RG AT I V I T Y   Representing Ergativity  2. Ergativity in Discourse and Grammar  John W. Du Bois

23

3. Parameterizing Ergativity: An Inherent Case Approach  Michelle Sheehan

59

4. Accusative and Ergative in Hindi  Anoop Mahajan

86

The Nature of Ergative Case  5. On Inherent and Dependent Theories of Ergative Case  Mark C. Baker and Jonathan David BOBALJIK

111

6. The Locus of Ergative Case  Julie Anne Legate

135

7. Ergative need not Split: An Exploration into the TotalErg Hypothesis  Itziar Laka

159

8. The Structural Source of Split Ergativity and Ergative Case in Georgian  Léa Nash

175

 

vi   Contents

PA RT I I   C HA R AC T E R I ST IC S A N D E X T E N SION S   Characteristics  9. Split Ergativity in Syntax and at Morphological Spellout  Ellen Woolford

205

10. Split Ergativity is not about Ergativity  Jessica Coon and Omer Preminger

226

11. Ergativity and Differential Case Marking  Andrej Malchukov

253

12. Three-​Way Systems do not Exist  Gereon Müller and Daniela Thomas

279

13. Antipassive  Maria Polinsky

308

14. Remarks on the Relation between Case-​Alignment and Constituent Order  Tarald Taraldsen

332

Extensions  15. Ergativity in Nominalization  Artemis Alexiadou

355

16. Ergativity and Austronesian-​Type Voice Systems  Michael Yoshitaka Erlewine, Theodore Levin, and Coppe van Urk

373

17. On the Morphosyntactic Reflexes of Information Structure in the Ergative Patterning of Inuit Language  Alana Johns and Ivona Kučerová 18. Ergative Constellations in the Structure of Speech Acts  Martina Wiltschko

397 419

 

Contents   vii

PA RT I I I   A P P ROAC H E S TO E RG AT I V I T Y   Diachronic  19. Grammaticalization of Ergative Case Marking  William B. McGregor

447

20. Deconstructing Iranian Ergativity  Geoffrey Haig

465

21. Intransitivity and the Development of Ergative Alignment  Edith Aldridge

501

22. Developments into and Out of Ergativity: Indo-Aryan diachrony  Miriam Butt and Ashwini Deo

530

23. Ergativity and Language Change in Austronesian Languages  Ritsuko Kikusawa

553

24. Lexical Category and Alignment in Austronesian  Daniel Kaufman

589

Acquisition  25. The Acquisition of Ergativity: An Overview  Edith L. Bavin

631

26. The Role of Defaults in the Acquisition of Basque Ergative and Dative Morphology  Jennifer Austin

646

27. A Comparative Study of the Acquisition of Nominative and Ergative Alignment in European and Mayan Languages  Clifton Pye and Barbara Pfeiler

665

Experimental  28. Processing Ergativity: Behavioral and Electrophysiological Evidence  693 Adam Zawiszewski

 

viii   Contents

29. Experimental Approaches to Ergative Languages  Nicholas Longenbaugh and Maria Polinsky

709

PA RT I V   C A SE ST U DI E S   30. Correlates of Ergativity in Mayan  Judith Aissen

737

31. Ergative Case in Burushaski: A Dependent Case Analysis  Mark C. Baker

759

32. Ergativity in Basque  Ane Berro and Ricardo Etxepare

782

33. Hindi/​Urdu and Related Languages  Miriam Butt

807

34. Ergativity in Inuktitut  Richard Compton

832

35. Ergativity in Nakh–​Daghestanian  Diana Forker

851

36. Ergativity in Neo-​Aramaic  Geoffrey Khan

873

37. Ergativity in Africa  Christa König

900

38. Ergativity in Tibeto-​Burman  Shobhana Chelliah

924

39. The Ergative in Warlpiri: A Case Study  Mary Laughren

948

40. Ergative–​Absolutive Patterns in Tongan: An Overview  Yuko Otsuka

989

41. Alignment across Tsimshianic  Tyler Peterson

1007

 

Contents   ix

42. What being a Syntactically Ergative Language means for Katukina-​Kanamari  Francesc Queixalós 43. Ergativity in Jê Languages  Andrés pablo Salanova 44. Interaction of Ergativity and Information Structure in Jaminjung (Australia)  Eva Schultze-​Berndt

1035 1065

1089

45. Alignment and Orientation in Kartvelian (South Caucasian)  Kevin Tuite

1114

References  Author Index  Language Index  Subject Index 

1139 1241 1245 1251

 

 

Notes on Contributors

Judith Aissen  is Research Professor of Linguistics at UC Santa Cruz. The primary focus of her work since the 1970s has been on the morphology, syntax, and information structure of the Mayan languages, especially Tsotsil, Tz’utujil, and K’ichee’. She has always been particularly interested in the properties and analysis of Agent Focus constructions, and their position at the interface of these three modules. Edith Aldridge is Associate Professor in the Department of Linguistics at the University of Washington. Her work focuses on syntax, particularly diachronic change, of Austronesian languages, Chinese, and Japanese. Her work on ergativity has appeared in the journals Language and Linguistics, Compass, Lingua, Linguistics Vanguard, Sophia Linguistica, as well as the collected volumes Grammatical Change:  Origins, Nature, Outcomes and Encyclopedia of Chinese Language and Linguistics. Artemis Alexiadou  is Professor of English Linguistics at the Humboldt University of Berlin. She received her PhD in linguistics in 1994 from the University of Potsdam. Her research interests lie in theoretical and comparative syntax, morphology, and most importantly in the interface between syntax, morphology, the lexicon, and interpretation. Jennifer Austin  is Associate Professor in the Department of Spanish and Portuguese Studies at Rutgers University, Newark. She received her PhD in linguistics from Cornell University with a minor in cognitive science. Her research interests include language acquisition, bilingualism, and the effects of language contact. She is a co-​author of the book Bilingualism in the Spanish-​speaking World (2015) and has published articles on the acquisition of Basque, English, and Spanish. Mark C. Baker is Distinguished Professor of Linguistics and Cognitive Science at Rutgers University. He received his PhD in linguistics in 1985 from MIT, and previously taught at McGill University. He specializes in the syntax and morphology of less-​studied languages, seeking to bring together generative-​style theories and data from fieldwork and typological comparison. He has written five research monographs, including Case: Its Principles and Its Parameters (2015). Edith L. Bavin  received her PhD from the University of Buffalo and taught at the University of Oregon before moving to Australia, where she conducted fieldwork on Nilotic languages and on the acquisition of Warlpiri. She switched to experimental work focusing on both typical and atypical language development. She was editor

 

xii   Notes on Contributors of the Journal of Child Language (2006–​12), edited the Cambridge Handbook of Child Language (2009) and a second expanded edition (2015) together with Letitia Naigles, and co-​edited with Sabine Stolle The Acquisition of Ergative Languages (2013). Now an honorary professor at La Trobe University, she is still conducting acquisition research. Ane Berro is a post-doc researcher at the University of the Basque Country (UPV/EHU) and the laboratory Structures Formelles du Langage (UMR 7023, CNRS/ Paris 8). She did her doctoral dissertation ‘Breaking Verbs: From Event Structure to Syntactic Categories in Basque’ under the supervision of Beatriz Fernández (UPV/ EHU) and Ricardo Etxepare (CNRS IKER UMR 5478), and, currently, she is working on aspect, categorization, and the relation between syntax and morphology. Jonathan David Bobaljik  has held appointments at Harvard and McGill, and is currently Professor of Linguistics at the University of Connecticut. His theoretical research has focused primarily on morphology and syntax, and he has conducted descriptive fieldwork on the Itelmen language, Kamchatka, Russia. Publications include Universals in Comparative Morphology (2012). Miriam Butt is Professor for Theoretical and Computational Linguistics at the University of Kontanz. Her research interests include morphosyntax, historical linguistics, and computational linguistics. The bulk of her research is on South Asian languages, with a special emphasis on Urdu, though she has also worked on English and German. She has written and edited several books on syntax, semantics, and computational linguistics, including a textbook on theories of case. Shobhana Chelliah  is  a documentary linguist working primarily on the Tibeto-​ Burman languages of Manipur state in northeast India. Her interests lie in morphosyntactic issues such as case morphology, referent tracking, and affix ordering. Her publications include A Grammar of Meithei (1997) and the Handbook of Descriptive Linguistic Fieldwork (co-​authored with Willem de Reuse (2011). With funding from the National Science Foundation, she is currently working with members of the Lamkang community (Tibeto-​Burman) to develop a practical orthography and create an online dictionary for their language. She is also working towards the creation of a language archive for Tibeto-​Burman languages. Chelliah served as the Program Director for the Documenting Endangered Languages program at the National Science Foundation (2012–​15), and is currently a professor of linguistics at the University of North Texas. Richard Compton is Assistant Professor in the Department of Linguistics at the Université du Québec à Montréal in Canada. His works examines polysynthetic word-​ formation, nominal and verbal incorporation, lexical categories, modification, and agreement in Eastern Canadian Inuktitut. He has conducted fieldwork in the communities of Iqaluit and Baker Lake in Nunavut and is currently co-​editing a new edition of a dictionary of the Kangiryuarmiut dialect of Western Canadian Inuit.

 

Notes on Contributors    xiii Jessica Coon  is Associate Professor of Linguistics at McGill University. She finished her PhD at MIT in 2010 and then spent one year as a postdoctoral fellow at Harvard University. Jessica has worked on topics including ergativity, split ergativity, verb-​initial word order, and agreement, with a special focus on Mayan languages. Her book Aspects of Split Ergativity was published by Oxford University Press in 2013. Ashwini Deo  received a Master’s in Sanskrit grammar and linguistics from Pune, India, followed by a PhD in linguistics from Stanford University (2006). She is an associate professor at Yale University. Her main research interest is in systematic semantic change phenomena—​particularly in the ways in which functional morphemes like tense–​aspect, negation, possession markers change over time in the ways that they do. Within semantics–​ pragmatics she also works on phenomena in the domains of aspect, temporal reference, lexical semantics of verbs, and genericity. Her empirical focus is on the Indo-​Aryan languages. John W. Du Bois  is Professor of Linguistics at the University of California, Santa Barbara. A specialist in functional linguistics, discourse, Mayan languages, and sociocultural linguistics, his work centers on the interaction between discourse and grammar. He has long been interested in the complex functional competitions that drive the emergence of grammar as a complex adaptive system, yielding the extraordinary typological diversity of argument structure constructions and syntactic alignments in the world’s languages. His publications include Competing Motivations (1985), The Discourse Basis of Ergativity (1987), Preferred Argument Structure (2003), Discourse and Grammar (2003), Motivating Competitions (2014), and Towards a Dialogic Syntax (2014). Michael Yoshitaka Erlewine is Assistant Professor at the National University of Singapore. He received his PhD in linguistics from MIT in 2014. His research interests are syntactic theory and the syntax–​semantics interface. Ricardo Etxepare  is a permanent researcher at the Centre National de la Recherche Scientifique, and director of the lab IKER (UMR5478), a research center devoted to the study of the Basque language and Basque texts in Bayonne, France. Diana Forker  teaches general linguistics at the University of Bamberg. She completed her PhD at the Max Planck Institute for Evolutionary Anthropology. Her main interests are languages of the Caucasus, typology, and morphosyntax. She currently works on the documentation of the Nakh-​Daghestanian language Sanzhi Dargwa and on a typological investigation of gender agreement. Among her publications are A Grammar of Hinuq (2013) and several articles on different aspects of Nakh-​Daghestanian languages. Geoffrey Haig  received his PhD in general linguistics from the University of Kiel in 1997. He is currently professor of linguistics in the Institute of Oriental Studies at the University of Bamberg. His research interests focus on the languages of the Middle East, in particular the syntactic features they have inherited, and those they have shared with their neighbors. He is also active in language documentation, and in corpus-​based approaches to language typology and areal linguistics.

 

xiv   Notes on Contributors Alana Johns  is Professor of Linguistics at the University of Toronto. Her research concerns the morphosyntax of complex words, especially in the Inuit language. She has published on ergativity, noun incorporation, and verbal inflection. One of her main research interests are syntactic differences between closely related dialects, where small distinctions can lead to a wide range of effects. Another area of involvement is language maintenance, where linguistics can help in capacity development within communities who are engaged with these issues. She has worked with the Inuit of Nunatsiavut for many years. Daniel Kaufman specializes in historical, descriptive, and theoretical issues in Austronesian languages with a focus on the languages of the Philippines and Indonesia. He is co-​founder and executive director of the Endangered Language Alliance, a non-​ profit organization dedicated to documenting and conserving the endangered languages of New  York City’s immigrant communities and is also Assistant Professor at the Department of Linguistics and Communication Disorders at Queens College, CUNY. Geoffrey Khan  is currently Regius Professor of Hebrew at the University of Cambridge. He was awarded his PhD by the University of London in 1984 for a thesis on extraposition in the Semitic languages. His published books include a series of grammars of Neo-​ Aramaic dialects, editions of medieval Judaeo-​Arabic grammatical texts and medieval Arabic documents and an introduction to the Hebrew Bible. He is general editor of the Encyclopedia of Hebrew Language and Linguistics. Ritsuko Kikusawa  is Associate Professor of the National Museum of Ethnology and the Graduate University for Advanced Studies in Japan. Her research interests include description of Fijian dialects and Betsimisaraka Malagasy, methodology of morphosyntactic comparison and reconstruction, and the linguistic prehistory of Oceania. Her recent publications include “The Austronesian language family” (The Routledge Handbook of Historical Linguistics, ed. Claire Bowern & Bethwyn Evans, 2015). Christa König  is Apl. Professor at the Institute of African Linguistics, University of Frankfurt. Her research interests include verbal aspect and case systems. She has carried out extensive field research on the following languages: Maa (Kenya, Tanzania), Ik (Northeastern Uganda), !Xun (Namibia), and Akie (Tanzania). Her publications include ‘Marked nominative in Africa’ (Studies in Language 30(4): 705–​782, 2006) and Case in Africa (Oxford University Press, 2008). Ivona Kučerová  is Assistant Professor of Linguistics at McMaster University. She specializes in theoretical syntax and semantics, and their interface. Her works explores information structure and its morphosyntactic correlates, definiteness systems and their relation to aspect, the morphosyntax and morpho-​semantics of case, agreement, and case splits, the syntax of null languages, and the syntax of copular clauses. She works mainly on Slavic, Germanic, and Romance languages.

 

Notes on Contributors    xv Itziar Laka  is Full Professor at the Department of Linguistics and Basque Studies and Director of The Bilingual Mind research group at the University of the Basque Country. Her research combines theoretical linguistics and experimental methods to inquire into the representation and processing of language, with a strong focus on syntax and bilingualism. Mary Laughren  is an honorary senior research fellow in the School of Languages and Cultures at the University of Queensland, where she taught linguistics (1993–​2009). Since 1975 she has studied the Warlpiri language and has been involved in the implementation of bilingual education programs in Warlpiri-​speaking communities. One focus of her research is the interplay between lexical and syntactic organization. She is currently collaborating in the documentation of traditional Warlpiri songs, and since 2000 of another Australian language, Waanyi. Julie Anne Legate  is Professor of Linguistics and the Chair of Undergraduate Studies in the Department of Linguistics at the University of Pennsylvania. She received her PhD from MIT in 2002. She is the author of Voice and v: Lessons from Acehnese, and is editor-​ in-​chief of Natural Language & Linguistic Theory. Theodore Levin  received his PhD in linguistics from MIT in 2015 and is a postdoctoral research associate in the Department of Linguistics at the University of Maryland. His primary research interests are in syntactic theory with specific interest in the interaction of case, agreement, and word order. Nicholas Longenbaugh  is a PhD student in linguistics at MIT, with a BA in computer science and linguistics from Harvard University (2014). His work focuses on issues in syntax,  with an emphasis on cross-​linguistic variation and universal principles. He has done original research and fieldwork on  Austronesian and Romance languages. Nicholas has also worked within the Tree-​Adjoining Grammar framework, and has explored topics concerning formal complexity in language. Anoop Mahajan  is Professor in the Department of Linguistics at UCLA. He received his PhD in linguistics from MIT in 1990 and has taught at UCLA since 1992. His research includes work in formal generative syntax with a special emphasis on how to account for typological variation across languages. He has published research on various topics in syntax that include word order and scrambling, agreement and case, ergativity, partial wh-​movement, and relative clauses. Andrej Malchukov  is a senior researcher at the St.-​Petersburg Institute for Linguistic Research (Russian Academy of Sciences), currently affiliated as Visiting Professor to the University of Mainz. Apart from descriptive work on Siberian (in particular, Tungusic) languages, his main research interests lie in the domain of language typology. He published extensively on the issues of morphosyntactic typology; in particular, he edited The Oxford Handbook of Case (together with Andrew Spencer; 2009), Studies in Ditransitive Constructions: A Comparative Handbook (together with Bernard Comrie and Martin

 

xvi   Notes on Contributors Haspelmath; 2010); Competing Motivations in Grammar and Cognition (together with Brian MacWhinney and Edith Moravcsik; OUP, 2014), and Valency Classes in the World’s Languages (together with Bernard Comrie; 2015). Diane Massam  (PhD MIT 1985) is Professor in the Department of Linguistics at the University of Toronto. Her research areas are argument structure, case, predication, and word order, with a focus on the Niue language (Polynesian), and an interest in register variation in English. She has edited volumes on Austronesian syntax, ergativity, and the count–​mass distinction, and was co-​editor of Squibs for Linguistic Inquiry. William B. McGregor  is Professor of Linguistics in the Department of Aesthetics and Communication, Aarhus University, Denmark. His primary research interest is in the languages of the Kimberley region of Western Australia; more recently he has begun work on a Khoisan language of Botswana. He has published widely on these languages, including descriptive grammars, and has a long-​term interest in ergativity. Gereon Müller  is Professor of general linguistics at Leipzig University, and head of the graduate program “Interaction of Grammatical Building Blocks” (IGRA). He got his Dr. phil. from Tübingen University in 1993, and his Dr. habil. from the same university in 1996, both with works on theoretical syntax. His main research interest is grammatical theory, with a special focus on syntax and morphology. Léa Nash  PhD (1995), Paris 8 University, is Professor of Linguistics at that university. She has published many articles in theoretical syntax, especially on argument structure, case theory, and ergativity. Yuko Otsuka is Associate Professor in Linguistics at the University of Hawai‘i at Mānoa. Her research interests are theoretical syntax and Austronesian languages. She has worked extensively on Tongan (Polynesian) with special focus on case and ergativity. Tyler Peterson  received his PhD from the University of British Columbia in 2010 and joined the University of Auckland in 2016. He has undertaken extensive fieldwork on the endangered indigenous language Gitksan (Tsimshianic, British Columbia), and has also worked with indigenous languages in the southwestern United States, the South Pacific, and the Brazilian Amazon. Barbara Pfeiler  is Professor of Linguistics at the National Autonomous University of Mexico in Mérida, Yucatán. She has conducted fieldwork on the acquisition of Yukatek and Teenek since 1997. She has published articles on the acquisition of these languages as well as on the sociolinguistics and dialectology of Yukatek. She edited the volume Learning Indigenous Languages: Child Language Acquisition in Mesoamerica (2007). Maria Polinsky  (PhD 1986) is Professor of Linguistics and Associate Director of the Language Science Center at the University of Maryland. Her main interests are in theoretical syntax, with an emphasis on cross-​linguistic variation. She is also interested in the

 

Notes on Contributors    xvii integration of experimental methodologies in linguistic research. She has done extensive work on ergative languages across several language families, namely, Austronesian, Kartvelian, Nakh-​Dagestanian, and Paleo-​Siberian. Omer Preminger  is Assistant Professor of Linguistics at the University of Maryland (College Park). He graduated from MIT in 2011, and was a postdoctoral associate at MIT and at Harvard. Before joining UMD, he was a faculty member at Syracuse University. Omer has worked on issues of agreement and case in a variety of languages, including Basque, Hebrew, Kaqchikel, and Sakha. His publications include Agreement and Its Failures (2014). Clifton Pye  is Associate Professor of Linguistics at the University of Kansas. His specialty is the documentation of the acquisition of Mayan languages with a special focus on the acquisition of K’iche’, Mam, Q’anjob’al, and Ch’ol. He has published numerous articles on the acquisition of phonology, morphology, syntax, and semantics in these languages. Francesc Queixalós  as a member of the Centre National de la Recherche Scientifique and, periodically, of the Institut de Recherche pour le Développement (France), Francesc Queixalós has conducted in-​depth research on two languages of lowland South America, Sikuani (Guahiban) spoken in the savanna area of the middle Orinoco, and Katukina-​Kanamari (Katukinan), spoken in the rain forest south of the middle Amazon, while addressing several issues in the morphosyntax of Tupi-​ Guarani languages. He has taught linguistics in several Universities in France and South America. Andrés Pablo Salanova is Associate Professor of Linguistics at the University of Ottawa. He has worked with the Mẽbengokre since 1996, totaling over one year in the field and writing on several different aspects of their language. Salanova holds a BA in mathematics from Brown University, an MA in linguistics from Campinas (Brazil), and a PhD in linguistics from MIT. Eva Schultze-​Berndt  is Professor of Linguistics at the University of Manchester, UK. She received an MA from the University of Cologne, and a PhD from the University of Nijmegen. Her research interests include complex predication, overt classification of verbs/​events, secondary predication, spatial language, parts of speech, information structure, language contact, and documentary linguistics. Michelle Sheehan is a senior lecturer at Anglia Ruskin University, specializing in comparative syntax. She has worked on null arguments, Control, word order variation, clausal–​nominal parallels, and case/​alignment. She is co-​author of Parametric Variation:  Null Subjects in Minimalist Theory (2010); The Philosophy of Universal Grammar (OUP, 2013); The Final-​over-​Final Constraint (forthcoming). Michelle is co-​ editor of Theoretical Approaches to Disharmonic Word Order (OUP, 2013) and of the journal IBERIA.

 

xviii   Notes on Contributors Tarald Taraldsen (PhD University of Tromsø, 1983)  has worked as Professor of Linguistics at the University of Tromsø since 1984. He was a senior researcher at the Center for Advanced Study in Theoretical Linguistics (2002–​12). He has worked on the syntax of Scandinavian and Romance languages and, since 2008, also on the syntax and morphology of Bantu languages. Daniela Thomas  gained her MA from Leipzig University in 2015, with a thesis on a harmonic grammar approach to scale effects in argument encoding that combines minimalist syntax with weighted constraints in post-​syntactic morphology. Her bachelor thesis from 2013 tackles split ergativity in subordinate contexts from a minimalist perspective. Lisa deMena Travis  completed her PhD in Linguistics at MIT in 1984, writing on parameters of word order variation.  She is currently Professor in the Department of Linguistics at McGill University where she has been teaching since 1984. Her research focuses mainly on phrase structure, head movement, language typology, Austronesian languages (in particular, Malagasy and Tagalog), and the interface between syntax and phonology. Recent publications include Inner Aspect: The Articulation of VP (Springer, 2010). Kevin Tuite  is Professor of Anthropology at the Université de Montréal. He directed the Caucasus Studies program at the Friedrich-​Schiller-​Universität Jena (2010–​14). Since 1985, he has been researching the languages and cultures of the Caucasus, with a focus on Georgia. He is presently working on a grammar of the Svan language, and a study of the cult of St George in the Caucasus. Coppe van Urk  received his PhD in linguistics from MIT in 2015 and is a lecturer at Queen Mary University of London. His research focuses on syntax, with specific interests in movement, case, agreement, and the structure of Dinka. Martina Wiltschko  is Professor of Linguistics at the University of British Columbia (Vancouver). She is interested in the range and limits of variation in the domain of syntax and its interfaces (syntax–​morphology; syntax–​semantics; syntax–​pragmatics). She has extensively published on several empirical domains pertaining to this question including her recent monograph The Structure of Universal Categories: Towards a Formal Typology. Ellen Woolford  received a BA from Rice University and a PhD from Duke University, with a dissertation on Tok Pisin based on fieldwork in Papua New Guinea. She is currently a professor of linguistics at the University of Massachusetts. She has published on a range of topics in syntax, including locality constraints on wh movement and the behavior of passives in double-​object constructions. Her recent work focuses on the theory and typology of case and agreement.

 

Notes on Contributors    xix Adam Zawiszewski  graduated in Romance philology (Adam Mickiewicz University), obtained a Master’s in Linguistics from the University of the Basque Country (UPV/​ EHU) where he also defended his PhD. He completed a postdoctoral training at the Max-​Planck Institute for Human Sciences and Cognition (Leipzig). In 2011 he was awarded a Juan de la Cierva Fellowship by the Spanish Ministry of Science and Innovation. Currently he is working as Assistant Professor at the University of the Basque Country (UPV/​EHU).

 

 

Chapter 1

Introdu c t i on Jessica Coon, Diane Massam, and Lisa deMena Travis

1.1 Introduction 1.1.1 Introducing Ergativity This volume tackles the phenomenon known as ergativity. The term “ergativity” has been most commonly used to refer to systems with one or both of the following properties: (i) transitive subject (A arguments in (1)) pattern differently from intransitive subjects (S arguments) and from transitive objects (P arguments); and (ii) transitive objects and intransitive subjects pattern alike (see Figure 1.1). In such a system, schematized in Figure 1.1(a), the A argument is referred to as the “ergative” argument, and the S and P arguments are the “absolutive” arguments. This type of system contrasts with a “nominative–​accusative” systems, shown in Figure 1.1(b). Just as there is more than one way to be “ergative,” it is important to point out that “ergativity” may refer to any characteristic which aligns arguments as in Figure 1.1(a)— this includes not only the more common morphological case marking and agreement, but also word order, discourse and information structure, or the extractability of arguments. A wide range of work across different traditions converges on the idea that “ergativity” is not a single unitary phenomenon, and is not realized in the same way across different languages. Dixon (1994: 219), for example suggests that “there is no necessary connection between ergative characteristics and any other linguistic feature,” and Johns (2000: 67) writes in a similar vein that there may be “little value in studying ergativity as a thing in itself.” In her recent survey of ergativity, Deal (2015b) suggests that “ergativity is not one but many phenomena.” Nonetheless, certain patterns and correlations emerge, suggesting that while there is certainly diversity, there is also some unity—perhaps motivating the existence of this additional volume on the topic. The general themes of unity and diversity in and among ergative systems are touched on in the chapters that follow.

 

2    Jessica Coon, Diane Massam, Lisa Travis (b) NOMINATIVE–ACCUSATIVE

(a) ERGATIVE–ABSOLUTIVE

transitive:

intransitive:

A

P

A

P

ERG

ABS

NOM

ACC

S ABS

S NOM

Figure 1.1  Alignment patterns

Before discussing the organization and content of the volume, a few disclaimers are in order. First, in this introduction, we do not attempt to provide a comprehensive survey of the wide range of existing work on ergativity. We refer readers to works such as Johns (2000), Aldridge (2008a), and Deal (2015) for overviews in the generative tradition; Dixon (1979, 1994), McGregor (2009), and Comrie (2013a, 2013b) for more typologically oriented overviews; and Coon and Adar (2013) for an annotated bibliography of work on this topic. Rather, we aim to highlight some of the topics that have emerged over the years which we feel reappear throughout this volume, and point to some common themes. Second, due to the broad reach of the term “ergative,” ergativity may be viewed from a number of different perspectives. We have done our best to include chapters representing a range of different theoretical and methodological traditions, though, as with any volume, imbalances reflecting the orientations of the editors nonetheless exist. Finally, many of the chapters contained here would fit naturally into more than one of the book’s four parts, and the reader should interpret these divisions as loose guidelines rather than a strict packaging.

1.1.2 Themes and Organization The volume is organized into four main parts. Part I, on accounting for ergativity, focuses on factors which distinguish ergative from non-​ergative systems and how these may be parameterized and formalized in the grammar. In Part II, common as well as less-​common characteristics and manifestations of ergative systems are discussed. Topics here include alignment splits, antipassive constructions, and word order correlations, as well as nominalization, voice systems, and connections to speech acts and information structure. Part III focuses on approaches which draw on data from a diverse range of methodologies; these chapters focus on ergativity through the lens of diachronic, experimental, and acquisition research. Finally, Part IV turns to case studies—in-​depth looks at ergativity and ergative phenomena in particular languages or language families.

 

Introduction   3 Throughout the four parts of this volume, several themes emerge. One such theme is the impressive diversity of languages which exhibit ergativity—languages from nearly every continent and an impressive number of language families are represented—as well as the wide range of phenomena that have been associated with the label “ergative.” In addition to diversity in the geographic and empirical landscape, the contributions to this volume also reflect the range of different analyses, views, and theoretical approaches of how to interpret these facts. Relatedly, it becomes clear that ergativity is not as fully isolated a phenomenon as it is sometimes made out to be. Some characteristics that have been argued to hold of ergativity do not hold in a uniform way, such as the existence of split systems, antipassives, or extraction restrictions, or continue to resist full explanation, such as correlations between word order and ergativity. As has been frequently noted, not only do we find non-​ergative patterns throughout languages traditionally labeled “ergative,” we also find ergative patterns in a number of language and domains normally considered “nominative–​accusative.” To pick just a couple of illustrative examples, this volume includes a number of chapters on split ergativity (see especially sections 1.2.2 and 1.3.1) that demonstrate not only the complexity of defining ergative splits and differentiating them from other types of differential argument marking systems, but also disagreement about how they should be formally represented: as simply morphophonological rather than syntactic; connected to something specific to the syntax of ergative languages; driven by competing functional discourse pressures; or even that they might just be the natural fallout of other structural properties, and hence not be a hallmark of ergative languages at all. In addition, the antipassive, once commonly thought to be exclusive to ergative languages, is argued to be found across other types of languages as well (see sections 1.3.1 and 1.5). The derivational origins and limits of ergativity are similarly unclear, being possibly based on information structure, or perhaps related to voice systems and nominalizations, and with a possible reach to other domains such as speech act structure (sections 1.2.1 and 1.4). Despite the diversity reflected here, a number of points of commonality or areas of agreement emerge. A look at the contributions in Part IV drives home the point that simply labeling a language as “ergative” or “accusative” is not enough. Many contributions here highlight the importance of careful, holistic investigations into individual languages. Just as a given language must be examined carefully, it may be examined from more than one angle. In this volume we see the benefits of increasing the diversity of approaches to the study of ergativity (see section 1.4), as well as an increase in cross-​collaboration in various disciplines—through studies of acquisition (Bavin; Austin; Pye and Pfeiler), experimental work (Longenbaugh and Polinsky; Zawiszewski), diachronic analyses (McGregor; Haig; Aldridge; Butt and Deo; Kikusawa; Kaufman), or through discourse and speech act structure (Wiltschko; Du Bois). The implications discussed in the chapters in this volume are similarly far-​ reaching, with consequences for the representation of case and agreement systems more generally, for argument structure, and the role of constraints in the grammar, to name just a few.

 

4    Jessica Coon, Diane Massam, Lisa Travis Another striking point of commonality across many of the formal accounts of ergativity presented here is that ergativity or ergative assignment occurs low in the structure. Whether this is formalized as a low, in situ licensing of ergative subjects (see Legate; Sheehan; Laka; Müller and Thomas; Woolford; Aldridge), or as the result of ergative being assigned configurationally to the higher of two nominals in some domain, perhaps by virtue of the ergative argument remaining low (Baker and Bobaljik; Baker; Nash; Coon and Preminger), the relevance of the structural height of subjects is discussed throughout many of the chapters in this volume. However it is formalized, the proposed “lowness” of ergative subjects may in turn have consequences elsewhere in the grammar, for example in word order (Taraldsen), or for connections between nominalization and ergativity (Alexiadou; Kaufman). We do not explore further the many threads of research in this volume, but will let the contributions—which we go on to introduce briefly in turn—speak for themselves. Far from being the last word on ergativity, we anticipate that this volume will serve to spark further interest and study of this topic, which we hope to have demonstrated has implications for linguists working in any discipline or subfield of linguistics.

1.2  Part I: Accounting for Ergativity The chapters in Part I share a common goal: to understand and model how ergativity arises either in a specific language, or cross-​linguistically. Part I has two sections: in the first of these, on representing ergativity, Du Bois, Sheehan, and Mahajan each discuss the “parameterization” of ergativity. Du Bois’ chapter focuses on functional motivations for ergativity, and the competing pressures which might result in ergative or accusative grammatical systems. Sheehan seeks to capture not only differences between ergative and non-​ergative languages, but also differences among what she identifies as subtypes of ergative languages, with a parameter hierarchy. Mahajan narrows in to discuss differences in how “absolutive” arguments are represented, with a focus on Hindi. Though they have different scopes and approaches, the chapters share a theme found throughout the volume and in other work, namely that languages may manifest ergativity in different ways and perhaps to different degrees. The chapters in the second section, on the nature of the ergative case, tackle a specific question in the formal representation of ergativity: What is special about ergative arguments? Two main approaches are presented, labeled in Baker and Bobaljik’s contribution as the “Inherent Case Theory” (ICT) and the “Dependent Case Theory” (DCT). In the former, ergative case is assigned to an external argument in its base position (e.g. specifier of vP). In the DCT approach, ergative case is assigned configurationally; it is not tied to a specific functional head, but rather is assigned to the higher of two nominals in some specified domain. Baker and Bobaljik introduce both options, presenting evidence in favor of DCT. In her contribution, Legate presents an overview of the behavior of ergative marking in a wide range of different languages.

 

Introduction   5 Despite the range of variation, she argues that what they share in common is a low source or ergative case, registering concerns for a DCT approach. Both Laka and Nash tackle the question of ergative case assignment and splits in specific languages— Basque and Georgian, respectively—coming down on different sides of the debate. Laka argues in favor of a consistently low locus of ergative case in Basque, providing a detailed analysis of the verb behar (‘need’). Nash, on the other hand, argues that the non-​ergative alignment in Georgian arises when the subject is outside of the vP domain, which she formalizes in terms of dependent case. We summarize each of the volume’s chapters in more detail in the following sections.

1.2.1 Representing Ergativity In Chapter 2, Du Bois discusses the relationship between ergativity and an “ergative discourse profile.” Du Bois argues that the presence of ergative alignment in discourse— specifically, an ergative alignment in terms of which roles arguments play when they are introduced and tracked throughout utterances—reveals motivations for grammaticalized patterns of ergativity cross-​linguistically. A  conflicting universal discourse–​ pragmatic pressure, namely for “topicality,” gives rise to accusativity. In this chapter, the ergative discourse profile is examined through the lens of typology, language acquisition, and language change. Additional functional factors which contribute to the grammaticalization of ergativity including verb semantics, aspects, and inherited morphosyntax, are also discussed. Chapter  3 by Sheehan outlines a parameter hierarchy to capture variation in alignment systems. The first parameter determines the presence of ergative vs. non-​ ergative alignment based on whether or not v0 is able to assign inherent ergative case. Further micro-​parameters within the ergative setting determine (i) the full distribution of ergative case (i.e. whether there are splits or active alignment); (ii) the presence or absence of extraction restrictions on ergative subjects; and (iii) the source of absolutive case in transitive contexts. Sheehan’s contribution allows for variation within ergative systems, while still restricting the range of possible alignment systems. She discusses how the rankings between parameters connect to the need to create convergent derivations. Mahajan (Chapter 4) tackles the mechanism by which direct objects are licensed in Hindi. Through the examination of the syntax of perfective and imperfective prenominal relative clauses, Mahajan argues that morphologically bare (“absolutive”) direct objects in Hindi are licensed by T. Specifically, Mahajan proposes that the restrictions on which arguments can be relativized in prenominal relatives provide evidence for how case licensing works in participial clauses; this in turn offers a window into licensing mechanisms in ergative constructions. These results contrast with recent work (e.g. Legate 2008 and others) which has argued that transitive objects in Hindi are licensed low by v0. Differential Object Marking is also discussed, and argued to not be a substitute for structural case licensing.

 

6    Jessica Coon, Diane Massam, Lisa Travis

1.2.2  The Nature of the Ergative Case Baker and Bobaljik (Chapter  5) discuss two approaches to how the ergative case is assigned in the grammar, labeled Inherent Case Theory (ICT) and Dependent Case Theory (DCT). In the former, ergative case is an inherent case assigned to the subject by v0, while in the latter, ergative is a dependent case assigned configurationally to the higher of two arguments in some local domain. Baker and Bobaljik discuss the predictions of the two accounts and argue in favor of DCT through an examination of languages such as Shipibo, Kalaallisut, and Chukchi. As evidence against the ICT, they present constructions in which non-​agents bear ergative case, and in which agents fail to receive ergative case. They also discuss the absence of active patterns of morphological case marking, argued to be predicted on the ICT. Legate argues in Chapter 6 that while the ergative case is not determined by a single factor cross-​linguistically, ergative-​assignment is governed by a consistent constellation of factors which share the property of occurring low in the clause, centered around vP. The factors Legate identifies include: theta-​position and theta-​role of the subject, the presence of a complement, the presence of a DP object, the theta-​role of the object, the case of the object, the presence of object agreement, the lexical predicate, the light verb, and the aspectual head which selects vP. A wide range of languages are discussed, including two for which ergative initially seems to have a higher locus (TP or CP): Kurmanji Kurdish and Yukulta. Legate concludes that even here, ergative is assigned low and that high ergative languages may not exist. Laka examines ergativity in Basque in Chapter 7. In particular, she discusses what she calls the “TotalErg” hypothesis: the hypothesis that ergative is an inherent case, and that ergativity does not actually split. Apparent splits, under this account, are epiphenomenal, resulting from different structures rather than from different case-​assignment properties of functional heads. She examines the Basque predicate behar (‘need’), which shows a split in the assignment of case to subjects and has been recently argued to provide evidence in favor of structural assignment of ergative by T (Rezac, Albizu, and Etxepare 2014). Laka argues instead that predicates like behar are nominals, and not raising modals. She concludes that there is no raising-​to-​ergative in Basque, and that ergative case is uniformly assigned by transitive v0. Nash’s contribution (Chapter 8), on the structural source of split ergativity and ergative case in Georgian, examines split ergativity in Georgian in order to understand the difference between ergative and nominative behavior within a single language, with cross-​linguistic implications. Specifically, Nash argues that nominative alignment arises when the transitive subject is case-​licensed in a position outside of vP. In an ergative system, on the other hand, the transitive subject is licensed inside of vP. Nash formalizes this variation in terms of a configurational approach to case assignment, in which the low subject is local enough to the object to receive dependent ergative case. Nash attributes the differences in licensing patterns to the presence or absence of an aspectual category, “Event,” which has both semantic and syntactic consequences.

 

Introduction   7

1.3  Part II: Characteristics and Extensions Part II has two sections. The first section includes chapters that examine some of the key grammatical characteristics that are commonly considered to be correlated with ergativity, and the second includes chapters that extend our notion of ergativity in one of two ways. Some chapters extend our view of ergativity empirically, by looking at phenomena that are usually considered to lie outside ergativity but which have been argued to be related to it, while other chapters extend the usual discussion of ergativity theoretically, by tying ergativity to theories of speech acts and information structure.

1.3.1 Characteristics Languages with ergative systems are said to exhibit properties and constructions that are characteristic of ergativity. In this first section of Part II, some of these characteristics are explored. One common claim about languages with ergative case systems is that they are never uniformly ergative—​rather, they always exhibit other case patterns as well. This property is referred to as split ergativity, in which the ergative pattern is lost in certain contexts, often in non-​perfective aspects or in contexts with “highly ranked” (e.g. first-​and second-​ person) subjects. Several chapters in this section explore split ergativity, as we now outline. In Chapter  9, Woolford focuses on types of split ergative languages, providing an overview of conflicting definitions in the literature. She argues that a consistent definition is important in evaluating claims about whether all ergative languages exhibit splits. She discusses familiar triggering factors such as person and aspect (e.g. Marathi, Chol) and lesser-​known triggers such as stage or individual predicates (Nepali) and social conventions (Folopa, Mongsen Ao). She includes languages where ergativity depends on object properties (e.g. Niuean), and she also examines languages with “active” alignments, arguing that while some are split (Choctaw), others are fully ergative (Laz), since all verbs that can license ergative case do so in all contexts. In their contribution, Coon and Preminger (Chapter 10) examine both aspect-​based splits (as in Basque and Samoan) and person-​based splits (as in Sakha and Dyirbal), arguing that split ergativity is epiphenomenal, and that it is not in fact limited to ergative languages. They consider that case splits are due to structural factors, with the non-​ ergative pattern arising as the result of a bifurcation of the clause, so that the clause becomes intransitive, hence straightforwardly not ergative. They argue that bifurcation can be the result of non-​perfective (i.e. locative) syntax, or of first-​and second-​person licensing requirements, thus accounting for the universal directionality of the splits. They conjecture that because all subjects pattern similarly in nominative languages, such splits are not as apparent as they are in ergative languages.

 

8    Jessica Coon, Diane Massam, Lisa Travis Malchukov (Chapter 11) takes a functional–​typological approach to splits, examining Differential Object Marking (DOM), which is widely attested, and differential subject marking (DSM), which is less discussed, and is found mainly in ergative languages. He shows that while DOM can be uniformly explained via markedness, DSM cannot be so explained, as many patterns mark subjects that are higher on the hierarchy (e.g. Hindi). He explores two views about case: indexing and distinguishability. Though these are sometimes taken to be in conflict, he argues that both are needed, as together they can account for the varying patterns of DSM and other case patterns. He presents an OT (Optimality Theory) analysis, showing that two unranked constraints, DIFF and INDEX, can converge or not, allowing for the existence of different patterns. Müller and Thomas (Chapter 12) discuss three-​way systems, arguing that such systems do not exist syntactically, but diverge from two-​way systems through scale-​driven optimization operations at the syntax–​morphology interface. They argue that such languages are actually either ergative or accusative, with case markers that disappear in certain contexts because of morphological processes. Through examination of a range of languages such as Kham, Djapu, Nez Perce, Upriver Halkomelem, and Dyribal, they propose adding a transitive scale to the standard definiteness, animacy, and person scales, which are also usually active in these systems. This allows the successful Principles and Parameters approach to syntactic case assignment to remain just as it is for nominative, ergative, and for three-​way systems. Ergative languages are often said to exhibit a particular construction, sometimes considered to be the ergative version of the passive construction, known as the antipassive. Polinsky, in her contribution (Chapter 13), examines this construction across languages (e.g. Chukchi, Diyari, Labrador Inuit, Warlpiri), summarizing its properties and key approaches. She defines it as a construction where the logical object of a predicate is not realized as a direct object but as a non-​core argument or is left unexpressed. She demonstrates various realizations including some less typical, such as (pseudo) noun incorporation and bi-​absolutive constructions. She argues that there are interpretative effects, but that none is a defining factor across all antipassives cross-​linguistically. She shows that antipassive and passive constructions are not mutually exclusive, and that antipassive is not limited to ergative languages, though it is more noticeable in such languages. Another proposed characteristic property of ergativity is word order; in particular, it has been claimed that SVO order and ergativity do not coincide. Taraldsen (Chapter 14) examines this generalization, demonstrating that SVO can be derived in a multitude of ways, as can ergative case marking. He questions whether the generalization holds of all these possible derivations, and argues that we would expect counterexamples, hopefully within certain types of well-​defined languages. He examines tripartite and neutral languages with ergative agreement patterns but he finds that no conclusion can be drawn, due to lack of data. The chapter also examines key proposals about ergativity, pointing out necessary modifications in order to account for the word order restriction. The chapter richly illustrates the complexity involved in developing detailed analyses of broad generalizations.

 

Introduction   9

1.3.2 Extensions In the second section of Part II, our familiar view of ergativity as a sentential argument indexing system is extended to allow for consideration of the role of ergativity in other domains such as nominalizations, voice systems, information structure, and speech act theory. The authors of the four chapters in this section take different positions on how and whether these extensions can be posited. Alexiadou argues that ergativity is linked to nominalization, and Erlewine, Levin, and van Urk argue against a currently widespread view that Austronesian-​style voice systems are an expression of ergativity. Johns and Kučerová argue that ergative patterns stem from structural properties of information structure, while Wiltschko argues that at the thematic level, ergative patterns are basic, and that they can be extended into speech act theory. In the following paragraphs, we outline each of these chapters. It has been noted that ergativity is related to nominality, both because nominalizations often exhibit ergative case, and because verbs in ergative languages seem to exhibit fewer verbal properties than verbs in nominative systems. Alexiadou explores these issues in Chapter 15, noting that many authors have attributed ergativity to the presence of a defective v or Voice head, which yields a more nominal clause. She observes that ergativity only arises in a subset of nominalizations in languages that have more than one nominalization pattern, and that these are cases that contain an n head. She argues that n-​based nominalizations allow only one structural case and do not contain an external argument. She also includes a discussion on the nature of unergative subjects. In her chapter, ergativity is related to characteristics that enable it to extend to other construction types. In Chapter 16, Erlewine, Levin, and van Urk examine recent extensions of ergativity to Austronesian voice system languages. They begin by reviewing and critiquing ergative analyses of voice systems as in Tagalog, Malagasy, and Atayalic languages, and they bring in new data from Balinese and (non-​Austronesian) Dinka. These languages have similar voice systems to the other languages, but they do not exhibit ergativity, thus they demonstrate the necessary dissociation of these two phenomena. They argue that there must be mechanisms other than ergativity that yield the behavior of Austronesian-​style voice systems. Their chapter thus suggests that there are limits on extensions of ergativity to explain other grammatical phenomena. Johns and Kučerová show in Chapter 17 that there is variability in the presence of object agreement in the ergative–​antipassive alternation in the Inuit language. They argue that this is related to information structure, and, given this, the case and agreement patterns fall out from familiar principles. They propose that absolutive object “agreement” is in fact cliticization, and that such cliticization is tied to the fact that absolutive objects are always “aboutness” topics. Such topics must be at the edge of a phase in order to be assigned a referential address, and this affects the locality relations of the arguments, yielding an ergative pattern. They also touch on dialect variation across the Inuit languages. Their chapter thus raises a new perspective on the nature of case splits.

 

10    Jessica Coon, Diane Massam, Lisa Travis Wiltschko argues in her programmatic contribution (Chapter  18) that ergativity extends beyond the familiar argument structure domain into the domain of speech act structure. She argues for the existence of this domain and shows that, as with argument structure and case structure, ergativity is a possible and indeed, expected pattern at this level. She overviews proposals about speech act structure, touching on assertions, imperatives, and presentatives, and on particles such as eh, and German jo. She argues that speech act structure consists of a grounding layer and a response system layer, and that ergative constellations can be detected in each of these.

1.4  Part III: Approaches to Ergativity The chapters in Part III show how diachronic, acquisition, and experimental work can probe data points and theoretical questions in ways that can both complement and support the work reported on in other parts of this volume—​and is divided into three sections, accordingly. Recurring themes in these chapters involve issues such as the amount of variation that is found in the instantiation of ergativity and the possibility of reanalysis and/​or grammaticalization of structure. There is the basic question of how closely-related languages can come to have very different grammatical systems, thereby raising further questions concerning how languages change, what the influence of language contact is, and what parts of language are susceptible to reanalysis. Diachronic work takes these puzzles as the starting point. But questions of language change and reanalysis lead to questions about acquisition. Acquisitionists explore what might be subject to reanalysis, what might be a default setting for a parameter, what triggers are salient, and what structures are learnable. Experimental work outside the domain of acquisition looks at related areas where similar questions are investigated, such as what systems are more easily processed, and what elements in the linguistic string aid intelligibility. Specifically, in the context of ergativity, we can ask whether ergative systems are stable or are prone to reanalysis; how one arrives at an ergative system; what the paths out of such a system are; and whether there is any evidence that an ergative system is either more or less complex than a nominative system.

1.4.1 Diachronic A striking characteristic of ergativity is how differently it may present itself from language to language. This is particularly noticeable where microvariation appears within language families. In this section, six chapters tackle the problem of variation by investigating paths of change. Several different types of focus are evident in these chapters— variation vs. commonality and description vs. theory. As more and more details about variation are uncovered (see also Part IV, the case studies, which we discuss shortly), the

 

Introduction   11 puzzle of what we mean by ergativity and what a theoretical account for ergativity might be becomes more complex. Some of these chapters stress the fact that the paths to ergativity are more varied than previously thought (McGregor, Haig), others try to reduce the number of possible paths (e.g. Aldridge). Without getting down to the details of specific mechanisms, a larger question can be raised as to whether change to an ergative system can be tied to a shift of one language-​particular characteristic, or whether a general flavor of ergativity is constructed by coinciding but logically independent changes. It is no surprise that these issues appear in Parts I and II of this volume as well when discussing what the parameters of ergativity are and how the particular characteristics of ergativity are accounted for. McGregor (Chapter  19) traces the creation and loss of ergative case morphemes cross-​linguistically, arguing that the range of sources for this case marker is wider than what is often assumed. He outlines and evaluates various proposals in the literature, critiquing the viability of lexical sources, but giving multiple examples where markers of other cases, indexical items, and directional elements have been reanalyzed to produce ergative case morphemes. He also discusses instances where ergative case markers themselves are reanalyzed as other case markers or grammatical categories. In the final section of the chapter he discusses the role of language contact in the development of ergative case markers and ergative systems. Haig focuses on ergativity in Iranian languages in Chapter 20. He introduces three case studies, Kurdish, Balochi, and Taleshi, to illustrate the extent of the micro variation of ergativity within Iranian languages. He focuses on the path of the emergence of these systems, supporting the claim that this micro-​variation stems from independent changes in interrelated subsystems such as case, agreement, and pronominal clitic systems. These findings result in raising doubts for any proposal that ergativity is best represented by a monolithic alignment parameter. Aldridge (Chapter 21) takes a different tack from the previous two chapters, emphasizing what characteristics paths to ergativity might share. Couched in a generative syntax framework, she explores data from several languages and language families, e.g. Indo-​Aryan, Iranian, Inuit, and Austronesian, arguing that ergativity is derived from nominal structures. She examines two cases (Indo-​Aryan and Inuit) where it has been claimed that the ergative structures came from earlier passives but advances an argument that even in these cases, the ergative was originally a possessive, supporting the hypothesis of a nominal base. The next three chapters focus on two language families that have been introduced in Aldridge’s chapter, but offer different viewpoints. Butt and Deo (Chapter 22) take a close look at four stages in the rise and fall of ergativity from Early Old Indo-​Aryan to New Indo-​Aryan, starting with the development of ergativity from participial constructions. Within New Indo-​Aryan, they describe three major innovative patterns. In Hindi and Nepali, ergative case marking is strengthened with new morphology, in Bengali and Oriya, both ergative case and ergative pattern agreement is lost, and in Marathi and Gujarati, ergative agreement remains in spite of complications, such as surface syncretism of morphology and differential subject case marking.

 

12    Jessica Coon, Diane Massam, Lisa Travis Kikusawa (Chapter 23) uses the Comparative Method to reconstruct the direction of change in various Austronesian languages and to explain the typological diversity found within this language family. More specifically, she outlines three paths of change: (i) the shift from a morphologically marked ergative system to a fixed word order voice system (Ibaloy, Pendau); (ii) the development of an accusative clitic system (Tongan, Samoan); and (iii) the development of a system of lexically marked NPs that can be analyzed as ergative (Tongan) or accusative (Maori). In Chapter 24, Kaufman focuses on Western Malayo-​Polynesian languages. He starts by comparing three theoretical accounts of Tagalog, posing problems for two of them—​ an ergative analysis and a case agreement analysis. He argues rather that Tagalog is a symmetric language where predicates are nominal rather than verbal. In order to support this account, he compares the structure of Tagalog to that of another Western Malayo-​Polynesian language, Mamuju, a canonically ergative language. Kaufman shows how ergative structures found in Mamuju, as distinct from the structures of Tagalog, are developed through the reappearance of verbal predicates.

1.4.2 Acquisition It is difficult to talk about language change without invoking questions about language acquisition. The next three chapters raise many of the relevant questions for the acquisition of an ergative language, the answers for which have an impact on how language change should be viewed. As is pointed out, ergative languages are often split systems creating complex input for the language learner, which makes this a particularly interesting and informative field of study. The problems addressed include issues of methodology, the status default cases, and the use of acquisition data to support theoretical claims. Bavin (Chapter 25) outlines various issues that arise in the study of the development of an ergative system in child language. By summarizing studies from the literature representing a range of languages and language families, she highlights several possible confounds in the input data that could create problems for acquisition. These include split systems, multiple uses for the same case marker, and the contribution of pragmatic function to the choice of construction. She also discusses potential hurdles such studies face, such as drawbacks in using naturalistic data and potential ambiguities in the acquisition data. In spite of this, the cross-​linguistic data show similar results of timely successful acquisition with very little overgeneralization of ergative case marking. Austin (Chapter 26) presents data from previous studies on the acquisition of the verbal morphology and case in Basque. She shows that children resort to a default morphological system where forms that encode fewer features substitute for more complex forms, for example absolutive case is produced rather than ergative or dative. Austin argues that this repair strategy is not surprising given a Distributive Morphology analysis of Basque morphology. In a morphological system where morphemes compete to

 

Introduction   13 realize a set of features, the notion of “best fit” will ensure that a less marked morpheme will appear in instances where the more complex form has not yet been acquired. Pye and Pfeiler (Chapter 27) use acquisition data to probe the status of person marking in Mayan languages by comparing the acquisition of nominative person markers in French (clitics) and Spanish (agreement) with the acquisition of both absolutive and ergative person marking in four Mayan languages: Wastek, Yukatek, Ch’ol, and K’iche’. The acquisition of the ergative person marking in all four Mayan languages followed neither the French nor the Spanish pattern; while the acquisition of the absolutive person marking produced mixed results. There were insufficient Ch’ol data, but the acquisition patterns of Yukatek absolutive personal marking were similar to those of Spanish agreement. Those of Wastek and K’iche’, however, differed from both those of French clitics and Spanish agreement markers. They argue that the grammaticalization of person markers as determined by their specific combination of clitic and affix properties predicts children’s production of the person markers more accurately than their categorical status as absolutive or ergative.

1.4.3 Experimental Experimental work on ergative languages is relatively new but clearly very important to our understanding of ergativity as a typological language category and as a theoretical construct. The existence of ergative languages raises questions concerning markedness, ease of acquisition, and ease of processing. Whatever the answers to these questions may be, it is clear that no universal pronouncements about language can be made without including data from ergative languages. Both chapters in this section give overviews of experimental research on ergative languages. They discuss the methodologies used, the importance of the work, and some ideas for future research. Zawiszewski (Chapter 28) presents an overview of current experimental studies on ergativity using a variety of methodologies (self-​paced reading, ERP, fMRI, grammaticality judgments) on a variety of languages (Basque, Hindi, Avar). After an introduction of the different experimental methods, he summarizes the studies and shows how they can be used to further probe the results from earlier experiments on nominative–​accusative languages investigating issues such as the distinction of syntactic vs. semantic processing, subject–​object asymmetries, and the effect of L1 and L2 acquisition. He concludes with a discussion of the overall results and directions to be explored further. Chapter 29 is a review by Longenbaugh and Polinsky of recent experimental work testing ergative-specific questions involving alignment, long-​distance relations, and agreement in a range of languages including Hindi, Basque, Niuean, and Avar. They stress the importance of doing experimental work on ergative languages to resolve some confounds that are found in the existing literature on accusative languages. More specifically, the alignment of grammatical case with grammatical function can be teased apart in ergative languages. Their chapter ends with a suggestion that further experimentation can probe the heterogeneous nature of ergative languages.

 

14    Jessica Coon, Diane Massam, Lisa Travis

1.5  Part IV: Case Studies This section of the book contains sixteen case studies of a range of languages from a range of language families. While the approaches of the authors and the scope of the studies vary considerably, the common threads that have appeared throughout the book reappear here. On one hand, ergative languages vary from one another to such an extent that one might suspect that they have no unifying feature, yet they also evidence enough recurring features to confirm their membership in the class. There are SOV, VOS, VSO, and SVO exemplars. Both morphological and syntactic ergativity are explored, and a variety of types of ergative systems are outlined in families such as Nakh–​Daghestanian, Tibeto-​Burman, and Kartvelian, as well as some apparently emergent systems in African languages. Several languages that are discussed exhibit both ergative case and ergative agreement systems, leading to insights about the relation between case and agreement, and to the relation between ergativity and features such as animacy, gender, number, and person, as well as tense and aspect. Most of the languages explored have split systems—some sensitive to aspect, some to person, while some are Split-​S systems, in which different types of intransitive verbs are marked differently for case. The relation between ergativity and other constructions is also explored, focusing on constructions such as the antipassive and control structures, relative clauses, coordination, and non-​ finite sentences and nominalizations. Aissen (Chapter  30) examines ergative characteristics of Mayan languages, with a focus on constraints on extraction. Like some of the other languages discussed in this volume, some (but not all) Mayan languages restrict the extraction of ergative subjects, which Aissen calls the Ergative Extraction Constraint (EEC). In this contribution, Aissen reviews the empirical facts and discusses two main approaches to the EEC in the recent literature on Mayan languages: (i) a Case-​based approach, in which restrictions are attributed to abstract Case assignment configurations; and (ii) a morphosyntactic approach which attributes extraction asymmetries to special morphology, in particular the “Agent Focus” morphology used when transitive subjects are extracted. Through a detailed look at a range of constructions, Baker (Chapter 31) provides a dependent case analysis of ergative case in Burushaski, a language of Northern Pakistan. To understand the distribution of ergative marking, Baker investigates three environments in which the canonical ergative pattern of the language disappears: (i) verbs with two absolutive arguments; (ii) verbs with an ergative argument and a dative argument; and (iii) future-​tense clauses which permit absolutive transitive subjects. Baker argues that the syntax of each of these constructions is more complex than surface appearances show, lending support to the proposal that ergative case is assigned only when one NP (the ergative) c-​commands another NP in the same local domain. Berro and Etxepare (Chapter  32) on ergativity in Basque, explore an ergative system that is manifested by both case and agreement morphology. They provide a thorough and detailed overview of the case and agreement systems in Basque and of their

 

Introduction   15 interaction across both the nominal and verbal inflectional systems. They also demonstrate how ergative marking interacts with other systems such as number, person, and tense. They present a cross-​dialectal study of the marking of ergative case on subjects of unergative predicates, which has been referred to as Split-​S system, while critiquing some of the claims that have been made about this system, such as the positing of implicit objects and light verb structures. They discuss claims that have been made that ergative is an inherent case linked to causation, by considering a range of construction types, including nominal and adjectival predicates, perception verb complements and raising verbs. In their closing section they discuss the notion of “marked case” in relation to case marking in Basque. Butt (Chapter 33) gives an overview of ergativity in Hindi/​Urdu but crucially sets the Hindi/​Urdu facts against a background of other South Asian languages such as Nepali, Gujarati, Marathi, and Bengali. She highlights the range of variation and also details the different roles that differential case marking plays in each of the languages. Butt argues that the variation in the behavior of case and agreement in these languages, and the variation that role of differential case marking in agreement patterns, makes a tight link between case and agreement difficult to maintain. Compton (Chapter 34) focuses on how ergativity is realized both in morphological case marking as well as in the rich agreement system of the language. After reviewing basic characteristics of Inuktitut, Compton discusses the various approaches to ergative and absolutive case assignment in the literature. Finally, he turns to antipassive constructions and their relationship to Differential Object Marking and aspect. Forker (Chapter 35) surveys the range of ergative alignment patterns found in the Nakh–​Daghestanian (or East Caucasian) language family, concluding that the main correlates of ergativity in these languages are morphological. In particular, Forker discusses the system of gender and person agreement on verbs and the morphological case marking found on nominals. Biabsolutive constructions—in which both A and P arguments are marked absolutive—are reviewed, as well as valence-​changing operations (causative, antipassive). Forker also provides an in-​depth discussion of control constructions, noting that there is a general tendency for syntactic accusativity in this domain. Kahn (Chapter 36) focuses on ergativity in Neo-​Aramaic. He organizes and presents a complex set of patterns of ergativity in modern spoken form of Aramaic (Neo-​Aramaic) split into four subgroups: Western, Turoyo, Northeastern, and Mandaic. Khan discusses the nature of split ergativity evidenced in the patterns of verbal suffixes across a number of dialects, which are described in detail. He argues that the influence of Iranian languages on Eastern Aramaic explains both why Neo-​Aramaic differs from other Semitic languages in its expression of ergativity and the non-​canonical type of ergativity that it displays. While African languages are generally left out of any discussion of ergativity, König (Chapter 37), describes ergative patterns that appear in the West Nilotic family of Nilo-​ Saharan, in particular Shilluk. She points out four features particular to ergativity in African languages—​marked nominative, no-​case before the verb, OVA word order, and its relationship to pragmatically marked word order. She argues that ergative case

 

16    Jessica Coon, Diane Massam, Lisa Travis developed in these languages through reanalysis of either determiners, genitive markers, or prepositions. Chelliah (Chapter  38) surveys morphological case marking in several representative languages of the Tibeto-​Burman family:  Dolakha Newar, Chintang, Tibetan, Meitei, and Burmese, using these to demonstrate four possible alignment patterns for core arguments. The first two languages exemplify a typical ergative alignment pattern. In Tibetan, there is a general pattern of ergativity but one in which transitivity factors influence whether the transitive subject receives ergative marking. Burmese shows an accusative alignment in which information structure (topicality, contrastiveness, and theticity) influences when “subject marking” occurs. Finally, Chelliah discusses Meitei, demonstrating that it falls somewhere between Tibetan and Burmese insofar as both transitivity and information structure considerations affect the marking of core arguments. Laughren (Chapter 39) focuses on the ergative in Warlpiri, and examines what has been claimed to be a morphologically ergative case system in a syntactically nominative–​ accusative language. Laughren begins with an overview of ergativity in Australian languages, then focuses on Warlpiri, which has certain verbs which take ergative subjects in finite clauses and other verbs which take unmarked or absolutive subjects. This chapter examines the distribution of the ergative morpheme, including on body parts and circumstantial adjuncts, and the functions of the ergative DPs in both finite and non-​finite clauses, with a focus on the relation between subject marking and instrument marking. Otsuka (Chapter 40) demonstrates that Tongan has an ergative pattern in both morphology and syntax, but that this pattern is not consistent throughout the language, as nominal morphology is split between clitic pronouns and other nominals. There are three syntactic manifestations of ergativity in the language:  relativization strategies, coordinate reduction strategies, and anaphoric antecedence patterns. Interestingly, these cannot be accounted for in a unified manner, and Otsuka argues that the first two are in fact PF phenomena. She claims that this necessitates a view of ergativity as a construction-​specific rather than a language-​specific phenomenon. Peterson (Chapter 41) demonstrates that the Tsimshianic languages have fully ergative agreement systems. Although there are splits, conditioned by clause type and person hierarchy, all sides of the splits exhibit ergativity. He describes the agreement patterns across the family, including a discussion of connectives, which are determiner-​ like particles that appear to contribute a further split. The more conservative languages are purely ergative, while other branches also exhibit transitive, contrastive and neutral alignments. He considers all to be expansions of ergativity, since A and S are never grouped together. Queixalós (Chapter  42) presents a detailed examination of alignment, and grammatical relations more generally, in the Amazonian language Katukina-​Kanamari (KatKan)—a language which Queixalós describes as “remarkably suited for raising pivotal issues on grammatical relations.” KatKan is shown to have two patterns of bivalent clauses: ergative and accusative. The latter, Queixalós shows, is more highly restricted in its distribution, and is found with unindividuated patient arguments. Queixalós’

 

Introduction   17 contribution includes a thorough survey of the empirical facts surrounding the two types of construction, as well as more general discussion of the interactions among grammatical roles, argument structure, and alignment. Salanova (Chapter 43) describes the distribution of ergative case marking in Jê languages in general, and Mẽbengokre more specifically. In these languages, the link between the ergative constructions and nominalization is clear, where the subject DP is marked with a postposition when it occurs with the nominal/​adjectival form of the verb. Further, he shows that the use of the nominalized structure is pervasive, appearing not only in embedded contexts, but in independent clauses as well depending on other considerations including aspect and the presence of post-​verbal modifiers. Schultze-​Berndt (Chapter 44) tackles the problem of a system where ergative case-​ marking appears to be optional, alternating with zero-​marking, and, less frequently, ablative case. She describes several factors that influence the choice, factors which include animacy, verb class, tense/​aspect, and information structure. Schultze-​Berndt shows that factors that categorically determine morphological marking in other languages show up only as tendencies in Jaminjung, connecting it to differential case-​ marking of subjects. Finally, Tuite (Chapter 45) traces the history of linguistic accounts of Georgian ergativity from the seventeenth century. This history is followed by a detailed description of the different case and agreement patterns found in Georgian, as well as in Laz, Mingrelian, two members of the Zan branch of the family, and in Svan, an outlier. Tuite further outlines the role of verb classes in determining these patterns. Once the present variation has been established, an overview is given of case, agreement, verb classes, and morphosyntax of Proto-​Kartvelian.

Acknowledgments In concluding this introduction, the editors would like to acknowledge the valuable help of several people. We would like to thank Crystal Chen, Patrick Murphy, and Rebecca Tollan for editorial assistance, Justin Royer for help with proofing, and Hisako Noguchi for work on the index. We would also like to thank the editorial team at Oxford University Press, in particular Vicki Sunter and Julia Steer, who has provided invaluable help and advice throughout the process of producing this volume. Finally, we would like to thank all the authors for their work not only on their own contributions, but also in the reviewing process.

 

 

PA RT  I

AC C OU N T I N G F OR E RG AT I V I T Y

 

 

Representing Ergativity

 

 

Chapter 2

Ergativit y in di s c ou rse and gramma r John W. Du Bois

Every language provides its users with systematic ways of organizing the core arguments of the clause, establishing a more or less stable and consistent framework for the foundations of its grammar. Remarkably, languages differ even in this most basic level of structural organization. Yet certain configurations of arguments tend to recur, emerging again and again in the grammars of the world’s languages. For syntactic alignment what proves pivotal is how the grammar treats the sole argument of a one-​place predicate (S), aligning it with one or the other argument of a two-​place predicate, with respect to case-​ marking, agreement, word order, extraction, and so on. Some languages treat the S like the object of a two-​place predicate (O), yielding ergative alignment (S=O vs. A), while others treat S like the subject of a two-​place predicate (A), yielding accusative alignment (S=A vs. O). Still other languages are sensitive to the semantic variability inherent in the population of one-​place predications, aligning the more agent-​like subset of S (Sa ) with A, and the more patient-​like subset of S (So) with O, giving active alignment (Sa=A vs. So=O). Yet languages are not simply ergative, accusative, or active. There is great diversity, as well as convergence, across the world’s languages with respect to the various systematizations of basic grammatical relations (Silverstein 1976; Comrie 1978, 2013a, 2013b, 2013c; Dixon 1979, 1994; Mithun 1991; Malchukov 2005; Bickel & Nichols 2009; Bickel 2010; Siewierska 2013). This diversity can penetrate into the grammar of a single language, where a mix of distinct alignments is often found in different parts of the same grammar (Silverstein 1976; Comrie 1978; Mithun 1991; Coon 2010a, 2013a; Malchukov 2014; see also Coon and Preminger, Chapter 10 in this volume). Accounting for diversity and convergence within and across languages remains a compelling yet elusive task for linguistics (Evans & Levinson 2009). Argument structure configurations (Goldberg 1995) represent the very foundations of the clause—​what may be considered the “basic body plan” (Mayr 2001)  of the utterance. No theory of grammar can be considered explanatory without contributing to an understanding of how such a diversity of basic

 

24   John W. Du Bois plans could have emerged in the world’s languages. Yet the challenge of accounting for structural variability at a foundational level has proved baffling, such that many leading linguists have postponed the day of reckoning with ergativity. Fillmore considered his principles of subject selection universal, “given certain qualifications for the interpretation of ergative systems” (1977: 61). Dowty acknowledged that “argument selection in ergative languages” (1991: 581) was relevant to his proto-​role model, but invoked an “inverse” model of ergativity which “means in effect treating the transitive ‘Patient’ as a grammatical subject and the transitive ‘Agent’ as analogous to an object” (1991: 582). Ergative languages are said to employ the same proto-​agent and proto-​patient roles as in accusative languages, but “merely reverse the syntactic association” with subject and object (1991: 582). Often it seems that ergativity is taken up only after commitments to basic theoretical assumptions are set (Ackerman & Moore 2001: 1, fn. 1). But ergativity is unlikely to reveal its secrets to those who approach it superficially or too late, whether with afterthoughts or mere mirror-​image models. This chapter explores the connection between the well-​known ergative pattern in grammar and a pattern in discourse that is isomorphic to it, with the goal of providing a functional explanation for ergativity. The ergative discourse pattern holds the key to the grammaticization of ergativity, perhaps—​or is just a piece of the larger puzzle. The specific approach presented here is known informally as discourse and grammar (Givón 1979; Hopper & Thompson 1980; Thompson & Mulac 1991; Du Bois 2003b, 2014), which seeks to understand grammar in light of discourse, and discourse in light of grammar. Patterns of grammatical form are linked to communicative functions on the evidence of naturally occurring language use, in order to shed light on why grammars are the way they are. Crucially, “grammars” is plural, inviting attention to typological diversity. Not only must the broad alignment types of ergative, active, and accusative be distinguished, but also such cross-​cutting typological features as head-​marking vs. dependent-​marking, optional vs. consistent, aspect-​based vs. person-​based splits, and so on (DeLancey 1981; Nichols 1986; Garrett 1990; Bickel & Nichols 2009; Malchukov 2014; Nichols & Bickel 2013a; van de Velde 2014). The discourse-​and-​grammar approach accords equal importance to discourse and to grammar. In this chapter, however, I will devote more space to the discourse side of the equation. This is feasible because ergative grammar is well documented in the literature (including this volume); it is necessary because ergative discourse is not. That said, a key task will be to bring together the facts of ergative grammar and ergative discourse, and to clarify the connection between them. If ergativity is seen as a problem, it’s one that is not going away any time soon. That may be a good thing for linguists, who have a lot to gain by taking up the challenge of explaining ergativity. But one group for whom ergativity has never been a problem is the speakers of ergative languages. Ergative speakers do just fine, learning their language with equal ease (Ochs 1982; Narasimhan, Budwig, & Murty 2005; Bavin & Stoll 2013; Brown et al. 2013; Pye et al. 2013), and using it to perform the full range of functions that every language serves (de León 1999, 2000). Linguists seeking explanations might take a cue from the language users, and treat ergative

 

Ergativity in discourse and grammar    25 grammar as a system that works. The idea is to see ergativity as the solution, and ask what the problem is. This is not to suggest that there will be easy answers, nor that every aspect of ergative grammar will be transparently motivated or directly functional. Nonetheless, how a grammar works for its users is one of the most productive questions a linguist can ask. The view that grammars solve problems in new and creative ways accords well with the approach advocated by Evans and Levinson (2009), who, while arguing for massive cross-​linguistic diversity, also maintain that languages tend toward “evolutionarily stable strategies,” representing “recurrent solutions across time and space.” These strategies result from myriad interactions between communicative, cognitive, and processing constraints which reshape existing structures through use. A major achievement of functionalist linguistics has been to map out, under the rubric of grammaticalization, the complex temporal sub-​processes by which grammar emerges as frequently used patterns sediment into conventionalized patterns (Bybee 2000; Givón 2009). (Evans & Levinson 2009: 444–​445)

From the discourse-​and-​grammar perspective, the enterprise begins in discourse with the search for “frequently used patterns”; it continues by showing how the “recurrent solutions” resolve universal functional needs; and it ends, if successful, by elucidating the emergence of the “conventionalized patterns” known as grammar—​including the grammar of ergativity and its competitors. One seeming paradox is that grammar is already present in discourse from the start. Discourse is never found without its grammatical clothing; but by the same token, grammar is never realized except in discourse. A basic task for discourse-​and-​grammar research, then, is to tease discourse and grammar apart. With ingenuity and a little typology, the problem is solvable, as will become clear. This then sets the stage for investigating the connection between the respective ergative patterns in discourse and in grammar. I begin (2.1) with a look at a stretch of discourse in an ergative language, identifying a recurrent pattern which is isomorphic with the ergative–​absolutive pattern of grammar. The next section (2.2) documents this pattern quantitatively as the ergative discourse profile, based on evidence from a typologically diverse array of ergative, active, and accusative languages. I propose that the ergative discourse profile is shaped by a set of soft constraints known as Preferred Argument Structure (Du Bois 1987b, 2006; Du Bois, Kumpf, & Ashby 2003). The next section (2.3) explores whether the ergative discourse profile represents a discourse universal, examining evidence from child language, typology, genre, and diachrony. In the following section (2.4) I take up the functional explanation for ergativity, having introduced the analysis of competing motivations (Du Bois 1985, 2014; MacWhinney, Malchukov, & Moravcsik 2014; Malchukov 2014) that favor either ergative or accusative alignment. Next (2.5) I respond to some common objections to the discourse explanation for ergativity. Finally, I present some directions for future research (2.6) and conclusions (2.7).

 

26   John W. Du Bois

2.1  Ergativity in Discourse To understand ergativity, it is important to look at how discourse connects to grammar and to meaning. As Dixon says The most important task for future work on ‘why some languages are ergative in a certain way and others are not’ is to investigate the semantic and discourse-​pragmatic makeup of each of a sample of languages, and study the way in which this determines (or partly determines) its grammatical profile. (1994: 219–​220)

If discourse has the power to affect grammar, it is because “discourse, clause structure, and verb semantics are all intimately interwoven” (Foley & Van Valin 1984: 373). But to move beyond broad generalities about the interdependence of form and function, it is necessary to tease apart these three forces, if only to show how they come together again to shape the fundamental grammar of the clause. To make good on the promise that “recurrent patterns” shape grammar (Evans & Levinson 2009), it is necessary to do the empirical work to document the specific “discourse profiles” (Du Bois 2003a: 40–​44) that are linked to the grammatical constructions of interest, and their functional niches. The relevant work on discourse profiles focuses on “discourse inside the clause” (Du Bois 2003a: 13; 2003b: 83), seeking to identify the distinctive discourse correlates of the clause, its arguments, and other aspects of structure. While researchers sometimes speak broadly of the discourse profile of a language, akin to whole-​language typology (Nichols & Bickel 2013b), it is often more useful to target the discourse profile at a more specific level—​on a par with a specific argument structure construction, for example. Thus one can ask about the discourse profile of the intransitive, transitive, or ditransitive clause; or the agentless vs. agentive passive, and so on. One challenge in working with discourse is its evident variability, born of freedom. Seemingly, speakers exercise the absolute liberty to construct their utterances as they wish, within the broad limits circumscribed by the rules of grammar. The result appears, to some, as unpredictable variability. Yet a closer look reveals recurrent regularities in discourse, including some which are not strictly required by any grammatical rule. Understanding ergativity depends on sorting out how argument structure constructions balance the multiple demands of information processing, anaphoric reference, topic continuity, event semantics, and other factors, including the inherited morphosyntax of the language at hand. To make these matters more concrete, it will be useful to examine a sample of discourse from an ergative language. The following narrative is in Sakapultek, an ergative, head-​marking, verb-​initial language of the Mayan family, spoken in highland Guatemala (Du Bois 1981). The narrative was produced as a telling of a short film, the Pear Film (Chafe 1980; Du Bois 1980). To highlight the grammatical elements that matter for the discourse profile, the following conventions are used: underscore represents a

 

Ergativity in discourse and grammar    27 referent mention marked by a reduced referential form (pronoun, agreement, or zero);1 boldface indicates a lexical noun phrase; italics in the free translation indicates a verb or preposition (head of its phrase). The distribution of lexical vs. reduced forms is also indicated schematically, with capital letters (A S O) indicating a lexical noun in the designated grammatical role, while small letters represent a reduced form (a s o). A clause-​ by-​clause analysis of this narrative’s ergative discourse profile is published elsewhere (Du Bois 2006); here I present just the first 13 lines. (1)

Pear Story (Sakapultek)

1

… (H) Ee    x-​Ø-​inw-​il-​anh, foc  cp-​3.abs-​1.erg-​see-​tv … (H) What I saw was,



oaV

2

..  x-​ Ø-​ aq’an    jun    achenh, cp-​3.abs-​ascend one man a man climbed up,



VS

3

..   ch-​u’  chee’, at-​top   tree ..   in a tree,



PX

4

… (H) x-​Ø-​a-​r:       … ch’up-​ o’ niky’aj   péera-​s.      cp-​3.abs-​lat-​3.erg     pick-​dep some  pear-​ pl … (H) he went and … picked some pears.



aVO

5

… Tik’ara’  Ø-​Ø-​qaaj-​uu: l, then     cp-​3.abs-​descend-​hither … Then he came dow: n,



sV

6

… Ø-​Ø-​r-​ su’         r-​ iij    juu: n, cp-​3.abs-​3.erg-​wipe  3.erg-​back one … he wiped the surface of one,



aVO

7

… (H) ii  despwee: s, and then … (H) and the: n,

8

… (H) x-​Ø-​r-​ya’2         qaj   p    l  chikech, cp-​3.abs-​3.erg-​put down at the basket … (H) he put it in the basket,

..

oaV PX

1  Note that this annotation focuses on referent mentions and how they are expressed. Thus a referent expressed with a lexical noun phrase plus a cross-​referencing affix in the same clause is treated as one mention, not two (Du Bois 1987b: 813). Here it always the heavier form (noun phrase) that is marked (with boldface). 2  The underlying r-​is devoiced in this phonological environment, coalescing with preceding voiceless fricative x-​and effectively disappearing in the surface form, yielding xya’ ‘he gives it.’

 

28   John W. Du Bois  9

Ø-​Ø-​r-​ alsa-​ aj         p  l    r:-​…  m  komo  ber  gabaacha. oaV PX cp-​3.abs-​3.erg-​remove-​tv at the 3.erg um like   see apron he removed it from his: … um like apron.

10

… Despwees  tik’ara’, then   then … Then,

11

… Ø-​ Ø-​ pee       jun  aj-​laab’, cp-​3.abs-​come one dim-​boy …  a little boy came,



VS

12

..   ch-​ij  bisikleeta,  at-​back    bicycle ..   on a bicycle,



PX

13

… (H) xaq x-​Ø-​a-​r"-​ k’am-​ a’         jun chkech peera. just cp-​3.abs-​lat-​3.erg-​take-​dep  one  basket    pear … (H) he just came and took a basket of pears.



aVO

Viewed in grammatical terms, the data exhibit the hallmarks of ergative alignment in the grammar of verbal agreement (pronominal clitics). Focusing on third person singular referents, transitive subjects (A) are cross-​referenced with r-​‘3rd person singular ergative’ (lines 4, 6, 8, 13). In contrast, intransitive subjects (S) are unmarked, i.e. cross-​ referenced with Ø-​‘3rd person singular absolutive’ (lines 2, 5, 11). Transitive objects (O) receive the same treatment as S (lines 4, 6, 8, 13). The ergative–​absolutive pattern holds throughout the agreement paradigm (Du Bois 1981, 1987a: 210; 1987b: 809–​810), as in virtually all Mayan languages (Larsen & Norman 1979; England 1983; Robertson 1983; Kaufman & Norman 1984; Law 2009). Viewed in functional terms, the data illustrate some common patterns in the realization of basic discourse functions. For example, a new human referent is introduced using a full lexical noun phrase in the S role (lines 2, 11), and is subsequently tracked through the discourse with reduced forms (e.g. zero). The tracking of the most topical referents occurs most often in reduced A’s (lines 4, 6, 8, 13), but also in a reduced S (line 5). Inanimate referents are introduced here with a lexical mention in O role (lines 4, 6, 13), and tracked in subsequent discourse using a reduced O (lines 8, 9). Locative prepositional phrases also serve for the introduction of lexical and New inanimates, which may occur as adjuncts in separate intonation units (lines 3, 12), or in more tightly integrated verbalizations within the same intonation unit as the verb (lines 8, 9). The point of this exercise is to show what discourse looks like when viewed through the lens of grammar. To generalize from this tiny sample, lexical mentions occur mostly in absolutive argument positions (S and O roles), expressing New information. A’s are mostly reduced forms, and express given information. In this verb-​initial language, the preferred order of overt lexical nouns is VO and VS (in the four-​way typology of Dryer 1997, 2013a, 2013b), which can be generalized as V-​Lex word order: verb followed by a

 

Ergativity in discourse and grammar    29 lexical noun.3 The attentive reader will have noticed that the discourse distribution of lexical arguments (and of new information) corresponds to the absolutive category in the grammar of ergative languages, while topically continuous elements are found in what would be the subject in accusative languages. The latter reflects, perhaps, the common wisdom that subject is a grammaticization of agent and topic (Givón  1983a; Comrie 1988). The ergative and accusative discourse patterns coexist in one and the same stretch of narrative, in a language whose grammar is basically ergative. The tension between these two discourse patterns will prove fruitful for understanding the role of competing motivations (Du Bois 1985, 2014) in the discourse motivation of ergativity (1987a; Du Bois 1987b, 2006); see section 2.4. But we are getting ahead of ourselves. First we must ask: Do these observations characterize the discourse of one speaker, or do they represent something broader? This calls for a quantitative perspective, which is addressed in terms of discourse profiles.

2.2  The Ergative Discourse Profile The first systematic study of information structure to distinguish A, S, and O was by Du Bois (Du Bois 1981, 1985, 1987a, 1987b, 2003a, 2003b, 2006; Du Bois, Kumpf, & Ashby 2003). Previous studies had commonly been framed in terms of subjects, documenting contrasts in information structure that naturally seemed to favor accusative languages. But as long as S and A were collapsed within the all-​too-​familiar subject category, the differences between them remained effectively hidden. Adopting the grammatical terms A, S, O (Dixon 1979) or P (Comrie 1978), originally designed for typologically neutral comparisons encompassing ergative and accusative languages, made it possible to recognize a new kind of pattern in discourse. Early evidence from Sakapultek, illustrated in example (1), revealed a skewed distribution of lexical arguments across A, S, and O. A similar skewing was discovered for New information. Specifically, most lexical mentions occur in absolutive argument positions (S or O), but are avoided in the ergative (A) slot, which is mostly restricted to reduced forms (pronoun, agreement, zero). Correspondingly, most new mentions occur in S or O, with few occurring in A. Many more languages have since been investigated; see Table 2.1 for a selection. In most cases the findings reported tend to fit the generalizations of Preferred Argument Structure. This holds true whether the language is ergative, active, or accusative. Of course, not all scholars agree on the interpretation of the findings; for a discussion of controversial issues, see 2.5. It is useful to distinguish between discourse profiles, understood as observable patterns of linguistic behavior in discourse, and Preferred Argument Structure, understood as the functional constraints that govern them. The ergative discourse profile

3 

The implications of V-​Lex as a kind of ergative–​absolutive word order (Dryer 2013b) are developed below in “Typology” and “Diachrony” (2.3).

 

30   John W. Du Bois Table 2.1 Selected studies of Preferred Argument Structure Language

Family

Region

Type

Genre

Studies

Sakapultek

Mayan

Guatemala

ergative

narrative (PF)

(Du Bois 1987b)

Mam

Mayan

Guatemala

ergative

narrative

(England & Martin 2003)

Tektiteko

Mayan

Guatemala

ergative

narrative

(England & Martin 2003)

Mochó

Mayan

Mexico

ergative

narrative

(England & Martin 2003; Martin 2003)

Q’anjob’al

Mayan

Guatemala

ergative

narrative

(England & Martin 2003)

Itzaj

Mayan

Guatemala

ergative

narrative

(Hofling 2003)

Tsotsil

Mayan

Mexico

ergative

narrative

(Martínez Álvarez 2012)

Hieroglyphic Maya

Mayan

Guatemala, Mexico

ergative

dynastic history

(Mora-​Marín 2004)

Inuktitut

Eskimo-​Aleut Canada

ergative

child

(Allen & Schröder 2003)

Hindi

Indo-​Iranian

India

ergative

child

(Narasimhan et al. 2005)

Nepali

Indo-​Iranian

Nepal

ergative

narrative

(Genetti & Crain 2003)

Hawrami

Indo-​Iranian

Iran

ergative

narrative (PF)

(Mahand & Naghshbandi 2014)

Gooniyandi

Australian

Australia

ergative

narrative

(McGregor 1998)

Roviana

Austronesian

Solomon Is.

ergative

monologue

(Corston-​Oliver 2003)

Chamorro

Austronesian

Guam

ergative?

narrative

(Scancarelli 1985)

Acehnese

Austronesian

Indonesia

active

narrative

(Durie 1988, 1994, 2003)

Chol

Mayan

Guatemala

active

narrative

(Vázquez Álvarez & Zavala Maldonado 2013)

Yagua

Yaguan

Peru

active

narrative

(Thomas Payne 1993)

Mapudungun

Araucanian

Chile

hierarchical

narrative

(Arnold 2003)

Tohono O’odham

Uto-​Aztecan

Arizona

narrative

(Doris L. Payne 1987)

 

Ergativity in discourse and grammar    31

Language

Family

Region

Type

Genre

Studies

Jarawara

Arauan

Amazon

accusative

narrative

(Dixon 1994: 209)

Japanese

Japonic

Japan

accusative

conversation

(Matsumoto 1997, 2000)

Korean

Koreanic

Korea

accusative

child

(Clancy 2003)

To’aba’ita

Austronesian

Solomon Is.

accusative

narrative

(Lichtenberk 1996)

Finnish

Uralic

Finland

accusative

conversation

(Helasvuo 2003)

Hebrew

Semitic

Israel

accusative

narrative

(Sutherland-​Smith 1996)

English

Indo-​ European

United States

accusative

narrative (PF)

(Kumagai 2001, 2006)

English, diachronic

Indo-​ European

England

accusative

drama

(Shibasaki 2006)

French

Indo-​ European

France

accusative

conversation

(Ashby & Bentivoglio 1993)

French, diachronic

Indo-​ European

France

accusative

epic poetry

(Ashby & Bentivoglio 2003)

Spanish

Indo-​ European

Venezuela

accusative

conversation

(Ashby & Bentivoglio 1993)

Spanish, diachronic

Indo-​ European

Spain

accusative

epic poetry

(Ashby & Bentivoglio 2003)

Portuguese

Indo-​ European

Brazil

accusative

narrative

(Dutra 1987)

Portuguese

Indo-​ European

Brazil

accusative

various

(Brito 1996, 1998; Everett 2009)

multiple

various

various

various

various

(Haig & Schnell 2016)

Notes The languages are arranged by alignment type, then by family. The designation as ergative or accusative is not meant to dichotomize, as its necessarily glosses over details of split ergativity, etc. “PF” indicates Pear Film narratives (Chafe 1980).

can be observed in a corpus as a skewed distribution of new and lexical noun tokens across the argument slots of the clause. Preferred Argument Structure represents the constraints or generalizations which—​as a first approximation—​appear to govern the skewed distribution of utterance tokens. Whether the four constraints should be considered mere generalizations about linguistic practices, or whether there are deeper cognitive motivations behind them, is a question that remains open to alternative theoretical

 

32   John W. Du Bois interpretations (see section 2.5). Though closely related, the two perspectives are useful to distinguish, at least for the moment. Specifically, Preferred Argument Structure comprises four soft constraints, which collectively influence the discourse distribution of grammatical expressions and pragmatic statuses. In the grammatical dimension, the Quantity constraint limits the number of lexical arguments in the clause core to a maximum of one. The Role constraint specifies where in the clause the single lexical argument may appear, excluding it from the A role. Paralleling this pair of constraints is a second pair in the pragmatic domain. Here the Quantity constraint limits the number of new information arguments in the clause core to a maximum of one. The Role constraint specifies where in the clause this new argument will appear, again excluding it from the A role. In contrast to the generalizations of Preferred Argument Structure, the ergative discourse profile represents the empirically observable pattern of recurrent linguistic behavior, characterized by a statistical skewing in the distribution of lexical argument and new information tokens in a population of utterances. Broadly speaking it is the difference between rules and consequences—​or generalizations and facts on the ground. Table 2.2 summarizes the relation between the four constraints of Preferred Argument Structure and the corresponding consequences observable in the ergative discourse profile. To be more precise, the pragmatic constraints should be described as applying, not to new information, but to low accessibility referents (Ariel 1990, 2001), i.e. to referents whose cognitive status motivates a verbalization with a relatively informative (high surprisal) form. For present purposes, the looser formulation in terms of new information is adequate. But the more precise theoretical framing has important implications for research design, including the need for a continuous scale of accessibility/​surprisal (Ariel 2001) and, correspondingly, a continuous variable for the size of linguistic forms. For future research, such terms of analysis are to be preferred. Preferred Argument Structure is neither a syntactic structure nor a discourse structure, but a preference in discourse for the use of certain configurations of grammar. All four constraints are soft constraints (Givón 1979; Bresnan, Dingare, & Manning 2001), which can be violated without producing ungrammaticality. Yet in spontaneous language use, speakers tend to follow them. Together they influence the shape of discourse productions, yielding the distribution of argument structure constructions recognizable as the ergative discourse profile. Figure 2.14 presents findings from several languages regarding lexical argument quantity, that is, the frequency of clauses with zero, one, or two lexical arguments. (For Figures 2.1–​2.4, the languages are sorted in the same sequence as Table 2.1, with ergative languages presented to the left, and accusative languages to the right.) While there is considerable variability, the key finding here is that clauses with two lexical arguments are consistently rare across languages of different typologies, regions, and genealogies.

4 

Figures 2.1–​2.4 are based on selected studies in Table 2.1 (see also t­ ables 2.1–​2.4 of Du Bois 2003b: 62–​69).

 

Ergativity in discourse and grammar    33 Table 2.2 Preferred Argument Structure and the ergative discourse profile Preferred Argument Structure

Ergative discourse profile

Grammar

Avoid more than 1 lexical core argument

lexical arguments ≤ 1

Pragmatics

Avoid more than 1 New core argument

New arguments ≤ 1

Grammar

Avoid lexical A

free use of lexical S & O

Pragmatics

Avoid new A

free use of New S & O

Constraint

Domain

Quantity

Role

100 0 1 2

90 80 70 60 50 40 30 20 10

w

ish gl En

re eb

se

H

ne

m ’o O

no

Ja

da

an iy To

ho

oo n G

pa

di

di in H

ut In

uk

tit

aj Itz

al b’ jo ’an

Q

ko

hó oc M

ite

am M

kt Te

Sa

ka

pu

lte

k

0

Figure 2.1  Lexical argument quantity: frequency of clauses with 0, 1, or 2 lexical core arguments

Noun phrases are not produced in a functional vacuum, of course, but are tied to cognitive-​pragmatic processes of information management. The use of a lexical noun is linked to, but far from equivalent to, the presenting of new information. Thus the Quantity constraint on lexical core arguments is paralleled by a similar constraint on new core arguments. Figure 2.2 presents findings across several languages regarding New argument quantity, that is, the frequency of clauses containing zero, one, or two new arguments. Some language samples show no instances at all of clauses containing two new core arguments (Sakapultek, Roviana), while others show very few (Inuktitut, English).

 

34   John W. Du Bois 100 90 80 70 60 0 1 2

50 40 30 20 10 0

Sakapultek

Inuktitut

Roviana

English

Figure 2.2  New argument quantity: Frequency of clauses with 0, 1, or 2 new core arguments

While the Quantity constraint allows speakers to freely deploy One Lexical Argument per clause core, not all syntactic positions in the clause are equally good candidates for realizing this mention. This is where the Role constraint comes in, constraining the use of lexical arguments in A role, while freely allowing their use in S or O roles. Figure 2.3 summarizes findings from a number of languages regarding where lexical arguments are realized within the clause, showing how lexical core arguments are distributed across A, S, and O. While there is considerable variation in some dimensions (e.g. the relative preference for S or O), again what matters most is what is specifically constrained: the A role, which shows relatively low frequencies of lexical mentions as a recurrent pattern across many of a typologically diverse array of languages. Nonetheless, variation here is substantial, raising interesting typological questions (see below). For one-​place predicates, it is easy to satisfy the Quantity constraint, since the S role is unconstrained, freely allowing the introduction of a new referent. Indeed, this may be one reason language users select a one-​place predicate over a transitive alternative—​ for its pragmatic, rather than semantic, affordances. But for two-​place predicates, the Quantity constraint on new information is more directly consequential. Thus the Role constraint establishes an opposition within the transitive clause between A and O, constraining new arguments in A, while freely allowing them in O. Figure 2.4 presents the distribution of new core argument realizations across A, S, and O. A close examination of the frequencies in Figures 2.1–​2.4 shows considerable cross-​ linguistic similarities in some dimensions, along with substantial variability in others.

 

Ergativity in discourse and grammar    35 100 90 80 70

A S O

60 50 40 30 20 10

Sa

ka pu lte k M a Te m kt ite k M o oc Q hó ’an jo b’a l Itz In aj uk tit ut H in d N i ep al M Rov i ap ia To ud na ho un no gu O n ’od Ja am pa ne s Ko e re H an eb Po re rtu w gu En g es l i e ( sh Br az i Fr l) en Sp ch O anis ld h O fren ld Sp ch an ish

0

Figure 2.3  Lexical argument role: Distribution of lexical arguments across A, S, and O

This is why it is important to evaluate the discourse evidence in light of a theoretical framework, such as Preferred Argument Structure, which offers specific hypotheses about which aspects of argument realization are constrained, and which are not. One misunderstanding that often arises in the literature involves an attempt to compare raw frequencies between two languages (one of which is usually Sakapultek). But whether frequencies match or not is beside the point (and no one language of the many surveyed has a privileged position). What matters instead is the testing of specific, theoretically motivated hypotheses about constraints on syntactic positions in argument structure constructions, e.g. with respect to information structure. Another common temptation is to attribute the difference in frequencies found in two studies to the structural attributes of the languages in question. This may prove to be correct in some cases, and is certainly a question worth asking. But there are other candidates that should be considered as well in accounting for variance, especially genre differences. As the field of Preferred Argument Structure studies develops further, these issues are likely to become more visible, and we can anticipate new findings that tease apart the subtle variables involved. Now that so many typologically diverse languages have been studied, new questions arise about universality and variability of Preferred Argument Structure. Especially interesting is the potential for two-​way interactions between the grammar of a language and its discourse profiles, occasioned by the never-​ending cycle of grammaticization and language use. Are there different Preferred Argument Structures for

 

36   John W. Du Bois 90 80 70 60 50 A S O

40 30 20 10

Sa

ka pu lte k M a Te m kt ite ko M oc Q hó ’an jo b’a l Itz In aj uk tit u Ro t vi M an ap a ud un gu n Fi nn ish H eb re w En gl ish Fr en ch Sp a O nish ld Fr en O ch ld Sp an ish

0

Figure 2.4  New argument role: Distribution of new arguments across A, S, and O

different languages, and does this correlate with the ergative–​active–​absolutive typological contrast (Durie 1988, 2003)? The way is open for a broader theoretical framing of questions about why these recurrent patterns arise across such a broad typological array of languages, yet vary within seemingly well-​defined limits. There is much that remains to be discovered.

2.3  A Discourse Universal? Is the ergative discourse profile universal? To answer this question, I begin with the child, and go on to examine the evidence from typology and diachrony.

2.3.1 Child Language As might be expected, children who are exposed to ergative languages exhibit the ergative discourse profile. This has been shown for Tzeltal (Mayan, Brown 1998, 2008; de León 1999), Inuktitut (Eskimo-​Aleut, Allen 2000; Allen & Schröder 2003), and Hindi

 

Ergativity in discourse and grammar    37 (Indo-​Iranian, Narasimhan 2013; Narasimhan et al. 2005), and similar observations have been made for Samoan child language (Austronesian, Duranti, & Ochs 1989; Ochs 1982). But if the question is about universals, what is more telling is what happens in the discourse of children learning accusative languages. Clancy’s extensive studies of Korean child language show that even when exposed to an accusative grammar, children produce the ergative discourse profile (Clancy 1993, 1995, 1996, 1997, 2003). For Spanish, Bentivoglio argues for late acquisition of the ergative discourse profile (1998), while other studies of Spanish and French show a more complex picture (Khorounjaia & Tolchinsky 2004; Salas 2010). Older English-​speaking children with autism show the ergative discourse profile (Weber 2003), as do adults with aphasia (Kohn & Cragnolino 2003). How deep does the ergative discourse profile go? For Goldin-​Meadow this leads her to ask what the child’s discourse can reveal about the mind, and she goes to great lengths to find out. Seeking out “people who have had no exposure to any conventional language whatsoever” (2003: 493), Goldin-​Meadow works with profoundly deaf children who have had no access to either spoken language or sign language. What these children do is invent their own gesture systems, with revealing implications for cognition: The thoughts of these individuals cannot possibly have been shaped by language. As a result, whatever categories they express reveal thoughts that do not depend on language—​thought before language. (Goldin-​Meadow 2003: 493)

One of Goldin-​Meadow’s most compelling discoveries is that in the invented gesture systems, discourse tends to “pattern like ergative languages:  intransitive actors and patients are treated alike (produced), whereas transitive actors are treated differently (omitted)” (2003: 502). She observes: We have found the ergative pattern to be robust in communication situations. Deaf children of hearing parents who are inventing their own gesture systems tend to organize their gesture sentences around an ergative pattern. Equally striking, we found that when asked to describe a series of action vignettes using their hands rather than words, English-​speaking adults invented an ergative structure identical to the one developed by the deaf children, rather than the accusative pattern found in their spoken language. These findings suggest that the ergative pattern may reflect a robust solution to the problem of communicating information from one mind to another, be it an adult or a child mind. (Goldin-​Meadow 2003: 516)

While Goldin-​Meadow finds evidence to support the absolutive distribution of lexical arguments, she does not find evidence for the role of new information (2003: 516; Schulman, Mylander, & Goldin-​Meadow 2001). Pointing to eye-​tracking research by Griffin and Bock (2000), Goldin-​Meadow suggests that “focusing on patients may be a default bias found in both processing and acquisition tasks” (2003:  517). While their proposed “patient focus” motivation must be balanced against other

 

38   John W. Du Bois factors—​competing motivations—​deriving from pragmatics, semantics, etc., Goldin-​ Meadow and associates have made a compelling case for the presence of an ergative discourse profile at the earliest stages of communication. This begins even before exposure to language, arguing for a deep-​rooted cognitive basis for “thinking ergative” (Goldin-Meadow, Yalabik, & Gershkoff-Stowe 2000; Goldin-Meadow 2003).

2.3.2 Typology From a typological perspective, the first question is whether the ergative discourse profile is restricted to languages with ergative grammar, or rather represents a property of the discourse of all languages, independent of syntactic alignment. If it is restricted to ergative languages, it could be a consequence of ergative grammar. But if it is a discourse universal, this is no longer a tenable position, and it becomes a potential factor in the functional motivation for ergativity (2.4). As already noted, studies of Preferred Argument Structure have been carried out for a number of languages (2.2, Table 2.1, and Figures 2.1–​2.4). While a truly random typological sample is not possible, the available languages appear to be reasonably diverse in typology, region, and genealogy, covering all major alignment types. Based on these languages, the preponderance of evidence supports the view that most languages, whether ergative, active, or accusative, tend to display an ergative–​absolutive patterning in the discourse distribution of lexical (or “heavy”) nouns, and of new (or low-​accessible) information, in the core argument positions of the clause. To be sure, the interpretation of the discourse patterns, and even their existence, remains controversial for some scholars; I address these issues in 2.5. I believe that when the evidence from discourse, typology, child language, and diachrony are all taken into account, the picture that emerges is of an ergative discourse profile pervasive across languages, independent of the typology of syntactic alignment. That said, one of the most interesting questions raised by cross-​linguistic evidence like that in Figures 2.1–​2.4 is whether ergative, accusative, and other language types may have their own distinctive variants of Preferred Argument Structure, maintaining some universal aspects while also fine-​tuning to the grammar at hand, through cyclic interactions between grammar and use.

2.3.3 Diachrony If the ergative discourse profile represents a universal, found in modern languages regardless of their type, the uniformitarian principle (Hock 1991)5 predicts that older

5 

“The general processes and principles which can be noticed in observable history are applicable in all stages of language history” (Hock 1991: 630).

 

Ergativity in discourse and grammar    39 stages should share this feature as well. While few languages can provide direct textual evidence for older stages, some do. Diachronic studies conducted so far, though few, have consistently found the ergative discourse profile in both earlier and later stages of the same language. Among accusative languages, diachronic studies of Old Spanish, Old French, and their modern descendants show that “despite a gap of seven centuries … the medieval and modern forms of French and Spanish are remarkably similar in their manifestations of Preferred Argument Structure” (Ashby & Bentivoglio 2003: 73; see also Bentivoglio 1994). In a study of English spanning six periods from Old English to Modern English, Shibasaki concludes that, despite higher lexical densities6 in older stages, each period largely conforms to the constraints of Preferred Argument Structure (Shibasaki 2006). Among ergative languages, the Mayan family offers especially rich historical implications, given the available comparative and philological evidence. With 30 modern descendants, almost all of the languages have remained consistently ergative over four millennia (Norman & Campbell 1978; England 1983, 1991, 1990; Robertson 1983, 1992; Kaufman & Norman 1984; Law 2009, 2014).7 In the domain of discourse, the ergative discourse profile has been documented in detailed case studies covering four of the six major branches of Mayan, extending across a wide expanse of the Mayan territory: Mamean (Mam and Tektiteko, England & Martin 2003); Q’anjob’alan (Q’anjob’al and Mochó, England & Martin 2003); K’ichean (Sakapulteko, Du Bois 1987b, 2006); and Yucatecan (Itzaj, Hofling 2003).8 While at least one nineteenth century scholar claimed to reconstruct the text of a proto-​Indo-​European myth, such feats were ultimately deemed beyond the reach of the comparative method. But a discourse profile is not a text. Its wide distribution across the Mayan family makes the ergative discourse profile a plausible candidate for reconstruction to older stages. Support comes from early textual evidence in yet a fifth branch of Mayan (Ch’olan-​Tzeltalan). Mora-​Marín shows that Maya hieroglyphic inscriptions, written in an early Ch’olan language, display the ergative discourse profile (Mora-​Marín 2004). While areal diffusion is a theoretical possibility, the family-​wide distribution and textual evidence taken together point to a 6 

It bears noting that in several language families, older texts show higher lexical densities than modern texts (Ashby & Bentivoglio 2003), while still conforming generally to Preferred Argument Structure. This may reflect differences in genre of the older texts (epic poetry for Old French and Old Spanish, dynastic history for hieroglypic Maya), for which the modern languages lack a common counterpart. 7  The main exception is Chol, which has innovated an agentive system (Law, Robertson, & Houston 2006; Coon 2013; Vázquez Álvarez & Zavala Maldonado 2013). A few Mayan languages show some split ergativity, but this is a late and partial development (England 1983; Law 2009, 2014), which coexists with ergative morphosyntactic alignment. Yet one line is never crossed: No Mayan language has ever been accusative, from Proto-​Mayan till today. 8  Discourse evidence from a fifth branch is provided for Chol (Vázquez Álvarez & Zavala Maldonado 2013), which seems to have a discourse profile similar to that identified by Durie for Acehnese (Durie 1988, 1994, 2003), with A=Sa opposed to So=O in both grammar and discourse profile. Whether Durie’s conclusion, that Acehnese is compatible with an (extended) Preferred Argument Structure account, can be applied as well to the Chol data is beyond the scope of this chapter.

 

40   John W. Du Bois more likely conclusion: that the ergative discourse profile was present in early stages of Mayan, even Proto-​Mayan. This conclusion is also consistent with the uniformitarian principle, given the typological evidence for universality. The wealth of evidence for the Mayan family offers unique opportunities for tracing the history of ergativity in grammar and discourse. The implications are worth dwelling on here, as they are relevant not only for discourse-​and-​grammar reconstruction, but also for tracking the development of ergativity down through the daughter languages. One point of intersection between studies of comparative Mayan and the ergative discourse profile comes in the work of Nora England and Laura Martin (England 1983, 1991; England & Martin 2003). Drawing on their own research on the grammar and discourse of four languages from two branches of Mayan, they find the Given A constraint in both Mamean and K’ichean, and consider how discourse tendencies may become grammatical rules: It may be the case that, in K’ichean languages in particular, a grammatical restriction against indefinite subject NPs exists or is developing. This would presumably be a grammaticalization of the discourse constraint noted by Du Bois (1987b), that agents (ergators) are typically not used to convey new information. His analysis and analyses on other languages by England and Martin (1989) [=(England & Martin 2003)] show that, in texts from five different Mayan languages, lexical new mentions in agent (transitive subject) role typically occupy about three percent of the total lexical new mentions. Therefore, it is very rare to encounter an indefinite transitive subject noun, for discourse reasons. K’ichean languages appear to be creating a syntactic rule that reflects the same constraint. (England 1991: 484)

Thus the ergative discourse profile, present in Mamean and K’ichean languages and presumably in their common ancestor, is undergoing incipient grammaticization in K’ichean but not Mamean. This means the K’ichean change cannot be considered deterministic. Nevertheless, in moving from soft constraints to hard constraints, it traverses a path from ergative discourse to ergative grammar.

2.3.4 Interim Conclusions Is Preferred Argument Structure universal? The evidence so far paints a complex picture (2.5), and a full answer must await further inquiry (2.6). But what seems clear is that something like the ergative discourse profile represents a “recurrent pattern,” appearing again and again in the discourse of children just learning their language—​or exposed to no language at all—​and remaining constant in languages widely separated by space, time, genealogy, and typology. Whether the complexity and variation evident in the research record can be subsumed under a single set of generalizations remains to be seen. For now, it seems productive to explore the view that the ergative discourse profile,

 

Ergativity in discourse and grammar    41 and the Preferred Argument Structure constraints that motivate it, reflect enduring properties of language in use, with consequences for cognitive processing, acquisition, typology, diachrony, and grammaticization. Having introduced several parts of the problem, it is time to see how they fit together, and to assess what it would take to provide a functional explanation for ergativity.

2.4  Explaining Ergativity Preferred Argument Structure is claimed to be universal, and to motivate the grammaticization of ergativity. So far the functional explanation for ergativity seems on track—​until we stop to think about accusative languages. If the ergative discourse profile is present as a motivating force in all languages, why aren’t all languages ergative? “A system-​external functional force, once appealed to, cannot simply be turned off at will” (Du Bois 1985: 353). To explain ergativity, one more piece of the puzzle must be introduced: competing motivations (Du Bois 1985, 2014; MacWhinney et al. 2014; Malchukov 2014). In the arena of discourse, the two top competitors are powerful and pervasive, even universal: topic continuity and new information. The grammaticizing power of topicality is widely acknowledged. Subject is said to be a grammaticization of topic, or topic-​cum-​agent (Givón 1983a; Comrie 1988). But what about languages that don’t have subjects, or at least don’t grammaticize them in the central way that accusative languages do? Speakers of ergative languages have been claimed to differ from speakers of accusative languages in having “different conceptions of prototypical agenthood” and “different basic topicalizations” (Plank 1979: 28). Curiously, these psychological claims were made in the absence of any experimental or corpus evidence. To put such speculations to the test, a corpus-​based study of topic continuity was devised (Du Bois 1987b: 842–​843), which showed that in the ergative language Sakapultek, topical referents (appearing in two successive clauses) overwhelmingly favor S=A continuity over S=O continuity, 80 percent to 20 percent. The idea that ergative speakers think differently about topicality is debunked. But this leaves us with a quandary. Admitting that ergative speakers track topics just like accusative speakers do, it is still only accusative languages that grammaticize topic in an all-​encompassing syntactic and morphological subject. If speakers of all languages track topic continuity in the S=A groove that motivates accusative alignment, why aren’t all languages accusative? The typological question comes full circle. Both questions receive the same answer:  competing motivations. Ergativity and accusativity are both motivated, each with its own dedicated motivation operating at all times in all languages. But only one motivation can prevail at a time, in organizing the basic structure of a grammar (or part of a grammar) in a given language. The discourse profiles that drive this eternal competition were hinted at already, in the discussion of the discourse excerpt in (1) (2.1). Within a single short stretch of discourse, two

 

42   John W. Du Bois recurrent patterns are observed to coexist: first, introduction of new referents is managed in O and S (absolutive); and second, topic continuity is managed in A and S (subject). Quantitative cross-​linguistic evidence shows the two discourse profiles are widely distributed, found together in diverse genres and across languages of all types. As usual, S is Janus-​faced: Variable in discourse, it becomes pivotal in grammar. Having seen some key pieces of the ergative puzzle, while acknowledging the competition from accusativity, we can now ask how it all fits together—​to articulate the discourse motivation for ergativity, at least, if not a complete explanation for ergativity. The Quantity and Role constraints of Preferred Argument Structure set broad limits on the information structure of the clause. In effect they define a gross template for any argument structure construction, including intransitive, transitive, and ditransitive clauses. In the arena of discourse, utterance tokens realizing such constructions undergo selection to satisfy the constraints, producing the “facts on the ground” of the distribution of argument structure configurations in the utterance population. The ergative discourse profile represents a generalization of the statistical distribution of recurrent patterns, as candidates for grammaticization in the grammar of a given language. At the same time, the accusative discourse profile, lurking in the very same utterance population, presents its own patterns as alternative attractors for grammaticization. In general, the distribution of New information (reflecting cognitive processing demands) motivates a discourse pattern isomorphic to ergative–​absolutive grammar. The distribution of topic continuity motivates a discourse pattern isomorphic to nominative–​accusative grammar. Both functional pressures are present in the discourse of all languages, but at any given point only one can grammaticize, determining the syntactic alignment of a specific argument structure construction. If just one syntactic alignment is to prevail, the competition must be resolved. And this is just what grammaticization does: It resolves competitions, converting functional motivation into normative motivation (Du Bois 2014: 280). What emerges is a grammar that may seem arbitrary in its specific forms and normative rules, but that works for its users, serving as a unified framework for communicative practice and cognitive affordance. Consider the Mayan case, which shows what can happen when the ergative discourse profile interacts with word order, setting up the conditions for the emergence of ergativity. First, in line with the ergative discourse profile, the single lexical argument (reflecting the Quantity constraint) typically occurs in either O for transitives, or S for intransitives (reflecting the Role constraint). Second, for Mayan languages going back to Proto-​Mayan, the dominant word order (Dryer 2013a) is verb-​initial. Taken together, these two factors place both lexical arguments to the right of the verb, producing a structurally consistent V-​Lex{s/​o} or V-​Absolutive word order (an ergative order in the sense of Dryer 2013b). This constitutes the gross structure of verb and noun, maintained consistently in the discourse profiles of most if not all Mayan languages from Proto-​Mayan to now. But Mayan languages also have a fine structure of pronominal clitics, implemented in head-​marking of both ergative and absolutive on the verb. How does the fine structure of agreement morphology interact with the gross structure of word order syntax?

 

Ergativity in discourse and grammar    43 As noted earlier, different parts of the cross-​referencing paradigm behave differently. First and second person referents are given information, and thus are regularly marked by overt (non-​zero) pronouns/​clitics, in both ergative and absolutive. For 3rd person, referents in A are typically given, and thus reduced, and potentially cliticized. But S and O are often new, hence lexical, and so are not likely to be reduced or cliticized. The absolutive mention, being expressed overtly in the clause with a lexical rather than pronominal mention, is not itself cliticizable. The absence of a clitic may then be interpreted as absolutive zero agreement. Once the pronoun-​to-​agreement shift takes places (presumably in pre-​Proto-​Mayan), the features that define the Mayan ergative complex (V-​Lex word order, head-​marking, absolutive zero, and the ergative discourse profile) create a kind of V-​Lex lock-​in. This proves to be an evolutionary stable strategy, resistant to change over four millennia. Along with the gross structure of information, there is also the fine structure of inflectional detail, semantic nuance, and the specificity of pragmatic interpretation. The gross structure defined by Preferred Argument Structure sets broad constraints, but leaves speakers and languages plenty of room to maneuver at the level of fine structure. This is where the precise details of morphosyntactic analysis and historical development become critical, as processes of utterance production, interpretation, analogy, reanalysis, and grammaticization deploy and reconfigure the fine points of grammar to serve the exacting needs of language users. The grammarian’s, semanticist’s, and historical linguist’s attention to detail comes into its own here, analyzing the fine structure to elucidate the precise accounting of grammar and meaning, which is indispensable to the functionality of language. But the gross structure matters too. In the end there is no need to choose between them. Gross and fine work together in all of human action, and language is no exception. Languages often show an apparent harmony between gross structure (e.g. information structure, topicality, word order) and fine structure (e.g. inflection, agreement). But what if a discord or disruption arises between gross structure and fine—​how is it to be bridged? Must it be? To locate the critical arena where the crisis comes to a head, look to word order in use. This is where discourse-​and-​grammar research may contribute to resolving certain mysteries about the grammaticization of syntactic alignment. Preferred Argument Structure motivates the gross configuration of lexical nouns, e.g. the ergative discourse profile. If these nominal elements are arrayed on the same side of the verb as in most Mayan languages, perhaps motivated by a formal and/​or functional analogy, the result is a recurrent word order configuration (e.g. V-​Lex{s/​o} or Lex{s/​o}-​V) that constitutes a potential model for ergativity. If another historical development (say a phonological merger or analogical leveling) collapses key distinctions between elements in the agreement system, this may undermine the fine structure of morphological agreement. Yet what remains is the gross structure. Now the ergative distribution of gross structure becomes a potential model upon which to rebuild a fine structure, this time along new lines, perhaps ergative. Whether such a trajectory can be confirmed or disconfirmed in a language family with a long written history remains an open question. But the only way to find out is to follow the patterning of language in use. This

 

44   John W. Du Bois means documenting the discourse profiles that define the gross structure of the relevant utterances. It may be rare to find suitable conditions for inquiry into both gross and fine structure at the required level of detail, but it is well worth seeking out. The analysis of the Mayan ergative complex in relation to the ergative discourse profile provides an important case study illustrating certain general principles of functional explanation, insofar as it combines universal generalizations about gross structure of the ergative discourse profile with locally specific generalizations about fine structure of the inherited morphosyntax of a particular language family. Extending this approach to other language families, each with its own unique history, one should expect three things. First, the layer of gross structure produced by principles like the ergative discourse profile has a discourse-​based coherence of its own, and is likely to remain relatively stable; yet word order developments can introduce fundamental changes even here. Second, the layer of fine structure (inflection, agreement, and so on) linked to the language-​specific inherited morphosyntax has its own logic of continuity and change, which operates in part independently of that of gross structure. Third, interactions between gross structure and fine structure may trigger a dynamic of change that disrupts the ecology of grammar, setting in motion events that lead to restructuring the system of grammar. The story of the interaction among gross and fine must be discovered anew in the history each new family. Yet even here, analogy, reanalysis, and adaptive selection (based on cognitively motivated information processing constraints, for example) offer general principles for a theory of grammaticization. The roster of motivations with the power to shape ergative and other syntactic alignments is not exhausted by the two considered here. A more complete account will have to incorporate interactions with competing and converging factors such as event structure, causal chains, verb semantics, tense and aspect, voice, the constructional repertoire, evidentiality, analogy, and more (DeLancey 1990; Croft 1998; see also the chapters in this volume). Many of these involve morphosyntactic, semantic, and pragmatic fine structure, which is essential to complete the functional explanation of ergativity. A critical task for future research is show how multiple layers of fine structure interact with layers of gross structure to shape the grammaticization of ergative and other argument structure configurations.

2.5  Objections and Refutations The claim of a discourse basis of ergativity (Du Bois 1987b) has generated a certain amount of controversy, which I address in this section. Some objections question the existence of the ergative discourse profile, or try to explain it away, while others accept it but doubt it supports a functional explanation for ergativity. While some studies raise useful points that warrant attention in future research, others reveal a misunderstanding of the nature of discourse-​based explanation. Common conceptual errors are essentialism, reductionism, and epiphenomenalism, all of which involving idealization of the

 

Ergativity in discourse and grammar    45 facts of language use. In the following I will try to distinguish the useful critiques from the dead ends; and, for the latter, suggest an alternative approach. A recurrent theme is that generalizations about language must be grounded in linguistic realism rather than idealization.

2.5.1 Diversity One of the earliest and most productive critiques was developed in a series of discourse-​ and-​ grammar studies by Mark Durie (1988, 1994, 2003). Preferred Argument Structure studies had adopted the A–​S–​O grammatical categories (Dixon 1979) for their neutrality with respect to typological diversity in syntactic alignment, to avoid imposing alien categories on ergative languages. But Durie argued they were doing just that to active languages. He rejected the view of S as a universal category for linguistic analysis, arguing that it obscured important diversity in both discourse and grammar, in active languages like Acehnese (Austronesian, Indonesia). Here, intransitive subjects are internally differentiated, both in grammar and in discourse profile, between Sa and So. A better analysis of Acehnese discourse could be achieved by respecting the alignment typology evident in Acehnese grammar, which distinguishes Actor (Sa=A) and Undergoer (So=O). By investigating categories relevant to the grammar at hand, Durie was able to show that each discourse profile was both internally consistent and distinct from its counterpart. Subsequent studies in ergative, accusative, and especially active languages have confirmed and extended these findings. For example, Chol, a Mayan language once characterized as “split-​ergative” but now recognized as active, makes a similar distinction between Actor and Undergoer in both grammar and discourse profiles (Vázquez Álvarez & Zavala Maldonado 2013). These studies challenge the assumption that S is uniform in language use, and provide a useful corrective to the A–​S–​O schema. More generally, they remind us that the categories of the language being investigated are a potent guide to new discourse profiles waiting to be discovered. This raises the question of whether there are different Preferred Argument Structures for active, ergative, and accusative languages, or even finer-​grained distinctions; and, if so, whether they can be interpreted as principled variations on a single unifying theme. Durie nevertheless saw value in framing his analysis in terms of Preferred Argument Structure, once the necessary adjustments were made to accommodate the diversity of grammar and discourse profiles of the language being analyzed. Durie recognizes a key point that befuddles some critics, regarding the role played in discourse-​functional explanation by the discourse profile (Du Bois 2003a: 40–​44), what I have previously called “recurrent clusterings in parole” (Du Bois 1985: 357). Discourse profiles are generalizations about recurrent patterns of linguistic behavior, including recurrent co-​ occurrences of pragmatic, semantic, and grammatical features. As statements about language use, they are not to be confused with the grammatical categories they motivate, which have a normative status in the linguistic system. Durie nicely points up a

 

46   John W. Du Bois common misconception about discourse-​functional motivation of grammatical categories in a response to Herring (1989) which remains relevant today: Herring (1989), misunderstanding the concept of motivation, regards the mismatch in Sakapulteko as a logical flaw in Du Bois’s argument. A further distortion in Herring’s argument derives from her demand that the grouping of S and [O]‌ should involve functional specialization in terms of cognitive or semantic factors. In Du Bois’s account, the cognitive/​semantic motivations do not themselves directly impact upon coding choices, but only proximately, through the “recurrent clusterings in parole” (Du Bois 1985: 357) which they produce. This is of course a key difference between the Du Bois account of ergativity and more directly cognitively or semantically based approaches. (Durie 2003: 192, fn. 193)

Discourse profiles are general statements about the “facts on the ground” of language use. They arise as a result of multiple factors, including factors such as cognitive processing, salience, and so on. As system-​external motivations they interact with system-​ internal factors to shape the emergence of grammar. But the process is not deterministic. While discourse profiles influence the grammaticization of linguistic categories, the profile is not the category.

2.5.2 Essentialism Less productive are critiques which, while bringing no empirical research to bear on the question at hand, translate originally statistical observations into the language of categorical statements, the better to draw logical deductions—​or contradictions. Needless to say, the meaning of a discourse profile may be severely compromised if its empirical variability is disregarded. When soft constraints are restated as hard constraints, the result is unlikely to be faithful to the original. Such a confusion mars the critique of Harris and Campbell (1995), who paint a portrait of the discourse basis of ergativity (Du Bois 1987b) that is almost unrecognizable to someone familiar with the theory. They confidently appeal to the “spirit” of Preferred Argument Structure constraints, which they feel are “blatantly violated” by the passive origin of ergativity (1995: 253). They enthusiastically assert that transitive “A [is] prohibited almost totally from introducing new information” (1995: 253), then quietly delete the “almost” as they substitute a categorical rule for the original soft constraint. They perceive “an unpleasant circularity in Du Bois’ picture” (1995: 254), based on their questionable reframing of Nora England’s findings of incipient grammaticization of certain Preferred Argument Structure constraints into grammatical constraints on the ergative in K’ichean (England 1991; England & Martin 2003). Their story of the passive origins of ergativity begins with a proto-​language agented passive, represented as containing two proper nouns (Mary was.hit by Jane) (1995: 252). This contrasts with the attested

 

Ergativity in discourse and grammar    47 examples they themselves cite from Old Persian to illustrate the passive to ergative change, which contain a pronoun rather than a lexical noun in the agentive by-​phrase (1995: 244, 255). Though somewhat murky, the reasoning behind their argument seems to depend on the following assumptions and inferences:





1. To get from accusative to ergative alignment, a well-​known diachronic pathway is via the reanalysis of an intransitive passive construction as a transitive active construction. 2. To express a two-​place predication fully explicitly, the passive sentence cannot be agentless, but must include the agent. 3. Thus the agented passive construction is used, with the agent expressed in an oblique role, e.g. a prepositional phrase adjunct (by-​phrase). 4. Obliques (prepositional phrase adjuncts) often contain new information and lexical nouns (by Preferred Argument Structure). 5. Therefore the oblique agent must have been new and lexical. 6. The two-​participant predication, being a passive, is intransitive. 7. Therefore the subject of the two-​participant predication (the semantic Patient) must be an S. 8. Intransitive S arguments are often new and lexical (by Preferred Argument Structure). 9. Therefore the S role argument must have been new and lexical. 10. By reanalysis, the by-​phrase (originally an optional adjunct) is reinterpreted as an obligatory core argument (transitive subject A), and the S becomes O. 11. The formerly rare agented passive intransitive construction undergoes a massive change in frequency to become the new normal for transitive constructions, but speakers make no changes in their use of pronouns vs. nouns, or given vs. new information. 12. Now there are two new and lexical nouns in the core arguments of the clause (violating Preferred Argument Structure). 13. Now there is a new and lexical A (violating Preferred Argument Structure).

The problems with this account are several, arising on multiple levels. First, the logical flaws. The reasoning depends on converting a statistical tendency to a categorical rule. This must be done twice (in deriving 5 from 4, and 9 from 8), in order to generate the desired contradiction. Second, language users are presented as being sufficiently creative to introduce major changes to the structure of their grammar, reanalyzing an optional oblique as an obligatory core argument (step 11), but they were apparently too timid to begin using a pronoun in place of a lexical noun—​during the decades or centuries it must have taken for the frequency shift and ergative reanalysis to be completed. Ignored is the fact that there are no constraints against using pronouns or given information: Not only are the Preferred Argument Structure constraints always soft constraints, the limits are always upper limits, not lower (Du Bois 1987b: 834; 2003b: 73).

 

48   John W. Du Bois Thus there has never been any minimum requirement to fill a syntactic slot, whether argument or adjunct, with either new information or a lexical noun. One begins to understand why it was necessary for Harris and Campbell to appeal to the “spirit” of Preferred Argument Structure in making their argument, instead of employing the actual constraints. The result is unfortunately fairly typical of essentialism, which, faced with statistical evidence of diversity in argument realization, substitutes a categorical stereotype, and then uses the stereotype to work out the logic of its reasoning. But surely this is antithetical to the “spirit” of Preferred Argument Structure, if it must have one. Is there an alternative? Actually, very little needs to be changed in the above account to make it clear that not only is Preferred Argument Structure compatible with the grammaticization pathway in question, it actually facilitates it. Moreover, the only changes needed involve replacing the rigidly idealized conjectures of Harris and Campbell with common-​sense observations on how speakers talk—​as confirmed in corpus-​based studies of naturally occurring language use. Preferred Argument Structure allows speakers the flexibility to use a pronoun in any slot, whether argument or adjunct, and speakers routinely do just that (Du Bois 1987a, 2003b). Moreover, as Ariel et al. (2015) have shown, speakers use pronouns when innovating a new argument slot, precisely because pronouns fly under the radar of the Quantity constraint. This leaves speakers free, when extending the use of a structure like the passive, to choose a pronoun in the by-​phrase. In fact this is precisely what the textual evidence from Old Persian shows: (2)    ima tya manā kartam Parθavaiy “This is what was done by me in Parthia” (Darius the Great) (John R. Payne 1980: 151) (cited in Harris & Campbell 1995: 255) The genitive pronoun (manā ‘me.gen’) expresses the agent in a by-​phrase, initially an optional adjunct. But as this structure comes to be routinely used to express two-​ place predications, it undergoes reanalysis (step 11) as an ergative core argument. The new transitive structure easily satisfies the Quantity and Role constraints of Preferred Argument Structure.

2.5.3 Reductionism One attraction of essentialism is that it feeds reductionism. By replacing a complex empirical generalization with a handy rule of thumb, a theoretical economy is achieved. If the rule is not only categorical but universal, the conclusion may be drawn that empirical investigation of the language at hand is unnecessary. But when statistical generalizations are replaced with categorical statements, the likely result is a false economy. Such a chain of essentialist substitutions is found in the reductionist proposal of Haspelmath (2006). Where I  had pointed out the functional motivation linking the avoidance of lexical nouns with avoidance of new information (Du Bois 1987b: 829–​830), Haspelmath wants to go further, suggesting that “since new information is mostly coded

 

Ergativity in discourse and grammar    49 by lexical NPs, … the [lexical] tendencies … could be reducible to the [new information] tendencies.” Moreover: [T]‌ he ‘quantity’ tendencies … [may] follow straightforwardly from the ‘role’ tendencies. (Haspelmath 2006: 910) [I]t appears that the strong tendency to avoid clauses with two new/​lexical core arguments can be reduced to the strong avoidance of new/​lexical As … . [W]‌e simply like to talk about human beings and their actions, so animates tend to be topical. … Since the A argument is almost invariably animate, it follows that it is typically topical and hence given. (Haspelmath 2006: 911, emphasis added)

Haspelmath concludes that “the whole notion of preferred argument structure may be reducible to other, well-​established tendencies and generalizations” (2006: 911, emphasis added); (see also Dahl 2000: 50; Everett 2009; Haig & Schnell 2016). This style of argument slides easily from most to all, from it tends to it follows, and draws inferences accordingly. While it overlooks the fact that the Quantity constraint probably has a better independent motivation as a cognitive limitation on information processing resources, Haspelmath’s proposal to dispense with the Quantity constraint, and indeed all of Preferred Argument Structure, makes a certain sense—​if we accept a series of inferences about language use, each apparently well motivated on its own:

1. 2. 3. 4.

Humans are interesting to humans. Therefore, humans are topical. Therefore, humans are given. Therefore, humans are expressed by reduced forms (pronouns or zeroes).

Leaving aside the inferential gaps, this is still not enough to make viable predictions about argument realizations in syntactic argument slots. If you add the further assumption that humans are agents, and that the subject role expresses agent-​cum-​topic, you can, seemingly, derive the inference that subjects are expressed by pronouns. In contrast, direct objects are said to encode inanimates. Inanimates are less interesting, therefore not topically continuous, therefore not accessible, therefore new, therefore lexical. But do these conclusions actually follow? Goldberg evaluates a similar reductionist proposal and rejects it: “[T]‌he Given A constraint does not follow directly from the prevalence of animate topics. … [T]he Given A constraint is not simply epiphenomenal” (Goldberg 2004: 431). The problem with the essentialist–​reductionist line of reasoning is that the conclusions follow only if each of a long chain of assumptions is valid—​specifically, only if each generalization is categorical. But none of them are. And because each step in the chain represents at best a statistical tendency, the inferential failure compounds with each step. Flawed logic aside, the most serious problem is empirical: the observable facts of discourse do not confirm the logical deductions about what discourse “should” look like. The fact is, only some

 

50   John W. Du Bois humans are topical, and therefore pronominalizable, and this makes a critical difference for grammaticization pathways, e.g. in the role of pronouns in the emergence of innovative argument structure constructions (Ariel 2000; Ariel et al. 2015). Is there another way? A recent corpus-​based study of Hebrew datives (Ariel et al. 2015)  offers a relevant model, even if the construction involved is different. Ariel et al. compare datives (syntactic arguments) with adjuncts headed by the preposition bishvil ‘for.’ Both introduce mostly humans, and both express the same thematic role (roughly, benefactive). Yet they part company in their discourse profiles: Only 5 percent of the datives are lexical, but as many as 23 percent of the bishvil adjuncts are. The difference cannot be attributed to animacy, but only to the difference in syntactic status: argument vs. adjunct. This accords with the predictions of constructional Preferred Argument Structure (Ariel et al. 2015: 270–​272; Du Bois & Lester in progress), which freely allows lexical mentions in adjuncts, but not in core argument slots other than S/O. More generally, pronouns sidestep any problems with the One Lexical Argument constraint, and thus are exploited as pioneers in an incipient grammaticization introducing an additional argument position into the clause, allowing a second human participant (in addition to the agent) to be expressed in a benefactive-​like role. What are the implications for ergativity? While the alignment types and constructions in question differ, interesting parallels can be drawn nonetheless. In both cases, speakers modify an existing argument structure construction, adding a new argument role (ergative or dative). And the strategy they adopt to accomplish it without violating Preferred Argument Structure constraints is essentially the same: cognitive containment (Ariel et al. 2015: 270–​272; Du Bois & Lester in progress). The safe strategy is to use a pronoun in the innovative argument slot, to avoid violating the One Lexical Argument constraint—​or, to put it in cognitive–​functional terms, to avoid overloading limited processing resources.

2.5.4 Historical Accident Claims of functional motivations have implications for language change, but are sometimes at odds with the findings of historical linguists (Cristofaro 2014). For example, Næss (2015) points to a series of seemingly random historical changes in rejecting, not only the competing motivations analysis of ergativity (citing Du Bois 1985; Du Bois 1987b), but any functional explanation for the ergative structure of Äiwoo (Oceanic, Solomon Islands). After detailing the complicated historical processes that produced the ergative verb phrase, he states: What the [ergative] VP structure of Äiwoo demonstrates is … that any linguistic system at any point in time is the outcome of a number of interacting factors which do not add up to either a universal structural template or a set of functionally-​based competing motivations. (Næss 2015: 102)

 

Ergativity in discourse and grammar    51 This statement about grammar as “the outcome of … interacting factors” is fine up to the conclusion, which does not follow. The assumption that the claimed VP merits its own “set of functionally-​based competing motivations” seems premature, in the absence of the relevant discourse profile information. Be that as it may, it is surely unwise to conclude that the randomness of historical changes precludes an adaptive account of the historical evolution of linguistic structure. From the perspective of linguistic as well as biological evolution (Beckner et al. 2009; Lane 2015: 172–​204), there is no principled contradiction between the fact that a system is subject to historical processes, which may randomly introduce partial arbitrariness, and the applicability of the evolutionary processes of adaptive selection, which yield partial motivation. The origin of ergativity seems particularly prone to invite frustration and despair, leading some historical linguists to go so far as to question whether there is any functional motivation for ergativity: The absence of a clear case of extension creating ergativity argues against a clear functional motivation unique to the ergative pattern. With the exception of the passive > past/​perfective ergative, it appears that ergativity is a historical accident that has come up again and again in many parts of the world. (Gildea 2004)

To conclude that “ergativity is a historical accident,” while acknowledging that it is a recurrent pattern arising independently in languages around the world, seems a contradiction. Anything in language can be made to look like a historical accident—​even the grammar of accusative languages. But such a stance appears valid only if one restricts the inquiry to tracing etymological sources and describing the arbitrary signs that result. What looks like a “historical accident” may well turn out to involve adaptive selection, given an evolutionary account. What is the alternative? The evolutionary development of any aspect of language can be seen as the result of many small, locally motivated actions, taken by speakers who lack an overarching view of the system (Keller 1994). But in this respect linguistic evolution is no different from biological evolution (Dediu et al. 2013; Lane 2015; Mayr 2001). Evolutionists don’t give up on adaptive explanation just because random mutations introduce one accident after another. On the contrary: Such accidents (mutations) provide the necessary variability (Bybee 2007) for selection to act on. On one interpretation, functional constraints play a role in grammaticization by acting as selective processes that winnow the variability of naturally occurring discourse. The forms and constructions that survive and reproduce in the longer term will be those that satisfy cognitively motivated constraints, like Quantity. In general, constraints on the selection of functionally viable linguistic structures can act over time to constrain the possible outcomes of grammaticization processes, leading to functionally motivated structures (Hopper & Traugott 2003; Traugott & Trousdale 2014).

 

52   John W. Du Bois

2.5.5 Epiphenomenalism Epiphenomenalism may seem an arcane philosophy, originating as it does in nineteenth-​ century mind–​body dualism (Walter 2009: 1137). But despite its abandonment in most modern sciences, in linguistics (and in sociobiology, Searle 2013) the epiphenomenon is invoked surprisingly often (Hopper 1987; Jaeger & Snider 2008; Malchukov 2008). This includes in claims about ergativity (Everett 2009; Haig & Schnell 2016). In practice, labeling an empirical observation as epiphenomenal typically prefigures a move to dismiss it as inconsequential. “An epiphenomenon is a secondary symptom, a mere “afterglow” of real phenomena” (Walter 2009:  1137). While the supposed epiphenomenon is acknowledged to have a real cause, it is claimed to have no further causal consequences in the world (Walter 2009:  1137). By setting up a disconnect between mind and materiality—​or langue and parole—​epiphenomenalism inherits the problems of dualism (Searle 2013). Not surprisingly, attempts to partition facts into real phenomena and epiphenomena tend to suffer from a lack of consensus about criteria for deciding which is which: One scholar’s epiphenomenon is another’s phenomenon. But the real problem lies in the idealization that is introduced into otherwise empirical questions: a dualism, not of mind and body, but of grammar and use. Appeals to epiphenomenality often arise in response to claims about language use as an influence on grammar. An empirically observable pattern in use is said to be epiphenomenal—​in effect, not a phenomenon, but only illusory—​to the extent that it can be explained away as caused by other factors—​in a word, reduced (Haspelmath 2006). But this neglects the first question that should be asked: Does the observable pattern in language use have downstream causal consequences? Specifically, the epiphenomenal charge has been made regarding discourse patterns identified by Du Bois (1987b) as consequential for the grammaticization of ergativity (Everett 2009; Haig & Schnell 2016). In a study otherwise notable for its careful analysis and impressive multilanguage database,9 Haig and Schnell, speaking of the Given A Constraint, maintain that: [T]‌he apparently marked behavior of the A role, another cornerstone of the ergativity claims, … is an epiphenomenal by-product of two well-​documented and robust tendencies: the pervasive tendency for transitive subjects to be [+hum], and the pervasive tendency for all subjects (S or A) to be topical, hence given information. (2016: 612)

Their claim to distinguish some patterns as epiphenomenal (while others, presumably, are not) leads them to conclude, somewhat surprisingly, that ergativity itself is epiphenomenal: 9  The massive study by Haig and Schnell (2016), encountered online as this chapter was going to press, makes a valiant effort to raise the bar statistically, and merits a more complete assessment than can be given here. Nevertheless, it introduces conceptual problems of its own, touched on here. Note that they cite different numbers of subjects and objects for transitive clauses (e.g. for English they report 1,111 transitive objects but only 422 transitive subjects; see their appendix 2). The gap reflects their omission of 1st and 2nd persons, making it difficult to compare with the findings reported here.

 

Ergativity in discourse and grammar    53 In Iranian for example, a shift from accusative to ergative alignment (restricted to past tenses) … was a particular, and highly contingent, combination of … changes that conspired to yield ergative alignment … . These and similar diachronic developments speak of a more contingent approach to ergativity, according to which ergativity arises as an epiphenomenal and construction-​specific constellation, through the combination of essentially independent morphological and phonological processes. (Haig & Schnell 2016: 614–615)

Here the idea of epiphenomenality begins to collapse on itself. Against claims of functional explanation, the epiphenomenalist proposes an absence of explanation: “ergativity is a historical accident” (Gildea 2004) or “epiphenomenal” (Haig & Schnell 2016: 615). True, a grammatical pattern may seem accidental, when viewed exclusively in terms of its etymological source materials. But some historical accidents are accidents waiting to happen. As a typologically recurrent pattern, ergative syntactic alignment must be recognized as a powerful attractor state, that is, an evolutionary stable strategy. Accusativity too is a potent attractor. But there is no contradiction here. Both recurrent patterns are attested worldwide, and each has its motivations. Indeed, the two motivations coexist within the discourse of every language. The eternal tension between ergative and accusative motivations, evident in split ergativity, is best understood in terms of the theory of competing motivations (Du Bois 1985, 2014; MacWhinney et al. 2014). But to dismiss a linguistic pattern—​even ergativity—​as epiphenomenal is to cut off inquiry prematurely. It would be better to drop the epiphenomenal stance altogether, and take up the very real challenges of building theory and method capable of accounting for recurrent discourse patterns and their downstream consequences for grammar. In the study of ergativity, the facts on the ground of discourse hold much promise for understanding split ergativity (Silverstein 1976; Comrie 1978; Dixon 1979, 1987; DeLancey 1981; Durie 1988, 2003; Malchukov 2005, 2014; Law et al. 2006; Mahand & Naghshbandi 2014; see also Laka, Nash, Coon & Preminger, and Woolford Chapters 7, 8, 9, and 10, this volume).

2.5.6 Interim Conclusions I have considered various objections to Preferred Argument Structure and the ergative discourse profile. The most productive tend to come from researchers who combine in-​depth first-​person research on the grammar in question with detailed and sensitive empirical investigations into the discourse distribution of grammatical elements (Durie 1988, 1994, 2003; Vázquez Álvarez & Zavala Maldonado 2013). Others were found wanting: mired in essentialism, reductionism, and epiphenomenalism. What is the problem? Not generalization per se, which is indispensable for understanding and explaining grammar. Rather, the problem lies in idealization, cutting the system of language off from the reality of its use. Idealization begins with a misplaced essentialism that reifies categories, obscuring the variability that characterizes

 

54   John W. Du Bois populations of naturally occurring utterances. It continues (sometimes) with a reductionism that creates an illusion of economy, without testing to see if the reduced principles can in fact reconstitute (predict) the facts of the world it claims to have reduced. On the other hand, sometimes scholars dwell on a maze of intricate historical details—​no lack of empirical facts here—​but when the time comes for an explanation they come up empty, claiming a historical accident. Finally, the apotheosis of idealization is reached in epiphenomenalism, which dismisses certain facts as not phenomena, banishing them to the black hole of epiphenomena—​from which no downstream causal consequences can escape. What is the alternative? First, linguists must commit to linguistic realism. The facts on the ground of discourse are not to be dismissed, lest their consequentiality in the world of language be overlooked. Nor are they to be shunted aside as epiphenomena. Rather, language use is firmly located in the world. Here it coexists and interacts with the practices, norms, and knowledge of language, even if all these have somewhat different ways of being in the world. Linguistic realism urges documenting the empirical generalizations about language use that define its capacity to shape grammar: the discourse profile. Second, it is equally important to follow through on the theoretical end. Discourse profiles have downstream causal consequences. Identifying the consequences serves at the very least to verify that a meaningful discourse profile has been identified. The combination of linguistic realism and theoretical generalization is critical for explaining ergativity, and for all questions of the interinfluence of discourse and grammar. While some objections have proved lacking, even so they serve to elicit clarification of issues left murky in previous formulations, and point to gaps in our knowledge that call for further research. Important issues have been broached regarding the relation of language use to grammar, and of linguistic realism to functional explanation. Certainly some aspects of the discourse approach to ergativity are likely to remain controversial, until resolved by further research. It remains for new collaborations among researchers, bringing together corpus-​based methodologies, multifactorial statistical techniques, grammar, typology, competing motivations, and functional theory, to advance our understanding of the outstanding questions.

2.6  Directions for Future Research What new possibilities does discourse research bring to the study of ergativity? Whether the topic is pragmatics, syntax, semantics, constructions, splits, morphology, prosody, cognition, typology, diachrony, or grammaticization, all can benefit from incorporating a focus on language in use. The study of discourse integrates well with a wide range of research questions, bringing new perspectives on how a targeted set of grammatical resources serves the communicative goals of its users. For ergativity the biggest payoff is likely to come from “discourse inside the clause” (Du Bois 2003a: 13; 2003b: 83), an approach that seeks to map out the distinctive functional

 

Ergativity in discourse and grammar    55 correlates of the various structural components of the clause. A useful research strategy is to follow the trail of difference, in both discourse and grammar. Differences in linguistic structure are shaped by the multiplicity of functional needs, modulated by complex interactions among competing motivations, and driven by the dynamics of grammaticization. Interesting structural differences arise between contrasting elements in split grammar, variation, typology, child language, diachrony, and grammaticization, all potentially linked to distinctive functions. The world’s languages offer a rich set of natural experiments (Evans & Levinson 2009), where each case represents a potentially novel way of linking contrasts in form to contrasts in function. Integrating a discourse methodology into cross-​disciplinary research can bring a key piece of the puzzle, helping to trace out the similarities and contrasts that manifest in such fundamental differences as, for example, ergativity vs. accusativity. For discourse to become an integral and valued component of such research, it must identify the specific discourse profiles that are relevant to the problem at hand. One ongoing challenge is to tease apart grammar and discourse: grammar with its seemingly static structure, and discourse with its free plasticity. The challenge is compounded, yet all the more important, because grammar and discourse are so intimately intertwined. For untangling the multiplicity of factors that impinge on the discourse distribution of referential forms across syntactic roles, a much-​needed development is the application of newer multivariate statistical techniques (Du Bois & Lester in progress). The increasing popularity of soft constraints (Bresnan et al. 2001) as a way of describing morphosyntactic and functional variation makes the development of statistical models all the more urgent for corpus-​based ergativity research. Such techniques are needed to address questions that have been raised about functional correlates of variation in ergative marking, for example in exploring Duranti’s observations about how the so-​called “optional ergative” is exploited in discourse to index social power and access to agency (Duranti & Ochs 1990; Duranti 2004). Similarly, Dixon, observing that discourse profiles differ in detail from one ergative language to the next, asks how this may reflect differences in cognitive style between their respective cultures (Dixon 1994: 219–​220). Yet perhaps the most productive questions, offering a combination of the most effective methodological purchase and the greatest theoretical rewards, will be those that explore the diversity of constructions that characterize the grammar of ergative and other languages. The constructional approach is particularly well positioned to shed light on issues of split ergativity, with large theoretical implications. The exploding interest in the grammar of constructions provides a useful framing for new investigations of the discourse profiles of a wide variety of constructions, each potentially revealing some new aspect of the structural and functional diversity of language. Constructions combining aspects of argument structure, valence, reference, person, voice, tense, aspect and so on are promising topics for new research on syntactic alignment that incorporates a discourse-​and-​grammar dimension. For the targeted argument structure construction, it will be important to map out its information structure, as revealed in its distinctive discourse profile, operating within the relevant functional niche (Du Bois 2014).

 

56   John W. Du Bois A further step is to work out how such construction-​level discourse profiles interact with the semantics of the verbal repertoire; inherited morphosyntax; functional strategies for utterance processing; and other factors that combine to motivate the grammaticization of ergative and other alignments. A particularly incisive version of this problem concerns the three-​way interaction between split ergativity, split function, and split discourse, where each new language’s “natural experiment” differentiates contrasting elements within the scope of a single language. Insights gained from split ergativity can extend even to languages that do not themselves overtly exhibit these splits, insofar as they provide clues to concealed patterns and discourse profiles that reveal the incipient fault lines of potential change. For linguists who are prepared to use the combined tools of discourse and grammar to investigate the diversity of natural experiments in the world’s languages, the potential for discovery is open-​ended. A valuable research strategy is to (1) document a construction, or better, a contrasting set of grammatical constructions; (2) identify their respective discourse profiles; (3) clarify the connection of each to its corresponding functional niche; and (4) explore the ramifications for grammaticization. Such a research agenda is well positioned to expand and refine our understanding of the dynamic processes that motivate the grammaticization of the structures of ergative languages in all their diversity, with broad implications for understanding all forms of grammar.

2.7 Conclusions Amidst increasing interest in the complexity, diversity, and heterogeneity of grammar in all its forms and functions (Beckner et al. 2009; Page 2010; Evans 2013), the challenge of coming to terms with ergativity takes on special significance. Ergativity has long revealed an uneasy fit with the conventional categories of standard theories of grammar, including the supposedly fundamental categories of subject and object, or even A, S, and O. Ergativity calls into question the universality of such preconceived categories; yet neither can its own indigenous categories of ergative and absolutive be set up as universal structures in their stead. Ergativity and its variations and competitors challenge the linguist’s desire for easy answers. One way to engage with this complexity is by attending to the variability, and the recurrence, of patterns of language use. For example, the ergative discourse profile has been documented in a typologically diverse array of languages around the world: ergative, active, and accusative. To be sure, it does not stand alone: competing discourse forces of topicality strongly motivate ergativity’s main competitor, the accusative type. Yet whether expressed in grammar or only in discourse, the robustness of the ergative discourse profile remains, attested in its continuity across grammatical typology, historical change, child language, and genre. Its presence, sometimes variable and often submerged, is nevertheless revealed in small ways, shaping the child’s earliest productions and the occasional grammaticized construction even in accusative languages. While it

 

Ergativity in discourse and grammar    57 cannot constitute a complete account of ergativity on its own, neither can any explanatory account of ergativity be complete, realistically, without incorporating the “facts on the ground” of the ergative discourse profile, with all its consequences. As a counter to reductionism, essentialism, and idealization, the empirical trend of modern linguistics invites attention to the explanatory power of corpus-​based evidence. Typologically aware research on discourse profiles documents the complexity and diversity of vast populations of utterances—​but also the consistency that gives them power to shape the adaptive evolution of grammars. Surely the functional, structural, and historical basis of ergativity is more complex than is envisioned in any one current analysis. A full explanation of ergativity and its variants and alternatives will require the corpus-​based identification of multiple competing and converging motivations, and their integration into a larger explanatory framework. Relevant forces include the distribution of given and new information across clausal arrays of argument slots, the semantics of force dynamics in the clause, the lexicon of verbs and argument structure constructions, the learning and reanalysis of inherited morphosyntax, recurrent pathways of change and grammaticization, and more. All impact the dynamics of the discourse profile, and all come together to shape the emergence of the system of grammar. Ergativity is too important to be left to the specialists of ergative languages. It is a problem for all linguistics, and a useful one at that. Ergativity invites linguists to investigate the most fundamental structures of grammar, not only in ergative languages but in every language; and to revisit questions of function and structure, of universality and diversity, that were once thought to be settled. As the field of linguistics turns more and more to evidence-​based analysis, traditional methods of elicitation are increasingly supplemented by the empirical tools of corpus-​based and experimental methods. There is much work to be done to document the discourse profiles that shed light, not only on the syntactic alignments that broadly define ergative, accusative, and active languages, but also on the details of the rich constructional repertoires of more specialized argument structure constructions, such as passives, antipassives, perfectives, and nominalizations, that may serve as bridges for diachronic realignment from one structural type to another. Ergativity has the potential to disrupt conventional thinking and existing explanations for grammar. Let the disruption begin.

Acknowledgments I wish to thank Jessica Coon and Shobhana Chelliah for comments on an early draft that have greatly improved this chapter. For helpful input on specific points I thank Matthew Dryer and Danny Law. For conversations that have deepened my thinking on this and other questions of linguistic explanation, I am profoundly indebted to Mira Ariel.

 

58   John W. Du Bois

Abbreviations The following symbols and abbreviations are used in glosses: CP, completive aspect; DEP, dependent; DIM, diminutive; ICP, incompletive aspect; FOC, focus; LAT, lative; PL, plural; TV, transitive verb; 1.ERG, ergative 1st person singular; 3.ERG, ergative 3rd person singular; 3.ABS, absolutive 3rd person singular. In addition to the standard A–S–O symbols introduced in the first paragraph, the following are used in the syntactic schemas: X, oblique/adjunct NP; V, verb; P, preposition. Transcription symbols (Du Bois 2014b; Du Bois et al. 1993) include: ; [semi-​colon] speaker label; … pause; .. micro-​pause; : [colon] prosodic lengthening; (H) in-​breath; , [comma] continuing intonation; . [period] final intonation. Intonation units are indicated by a carriage return (one IU per line). Sakapultek data are presented in a practical orthography (Du Bois 2006: 198), in accordance with standards of the Academy of Mayan Languages, with the following special values: x voiceless laminopalatal fricative; j voiceless velar fricative; q voiceless uvular stop; tz voiceless apicoalveolar affricate; nh voiced velar nasal stop (engma). Apostrophe (’) following a consonant represents a glottalized consonant; following a vowel it represents a glottal stop. Double vowels represent phonemic length.

 

Chapter 3

Parameteri z i ng ergativit y: An i nh e re nt case approac h Michelle Sheehan

3.1  Introduction: Theta-​Roles and Inherent Case Many recent (and not so recent) approaches argue that ergative is an inherent case associated with the specifier of little v (see Levin 1983; Mahajan 1989; Woolford 1997, 2006; Massam 1998, 2006; Aldridge 2004; Anand and Nevins 2006; Legate 2006, 2012a; among others):1 (1) 

vP DPERG

v' v

...

GB versions of this proposal took inherent case to be assigned at D-​structure, as opposed to structural case, which was assigned at S-​structure. In Minimalist terms, inherent case can be thought of as a K-​projection dominating DP, which is s-​selected by a class of thematic heads, or simply as case-​valuation coupled to theta-​role assignment (Woolford 2006). The inherent-case approach to ergativity is attractive because 1  In the following discussion, I use the term ‘accusative’ to denote languages which lack anything which could be classified as ergative case and ‘non-​accusative’ to denote those languages which use ergative case in some way. This is intended to avoid the problem of referring to a rather heterogeneous class of languages as ‘ergative’ (complete with scare quotes).

 

60   Michelle Sheehan (i) inherent (theta-​related) cases appear to be independently needed in order to model the case/​agreement properties of accusative languages, making ‘ergativity’ (i.e. non-​ accusativity) less exotic, and (ii) if ergative is an inherent case, this immediately explains Marantz’s (2000 [1991]) much discussed generalization that non-​thematic subjects do not bear ergative, as many others have noted (Woolford 2006; Legate 2012a). Essentially, the inherent approach to ergative case makes four distinct kinds of predictions not made by structural or dependent case approaches: A. Ergative will only occur on (a subset of) arguments externally merged in spec vP. B. The presence of ergative may be independent of transitivity, so we might find ergative subjects without absolutive objects. C. There will be no derived/​non-​thematic ergative subjects (no ergative expletives, raising to ergative or ergative subjects of passives, ditransitives, or otherwise). D. Ergative case will not be lost in contexts where structural case is not available (no change of case under ECM, no loss of ergative under raising). While C and D can only be evaluated via in-​depth language-​specific consideration of raising, passives, applicatives, and ECM in non-​accusative languages (to the extent that they exist—​see Rezac et al. 2014; Berro and Extepare, Chapter 32 and Laka, Chapter 7, this volume, on Basque; Baker and Bobaljik, Chapter 5, this volume on Burushaski), A and B should be much easier to evaluate on a broad cross-​linguistic basis as they concern the surface distribution of ergative case.2 Nonetheless, to my knowledge, no systematic cross-​linguistic survey of the distribution of ergative case has been given in favour of the inherent-case proposal. The main aim of this chapter, then, is to fill this gap, in the context of a broader parametric account of basic alignment. The obvious challenge in relation to A is that the inherent-case proposal only has predictive power inasmuch as there is an independent theory of theta-​roles, distinguishing those arguments introduced by little v from those introduced by other (lower) thematic heads such as Appl and V. While there has been rich cross-​linguistic research on argument structure, it is still much debated exactly how many thematic distinctions need to be made syntactically and even how best these thematic distinctions should be described: by distinct theta-​positions (Baker 1988, 1997) or by the combination of proto-​roles/​features (Dowty 1990; Reinhart 2003), possibly accumulated via movement (Ramchand 2011). Nonetheless, certain proposals regarding theta-​roles have become widely accepted. There is a general (empirically grounded) consensus, for example, that certain (proto) theta-​roles are subject-​like (agent, causer, initiator) while others are more object-​like (theme, patient, undergoer, result), or lie somewhere in between (experiencer, recipient, process) (Baker 1997, 2009; Platzack 2009, 2011; Ramchand 2011;

2  Of course, the very real possibility exists that ergative has a different inherent/​structural/​dependent status in different languages. The null hypothesis, though, should be that it has the same status cross-​linguistically.

 

Parameterizing ergativity   61 but see also Bowers 2013).3 Here, I will adopt the conservative position, stemming from Baker (1988), that there are distinct theta-​roles, which are configurationally determined.4 As Folli & Harley (2007) point out, there is good reason to take agents and (animate/​ inanimate) causers to be introduced by distinct little vs: vdo and vcause respectively. The evidence for distinguishing these two theta-​roles comes from the fact that some predicates require agent subjects, and cannot take causers (Oehrle 1976; Hale & Keyser 1993; Folli & Harley 2007, among others). For example, as Oehrle (1976) showed, prepositional datives require agent subjects but the double object construction takes a causer subject, at least with verbs of transfer of possession (see also Pesetsky 1995): (2) a.  My relationship with him gave/​brought me a daughter. b. *My relationship with him gave/​brought a daughter to me. The fact that the same lexical verb occurs in both examples shows, moreover, that this is not a lexical but a syntactic effect. A similar contrast holds of certain transitive verbs in English: those derived from unaccusative change of state verbs allow causer subjects, unlike unergative activity verbs like read, which take only agents (Hale & Keyser 1993): (3) a. *The homework assignment read several books. (Intended. caused the reading of several books) b. The snowfall closed several roads. It has also been observed that in many languages anticausatives permit causers but not agents to be overtly expressed as PPs (headed by from in English) (Alexiadou & Schäfer 2006: 41): (4) a. The window broke from the pressure /​from the explosion /​from Will’s banging b. *The window broke from John The opposite pattern is observed with the by-​phrase in passives in some languages, as Alexiadou & Schäfer also note. Finally, certain ‘causative’ constructions are actually ‘agentive’, requiring an agent and not a causer subject (see Folli & Harley 2007 on Romance ‘causatives’): (5) a. The fact that it was hot in the room made/​?let/​*had Mary take off her jacket. b. The teacher made/​let/​had Mary take off her jacket.

3  Assuming that the goal theta-​role in prepositional datives is distinct from the recipient theta-​role in double object constructions, with recipients being externally merged higher than themes, unlike goals (see Pesetsky 1995; Harley 2002). 4  For our purposes here, it is not important whether theta-​roles can be acquired only by external merge or by either external or internal merge (Hornstein 1999), but see Sheehan (2014a) for arguments in favour of the second possibility.

 

62   Michelle Sheehan The fact that both agents and causers can, nonetheless, be introduced as additional arguments in ‘causative’ constructions and (to varying degrees in different languages) appear as a by-​phrase in the passive, provides strong evidence that they are both ‘external arguments’ introduced by little v. A first prediction of the inherent approach to ergative case is that agents and/​or causers can appear as ergatives. In many accusative languages, there is also evidence that the subjects of some intransitive predicates are introduced by little v. As Burzio (1986) showed, intransitive verbs divide into those which have a thematic object or theme, externally merged as the complement of V (so-​called unaccusatives) and those whose single argument is a thematic agent, now usually taken to be externally merged in spec vP (unergatives), following Koopman & Sportiche (1991).5 A further prediction of the inherent case approach is therefore that, among intransitive subjects, only the subjects of unergatives should be able to surface with ergative case.6 In line with Alexiadou & Schäfer (2006), I  take instrument subjects to reduce to either agents or causers. Differences in binding possibilities strongly suggest that subject instruments are externally merged in a higher position than PP instruments: (6)  a.  *Mary hit himi on the foot with Johni’s baseball bat b.  Johni’s baseball bat hit himi on the foot (when it fell off the shelf). It does not seem, however, that there is a dedicated theta-​position for instruments high in the clause, but rather that subject instruments behave like either agents or causers. The main evidence for this comes from the fact that not all instruments can surface as subjects (examples adapted from Alexiadou & Schäfer 2006): (7) a. The doctor cured the patient with a scalpel/​chamomile tea. b. ??The scalpel cured the patient. c. Chamomile tea cured the patient. (8) a. The crane picked up the crate. b. *The fork picked up the potato. Alexiadou & Schäfer (2006: 42) claim that instruments make good subjects when they can be conceived of as ‘acting on their own without being (permanently) controlled by 5  There is some disagreement with respect to the correct label or semantic characterization of the argument of unergative subjects, partly due to the fact that there is non-​trivial cross-​linguistic (lexical) variation regarding the unaccusative/​unergative divide (see Rosen 1984; Sorace 2000). While these issues are of course interesting and relevant, we abstract away from them here for reasons of space (see Pesetsky 1995 for relevant discussion). 6  It has sometimes been claimed, in fact, that the inherent-case approach predicts that transitive and intransitive little v should pattern alike in this respect. As we shall see, though, this does not necessarily follow if the distribution of inherent ergative is subject to more intricate parameterization (see also Legate 2012a).

 

Parameterizing ergativity   63 a human agent’, as is the case with chamomile tea and cranes (in German, Greek, and Dutch as well as English). This means that there is no independent theta-​role ‘instrument’ introduced by little v, subject instruments are either agents or causers, depending on their semantics. In non-​accusative languages, then, we expect to find ergative instruments to the extent that they are semantically permitted, but there might be quite subtle restrictions on their use. Deciding which other kinds of arguments are introduced by little v is more difficult. The subjects of verbs of perception, though often labelled experiencers, actually pattern with agents/​causers and unlike other experiencers in certain respects: they can form so-​ called agentive nominalizations (hearer, feeler, sooth-​seer) for example, and can appear in by-​phrases in the passive. Note also that verbs of perception also give rise to causative-​ like constructions in English and other languages (see Guasti 1993), again making them syntactically similar to causers/​agents as ‘external arguments’. There are other experiencers too, which pattern like this (lover, cheese-​hater, ghost-​fearer). Broadly speaking, these are the subjects of stative psych-​predicates, which, in many accusative languages, surface as subject-​experiencer predicates, taking a nominative subject (Grimshaw 1990). Eventive psych-​predicates, on the other hand, often surface as object-​experiencer predicates, taking a dative or accusative experiencer (Belletti & Rizzi 1988; Pesetsky 1995; Landau 2009). The fear/​frighten contrast illustrates this difference in English: (9) a. I fear ghosts. b. *I am fearing (some) ghosts (right now). (10) a. Ghosts frighten me. b. Some ghosts are frightening me (right now). Building on Platzack (2011) and Ramchand (2011), I take the subjects of (stative) psych-​ predicates/​verbs of perception to be holders rather than experiencers, introduced as external arguments by vhold. The experiencers in eventive object-​experiencer constructions, I take to bear the true experiencer theta-​role, introduced by a lower thematic head (Appl) (in the spirit of Belletti & Rizzi 1988; Pesetsky 1995; Landau 2010). This means that we might expect to find a split in some non-​accusative languages between stative psych-​predicates which take an ergative holder and eventive psych-​predicates which take a dative or absolutive experiencer. As we shall see, this is exactly what we find in many non-​accusative systems. Additionally, the second argument in object-​experiencer predicates is persuasively argued by Pesetsky (1995) to be either (i) an internal theme/​ target (with the unaccusative piacere class) or (ii) an external causer (with the causative preoccupare class). A further prediction, then, is that the non-​experiencer argument of an eventive psych-​predicate will also be able to surface with ergative case, again something which is born out in many cases. Following Pylkkänen (2008) among others, I take recipients and benefactives to be introduced by low/​high Appl heads respectively. Finally, themes/​patients and goals, I assume to be externally merged low down inside VP. This gives the following

 

64   Michelle Sheehan range of thematic positions in the clause (not all of which can co-​occur—​see, again, Pesetsky 1995): (11)  [vP agent/​causer/​holder vagent/​vcauser/​vholder [ApplP benefactive Applbenefactive [VP V [ApplP recipient Applrecipient theme ([PP P goal])]]]]7 The prediction for ergative languages is therefore that only holders, agents, and causers (including instruments and the non-​experiencer arguments of eventive psych-​ predicates) should be able to bear ergative case. In the remainder of this chapter I  show that this prediction holds for a range of non-​accusative languages, but that languages vary as to exactly which subset of little vs assigns ergative case. In section 3.2 I consider the differing distribution of ergative case in Basque, Hindi, Tsez, Lezgian, Trumai, Cavineña, and Chamorro. I argue that the best way to account for this variation is via a series of parameters arranged in transitive dependencies in the general way proposed by Roberts (2012). The section further considers two additional dependent parameters active only in languages with transitive-​ sensitive ergative case, determining the presence of syntactic ergativity (understood in the narrow sense) and the source of absolutive case (T or v). Section 3.3 briefly discusses the resultant parameter hierarchy and its theoretical status as well as raising some questions for future research. Finally, section 3.4 concludes.

3.2  The (Parameterized) Distribution of Ergative Case This section considers variation in the distribution of ergative case across languages. It does not consider, for reasons of space, the kind of split-​ergativity that Sheehan (2015) calls ‘variable alignment’, whereby the same predicate in the same language displays different case/​agreement properties depending on syntactic context (root/​embedded and/​or tense/​aspect properties of the clause, person/​animacy of the subject). This kind of variability, I assume, can be attributed to independent facts about the languages in question which serve to obscure basic alignment in certain contexts, rather than to variable parameter settings (see Laka 2006a, Chapter 7, this volume; Coon 2013a; Coon & Preminger 2012, Chapter 10, this volume; see also Woolford, Chapter 9, this volume). As Hindi, Basque, and Tsez show aspect-​sensitive variability, Chamorro mood-​based

7 

Note that the theme appears as the complement of V where no recipient is present. Where a recipient is present, I assume that the theme is the complement of Appl, for the reasons discussed in Pylkkänen (2008). Likewise, where a PP goal is present, the theme is the specifier of P. For this reason, theta-​roles are configurationally determined in a relative rather than an absolute sense. Thanks to Lisa Travis for asking me to clarify this point.

 

Parameterizing ergativity   65 variability, and Yidiɲ personal sensitive variability, I focus on the distribution of ergative case in ergative contexts here, to control for this effect.8

3.2.1 Ergative Unergatives: Basque and Hindi In Western Basque (henceforth Basque), it has been claimed that the subjects of unergatives surface with ergative case, while the surface subjects of unaccusatives surface with absolutive (Laka 2006b).9, 10 (12)

a.  Txalupa hondora-​tu  boat.def.abs sink.perf ‘The boat sank.’ 

da.  is

b.  Klara-​k ondo  eskia-​tzen  Klara-​erg well ski.impf ‘Klara skis well.’ 

du. has 

[Basque] (Laka 2006b: 376)

(Laka 2006b: 379)

The fact that the Basque verb hondura ‘sink’ occurs with an absolutive subject in (12a) follows if the surface subject is base-​generated as the complement of V (i.e. it is unaccusative). The surface subject of the (unergative) verb eskia ‘ski’, on the other hand, surfaces as ergative because it base generated in spec vP. This structural difference is also illustrated by the different auxiliaries selected by the two kinds of predicates in Basque (‘be’ vs. ‘have’). While many unergative Basque verbs are N-​do compounds and hence might be considered transitive under some definitions (Bobaljik 1993a; Laka 1993b), Laka (2006b), and Preminger (2012) show that this is not true of all unergative verbs, as is obvious from (12b). It therefore seems to be the case that in Basque intransitive little v also assigns ergative case to its specifier.

8 

An anonymous reviewer asks about the status of tripartite systems in this approach. Sheehan (2014b) shows one option regarding how such systems might be accommodated in the hierarchy, based on the idea that absolutive on transitive objections in some languages is underlying equivalent to accusative Case (Legate 2006, 2012a). This would mean that whether a low ABS non-​accusative language is ergative or tripartite is purely a matter for the morphology. I leave a discussion of this complication to one side here for reasons of space (see also Müller and Thomas, Chapter 12, this volume for arguments that genuine tripartite systems do not exist). 9  The following languages have been argued to behave similarly in this respect: Georgian, Kartvelian (Harris 1982); Chol, Mayan (Coon 2013a); Lakhota, Siouan; Caddo, Caddoan (Mithun 1991b), Hindi (Bhatt 2003). We consider Hindi shortly. 10  Guaraní also displays a split regarding the behaviour of intransitive verbs. While Mithun (1991b) characterizes this as an active–​stative split, Velázquez-​Castillo (2002) suggests that matters are more complex than this. Like the other split-​S systems discussed here, Guaraní is not syntactically ergative (Velázquez-​Castillo 2002), but it is not totally clear at present how to fit this language into the parameter hierarchy presented here. I therefore leave this as a matter for future research.

 

66   Michelle Sheehan In fact, closer examination of Basque suggests that all arguments introduced by little v in transitive and intransitive contexts seem to bear ergative case. Consider first agents and animate/​inanimate causers:11 (13) Maddi-​k sagarr-​a jan du. Maddi-​erg  apple-​det  eat  3sabs.aux.3serg ‘Maddi ate the apple.’

[Basque] (Oyharçabal 1992: 313)

(14) (Nik) zuri lan egin  arazi dizut. 1.erg  you.dat  work  do caus  aux.2dat.1erg ‘I made you work.’ (Oyharçabal 1992: 332 fn.) (15) Haize-​a-​k ate-​a ireki du. wind-​det-​erg  door-​det  open  3sabs.aux.3serg ‘The wind opened the door.’ Even instruments can surface with ergative case in Basque, as noted by Woolford (2006). This follows if, as discussed, subject instruments are actually causers/​agents: (16) Giltz-​a-​k ate-​a ireki zuen. key-​det-​erg  door-​ det  open  3sabs.aux. pst.3serg ‘The key opened the door.’ (Woolford 2006: 124, citing unpublished work by Juan Uriagereka) Turning to experiencers, we find that the latter surface either with dative, absolutive, or ergative case, depending on predicate type (Etxepare 2003). These three options seem to be equivalent to the three kinds of psych-​predicates identified by Pesetsky (1995) and Landau (2010), building on Belletti and Rizzi (1988). Thus, unaccusative object experiencer verbs (the piacere class) surface with a dative experiencer and an absolutive theme/​target (see Etxepare 2003; Rezac 2008): (17) Ni-​ri zure oinetako-​a-​k gustatzen  zaizkit. I-​dat  your  shoes-​det-​pl  like 3plabs.aux.1sdat ‘I like your shoes.’ (Woolford 2006: 115, citing Austin and Lopez 1995: 12) Other verbs in this class include interesatu ‘to interest’ and dolutu ‘to repent’ (Etxepare 2003: 39). Causative object experiencer psych-​predicates follow an erg–​abs pattern, whether or not the causer is animate (and hence potentially agentive) or inanimate:

11 

Thanks to Maia Duguine for help with the Basque data.

 

Parameterizing ergativity   67 (18)

Mikel-​ek ni Michael-​erg  I.abs  ‘Michael angered me.’

haserretu nau. angry-​perf  1sabs.aux.3serg (Woolford 2006: 124, citing Manandise 1988: 118)

(19)

Berri-​ek (ni)  haserretu  naute. new-​det.pl.erg  I anger 1sabs.aux.3plerg ‘The news angered me.’

Verbs in this class, which often alternate with an intransitive form, include kezkatu ‘to worry, to become worried’, gogoratu ‘to remember’, zoratu ‘to madden’ (Etxepare 2003: 41). The ergative here is as expected if the non-​experiencer argument is a causer rather than a theme, as discussed. Moreover, the absolutive case on the experiencer is plausibly equivalent to the accusative case received by experiencers in the equivalent accusative structures (the preoccupare class) (but see Landau 2009 for complications). There is a class of experiencers, though, including the subjects of verbs of perception, which surface with ergative case (including miretsi ‘to admire’, gutxietsi ‘to despise’, desiratu ‘to desire’—​Etxepare 2003: 41) (20) Ni-​k asko ikusi  ditut I-​erg  many.abs  seen 3plabs.aux.1serg  ‘I saw many.’ (21)

Jon-​ek liburu hauek  nahi/​ezagutzen  Jon-​erg  book these want/​know ‘John wants/​knows these books.’

(Rezac et al. 2014: 6)

ditu. 3plabs.aux.3serg

These appear to be equivalent to subject-​experiencer (temere class) verbs in accusative languages. The case patterns observed in Basque are therefore exactly as expected if all little vs assign ERG to their specifiers and thematic structure in Basque follows expected universal patterns.12 Hindi shares many of these properties with Basque even to the extent that unergative but not unaccusative predicates can surface with ergative subjects. In Hindi, though, few intransitive verbs actually require ergative subjects, the verb ‘to bathe’ being a rare exception in this respect: (22)

Raam*(-​ne)  nahaayaa Ram-​erg bathe.perf ‘Ram bathed.’

(Mohanan 1994a: 71)

12  It may turn out, of course, that the inherent case approach to Basque fails for reasons C–​D (as Rezac et al. 2014 argue). If this is the case, the implication is merely that the approach to non-​accusative alignment here is incomplete as there are other kinds of ergative languages in which ergative is not an inherent case. I leave this as a matter for ongoing research. See also Laka, Chapter 7, this volume for a defense of the inherent ergative approach to Basque.

 

68   Michelle Sheehan The verbs which behave in this way appear to have a reflexive meaning. In most cases, intransitive verbs take ergative subjects only optionally. More precisely, a subset of the class of verbs which satisfy independent diagnostics for unergativity can surface with either an absolutive or ergative subject (Bhatt 2003; Butt, Chapter  33, this volume; Surtani, Jha, & Paul 2011, Surtani & Paul 2012).13 This includes jhool ‘swing’, dauR ‘run’, kood ‘jump’ naac ‘dance’, hans ‘laugh’ tair ‘swim’, gaa ‘sing’, kʰel ‘play’ and chillaa ‘shout/​ scream’ (Mahajan 1990; Mohanan 1994a; Surtani, Jha, & Paul 2011). As Mohanan (1994a) shows, however, the choice between an absolutive and ergative subject with these verbs is not purely optional but correlates with a semantic difference: ergative subjects imply volition in intransitive contexts: (23)  raam-​ko acaanak šer dikʰaa. vah/​us-​ne cillaayaa [Hindi] Ram-​dat suddenly lion.abs  appear.perf he/​he.obl-​erg  scream.perf  ‘Ram suddenly saw a lion. He screamed.’ (Mohanan 1994a: 71) (24) Us-​ne/​*vah jaan buujʰkar  cillaayaa He.obl-​erg/​he.abs  deliberately shout.perf  ‘He shouted deliberately.’

(Mohanan 1994a: 72)

Verbs which can be independently diagnosed as unaccusative, however, never take ergative subjects, even where volition is implied (gir ‘fall’ soo ‘sleep’, jaa ‘go’, phail ‘spread’): (25) Raam(*-​ne) giraa Ram-​erg fall.perf ‘Raam fell hard.’  The implication seems to be that unergative verbs in Hindi can, but need not, take ergative subjects; where they do, the subject is interpreted as volitional. These facts are consistent with ergative being an inherent case but they are not immediately explained by such an account. What is required is some further parameter differentiating intransitive little v in Basque from intransitive little v in Hindi.14 Because of these facts, Mohanan (1994a) proposes to associate ergative case directly with the semantic feature [volition] across the board. The problem with this idea,

13 

These diagnostics include (i) the possibility of a cognate object, (ii) participation in impersonal passives, (iii) non-​participation in participial relatives, (iv) compound verb selection of le ‘take’, de ‘give’, daal ‘did’ and not jaa ‘go’, and (v) inability to appear without genitive marking in non-​finite clauses. 14  Another complication in Hindi is that a small number of transitive verbs (verbs taking absolutive/​ ko-​marked objects) fail to assign ergative case (bolnaa ‘speak’ and laanaa ‘bring’) or do so only optionally (samajʰnaa ‘understand’ and jannaa ‘give birth to’) (see Mohanan 1994a). These appear to be idiosyncratic lexical gaps.

 

Parameterizing ergativity   69 though, is that ergative case assignment with transitive predicates is insensitive to [volition]. In fact, the distribution of ergative case across transitive predicates in Hindi follows very closely the Basque distribution. Non-​volitional animate and inanimate causers receive ergative just as volitional agents do. This is true of instruments as well as the causers in object-​experiencer constructions:15 (26) Havaa-​ne patte bikʰer diye tʰe Wind-​erg  leaves  scatter  give.perf  be.pst ‘The wind scattered the leaves.’

(Mohanan 1994a: 75)

(27) [Mina-​ke cillaa-​ne]-​ne sab-​ko Daraa  diyaa Mina-​gen  scream-​inf/​ger-​erg  all-​dom  scare give.perf Mina’s screaming scared everyone. (28) nayii  khabaroN-​ne  Sita-​ko dukhii  kar  diyaa new news-​erg Sita-​dom  sad do give.perf The new news saddened Sita. (29) caabhii-​ne  taalaa  khol-​aa key-​erg lock open.tr-​perf The key opened the door. Finally, holders of states also receive ergative with both verbs of perception and subject-​experiencer verbs: (30) tuṣaar-​ne caand dekʰaa Tushar-​erg  moon  see/​look.at.perf  ‘Tushar saw the moon.’ (31)

tuṣaar-​ne vah kahaanii  yaad kii Tushar-​erg  that  story memory  do.perf ‘Tushar remembered that story.’

(Mohanan 1994a: 141)

(Mohanan 1994a: 141)

Eventive object-​experiencers versions of these verbs surface with dative subjects: (32) tuṣaar-​ko caand dikʰaa Tushar-​dat  moon  appear.perf  ‘Tushar saw the moon.’ (lit. The moon appeared to Tushar.)

15 

Thanks to Rajesh Bhatt for help with the Hindi data.

(Mohanan 1994a: 141)

 

70   Michelle Sheehan (33)

tuṣaar-​ko vah kahaanii  yaad aayii Tushar-​dat  that  story memory  come.perf ‘Tushar remembered that story.’ (Mohanan 1994a: 141) (lit. The memory of that story came to Tushar.)

The correct characterization of the distribution of ergative case appears to be that it surfaces wherever v is transitive and additionally in intransitive volitional contexts (a non-​ natural class). The pattern observed in Basque falls out straightforwardly from a parametric approach along the lines proposed by Roberts (2012). Assuming that the basic alignment parameter concerns whether transitive v assigns a theta-​related case, an alignment of the Basque kind arises where ergative is generalized to all little vs via ‘input generalization’, an acquisition strategy. The Hindi system arises where instead of generalizing the transitive system to all intransitives, the system is simply extended to a subset of all possible contexts (i.e. volitional intransitive vs). The non-​natural class of ergative DPs arises as a result of this extension procedure. This can be represented via the following parameter hierarchy: (34) Parameter hierarchy (first version, to be extended and revised) P1: Does transitive v assign a theta-related case (ERG) in a language L? N - accusative

Y– non-accusative P2: is this generalised to all vs in L? Y – Basque

N – P3: is this extended to intransitive [volition] vs in L? Y – Hindi

N-…

The rather idiosyncratic Hindi system therefore reduces to an extension of the basic non-​accusative system. In the following sections, we see that further parameterization is also required in order to account for the distribution of ergative case in other contexts.

3.2.2 Instigators Only: Tsez and Lezgian The distribution of ergative case in Tsez and Lezgian suggests that a further kind of parameterization must be added to (34). In these languages, transitive agents and animate/​inanimate causers receive ergative case, just as in Basque and Hindi (Comrie 2004; Polinsky 2015. on Tsez; Haspelmath 1993 on Lezgian):16 16  Tsez also marks accidental causation via the poss-essive case, though this is optional (Comrie 2004: 118; Polinsky 2015: 144–146) and does not seem to occur on inanimate causers (Bernard Comrie and Maria Polinsky, p.c.).

 

Parameterizing ergativity   71 (35)

žek’-​ā bišwa r-​ac’-​xo man-​erg  food.abs.iv  iv-​eat-​pres  ‘The man is eating the food.’

[Tsez] (Comrie 2004: 115)

(36) žek’-​ā is b-​exu-​r-​si Man-​erg  bull.abs.III  III-​die-​caus-​pst.wit ‘The man killed the bull.’

(Comrie 2004: 116)

(37) C’i-​d-​ä ʕаˤɣur y-​iku-​r-​si. fire-​erg  mill.abs.II  II-​burn.intr-​caus-​pst.wit ‘Fire burnt the mill.’

(Polinsky 2015: 138)

[Tsez]

(38) Nes-​ä ža kaɣat kid-​be-​q t’et’r-​er-​si. dem.i-​erg  dem  letter.abs.ii  girl-​os-​poss.ess  read-​caus-​pst.wit ‘He made the girl read that letter.’ (Polinsky 2015: 160) An apparent difference between the languages concerns instruments. In Tsez, instruments can also surface with ergative case, as expected: (39) Yiła rek-​ä ħišimuku r-​aˤɣi-​x. dem  key-​erg  lock.abs.iv  iv-​open-​pres ‘This key opens the lock.’

[Tsez] (Polinsky 2015: 139)

While this was also previously true of Lezgian instruments, Haspelmath (1993: 84) claims that ergative instruments are ‘never used in the modern language’, raising some potential issues for the collapsing of instrument subjects with agents/​ causers. This is where the similarities with Hindi and Basque end, however. The subjects of intransitive verbs always receive absolutive case in Tsez, regardless of the unaccusative/​ unergative distinction, which is independently diagnosable in the language (Polinsky 2015: 121–123): (40) is b-​exu-​s bull.abs.III  III-​die-​pst.wit  ‘The bull died.’ (41) Ecru žek’u qoqoƛi-​s old man.abs laugh-​pst.wit  ‘The old man laughed.’

[Tsez]

(Comrie 2004: 115)

While this is also generally the case in Lezgian, there is a class of verbs derived from N-​ do compounds which take ergative subjects despite appearing intransitive:

 

72   Michelle Sheehan (42) ada k’walax-​zawa  she.erg  work-​impf ‘She was working.’

[Lezgian]17 (Haspelmath 1993: 284)

Based on a range of facts, Haspelmath proposes that such examples involve incorporation of the nominal into the (light) verb do, giving rise to absolutive absorption in an otherwise transitive structure.18 In other cases, though, Lezgian makes no distinction between unaccusative and unergative verbs, with both taking absolutive subjects (e.g. change of state verbs such as kisun ‘fall silent’, kusun ‘fall asleep’, ifin ‘become hot’, q̃un ‘become cold’, q’ežin ‘become wet’ and the verb of motion juzun ‘move’, which are presumably unaccusative, as well as verbs which are presumably unergative: q̃uǧun ‘play’, qʰürün ‘laugh’, q̃eq̃ün ‘walk around’—​Haspelmath 1993: 271): (43)  a. stxa k’wal.i-​z xta-​na brother.abs  house-​dat  return-​aor  ‘The brother came back home.’

[Lezgian] (Haspelmath 1993: 5)

b. q’if wiči-​n t’ekwen  galaj.pataqʰ kat-​na. mouse.abs  self-​gen  hole toward run-​aor ‘The mouse ran towards its hole.’ (Haspelmath 1993: 223) In both Lezgian and Tsez, then, intransitive v does not assign ergative, assuming that in examples like (42), little v is formally transitive.19 Another difference between Tsez/​Lezgian vs. Basque/​Hindi is that holders never seem to surface with ergative case in the former. The subjects of verbs of perception and psych-​predicates surface with dative in Lezgian and either lative or absolutive in Tsez: (44) Zamira.di-​z Diana aku-​na. Zamira-​dat  Diana  see-​aor  ‘Zamira saw Diana.’ (lit. Diana was visible to Zamia.)

17 

[Lezgian] (Haspelmath 1993: 270)

Further examples include the verbs meaning ‘howl’ and ‘dance’ (Haspelmath 1993: 284). This recalls Hale and Keyser’s (1993) analysis of unergative verbs as well as Bobaljik’s (1993a) account of Basque. 19  In Lezgian, verbs taking an oblique complement can also surface with either an absolutive or ergative subject. As is the case with Warlpiri (Legate 2012a) this appears to depend on the thematic status of the subject. Agents appear to surface with ergative (q̃arǧi  šun ‘curse’, ewerun ‘call’, ikramun ‘bow to’, taʔsirun ‘influence’, hürmetun ‘respect’) (Haspelmath 1993: 284), whereas non-​agents are absolutive alatun ‘falls off, passes, exceeds’, eläč’un ‘crosses’, agaq’un ‘reaches’ ac’un ‘becomes full of ’, gaw ‘is near’ (Haspelmath 1993: 272–​277). An apparent outlier is raxun ‘talks to’, which takes an absolutive subject, despite being apparently agentive. 18 

 

Parameterizing ergativity   73 (45) Elu-​r mašina c’aq’ b-​et-​äsi yoł. [Tsez] 1pl-​lat  car.abs.iii  very  iii-​want-​res.ptcp aux.pres  ‘We badly need a car.’ (lit.: car is wanted to us)   (Polinsky 2015: 156) (46) [Yedu kid] [meži-​z ɣwˤay-​q] y-​uƛ’-​xo. dem girl.abs.ii  2pl-​gen2 dog-​os-​poss.ess  ii-​fear-​pres ‘This girl is afraid of your dog.’ (lit.: fears on your dog)

(Polinsky 2015: 157)

Polinsky (2015) further shows that these verbs can be causativized in Tsez, giving rise to two different patterns. The first pattern is as expected: an external causer is added: (47) Eni-​y-​ä debe-​q yedu čorpa b-​et-​ir-​xo. mother-​os-​erg  2sg-​poss.ess  dem.ni  soup.abs.IV  IV-​like-​caus-​prs ‘The mother is making/​will make you like this soup.’    (Polinsky 2015: 165) In the second pattern, however, causativization serves merely to alter the case and theta-​ role of the subject, making it ergative and agentive: (48) Madin-​ä [gagali-​s maħ] b-​iy-​r-​si. Madina-​erg  flower-​gen  smell.abs.III  III-​know-​caus-​pst.wit ‘Madina smelled flowers.’ [ERG-​agent, ABS-​stimulus] (Polinsky 2015: 163) Based on other diagnostics such as binding and Control, it appears that the like-​type verbs are unaccusative and equivalent to the piacere class in Italian (see Polinsky 2015: 163–166), meaning that the smell-​type verbs are presumably subject-​experiencer verbs. If this is the case, then holders in these languages are not ergative. Finally, Tsez also has what appear to be object-​experiencer verbs of the preoccupare class which take an ergative causer (ambiguous between direct and indirect causation) and an absolutive experiencer: (49) Meži-​z ɣwˤay-​ä kid y-​uƛ’-​er-​xo. 2pl-​gen  dog-​erg  girl.abs.ii  ii-​fear-​caus-​pres  ‘Your dog frightens the girl.’

[Tsez] (Polinsky 2015: 168)

In this way, both Tsez and Lezgian limit ergative to (animate and inanimate) causers and agents, with dative case surfacing on holders.20 In order to capture the basic 20 

A further difference, not encoded in the hierarchy is that Lezgian but not Tsez can have ergative subjects in the absence of absolutive case with a small class of intransitives (derived from N-​do compounds) and with bivalent verbs taking oblique complements. It seems that this difference is due to an independent parametric difference between the languages regarding what counts for transitivity, a discussion of which would take us too far afield here.

 

74   Michelle Sheehan behaviour of these languages a further kind of parameter must be added to the proposed hierarchy: (50) (Parameter hierarchy (second version, to be extended and revised)

P1: Does transitive v assign a theta-related case (ERG) in a language L? N - accusative

Y – non-accusative P2: is this generalised to all vs in L? Y – Basque

N – P3: is this extended to intransitive [volition] vs in L? Y – Hindi

N – P4: is this restricted to [instigator] vs in L?

Y – Tsez, Lezgian

N…

The pattern in Tsez and Lezgian suggests that in addition to generalizing and extending the distribution of ergative case it must also be possible to restrict it to a subset of transitive vs. As discussed in section 3.3, the parameters in (50) are not intended to be pre-​specified by Universal Grammar, nor is their format intended to be fixed. Rather, the hierarchy, it is proposed, emerges as the result of acquisition based on the acquisition strategies of feature economy, input generalization and analogy (see Roberts 2012).

3.2.3 Default Transitive Ergativity: Trumai and Chamorro The languages discussed up to now are all morphologically rather than syntactically ergative in that they do not have syntactic operations sensitive to the transitive/​intransitive subject distinction. It is well known, however, that many non-​accusative languages do display a form of syntactic ergativity in that they prohibit (straightforward) A-​bar extraction of ergative DPs.21 Following the general approach in Aldridge (2004, 2008a, 2008b) and Coon et al. (2014), I assume that this property results from the presence of a movement-​triggering EPP feature on ergative-​assigning little v.22 21  An anonymous reviewer asks about the other apparent instances of syntactic ergativity such as topic chaining in Dyirbal. These apparent instances of syntactic ergativity remain controversial (see Legate 2012a). The ban on A-​bar extraction is a much more robust effect, attested in many unrelated languages, and I therefore limit the discussion to this narrower definition of syntactic ergativity here (see also Aldridge 2008a). The typologically and genetically diverse languages displaying this restriction include some Mayan languages (Assmann et al. 2012; Aissen, Chapter 30, this volume; Campana 1992; Coon et al. 2014; ), some Austronesian languages (Tongan, Otsuka 2006; Otsuka, Chapter 40, this volume; Tagalog, Seediq, Aldridge 2004; 2012b; Chamorro, Chung 1982), some Eskimo–​Aleut languages (Manning 1996) as well as Australian and Brazilian languages: Dyirbal (Dixon 1979, 1994), Trumai (Guirardello-​Damian 2010), Karitiana (L. Storto, p.c.), and Katukina (Queixalós 2012; Queixalós, Chapter 42, this volume). 22   As an anonymous reviewer reminds me, the idea that syntactic ergativity can be attributed to movement of the absolutive DP past the ergative DP was first proposed by Bittner and Hale (1996a).

 

Parameterizing ergativity   75 (51)

vP DPABS

v' DPERG

v' vEPP

VP V

t

This feature attracts the absolutive argument to the external specifier of v and serves to trap the ergative DP inside the vP phase.23 Implementations of this basic idea differ in Aldridge and Coon et al.’s work. I assume that it is an effect of anti-​locality (see also Erlewine 2016). This follows if only the outermost specifer of vP can avoid the phase impenetrability condition (PIC). In this way the absolutive DP is always available for further extraction and lower adjuncts/​locatives can raise to a further external spec vP, avoiding the PIC, past both the ergative and absolutive DPs. The ergative DP is, however, doomed to be trapped inside vP as it can never raise to the external specifier of vP, due to anti-​locality (one cannot raise from the internal specifier of a given head to its external specifier).24 In parametric terms then, syntactic ergativity of this kind arises where a dependent parameter associates an EPP feature with the ergative property, in the following way (see also Sheehan 2014b): (52) Parameter hierarchy (third version, to be extended and revised)

P1: Does transitive v assign a theta-related case (ERG) in a language L? N - accusative

Y – non-accusative P2: is this generalised to all vs in L?

Y – Basque

N – P3: is this extended to intransitive [volition] vs in L?

Y – Hindi Y – Tsez, Lezgian

N – P4: is this restricted to [instigator] vs in L? N – P5: does verg bear an EPP in L?

N – Warlpiri, Cavineña(?)

Y – Chamorro, Trumai

A direct consequence of (52) is that P5 (the syntactic ergativity parameter) is only active in languages which answer Y to P1 and N to P2–​P4. In other words, a prediction is that syntactic ergativity will be possible only in languages which have default (transitive) 23  Note that this movement must be covert in VSO languages (Aldridge 2004). In many languages, it nonetheless triggers a Diesing-​type effect on the absolutive DP. I leave a full exploration of the relationship between syntactic ergativity, object interpretation, and word order to one side here. 24  One assumption that needs to be made is that this movement does not lead to ‘tucking in’ (Richards 1997).

 

76   Michelle Sheehan ergative alignment rather than generalized (Basque), extended (Hindi), or restricted (Tsesz, Lezgian) ergative alignment. This is a one-​way implication, though, and there can be languages which are only morphologically ergative, such as Warlpiri, but which have default ergative alignment (ergative on transitive causers, agents, and holders—​Legate 2012a).25 Chamorro is an example of a syntactically ergative language displaying default non-​ accusative alignment, whereby ergativity is tied very closely to transitivity. In the realis mood, the verb in Chamorro displays ergative agreement with transitive subjects only, but this agreement is lost where such subjects are A-​bar extracted, and the infix -​um-​is added instead (Chung 1982). This kind of morphological compliance strategy is common in syntactically ergative languages (see Dixon 1994 on antipassives; Coon et al. 2014, Erlewine 2016, on Mayan Agent focus): (53) a. Ha-​fa’gasi  si Juan i kareta.  erg.3s-​wash  unm  Juan  the car ‘Juan washed the car.’ b. Hayi  f-​um-​a’gasi  i kareta?  who um-​wash the car ‘Who washed the car?’

[Chamorro]

(Chung 1982: 49)

Extraction of objects and intransitive subjects, however, can proceed straightforwardly:26 (54) Hayi  na  famalao’an  man-​ma’pus ?  who l women pl-​leave ‘Which women left?’

(Chung 1982: 46)

(55) Hafa ha-​fahan si Maria gi tenda?  what  erg.3s-​buy  unm  Maria  loc  store ‘What did Maria buy at the store?’

(Chung 1982: 51)

In addition to agents, (animate/​inanimate) causers trigger ergative agreement, including the subjects of object experiencer verbs, as do holders (chat-​‘hate’ patterns

25  An anonymous reviewer points out that some Mayan languages (Ixil and Chuj) are syntactically ergative in this sense but require unaccusative subjects to be co-​indexed by set A (ergative) marking on the verb in durative/​progressive aspect (see Assmann et al. 2015). Crucially, these apparently ergative intransitive subjects can be A-​bar extracted unlike their transitive counterparts. While a full discussion of these facts is beyond the scope of the present discussion, Coon (2013a) and Coon and Preminger (Chapter 10, this volume) argue convincingly that aspect-​sensitive splits of this kind in Mayan result from embedded nominalizations, so that the set A marking is actually genitive, rather than ergative, case (the two are often homophonous in Mayan). For this reason, these examples are only a superficial counterexample to the prediction. 26  Where obliques are extracted, the clause must be nominalized, however (Chung 1982: 51). It is not clear why this should be the case or to what extent this holds in other syntactically ergative languages.

 

Parameterizing ergativity   77 like ya-​ ‘like’, ga’o-​ ‘prefer’, gusto-​ ‘like’, ga’ña-​ ‘prefer’ in this respect—​Chung 1982, 1998, p.c.): (56) Hu-​punu’ i lalu’  ni niuis. erg.1s-​kill  the  fly obl  newspaper  ‘I killed the fly with the newspaper.’

[Chamorro] (Chung 1982: 51)

(57) a. Ha-​istotba ham [na malagu’ i lahi-​nmami  ni kareta]. erg.3s-​disturb us  comp  want  the son-​our obl  car ‘That our son wants the car disturbs us.’ b. Ha-​istototba yu’ si Juan. erg.3s-​disturb.impf  me unm  Juan ‘Juan is disturbing me.’ (58) Ha-​chatli’i’ yu’ atyu  na taotao. erg.3s-​hate  me  that lk  man ‘That man hates me.’

(Chung 1982: 54–​55)

(S. Chung, p.c.)

In causative object-experiencer constructions, experiencers are absolutive (i.e. they do not trigger agreement as in (57)). Where they take an oblique complement, psych-​ predicates also take an absolutive subject: (59) Ma’a’niao  yu’  ni ga’lagu.  fear I obl  dog ‘I am afraid of the dog.’

[Chamorro] (Chung 1982: 51)

The subjects of intransitive predicates never trigger ergative agreement, regardless of the thematic status of the subject (in realis mood). In all cases, then, it seems that v inflects for ergative agreement exactly where it is transitive, due to its positive setting of P1 (and negative setting of P2–​P4).27 Trumai is another syntactically ergative language which displays a strong connection between transitivity and ergative case (Guirardello-​Damian 2003, 2010). In Trumai where an absolutive argument (subject or object) is relativized, the verb is modified by the relativizer ke: (60) Ha  hu’tsa  chï_​in [axos-​a-​tl]i [Øi esa-​t’ ke] [Trumai] I see foc/​tense  child-​ev-​dat  dance-​nzr.past rel  ‘I saw the boy who danced.’    (Guirardello-​Damian 2010: 218)

27 

No data is available regarding the behaviour of instrument subjects (S. Chung, p.c.).

 

78   Michelle Sheehan Where a (transitive) ergative subject is relativized, however, the verb is modified by chïk:28 (61) Ha  hu’tsa  ka_​in [axos-​a-​tl]i [ha  aton  mud  husa-​t’ chï-​k Øi] I see foc/​tense  child-​ev-​dat  1 pet neck tie-​nzr.past  rel ‘I saw the boy who tied my pet.’ (Guirardello-​Damian 2010: 219) As in Chamorro, agents and animate/​ inanimate causers alike take ergative case (Guirardello-​Damian 2003), including the subjects of causative object experiencer verbs (Guirardello-​Damian 2010: 221): (62) [sud yi]-​k [pike xop yi] mahan.  wind yi -​erg  house mouth yi  close ‘The wind closed the door.’

[Trumai] (Guirardello-​Damian 2003: 201)

(63) [martelu yi]-​k [atlat]  mapa. hammer  yi-​erg  pan break ‘The hammer broke the saucepan (by falling on it).’ (Guirardello-​Damian 2003: 201) (64) hai-​ts Yakairu-​ø  sa ka 1-​erg Yakairu dance  caus  ‘I made Yakairu dance.’ 

(Guirardello-​Damian 2003: 210)

It is even possible to have two ergatives following the causativization of a transitive verb: (65) Hai-​ts  chï_​in Atawa-​k atlat-​ø mapa ka 1-​erg foc/​tense  Atawak-​erg  pan-​abs  break  caus ‘I made Atawak break the pan.’ (Guirardello-​Damian 2003: 210) Somewhat surprisingly, instruments are reported never to be ergative: (66) chavi  letsi     [ pike xop yi]  mahan.  [Trumai] key instr  house  mouth  yi close Lit. ‘pro closed the door with a key.’ (Guirardello-​Damian 2003: 201) An apparent difference between Chamorro and Trumai is that in the latter, holders are never ergative, but rather surface with absolutive case (with verbs of perception like see, hear, smell, feel as well as subject experiencer verbs: like, think, believe, forget, remember):

28 

Note, though, that this is true also where recipients are relativized (Guirardello-​Damian 2010: 219).

 

Parameterizing ergativity   79 (67) axos-​ø hu’tsa  de kasoro-​tl child-​abs  see already  puppy-​dat  ‘The child saw the puppy.’

[Trumai] (Guirardello-​Damian 2003: 204)

A crucial point here, though, is that the theme/​target complement in such cases is always dative, meaning that the verbs in question are not formally transitive. Note that this situation is different from that displayed in Tsez/​Lezgian where it is the holder which receives dative/​lative while the theme/​target is absolutive. In Trumai, unlike in Warlpiri and Lezgian oblique arguments can never count for transitivity. This is true even where the subject is an agent: if the complement of V is oblique, then the subject remains absolutive. This is the case with the heterogeneous class of verbs of routine events translating variously eat, drink, cook, roast, kiss, hunt, fish, which take absolutive subjects and dative complements. In fact, as Guirardello-​Damian notes, many abs–​ dat verbs are direct synonyms of erg–​abs verbs: kapan/​chuda ‘make/​produce’, disi/​ fa ‘kill/​beat’ tako/​make ‘bite’, tuxa’tsi/​dama ‘pull’, padi/​fatlod ‘wait’. As long as dative objects never count for transitivity in Trumai, then it shares with Chamorro the default ergative alignment whereby ergative case is simply tied to transitivty. Intransitive verbs, whether unergative or unaccusative always take absolutive subjects (Guirardello-​ Damian 2003: 196).29

3.2.4 Parameterizing the Source of Absolutive Case A final point of variation between non-​accusative languages concerns the source of absolutive case (Aldridge 2004, 2008a; Legate 2006, 2012a; Coon et al. 2014). Whereas in some languages it appears that absolutive case has a mixed source, coming from v in transitive and T/​Asp in intransitive contexts (mixed ABS), in other languages it appears to come from T/​Asp across the board, leading to what has been called a ‘high ABS’ system. The evidence for this comes from the distribution of absolutive in non-​finite contexts, where T/​Asp loses it structural Case-​assigning capabilities. It has been observed that at least in some accusative languages, non-​finite T/​Asp fails to assign nominative case and, for this reason, the only possible subject of such clauses is PRO (possibly derived via movement, at least in some cases—​Hornstein 1999; Sheehan 2014a). In some non-​accusative languages, we see that the absolutive on transitive objects is retained in non-​finite contexts, suggesting it does not come from T/​Asp, but from v. In other cases, we find that absolutive case in not straightforwardly available in non-​finite clauses at all, so that the transitive object must be licensed in some special way, if transitive control is possible. In these high ABS languages, then, it seems that absolutive always comes from T/​Asp. 29 

As noted, an independent parameter is needed to govern whether obliques count for transitivity.

 

80   Michelle Sheehan Chamorro appears to be a high ABS language in these terms. The evidence for this is that the infix -​um-​, which surfaces where ergative subjects are extracted, also surfaces where a transitive predicate appears in a control context (Chung 1982: 49, fn. 5): (68) Malagu’  gui’  b-​um-​isita  want he um-​visit ‘He wants to visit Rita.’

si Rita. unm Rita 

[Chamorro]

Plausibly, -​um-​serves to license the absolutive DP in (68), as is the case with the “crazy antipassive” in high ABS Mayan languages (Coon et al. 2014). In Tagalog, on the other hand, ABS is retained on the transitive object in non-​finite contexts because it is a mixed ABS language (see Aldridge 2004, 2013b): (69) Nagba-​balak ang babae-​ng    [PRO  intr.prog-​plan  abs woman-​lk ‘The woman is planning to help the man.’

tulung-​an ang lalaki]  [Tagalog] help-​appl  abs man (Aldridge 2013: 2)

There is some evidence that Trumai is also high ABS in these terms. Verbs like padi ‘wait’ can take reduced clausal complements which appear to give rise to ECM, whereby what would be the absolutive argument of the embedded clause surfaces as an enclitic on the matrix verb: (70) hai-​ts chï_​in [Kumaru-​k tïchï] padi-​n 1-​erg  foc/​tense Kumaru-​erg  scarify  wait-​3abs  ‘I waited for Kumaru to scarify her.’ (71) hai-​ts [huma] padi-​n 1-​erg  take.bath  wait-​3abs ‘I waited for her to take a bath.’

[Trumai]

(Guirardello-​Damian 2010: 220–​1)

The fact that this process applies uniformly to absolutive subjects and objects in Trumai, whereas ergative case is retained (as in (70)), suggests that both get case from T in finite contexts. In non-​finite contexts, T fails to assign absolutive case, but ergative, from v, is still available, as expected. High ABS languages appear to be a proper subset of syntactically ergative languages and given standard assumptions about intervention, there is a principled reason why this should be the case. In order for a lower DP1 to receive structural Case from a higher head where another DP2 intervenes, DP1 must first move past DP2. This is essentially the movement proposed to hold in syntactically ergative languages: v bears an EPP feature and attracts DP1 past DP2. No intervention obtains for this movement because DP2 occupies the specifier position of vP and so is not c-​commanded by v. To ensure that DP1 receives Case from this higher head (T), though, it must also be the case that it has not received Case from v, prior to movement. In this way, the parameter distinguishing mixed ABS and

 

Parameterizing ergativity   81 high ABS languages is that which determines whether v loses its ability to assign a structural Case. In a mixed ABS language, v retains this ability (which is the default option for transitive vs—​Burzio 1986). In a high ABS language, on the other hand, v loses the ability to assign a structural case (i.e. it bears no ɸ-​features) and so DP1 receives Case from T, after movement has occurred.30 This final dependent parameter is thus active only in syntactically ergative languages for principled reasons, giving the following parameter hierarchy: (72)

Parameter hierarchy (fourth version)31 P1: Does transitive v assign a theta-related case (ERG) in a language L? N–accusative

Y – non-accusative P2: is this generalised to all vs in L?

Y – Basque

N – P3: is this extended to intransitive [volition] vs in L?

Y – Hindi

N – P4: is this restricted to [instigator] vs in L?

Y – Tsez, Lezgian

N – P5: does verg bear an EPP in L?

N – Warlpiri, Niuean(?) Y – P6: are verg’s -features suppressed in L? N – West Greenlandic, Tagalog, Cavineña(?) Yidi , Kuikuro, Tongan

Y – Chamorro, Mam, Q’anjobal, Dyirbal, Seediq, Trumai

3.3  The Status of the Hierarchy The parameter hierarchy in (72) serves to model micro-​parametric variation among non-​accusative languages. What is given, I assume, are the functional categories themselves, formal features such as EPP and ɸ (leading to structural Case valuation) and a requirement for nominal licensing. The other properties of v are open to parameterization: whether v is overt/​covert; whether it assigns a theta-​related Case (always, never, 30  This raises the question what happens to T’s ɸ-​features in mixed ABS languages’ transitive clauses, where both arguments get Case from v. Either T must simply lack ɸ-​features in this context or we must assume, following Preminger (2011a) that unvalued ɸ-​features fail to crash the derivation. 31  With extra languages tentatively added for purposes of illustration: see Dixon (2010) on Yidiɲ; Massam (1998, 2006) on Niuean; Otsuka (2006) on Tongan; Guillaume (2008) on Cavineña; Aldridge (2004) on Seediq and Tagalog; Franchetto (2010) on Kuikuro; Coon et al. (2014) on Mam and Q’anjobal. It is not actually possible to say, as of yet, whether Cavineña is syntactically ergative or not, though there is suggestive evidence that it is (Guillaume 2008, p.c.). I have not been able to ascertain whether Niuean patterns with Warlpiri or Tsez/​Lezgian, but it is reported not to be syntactically ergative (Levin and Massam 1985).

 

82   Michelle Sheehan sometimes); whether it assigns a structural Case; and whether it bears an EPP feature. The upper end of the hierarchy remains fairly descriptive, and I have no deep explanation for the fact that P1 refers to transitive rather than intransitive v, though this is clearly empirically justified, given that all non-​accusative languages seem to share this property. It is possible that transitive v is the most salient instantiation of the category v and as such has a privileged status for acquisition. The format of P3–​P4 is intended to be open rather than fixed by UG, the idea being that the child acquiring a non-​accusative system can extend or restrict ergative case to any coherent class of vs, with Hindi and Tsez/​Lezgian being just two possibilities (see Roberts 2012). While this is a fairly powerful model, it appears to be empirically necessary. It is an empirical question, though, to what extent all potential extensions/​restrictions of ergative case are attested. I have found no language, for example, in which ergative is limited to agents, to the exclusion of causers, though there is at least one language which may limit ergative to animate DPs (Nepali).32 The model can therefore be seen as a working hypothesis. P5–​P6 are substantively different from P2–​P4. First of all, the dependencies between P2–​P4 are negative, so that they do not determine cumulative properties of a system but rather mutually exclusive properties. A  language either generalizes, extends, or restricts ergative case, but it cannot, by hypothesis, do more than one of these things. P5–​P6, however are different. Syntactically ergative high ABS languages are a subset of syntactically ergative languages. It is only positive dependencies of this kind which are truly dependent and hierarchical in this sense. P2–​P4 are non-​cumulative and so could be reordered without altering potential outputs, but the same is not true of P5–​P6. An anonymous reviewer points out that it is possible to rephrase P2–​P4 so that they too are cumulative in this sense, giving the following alternative parameter hierarchy: (73)

Parameter hierarchy (alternative fifth version) P1: Does transitive v assign a theta-related case (ERG) in a language L? N – accusative

Y – non-accusative P4’: is this generalised to all transitive vs in L?

N (some) – Tsez, Lezgian N – P5’: does verg bear an EPP in L?

Y – P3’: is this extended to intransitive vs in L? Y – P2’: is this generalised to all intransitive vs?

N (some) – Hindi N – Warlpiri, Niuean(?) Y – P6’: are verg’s -features suppressed? N – West Greenlandic, Tagalog, Cavineña(?) Yidi , Kuikuro, Tongan

Y – Basque

Y – Chamorro, Mam, Q’anjobal, Dyirbal, Seediq, Trumai

In (73), there is only one negative dependency: that between extension of ergative to intransitive vs and the association of an EPP feature. Again, there is a principled 32 

Thanks to Joe Perry for providing me with and discussing the Nepali data.

 

Parameterizing ergativity   83 reason why these two grammatical properties would be incompatible: a language which extended ERG to intransitive contexts would have no argument to satisfy an EPP feature in such contexts. With this exception, though, the parameters in (73) all involve positive dependencies. This has the advantage that, moving down the hierarchy, the output grammars stand in superset relations. The contexts in which ergative surfaces in Basque are a proper superset of the contexts where ergative surfaces in Hindi, which are a proper superset of the contexts where ergative surfaces in Tsez. Similarly, little v in Chamorro has all the properties of little v in West Greenlandic plus suppressed ɸ-​features (the marked option) and little v in West Greenlandic has all the properties of little v in Warlpiri plus an additional EPP feature. In this way, assuming that the hierarchy models acquisition, this is a process of selecting grammars of ever increasing complexity and size, providing a potential solution to the subset problem identified by Wexler & Manzini (1987) and Manzini & Wexler (1987). Essentially, as they point out, given the negligible role played by negative evidence in language acquisition, children face a superset trap, whereby if they posit a grammar consistent with the data they observe, but not restrictive enough, they may never be able to posit a grammar which is a subset of that initial hypothesis. The kind of parameter hierarchy in (73) addresses this problem head on by proposing that children start off by positing smaller grammars and only extend them in the face of positive evidence. Reordering the hierarchy in this way thus seems attractive, although it appears to involve some redundancy concerning the actual parameters required. A remaining question concerns the relative ordering between parameters. Is there any deeper rationale for the positive dependencies between parameters in (73)? The answer appears to be that these dependencies are due to the need for convergence. We have already provided a potential explanation for the fact that syntactic ergativity is compatible only with transitive-​sensitive ergative alignment. A similar account emerges for the dependency between P1, P5/​P5’ and P6/​P6’. In a language in which v fails to assign ergative and/​or lacks an EPP feature, there will be no way for both arguments to receive Case if v loses its ability to assign a structural Case (i.e. loses its ɸ-​features). In order for both DPs to get case in such as context: (i) the higher DP must get a non-​structural case; and (ii) the lower DP must scramble past the higher DP so that it is in a position to receive a higher structural case (from T) without defective intervention. It is only in such contexts, then, that Burzio’s generalization can be violated. All of this is implicit in the parameter hierarchies in (72)/​(73).

3.4 Conclusions This chapter has developed a parameter hierarchy for non-​accusative alignment based on the hypothesis that ergative is an inherent Case. It has been shown that while there are minimal differences in the distribution of ergative case across languages, there are also many similarities all of which seem to be broadly in line with the predictions of

 

84   Michelle Sheehan the inherent case account, based on what is known about theta-​roles in accusative languages. Of course, as mentioned in the introduction, the inherent case approach makes four different kinds of predictions and this chapter has focused mainly on the first of these (A): A. Ergative will only occur on (a subset of) arguments externally merged in spec vP. B. The presence of ergative may be independent of transitivity, so we might find ergative subjects without absolutive objects. C. There will be no derived/​non-​thematic ergative subjects (no ergative expletives, raising to ergative or ergative subjects of passives, ditransitives, or otherwise). D. Ergative case will not be lost in contexts where structural case is not available (no change of case under ECM, no loss of ergative under raising). In relation to B, it has been shown that Western Basque and Hindi both have ergative with unergative intransitives, but the possibility of ergative with oblique objects seems to be rarer (though it is observed in Warlpiri and Lezgian). The discussion of the source of absolutive case in Trumai and Chamorro touched on D, though, there is clearly much more to be said. Finally, C has not been discussed at all. In a sense, then, showing that ergative case occurs only on a (a subset of) arguments externally merged in spec vP is just the first step towards arguing that ergative is an inherent case. While the facts have been shown to be broadly compatible with the inherent case approach, the results are by no means conclusive (see especially Rezac et al. 2014 on Basque). Given the parameter hierarchy approach, though there are added advantages to the inherent case approach, not least because it enables us to conceive of accusative/​ergative variation in terms of variation of the properties of a single class of functional heads (little v), broadly in line with the Borer–​Chomsky conjecture. Parameter hierarchies, in these terms, are the pathways used by the child to acquire the properties of a class of functional heads, aided by the kinds of dependencies and acquisition strategies discussed. The basic case/​alignment properties of a language are thus effectively encoded on little v in systematically defined ways.

Acknowledgments Many thanks to the European Research Council for funding this research under the auspices of the project Rethinking Comparative Syntax (ReCoS). Thanks also to the other members of that project: András Bárány, Tim Bazalgette, Theresa Biberauer, Alison Biggs, Georg Höhn, Anders Holmberg, Ian Roberts, and Jenneke van der Wal for providing critical feedback at various stages. Different parts of this work were presented at West Coast conference on Formal Linguistics (Arizona), the Syntax of the World’s Languages (Dubrovnik), GLOW Biolinguistics Workshop (Lund), Formal Ways of Analyzing Variation (Reykjavik), Towards a Theory of Syntactic Variation (Bilbao), International Congress of Linguists (Geneva), Societas Linguistica Europeae (Split), Workshop on Building Blocks (Leipzig), and at What Happened to Principles and Parameters? (Arezzo). Thanks to the audiences at those venues

 

Parameterizing ergativity   85 as well as at the University of Cambridge for providing thought-​provoking questions and critiques, especially Joe Perry, Katya Pertsova, Gereon Müller, Edith Aldridge, Amy Rose Deal, Adam Ledgeway, and Bob Freidin. Finally, special thanks to an anonymous reviewer and to Lisa Travis for providing detailed comments on an earlier draft of this chapter which have (I hope) led to a great deal of improvements. As ever, I take full responsibility for the way I have interpreted and used comments and suggestions.

Abbreviations ABS, absolutive; AOR, aorist; APPL, applicative; AUX, auxiliary; CAUS, causative; COMP, complementizer; DAT, dative case; DEF, definite; DEM, demonstrative; DET, determiner; DOM, differential object marker; EPP, movement trigger; ERG, ergative; EV, evidential; FOC, focus marker; GEN, genitive; GER, gerund; I noun, class I agreement; II noun, class II agreement; III noun, class III agreement; impf, imperfect; inf, infinitive; INSTR, instrumental; INTR, intransitive; iv noun, class iv agreement; LAT, lative case; LK, linker; LOC, locative; NZR, nominalizer; OBL, oblique case; OS, oblique stem; PERF, perfect; PL, plural; POSS.ESS possessive case; PRES, present; PROG, progressive; PST, past; PTCP, participle; REL, relative marker; RES, resultative; S, singular; TENSE, tense marker; TR, transitive; UM, ‘um’ morpheme; UNM, unmarked; WIT, witnessed; YI, ‘yi’ morpheme.

 

Chapter 4

Ac cusative and e rg at i v e in Hindi Anoop Mahajan

4.1 Introduction This chapter deals with a specific issue of case licensing in ergative languages. The issue concerns the licensing of the case on the direct object (absolutive) argument in ergative constructions. This issue is of importance, since over the last few years the case of the direct object (DO) argument in ergative constructions has been argued to be licensed by a variety of conditions that include: case licensing by T; case licensing by an accusative licensing head, which could be either a little v or a lower agreement (AGR) head; or by some mechanism of case competition.1 The discussion is often complicated by the well-​known fact that the overt case realization of absolutive is often morphologically null. Furthermore, in a language like Hindi, the DO can have a differential object marking that is often equated with the overt manifestation of accusative case (as in Mohanan 1994a, for example) in opposition to the null case on a non-​differentially case marked object which is then sometimes labeled as nominative (Mohanan 1994a) or as accusative (Legate 2008). My objective in this chapter is to argue, employing data that has not been previously invoked within this domain, that the Hindi DO in ergative constructions does not have accusative case. This casts doubt over the universal validity of absolutive-as-accusative type theories, and also on Hindi specific proposals where it has been argued that the DOs in Hindi ergative constructions bear accusative case. Furthermore, it will once again highlight the question concerning why, at least in certain environments, the absence of accusative assignment is a prerequisite for the licensing of ergative subjects and perhaps other oblique subjects as well. The discussion in this chapter will

1 

I use the terms ‘case’ and ‘inherent case’ in this chapter to distinguish between ‘structural case’ and ‘inherent case.’

 

Accusative and ergative in Hindi    87 also emphasize that focusing on case licensing as structural argument licensing in the original Government and Binding sense is a more fruitful strategy to resolve disputes over what case the arguments in ergative constructions bear, rather than looking at the morphology on the arguments—which that can often be confusing.2 To set the background for this discussion, consider (1) in Hindi:3 (1) Kabiir-​ne vah laal gaaṛii jaldii-​se beč-​ii thii Kabir(masc)-​erg  that red car(fem)  quick-​with  sell-​perf.fem.sg  be.pst.fem.sg ‘Kabir had sold that red car quickly.’ The central problem to which we seek an answer concerns the case borne by the DO in (1). The case of the subject in (1) is identifiable as the ergative case by its postpositional case ending –​ne. However, the DO has no overt morphological case ending, therefore it is not a priori clear whether this DP has a case and if it does, what is the nature of that case. I am not aware of any proposal that asserts that the object DP in (1) does not have a case, and indeed if one assumes the validity of the traditional GB case filter, the DO in (1) must surely have a case in order to be visible.4 The usual practice in traditional literature on ergativity (for example in Comrie 1978 and in Dixon 1994) is to label the DO as having an absolutive case. However, the label ‘absolutive’ does not tell us how this absolutive case is assigned or licensed, and therefore the use of this label often obscures the formulation of the structural licensing conditions under which the object nominal is (case) licensed. My aim here is to suggest that the DO in (1) is case licensed by T, and within the GB-​minimalist type theories, the case that the DP bears in (1) should be labeled as the nominative case. The alternative that has been proposed (for Hindi) is that the DO in (1) is case licensed by a little v and should be labeled as accusative (Bhatt 2005; Legate 2008; also related general proposals in Murasugi 1991, Bobaljik 1993b, and others). I will argue that this alternative view is not correct. It should however be made clear that the case labels themselves are not so important. The GB-​minimalist theories are clear on this. The real issue concerns the nature of case licensing of the object, and 2 

In this respect (realization of the morphological case), I am in general agreement with distributive morphology based approaches such as that of Legate (2008). 3  The glossing I provide for the examples includes the features that are relevant to the discussion as well as for clarity. Not all morphological features are always identified in the glosses. For example, I gloss gender agreement for participle verbs and do not usually gloss the number agreement, even though it can be morphologically realized in many cases. 4  However, the possibility that the DO in (1) may be caseless should perhaps be evaluated in view of proposals concerning the case of DOs in antipassive constructions, where it has been suggested by some (including Aldridge 2012b and Mahajan 2012) that DO nominals may be caseless and perhaps be a special case of incorporation (or pseudo-​incorporation). I have tried to minimize that possibility in (1) by making the DO non-​adjacent to the verb and by including a demonstrative and a modifier. Furthermore, the DO in (1) is clearly referential and it can be shown that its scope properties are different from that of the DOs in antipassive constructions (for an outline of the relevant properties of DOs in antipassives, see Polinsky 2005; for relevant issues concerning Hindi noun incorporation, see Mohanan 1995 and Dayal 2011).

 

88   Anoop Mahajan indirectly of the subject, particularly the identification of the relevant case licensing heads and the environments of case licensing.

4.2  A Selective Overview of Proposals about DO Case Licensing in Ergative Constructions 4.2.1 DO Has ACC Case in Ergative Constructions Within the GB tradition, this is perhaps the earliest proposal, represented by Massam (1985) and Levin and Massam (1985), and followed up in Bobaljik (1993b).5 Taking Bobaljik (1993b) to be representative of this tradition, the general idea is that the sources of structural case for subjects and objects in ergative–​absolutive (ERG–​ABS) languages and nominative–​accusative (NOM–​ACC) languages are parallel. Specifically, ABS=ACC and ERG=NOM (Bobaljik 1993b: 46). Even more specifically, Bobaljik proposes that NOM and ERG are both assigned by a structurally higher head, while ACC and ABS are assigned by a structurally lower head. He labels those heads as AGR1 (the higher head that takes TP as its complement) and AGR2 (the lower head that takes VP as its complement). The crucial way in which ERG–​ABS languages differ from NOM–​ ACC languages in his system concerns the case that the sole argument of an intransitive clause is assigned. Bobaljik makes his proposal in the form of a parameter he calls the ‘Obligatory Case Parameter’, which essentially says that in NOM–​ACC languages, this sole argument has NOM (=ERG) case, while in ERG–​ABS languages, this argument has ACC (=ABS) case. Given the existence of split ergative languages, it is clear that this type of parameter cannot be a parameter distinguishing languages, though one can develop an implementation of this approach to include split ergativity (which I will not attempt here). I should note that I will follow Bobaljik and others in an important way in that I will assume that NOM is the label of the case licensed by a (higher) T head while ACC is the label of the case that is licensed by a (lower) v head under well-​defined locality conditions. An implementation of this would be to use the locality involved in the AGREE relation of the minimalist tradition and I will assume that for this chapter. It would perhaps be wise to develop a different terminology and relabel NOM as case1 and ACC as case2 or something similar so that we can get away from the confusion caused by the use of the traditional terminology in this domain. However, no one, including me at this point, is making this move, and therefore I want to make sure that we are talking about licensing relations as opposed to morphological forms, which may sometimes yield useful clues, but can also add unnecessary confusion. 5 

For Hindi, Bhatt (2005: 759–​760) also makes the assumption that the absolutive DO has ACC. He also assumes that Hindi differentially case marked objects have ACC.

 

Accusative and ergative in Hindi    89

4.2.2 DO Has Nominative Case in Ergative Constructions This view is advocated and developed in Bittner (1987), Mahajan (1990), Bok-​Bennema (1991), and Murasugi (1992) among others. The general idea in most of these approaches is that ergative licensing is handled by some head other than T (though see Bittner and Hale 1996a for a somewhat different perspective), leaving T to license NOM on the DO in ergative languages. Whether ergative is licensed as an inherent case or a structural case is a separate issue and I will not be directly concerned with that here (see among others, Marantz 1991, Bittner and Hale 1996a, Woolford 1997). My focus in this chapter is more on the nature of DO licensing than that of subject (SUB) licensing.

4.2.3 DO Has Sometimes NOM and Sometimes ACC in Ergative Constructions This view has been developed in Legate (2008) who argues that there are two typologically distinct groups of languages. In one group, the absolutive argument is NOM, while in the second group, the absolutive argument is ACC. Legate makes a specific proposal about Hindi, arguing that Hindi is a type of language in which absolutive objects have ACC. This approach agrees with the assumption made concerning the case of DO in Hindi ergative constructions in Bhatt (2005: 759–​760). I will argue in this chapter that Hindi DOs in ergative constructions do not have ACC. I will use evidence from a variety of prenominal relative clauses that are based on the same verbal form as the Hindi ergative constructions to substantiate my argument. If my argument is on the right track, it will cast doubt over Legate’s proposal about Hindi, though I do not evaluate the Hindi external data that Legate provides for her general proposal.6

4.2.4 The Source of Ergative Case Once again, there are differing views on the licensing of the ergative case. Marantz (1991) and Bittner and Hale (1996a) propose in different ways that the ergative case is structural. Bobaljik’s (1993b) proposal tying ERG to the higher AGR head also makes the ergative case look like a structural case, though in ways different from Bittner and Hale, and Marantz. However, much of the recent work on ergativity has developed the idea that the ergative case is inherently assigned by little v. This view is represented in Mahajan (1990, 2000, 2012), Woolford (1997), Anand and Nevins (2006), and in numerous other recent papers. In this chapter, I will assume that ergative in Hindi is assigned inherently by the little v that heads the complement of a perfective Asp head 6 

Another type of variation is discussed in Massam (1996) who argues that in Niuean the absolutive case patterns unlike both NOM and ACC.

 

90   Anoop Mahajan (more details in section 4.3). While this aspect of ergativity is not the focus of this chapter, there appears to be a connection between the unavailability of accusative case and the appearance of ergative case in Hindi. This raises the issue of whether the ergative case is some deviant form of the accusative case (Marantz 1991; Mahajan 2000). The unavailability of ACC in Hindi extends beyond the ergative subject construction to include dative subject constructions and a variety of Hindi passives (see Mahajan 2000 for relevant details). Therefore, the proper correlation in Hindi is between the unavailability of ACC and the appearance of an oblique case on the subject. I do not pursue this matter in this chapter.

4.3  Some Basic Hindi Ergativity Facts and the Basic Issues 4.3.1 Null Case Objects in Hindi Ergative Constructions As is well known, Hindi is a split ergative language. Ergative case appears on the subject of transitive perfective participle verbs as in (2) (on the connection between ergative and perfective in Hindi, see, among others, Porizka 1967, 1968, 1969; Kachru and Pandharipande 1978; Mohanan 1994a; Davison 2004b). The subjects of non-​perfective participle transitive clauses must not have an ergative ending; and indeed in (3), a transitive imperfective construction, and in (4), a transitive future construction, the subjects are unmarked and cannot have an ergative ending.7 (2) mε˜-​ne vah akhbaarẽ jəldii-​se  beç-​ĩĩ thĩĩ I-​erg(masc)  those newspapers (fem)  quickly sell-​perf.​fem.pl  be.pres.fem.pl ‘I sold those newspapers quickly.’ (3) mε˜ (*-​ne)  vah akhbaarẽ jəldii-​se  I(masc) those newspapers (fem)  quickly ‘I (habitually) read those newspapers quickly.’

paṛh-​taa hũ read-​imperf.masc.sg  be.pres.IP.sg

(4) mε˜ (*-​ne)  vah akhbaarẽ jəldii-​se  paṛhũ-​gaa I(masc) those newspapers (fem)  quickly read.IP-​fut.masc.sg ‘I will read those newspapers quickly.’ A lot of the later discussion in this chapter will focus on the verb forms in (2) and (3). At this point, what we need to note about (2)–​(4) is that: (i) the morphological shape 7 

Examples (3) and (4) are ungrammatical with a subject –​ne ending even if the agreement is changed to object agreement.

 

Accusative and ergative in Hindi    91 of the DO in all of these sentences is identical; (ii) the subject in (2) has an ergative case postposition,8 while the subjects in (3) and (4) must be unmarked, and are therefore labeled as nominative by most Hindi linguists; and (iii) the perfective participle and the copula in (2) agree with the object, whereas we get subject agreement in (3) and (4).9 The Hindi verbal cluster (verb and the copula/​auxiliary) always agrees with the same argument, though not exactly in the same features. Given that the DO in (2)–​(4) looks exactly the same, there are various logical possibilities. These include: (i) the DOs in (2)–​(4) are all NOM, since they are morphologically bare and are morphologically identical, and share the same bare case form as that of the subject in (3) and (4). The fact that the DO agrees with the verb in (2) but not in (3) and (4) is not relevant. This is the view held by Mohanan (1994a). She treats (3) and (4) as double nominative constructions.10 (ii) the DOs in (2)–​(4) are all ACC, since they are all objects and licensed in the same way, arguably by the same case licensing head. Once again, the fact that the DO agrees with the verb in (2) but not in (3) and (4) is not relevant. This is the view held by Bhatt (2005) and Legate (2008). (iii) the DO in (2) is NOM and is case licensed by T, while the DOs in (3) and (4) are ACC and are case licensed by a distinct head, presumably a little v. Agreement relations in (2) vs. (3) and (4) mirror case licensing. This view is proposed by Mahajan (1990) and is consistent with Chomsky’s (1993) idea about the relationship between case and agreement feature checking. (iv) the DO in (2) is ACC while it is NOM in (3) and (4). This possibility has not been explicitly suggested by anyone and I will not discuss it further. A variant of this idea is followed by Mohanan (1994a) in the context of differentially case marked (DOM) objects, which I will discuss later. If the evidence I present and discuss in this chapter is on the right track, the possibilities and proposals in (i) and (ii) are wrong. That is, the proposals that treat all unmarked objects alike, whether as ACC or as NOM, are inadequate in dealing with the data that I will discuss. I argue that the DO in ergative constructions in Hindi, whether unmarked or marked, is never licensed by an ACC licensing head. This essentially leaves us with the proposal in (iii) that the DOs in (2) vs. (3) and (4) are licensed by different heads and should therefore be labeled differently: NOM in (2) vs. ACC in (3) and (4). A sketch of the proposal concerning case licensing that I  defend is outlined in the structural configurations that represent perfective (Figure 4.1, corresponding to

8 

Hindi case postpositions are clitics (see Mahajan 1990 and Mohanan 1994a for some discussion). Object agreement in ergative constructions in Hindi does not include person features. The subject agreement in non-​ergative constructions (3) and (4) shows person agreement, though in (3) the person agreement shows up only on the auxiliary. In the present discussion, this fact will not be relevant. 10  Mohanan’s (1994a) discussion is within the LFG framework where the case linking conditions are quite distinct from the case licensing conditions of the GB-​minimalist tradition. Therefore, a direct comparison of her proposal and its possible counterpart in the GB-​minimalist tradition is difficult. However, Mohanan’s work does represent the tradition of taking morphological shape of the case endings (including null endings) seriously for linking/​licensing purposes. 9 

 

92   Anoop Mahajan sentence (2)) and imperfective (Figure 4.2, corresponding to sentence (3)) transitive clauses with a finite tense. For the sake of clarity, these configurations are depicted with pre-​movement structures that only show case relations and not the final word order of Hindi.11

selection … TFIN

[AspP

Aspperf

[vP SUB-erg

Vperf

[

V

DO ]]]

[

V

DO ]]]

inherent ergative NOM

Figure 4.1  Case relations in Hindi perfective transitive clauses

selection … TFIN [AspP

Aspimperf

[vP

SUB

Vimperf

NOM

ACC

Figure 4.2  Case relations in Hindi imperfective transitive clauses

4.3.2 DOM in Hindi Hindi displays differential object marking (DOM) that interacts with the characterization of object case. Examples (5) and (6) provide examples of DOM objects in Hindi ergative constructions, while (9) and (10) illustrate that DOM objects can be found in non-​ergative constructions as well. (5) Miiraa-​ne laṛkii-​ko kal dekh-​aa thaa Meera-​erg(fem)  girl(fem)-​DOM  yesterday  see-​perf.masc  be.pst.masc ‘Meera had seen the girl yesterday.’

11 

In this chapter, I omit discussion of why the ergative subject does not block case assignment by T to the DO in Figure 4.1. Various ways of handling this ‘non-​intervention’ effect have been proposed in the literature. For the purpose of this chapter, I will assume that ergative assignment makes the subject ‘inert’ and not visible to structural case assignment by T. See Mahajan (1990) and Bhatt (2005) for two of the possible analyses.

 

Accusative and ergative in Hindi    93 (6) Miiraa-​ne kitaab-​ko kal paṛh-​aa thaa Meera-​erg(fem)  book (fem)-​DOM  yesterday  read-​perf.masc  be.pst.masc ‘Meera read the book yesterday.’ (7) mε˜ (*-​ne)  laṛkii-​ko har roz bulaa-​taa hũ I(masc) girl-​DOM  every day  call-​imperf.masc.sg  be.pres.IP.sg ‘I call the girl every day.’ (8) mε˜ (*-​ne)  kitaab-​ko jəldii-​se  paṛhũ-​gaa I(masc) book-​DOM (fem)  quickly read.IP-​fut.masc.sg ‘I will read the book quickly.’ Broadly speaking, Hindi proper names and pronouns must always be followed by DOM and specific animate and inanimate objects are also followed by DOM. There is extensive literature on Hindi DOM that deals with various interpretational issues of DOM objects. (For formal proposals see, among others, Mahajan 1990, Butt 1993b, Mohanan 1994a, Bhatt and Anagnostopoulou 1996).12 The DOM marking in Hindi is the same as the case ending obligatorily carried by indirect objects. With a few exceptions, the Hindi DO -​ko ending is identified as the accusative ending (as in Mohanan 1994a and Butt 1993b, 1995). Given this common and pervasive view, the -​ko ending in (5) and (6) is taken to represent ACC. In view of the new data discussed in this chapter, the proposal is that the -​ko ending in (5) and (6) is not morphological realization of ACC but simply DOM marking. This view was originally put forward for Hindi in Mahajan (1990). The licensing of the DOM morphology is not an issue discussed here, though see Bhatt and Anagnostopoulou (1996), and a more recent idea in Kalin (2014). My contention is that the case licensing of DOs in (5) and (6), and also (7) and (8) is obscured by the surface appearance of the DOM marker. In particular, I will argue, contrary to much work in Hindi linguistics, that the DO in (5) and (6) is case licensed by the T head while the DO in (7), and also in (8) (though I do not directly argue for that), is licensed by a little v. I leave open the issue of what governs the appearance of the DOM marking. The point I want to make is that the DOM marking is not a substitute for the structural case licensing requirement. This in turn raises yet another interesting issue of whether non-​structural case marking can ever substitute for structural case licensing in terms of a condition of the sort envisaged by Vergnaud’s (2008) original proposal concerning case. My tentative answer to this would be that DOM does not substitute for structural case, or more specifically, DOM nominals must be structurally case licensed in the sense of the classical structural case requirement.13

12  Mohanan (1994a) disallows DOM endings on inanimate DOs. Other works such as Mahajan (1990) and Bhatt and Anagnostopoulou (1996) accept DOM inanimates. 13  This raises the issue of whether the inherent ergative case (for those who view it as inherent case) is sufficient to fulfill the case theory requirements. I will not discuss this here, though see Legate (2002) who

 

94   Anoop Mahajan

4.4  A New Empirical Domain in Resolving the Issue of Case in Hindi Perfective Environments Given that Hindi DOs can have null case ending or be DOM marked in various environments, it is hard to find crucial evidence that will help us decide the exact nature of case licensing in normal transitive clauses. My strategy here is to turn to a new empirical clausal domain where both perfective and imperfective verb forms that we see in Hindi ergative and non-​ergative constructions are employed and isolate the DO licensing outcomes in those environments. The clausal domains are similar to ergative constructions with respect to the verbal forms (the lower vP clausal domain) but crucially different in that they filter out the possibility of case licensing by T since finite T is systematically unavailable in these environments. This in turn provides us with a useful control in looking at how the DOs may be licensed in structures that contain sub-​parts of the ergative construction and provide a new insight into object case licensing. The empirical domain that we now turn to is that of non-​finite prenominal relative clauses.

4.5  Perfective Participle Prenominal Relative Clauses 4.5.1 The Case of DOs In this subsection, I will present the core argument of this chapter suggesting that the perfective participles in Hindi are simply incapable of case licensing DOs in Hindi. Since the ergative construction by itself does not provide conclusive evidence, I turn to a somewhat similar construction that helps us resolve the relevant issue. Hindi prenominal relative clauses come in various varieties—​the one we start with here is the form that is of most interest to us—​the relative clauses that use a perfect participle form of the verb which is identical to the one found in Hindi ergative constructions. Consider the data below. The brackets indicate the limits of the prenominal relative clause; the head of the relative clause is the NP that appears after the right bracket.

presents some Warlpiri data with ergative subjects in infinitive constructions that may have consequences for this issue. Hindi ergative case is however only found in tensed environments. I discuss in note 18 one environment where the assignment of the ergative case could be attributed to a non-​finite context, though the ergative nominal in that case can be argued to have raised to a finite clause.

 

Accusative and ergative in Hindi    95 (9) [Kabiir-​kii likh-​ii (huii)] kitaab Kabir-​gen  write-​perf.fem  be.perf.fem  book(fem) ‘a/​the book written by Kabir’ (10) [bazaar se aa-​yii (huii)] taazii  sabzii market-​from  come-​perf.fem  (be.perf.fem)  fresh vegetable(fem) ‘fresh vegetables (which) arrived from the market’ (11)

[mar-​ii (huii)] čhipkali die-​perf.fem  (be.perf.fem)  lizard(fem) ‘a/​the lizard that is dead’

(12)

[mez-​par  soy-​ii (huii)] billii table-​on sleep-​perf.fem  (be.perf.fem)  cat(fem) ‘a/​the cat sleeping on the table’

Examples (10)–​(12) are cases of subject relativization.14 Example (9) is an instance of an object relativization. In (9), the subject of the relative clause is followed by the genitive postposition. The perfective participle verb and participial aux be (which is optional) agree with the relativized head noun (and the genitive postposition). The subject can also be marked as a by-​phrase instead of with genitive (but can never be unmarked). (13)

[Kabiir-​dwaara  likh-​ii (huii)] kitaab Kabir-​by write-​perf.fem  (be.perf.fem)  book(fem) ‘a/​the book written by Kabir’

Example (9) is parallel to a normal perfective transitive ergative clause (14). Similarly, (10) is parallel to a normal intransitive perfective non-​ergative (15).

14  There are interesting restrictions on intransitive subject relatives. Some unergative subjects that can optionally take ergative subjects cannot be relativized (Mahajan 1990), as in (i)–(iii) below. Though so-​naa ‘to sleep’ can optionally take an ergative subject, it can appear in prenominal perfective relatives. This pattern is not fully understood and needs to be investigated.

(i)

*[tez  dauṛ-​aa (huaa)]    fast run-​perf.masc  be.perf.masc     ‘a/​the boy who ran fast’

laṛkaa boy

(ii)

*[zor-​se  bhõk-​aa (huaa)]    Loudly bark-​perf.masc  be.perf.masc     ‘a/​the dog that barked loudly’

kutta dog

(iii)  [zor-​se soy-​aa (huaa)] aadmii    soundly  sleep-​perf.masc be.perf.masc  man    ‘a/​the man who is sleeping soundly’

 

96   Anoop Mahajan (14) Kabiir-​ne kitaab likh-​ii (thii) Kabir-​erg  book(fem)  write-​perf.fem  (be.pst.fem) ‘Kabir had written the book.’ (15)

bazaar se taazii  sabzii aa-​yii (thii) market-​from  fresh vegetables(fem)  come-​perf.fem  (be.pst.fem) ‘The fresh vegetables had arrived from the market.’

The point that I wish to make is that the perfective prenominal relatives in (9)–​(12) are structurally parallel to normal non-​relative clauses such as (14) and (15) with a crucial difference: they use the same perfective participle verbal forms and have similar agreement patterns; but while (9)–​(12) lack a finite T, (14) and (15) have a finite T (though it can be null). Prenominal relative clauses in Hindi cannot be finite. This is illustrated by the ungrammaticality of (16). (16) *[Kabiir-​kii  likh-​ii thii /​ hɛ̃] kitaab    Kabir-​gen write-​perf.fem  be.pst.fem  /​ be.pres  book(fem)    ‘a/​the book that was/​is written by Kabir’ Also, while a genitive subject is possible in (9), and a by-​subject is possible in (13),15 an ergative subject is systematically impossible in prenominal relative clauses. (17) *[Kabiir-​ne  likh-​ii (huii)]      kitaab    Kabir-​erg write-​perf.fem  be.perf.fem  book(fem)    ‘a/​the book written by Kabir’ (18) *[Miira-​ne  Kabiir-​ko likh-​ii (huii)] kitaab    Meera-​erg Kabir –​dat  write-​perf.fem  be.perf.fem  book(fem)    ‘a/​the book written to Kabir by Meera’ The reason for the ungrammaticality of (17) and (18) is very likely due to the non-​ finiteness of the relative clauses, though the exact nature of the finiteness requirement (for ergative) in Hindi is not clearly understood.16 One possibility is that Hindi ergative

15 

I leave aside the issue of whether the relative clause in (13) is a passive, since it has a subject form that is also found in passives. It should be noted though that (13) itself does not have the usual passive auxiliary which would be based on the verb form jaa-​‘go.’ However, the be auxiliary of the examples like (9) and (13) can be replaced by the go auxiliary (though only in transitive prenominal perfective relatives) making them look more like passives even though (9) has a genitive subject, and regular clausal passives in Hindi do not mark the agent with a genitive. 16  Bhatt (2005: 767) includes finiteness as one of the three requirements for ergative licensing (the other two are perfectivity and transitivity).

 

Accusative and ergative in Hindi    97 subjects are only realized in finite environments and while it can be argued (as I do in Mahajan 2012) that Hindi ergative case is inherently assigned, there is a further connection between the ergative case and the finite T in this language.17, 18 It is also important to note that an unmarked subject is systematically excluded from the prenominal relative clause. (19)

*[Kabiir  paṛh-​ii (hui)] kitaab    Kabir read-​perf-​.fem be.perf.fem  book(fem)    ‘a/​the book read by Kabir’

(20) *[Miira  Kabiir-​ko likhii (hui)] kitaab    Meera Kabir –​dat  write-​perf.fem  be.perf.fem  book(fem)    ‘a/​the book written to Kabir by Meera’ The ungrammaticality of (19) and (20) is perhaps not surprising given the ungrammaticality of (16). Since prenominal relative clauses in Hindi are non-​finite, and cannot have a finite auxiliary as shown in (16), the unmarked assumption would be that T case licensing is not available in such clauses, forcing the subjects to be either marked by a genitive postposition or an agentive postposition (or be null).19 17  The assignment and realization of the ergative nominal is a somewhat complex matter. So while Hindi infinitives (as well as prenominal relatives) do not allow ergative subjects, perfective participle if-​conditionals do.

(i)  agar  tum-​ne vah kitaab  paṛh-​ii, If you-​erg  that  book read-​perf.fem  ‘if you read that book, then …’

to … then

If the generalization that finiteness is a requirement for ergative realization is true for Hindi, then one would have to argue that if-​conditionals have a hidden tense (and they do optionally allow an overt finite tense auxiliary). 18  The example below (somewhat marginal for some Hindi speakers) supports the idea that ergative licensing is itself not dependent on finite tense (the right bracket is provided to indicate the clausal boundary between the raising verb and its complement; the left bracket is left out on purpose since its placement depends on discussion beyond the scope of this chapter). (i) Kabiir-​ne bahut  baṛii galtii kar  d-​ii (*thii)] Kabir(masc)-​erg  very big mistake(fem)  do give-​perf.fem  be.pst.fem lag-​tii  thii seem-​imperf.fem  be.pst.fem  ‘Kabir seemed to have made a big mistake.’ The ergative is surely assigned in the complement clause since that is a transitive perfective clause, and the matrix clause is not. The complement clause must however be non-​finite, as shown by the inability to place a finite auxiliary inside it. This may provide evidence that ergative assignment itself is independent of finiteness. However, it is possible that the ergative phrase in (i) has undergone raising and is in the matrix finite clause thus meeting a possible requirement (in Hindi) that ergative subjects need to be in a finite clause. This type of raising is presumably blocked in (17) and (18) since that would involve raising out of a relative clause. 19  This does imply that the genitive and agentive subjects are distinct in their case requirements from the ergative subjects in the sense that they are more like real PPs.

 

98   Anoop Mahajan The important generalization seems to be that subjects and objects can be relativized in prenominal perfective relatives in Hindi. Furthermore, when an object is relativized, the subject cannot be unmarked—​it must either have a genitive case postposition or an agentive postposition. The crucial observation is that the DO, when present because the verb is transitive, must be relativized. It cannot appear inside the perfective prenominal relative clause.20 Examples (21)–​(24) involve attempted subject relativization with an overt DO inside the relative clause and they are ungrammatical. (21) *[kitaab paṛh-​aa (huaa)] laṛkaa    book(fem)write-​perf.masc (be-​part-​masc)  boy    ‘a/​the boy who has read the book’ (*even if agreement on V and aux is fem) (22) *[angrezii akhbaar khariid-​aa (huaa)]    aadmii    English newspaper  buy-​perf.masc  be.perf.masc man   ‘a/​the man who had bought the English newspaper’ (23) *[Kabiir-​ko  bahut  pɛse diy-​aa (huaa)] aadmii    Kabir-​dat lot money  give-​perf.masc  be.perf.masc  man    ‘a/​the man who gave a lot of money to Kabir’ (24) *[Miiraa-​se us-​kii kitaab  maaŋg-​aa (huaa)] aadmii    Meera-​from she-​gen  book ask-​perf.masc  be.perf.masc  man    ‘a/​the man who asked Meera for her book’

20 

An exception involving a DO inside the relative clause is given in (i). (i) [šaraab  piyaa (huaa)] aadmii Liquor drink-​perf.masc  be.perf.masc  man ‘a/​the man who drank liquor’ (= a drunk man)

However, (ii) is ungrammatical. (ii) *[duudh/​paanii/​dawaaii  pi-​yaa (huaa)] aadmii     milk/​water/​medicine drink-​perf.masc  be.perf.masc  man    ‘a/​the man who drank milk/​water/​medicine’ My hunch is that (i) involves noun incorporation of the DO and (therefore) has an idiomatic meaning. Since one does not usually get drunk drinking milk/​water/​medicine, (ii) is ungrammatical. Furthermore, attempts to modify the object in (i) yield ungrammaticality. (iii) *[desii/​videšii  šaraab  pi-​yaa (huaa)] aadmii    local/​foreign liquor drink-​perf.masc  be.perf.masc  man    ‘a/​the man who drank country/​foreign liquor’

 

Accusative and ergative in Hindi    99 Examples (21)–​(24) show that in Hindi an unmarked DO cannot be present inside a perfective prenominal relative clause while another nominal is relativized. We have already seen that unmarked subjects are not possible inside prenominal relative clauses in (19) and (20). I had suggested that (19) and (20) are ungrammatical because there is no source of NOM in the non-​finite prenominal relative clause. The ungrammaticality of (21)–​(24) can now similarly be attributed to the unavailability of any additional structural case inside the prenominal relative clause. In particular, the ungrammaticality of (21)–​(24) tells us that perfective prenominal relative clauses are unable to license any non-​PP arguments inside them. This can be taken to imply that a structural ACC is not available in these clauses.21 I don’t know of any other reason for the ungrammaticality of (21)–​(24). The fact that a DO, if present within a perfective prenominal relative clause, must be externalized (i.e. must be the head that appears outside the relative clause) indicates that the heads of prenominal relative clauses are externally case marked. Thus, relativized DPs ((21)–​(24)) must appear in a structural case position in a clause as in (25) (in a subject position) and in (26) (as an object of a preposition). The oblique ending of the plural head in (26) is due to the postposition -​se, supporting the idea that these heads are externally case marked. (25)  [Kabiir-​kii  paṛh-​ii (huii)] kitaab kho ga-​yii Kabir-​gen read-​perf.fem  be.part.fem  book(fem)  lost  go-​perf.fem ‘A/​the book written by Kabir was lost.’ (26)  Mohan-​ko [[Kabiir-​kii  likhii (huii)] Mohan-​dat  Kabir-​gen write-​perf.fem  be.​part.​fem  ḍar lagtaa hɛ fear feel-​imperf.masc.sg be.pres ‘Mohan is afraid from a/​the book written by Kabir.’

kitaabõ]-​se books.obl(fem)-​from

Further support for the proposal that the heads of the prenominal relative clauses are externally case marked comes from the relativization possibilities of phrases like locatives, manner phrases and instrumental phrases (and other oblique phrases) which must be case licensed by their own postpositions. The prediction is that such nominals cannot be relativized using the perfective prenominal relative clause construction. This prediction is borne out as illustrated in (27)–​(29). (27) a. Mohan  kũẽ-​mẽ ḍuub ga-​yaa Mohan well-​in drown  go-​perf.masc ‘Mohan drowned in the well.’

21  As pointed out by an anonymous reviewer, English perfective participle relative clauses like in [[a book [written by John]] is on sale] display restrictions similar to Hindi. Part of the analysis being developed here for the Hindi facts may have potential consequences for the syntax of reduced participle relatives in English.

 

100   Anoop Mahajan b. *Mohan-​kaa  ḍuub-​aa huaa kũãã/​kũẽ  (-​mẽ)   Mohan-​Gen drown-​perf.masc  be.perf.masc  well     in   ‘the well in which Mohan drowned’ (28) a. vo kaar-​se ga-​yaa he  car-​with  go-​perf.masc ‘He went with a car.’ (He used a car to go.) b. *us-​ki  ga-​yii  huii  kaar he-​ gen go-​ perf.fem be.perf.fem car ‘the car in which he went’ (29) a. čor-​ne laṛke-​ko čaaku-​se maar-​aa thief-​erg boy-​DOM knife-​with kill-​perf.masc ‘The thief killed the boy with a knife.’ b. *čor-​kaa laṛke-​ko maar-​aa huaa čaaku   thief-​gen boy-​DOM kill-​perf.masc be.perf.masc knife   ‘the knife with which the thief killed the boy’ The proposal that the head in a Hindi prenominal relative clause must be externally case marked now helps us make sense of the restriction that in perfective prenominal relative clauses in Hindi, only subjects and DOs can be relativized, since they are the only type of arguments that can be structurally case licensed by clause internal (spinal) heads like T and v. The inability of a DO to survive inside a perfective prenominal relative clause must therefore follow from the lack of structural case inside the relative clause. Given that a finite T is clearly absent in Hindi prenominal relative clauses, the only other plausible source for the structural case would have been an accusative assigning little v. On the basis of the evidence that we have seen so far, I suggest that the little v present inside these clauses is unable to assign ACC, thus making these clauses very similar in ACC case assigning property as transitive perfective ergative clauses.22 To complete this line of argumentation, the big difference between a prenominal relative clause in (30) and a normal finite transitive clause like (31) is the absence/​presence of finite T. The fact that the object is licensed in (31) and not in (30) must then be due to

22  The issue of why the subject of a transitive perfective relative clause cannot be ergative (and must be genitive, if present) remains unresolved here. If, as suggested by Bhatt (2005), the presence of a finite T is a further requirement for ergative case licensing (as mentioned in n. 17), then the lack of a finite T within prenominal relative clauses may be tied to the lack of ergative case. However, in view of the data in n. 17, it is not clear if this would be a desirable analytical direction. An alternative that I do not develop in this chapter could be that the little v that assigns case is distinct from the little v that introduces the external argument (as in Mahajan 2012). If one follows that analytical option, then one may be able to argue that the perfective prenominal relative clauses simply lack the case assigning little v head.

 

Accusative and ergative in Hindi    101 this finite T. Therefore, if finite T licenses NOM, then the object in (31) must have NOM, since we have already eliminated the possibility of the availability of ACC inside perfective clauses. (30)  *[kitaab paṛh-​aa (huaa)] laṛkaa    book(fem) read-​perf.masc  (be.perf.masc)  boy    ‘a/​the boy who has read the book’ (*even if agreement on V and aux is fm) (31)

laṛke-​ne kitaab paṛh-​ii boy-​erg book(fem) write-​perf-​masc ‘The boy had read the book.’

thii /​ (*huii) be.pst.fem /​ (be.perf.fem)

A schematic sketch comparing the analyses of finite perfective transitive clauses and prenominal perfective relative clauses is presented in Figures 4.3 and 4.4 (Figure 4.3 repeated from Figure 4.1 in section 4.3.1): selection … TFIN

[AspP

Aspperf

[vP SUB-erg

Vperf

[

V

DO ]]]

inherent ergative NOM

Figure 4.3  Case relations in Hindi perfective transitive clauses

selection … TNONFIN [AspP

Aspperf

[vP

NOM NOT AVAILABLE

SUB-gen

Vperf

[

V

DO ]]]

therefore DO must be externalized

Figure 4.4  Case relations in Hindi prenominal perfective relative clauses

A by-​product of this discussion is that it helps us make sense of why only certain kinds of grammatical function positions are accessible for relativization in the context of the typology of relativization as discussed in Keenan and Comrie (1977). Only subjects and (non-​ postpositional) DOs are accessible for relativization in Hindi perfective prenominal relative clauses because: (i) only those two need structural case licensing; (ii) structural case is not available in Hindi perfective prenominal relative clauses; and (iii) the relative clause head in Hindi perfective prenominal relatives is externally case marked. The fact that the indirect object and the obliques (PPs) in Hindi cannot be relativized in perfective prenominal relatives follows since they all receive a case from a postposition inside the relative clause.

 

102   Anoop Mahajan

4.5.2 PP DOs and DOM Objects Interestingly, at least some oblique DOs can appear inside the prenominal perfective relatives. (32) [dušman-​se mil-​ii (huii)] enemy-​with join-​perf.fem be.perf.fem ‘a/​the girl who has joined (with) the enemy’

laṛkii girl

(33) [dušman-​se ḍar-​ii (huii)] senaa enemy-​with fear-​perf.fem  be.perf.fem  army ‘a/​the army that is afraid of the enemy’ However, a differentially case marked object cannot appear within such relative clauses. This has obvious consequences for analyses of DOM. (34) *[laṛkii-​ko dekh-​aa (huaa)]   girl-​DOM  see-​perf.masc  be.perf.masc    ‘a/​the man who saw the girl’

aadmii people

(35) *[dhyaan-​se  tasveer-​ko dekh-​aa (huaa)] aadmii   care-​with picture-​DOM  see-​perf.masc  be.perf.masc  man   ‘a/​the man who carefully saw/​examined the picture’ The contrast between (32)–​(33) and (34)–​(35) is interesting. Examples (32)–​(33) tell us that the constraint on not having a DO inside a perfective prenominal relative clause is not about DOs per se but is about whether that DO is a PP or not. A PP DO does not need PP external case licensing, and we would expect it to survive inside a prenominal relative clause, which it does. The ungrammaticality of (34) and (35) appears to be telling us that the differential object marker -​ko is treated differently from normal postpositions. In particular, it looks like DOM -​ko objects require structural case licensing. Given that no structural case is available inside perfective prenominal relative clauses (as we have argued), the ungrammaticality of (34) and (35) can be attributed to the failure of case licensing of DOs in these examples. The consequence of this is that the DO -​ko marking itself cannot be the morphological realization of structural ACC case (as is often assumed in Hindi linguistics).23 Furthermore, it also tells us that -​ko marked objects in Hindi need to be structurally case licensed. Given that we have already argued that there are no structural case licensing heads inside perfective prenominal

23  See Bhatt and Anagnastapolou (1996) and Kalin (2014) for some relevant discussion, where it is argued that –​ko may be assigned, or enters the derivation, higher than the base position of the DOs. Both of these proposals are compatible with the current proposal.

 

Accusative and ergative in Hindi    103 relative clauses in Hindi, it now becomes possible to entertain the possibility that the -​ko objects in (36) and (37) (finite main clause counterparts of the prenominal relatives in (34) and (35)), are actually case licensed by the finite T, i.e. they have NOM structural case (Mahajan 1990). (36)  lõgõ-​ne laṛkii-​ko dekh-​aa people-​erg  girl-​DOM  see-​perf.masc  ‘The people had seen the girl.’ (37)

thaa be.pst.masc

aadmii-​ne  dhyaan-​se  tasveer-​ko dekha man-​erg care-​with picture-​DOM  see-​perf.masc  ‘The man had seen the picture carefully.’

thaa be.perf.masc

I am obviously not suggesting that (all) -​ko marked objects in Hindi always have NOM. In section 4.6, we will see that such objects may have ACC when ACC is available.

4.6  Imperfective Participle Prenominal Relative Clauses 4.6.1 The Case of DOs Hindi prenominal relatives can also be built on an imperfective participle and the properties of this type of relative clause provides more support for the proposal that I have developed so far. Recall that Hindi ergativity is crucially dependent upon perfectivity and that imperfective transitive constructions cannot have ergative subjects. This was shown in section 4.3.1 by the contrast between examples (2) and (3) (repeated below). (2) mε˜-​ne vo akhbaarẽ jəldii-​se  beç-​ĩĩ thĩĩ I-​erg(masc)  those  newspapers(fem)  quickly sell-​perf.fem  be.pres.fem ‘I sold those newspapers quickly.’ (3) mε˜ (*-​ne)  vo akhbaarẽ jəldii-​se  I(masc) those  newspapers(fem)  quickly ‘I (habitually) read those newspapers quickly.’

paṛh-​taa hũ read-​imperf.mas  be.pres.IP

I argued in section 4.5, contra Bhatt (2005) and Legate (2008), that the DO in (2) does not have ACC (but has NOM) and that ACC is systematically unavailable in transitive perfective clauses. The issue that I now take up is the case of the identical-​looking DO vo akhbaarẽ ‘those newspapers’ in the transitive imperfective clause in (3).

 

104   Anoop Mahajan For some relevant evidence, I once again turn to prenominal relative clauses. Examples (38)–​(40) are prenominal relative clauses that contain imperfective participle main verbs.24 (38) [bhuukh-​se mar-​taa (huaa) /​ (*hε˜)] aadmii hunger-​from  die-​imperf.masc  be.perf.masc / ​be.pres  people ‘a/​the man (who is) dying of hunger’ (39) [saṛkõ-​par  so-te (hue) /​ (*hε˜)] log roads-​on sleep-​imperf.pl  be.​perf.​pl/ ​ be.pres.pl  people ‘people (who are) sleeping on the roads’ (40) [tezii-​se bhaag-​tii (huii) /​ (*hɛ)] laṛkii quickness-​with  run-​imperf.fem  be.perf.fem  be.pres.pl  girl ‘girl (who is) running fast’ Examples (38)–​(40) are all cases of subject relativization. Given: (i) that prenominal imperfective relative clauses must be non-​finite, as shown by the ungrammaticality of attempts to insert a finite auxiliary inside them (compare with (3) where a finite auxiliary appears in a normal imperfective participle clause); and (ii) our discussion earlier showing that the relative clause head of prenominal clauses is case marked externally, we expect (38)–​(40) to be grammatical, since the non-​finite relative clause does not have the capability of licensing an argument with a case assigned by a finite T.  When the subject is relativized in (38)–​(40), under our assumptions, it can be externally case licensed, and therefore the grammaticality of (38)–​(40) is in line with our expectations. The crucial difference between the imperfective prenominal relative clauses and the perfective prenominal relative is that the former allow a lexical DO inside them while the latter do not (as we observed in section 4.5). (41) [kitaab  paṛh-​tii (huii)] laṛkii book read-​imperf.fem  (be.perf.fem)  girl ‘a/​the girl (who is) reading the book’ (42) [sarkaar-​se apne haq maaŋg-​te (hue)] government-​from  self ’s rights ask-​imperf.pl be.perf.pl  ‘people demanding their rights from the government’

24 

log people 

The morphological shape of the optional auxiliary in (38)–​(40) is the same as in prenominal perfective relatives. While I continue to gloss this auxiliary be.perf, the relative clauses in (38)–​(40) are not semantically perfective. Interestingly, though, they have a progressive reading and they lack the habitual interpretation that is available with main clause imperfectives in Hindi. The precise nature, function, and representation of the optional auxiliary remains an open question.

 

Accusative and ergative in Hindi    105 (43)

[Mohan-​ko lambe lambe  patr likh-​tii (hui)] paagal  laṛkii Mohan-​dat  long long letters  write-​imperf.fem  (be.perf.fem)  crazy girl ‘the crazy girl writing long letters to Mohan’

Given that a finite T is absent in these clauses, the DO nominal can only be licensed by a different head, and I suggest that this head is the little v of the imperfective vP. This also implies that the case of the DO in main imperfective clauses like (3) is ACC. A sketch of the analyses of finite imperfective transitive clauses and imperfective prenominal relative clauses is provided in Figures 4.5 and 4.6 (Figure 4.5 repeated from Figure 4.2 in section 4.3.1).

selection … TFIN [AspP

Aspimperf

[vP

SUB

[

Vimperf

NOM

V

DO ]]]

ACC

Figure 4.5  Case relations in Hindi imperfective transitive clauses

selection … TNONFIN

[AspP

Aspimperf

[vP SUB

Vimperf

[

V

DO ]]]

ACC NOM NOT AVAILABLE

subject must be externalized

Figure 4.6  Case relations in Hindi prenominal imperfective relative clauses

If my suggestion that prenominal relative clauses are externally case marked is correct, then a clear prediction can now be made. The prediction is that unlike in perfective prenominal relative clauses, the imperfective prenominal relative clauses will disallow DO relativization, since that configuration will lead to case marking the head of the relative clause twice (inside the relative clause and also externally). This prediction is fulfilled. (44) *[laṛkii-​kii  paṛh-​tii (huii)] kitaab girl-​gen read-​imperf.fem  be.perf.fem  book ‘the book that the girl is reading’ (45)

*[logõ-​kaa sarkaar-​se maaŋg-​taa (huaa)] people-​gen  government-​from ask-​imperf.masc be.perf.masc ‘the right that the people are demanding from the government’

apnaa haq self ’s right

 

106   Anoop Mahajan (46) *[paagal laṛkii-​ke  Mohan-​ko  likh-​te (hue)] lambe lambe  patr crazy girl-​gen Mohan-​dat write-​imperf.pl  (be.perf.pl)  long long letters ‘the long letters that the crazy girl is writing to Mohan’ In fact, the prediction is that only subjects can be relativized in imperfective relatives, given that the imperfectives have a little v that licenses DOs and oblique nominals like locatives and instrumentals are internally case marked PPs. These nominals are therefore internally case licensed and cannot move to a clause external case position. As shown below, PPs fail to be relativized using this strategy. (47) *[Mohan-​kaa  ḍuub-​taa (huaa)] kũãã   / ​kũẽ -​mẽ Mohan-​gen drown-​imperf.masc  (be.perf.masc)  well   well-​in ‘the well in which Mohan is drowning’ (48) *[Raam-​kaa  čuhaa  maar-​taa huaa] čaaku ​/čaaku-​se Ram-​gen rat kill-​imperf.masc.sg  be.perf.masc.sg  knife​ /knife-​with ‘a/​the knife with which Ram is killing the rat’ If the discussion in this section is on the right track, then we now have an answer to the question posed at the beginning of this subsection where we asked about the case of the DO nominal in (2) and (3) given that there is no morphological case distinction. The answer is that the DO in (3) has ACC, while the DO in (2) does not have ACC, but has NOM as suggested in section 4.5.

4.6.2 DOM Objects Following up on the parallel discussion on morphologically marked objects, we expect DOM objects to be fully well formed inside the imperfective prenominal clauses, since they will now be able to have ACC. This prediction is fulfilled. (49) [laṛke-​ko ghuur-​tii (huii)] laŗkii boy-​DOM  stare-​imperf.fem  be.perf.fem  girl ‘a/​the girl (who is) staring at the boy’ (50) [dhyaan-​se  tasveer-​ko dekh-​taa (huaa)] aadmii care-​with picture-​DOM  see-​imperf.masc  be.perf.masc  man ‘a/​the man who was carefully looking at the picture’ If the DOM objects in (49) and (50) have ACC, a further prediction is that they cannot be relativized in these constructions. This prediction also holds.

 

Accusative and ergative in Hindi    107 (51)

*[laṛkii-​ke  ghuur-​te (hue)] laṛke-​ko girl-​gen stare-​imperf.obl  be.perf.obl  boy-​DOM ‘the boy whom the girl is staring at’

(52)

*[aadmii-​kii  dhyaan-​se  dekh-​tii (huii)] tasveer-​ko man-​gen care-​with see-​imperf.fem  be.perf.fem  picture-​DOM ‘the picture that the man is carefully looking at’

To conclude this section, I hope to have shown that imperfective transitive constructions in Hindi always have an ACC available and that this case is assigned to both morphologically unmarked objects as well as to DOM objects. Thus, imperfective vPs in Hindi are crucially distinct from perfective vPs with respect to their case licensing capability.

4.7  General Discussion and Conclusions The basic question that this chapter addressed concerned the structural case assigned to the DO in Hindi ergative constructions such as (1) repeated below. (1)  Kabir-​ne vah laal gaaṛii jəldii-​se beç-​ii thii Kabir-​erg(masc)  that  red  car(fem)  quick-​with  sell-​perf.​fem.sg  be.pst.fem.sg ‘Kabir had sold that red car quickly.’ There is no morphological case on the DO in (1), a common situation in ergative languages where the absolutive case is often null. Legate (2008) suggests that it is possible to identify two types of languages with null absolutives. She distinguishes between ABS=DEF(ault) vs ABS=NOM type languages. She argues that Hindi is an ABS=DEF language. In particular, she proposes that the distinction between the two types of languages is located in the case assignment properties of little v. In ABS=DEF, little v assigns accusative case, while in ABS=NOM languages, little v does not assign accusative case (Legate 2008: 58). She further suggests that in ABS=DEF languages, nominative case is not assigned in transitive clauses and the subject receives an inherent ergative case while the DO receives ACC. Thus, with respect to (1), her specific proposal would be that: (i) the DO has ACC; and (ii) NOM is not assigned in (1). In this chapter, I have argued against both of these proposals by bringing in a new set of data that abstracts away from morphology and focuses on argument licensing in terms of abstract case licensing as originally envisaged within the GB framework (see Chomsky 1981 and Vergnaud 2008). I have argued that the DO in (1) does not have ACC, and that ACC is

 

108   Anoop Mahajan not licensed by perfective little v in Hindi. I have also argued that the case that the DO in (1) has is NOM.25 In trying to uncover the case licensing condition on DOs, we have observed that DOM objects behave like non-​DOM objects in their case licensing in Hindi participle prenominal relative clauses. I have suggested that this implies that DOM objects must be structurally case licensed, that they can have a NOM (in perfectives) or ACC (in imperfectives), and that the presence of DOM -​ko cannot be taken to reflect the same underlying structural case (usually suggested to be accusative). I have also argued that despite superficial appearances, the DO in Hindi imperfective constructions bears ACC. That is, the imperfective little v is an ACC licenser. Given the pattern of data that we have seen in this chapter, it appears that ERG and ACC in Hindi are in complementary distribution, though a satisfactory theory of this complementarity remains unclear.26

Acknowledgments Parts of this chapter were presented in the ‘Case by Case’ Workshop held at École Normale Supérieure, Paris, in October 2011, and at the Third Formal Approaches to South Asian Languages workshop at USC in March, 2013. My thanks to the participants of these workshops, including Adriana Belletti, Rajesh Bhatt, Jessica Coon, Amy Rose Deal, Hilda Koopman, Luigi Rizzi, Dominique Sportiche, and K. V. Subbārāo for their comments. I also thank an anonymous reviewer for helpful comments.

Abbreviations Abbreviations used in the glosses in this chapter: 1P, first person; ABS, absolutive; ACC, accusative; AGR, agreement node; AUX, auxiliary verb; DAT, dative; DO, direct object; DOM, differential object case marking; ERG, ergative; FEM, feminine gender; FUT, future tense; GEN, genitive; IMPERF, imperfective aspect; MASC, masculine gender; NOM, nominative; OBL, oblique; PERF, perfective aspect; PL, plural; PRES, present tense; PST, past tense; SG, singular; T, tense node.

25  In this chapter, I have not invoked the argument that the presence of object agreement on T is a reflex of NOM. For more discussion on that topic, see Mahajan (1990) and Bhatt (2005). 26  Marantz’s (1991) dependent case assignment account is a possible contender. Another possibility (a variation on Marantz’s proposal) is that a little v can license case once, either an inherent ERG or a structural ACC. It is not obvious how such accounts can be parametrized if ERG–​ACC languages do exist.

 

The Nature of Ergative Case

 

 

Chapter 5

On inherent a nd de pendent th e ori e s of ergative  c ase Mark C. Baker and Jonathan David Bobaljik

5.1 Introduction The phenomenon of ergative case, in which a specially marked case is found only on the subject of a transitive clause (see (1)) has long posed a special challenge for the ‘Case Theory’ of Government and Binding (GB) and its descendants. (1)  a.  Maria-​nin-​ra ochiti noko-​ke. Maria-​ERG-​PRT  dog      find-​PRF ‘Maria found the dog.’ b.  Maria-​ra Maria-​PRT ‘Maria went.’

(Shipibo, Baker 2014a: 342)

ka-​ke. go-​PRF

This is partly because in this theoretical tradition the structural configurations posited as being relevant for the assignment of structural case are the same as the structural configurations that characterize grammatical functions. For example, nominative case is assigned to the subject of a finite clause, interpreted as either a particular position (Spec,IP) or a configurational relation with a functional head (governed by Tense). The problem, then, is that ergative and absolutive cases simply do not align with the grammatical functions of subject and object. Nevertheless, many theories of ergativity that are broadly consistent with a GB-​style Case Theory have been offered. The most prominent such approach in recent years is one that treats ergative case as an inherent case. We call this the IC theory (ICT). This approach maintains a conservative version of Case Theory, but restricts its scope: the subject of a transitive clause in an ergative language

 

112    Mark C. Baker and Jonathan David Bobaljik receives case in a manner fundamentally different from either the subject or object in a nominative–​accusative language. An alternative approach to ergative alignment, initiated by Marantz (1991) and developed extensively in Baker (2015), among others, is the dependent case theory (DCT). The syntactic configuration that matters for the DCT is not the absolute position of an NP or its relationship to a functional head, but rather its position relative to other NPs within a particular syntactic domain: if there is only a single NP in the domain (e.g., an intransitive clause like (1b)), then the NP bears unmarked case (NOM/​ABS); if there are two NPs, then at least one of the NPs may bear a dependent case—​so called because its appearance on one NP depends on the presence of another NP in the same domain. On this view, NOM–​ACC and ERG–​ABS/​NOM systems are distinguished by a simple parameter: if the lower of two NPs in a domain is marked, then the dependent case is named accusative; if the higher is marked, as in (1a), it is named ergative. Languages may choose to mark one or the other (or both or neither; yielding further alignment types). For concreteness, we may express the DCT view of ergative as follows (see Baker 2015 for refinements; and the chapters by Nash, Coon and Preminger, and Baker, Chapters 8, 10, and 31 in this volume, for analyses making use of a DCT). (2) a. If NP1 c-​commands NP2 and both are contained in the same domain (say, clause), then value the case feature of NP1 as ergative. b. Otherwise NP is nominative/​absolutive. In this chapter, we put these two approaches to ergative case side by side, comparing some of their predictions and assessing them empirically.1 In doing this, we review some of the arguments that have been given for treating ergative as inherent case. We show that these arguments fail to generalize, and indeed provide evidence that ergative in some languages is not an inherent case; rather, it is best characterized under the dependent case viewpoint (at least for many canonical ergative languages). On the DCT side, (2) will already do as a starting point for comparison. There is, however, more to say about the roots of the ICT, so we begin by reviewing the defining characteristics of inherent case.

5.2 Inherent Case 5.2.1 The Roots of Inherent Case It is well known that in languages with even moderately rich systems of case morphology, there seems to be a rough division between structural cases, which are determined 1  There is, of course, a third major theoretical contender: the view that case is assigned to an NP by a functional head under a relationship of Agree. However, that sort of theory is known to be particularly hard to apply to ergative case (see, for example, Baker 2015: ch. 2), so we do not consider it here.

 

On inherent and dependent theories of ergative case    113 by the surface–​syntactic context that an NP finds itself in, and inherent cases, which are more restricted, linked either to particular semantic functions (theta-​roles) or assigned as a quirky lexical property of particular heads. Structural case is typically blind to thematic roles: a surface subject bears nominative, regardless of its exact theta-​role, and an object bears accusative. The ECM/​Raising-​to-​Object configuration (I believe him to be my friend) provides a canonical example of structural case: the subject of the non-​finite complement of a verb like ‘believe’ may bear no thematic relation to the matrix predicate, only a structural one, yet because of that structural relation, the NP has accusative case. Conversely, passive and unaccusative clauses show that the thematic object (patient) of a verb can bear either nominative or accusative case, depending on the larger structure that contains it. Inherent case, by contrast, is tied to theta-​roles and/​or to specific predicates. Verbs meaning ‘help’ in many languages assign dative to their object, rather than accusative. As theta-​roles are a property of an object’s base position, it is this base position, rather than any subsequent post-​movement configuration, that matters. As a result, inherent case NPs typically do not undergo case-​alternations. Thus, in a language like Icelandic, the internal argument of ‘help’ is dative not only in a simple active sentence, but also in the corresponding passive, in a passive under an ECM predicate like ‘believe,’ and in a passive under the passive of an ECM predicate (see Andrews 1990). Inherent case comes to this argument from the verb along with its thematic role.2 Since, in the normal instance, a given NP bears a unique theta-​role, unchanged throughout the derivation, inherent case cues this ‘base’ position of the NP, and remains constant. Structural cases, by contrast, appear to alternate among closely related derivations, such as active and passive, where the theta role of an NP remains constant but its surface grammatical relation (highest A-​position) varies.

5.2.2 Ergative as an Inherent Case Against this background, a prominent line of reasoning within the recent Chomskyan tradition holds that ergative is an instance of inherent case (see L. Nash 1996; Woolford 1997, 2006; Aldridge 2004, 2008a, 2012b; Anand & Nevins 2006; Laka 2006b; Legate 2006, 2008, 2012a; Massam 2006; Coon 2013a; Mahajan 2012; see also the chapters by Sheehan, Laka, Legate, and Woolford, Chapters 3, 6, 7, and 9 in this volume). These authors hold that ergative is assigned by v to the external argument in Spec,vP together with that argument’s theta role, much as dative case is assigned to the internal argument of ‘help’ along with the beneficiary theta-​role in Icelandic. The difference between

2  This holds at least in the canonical A-​movement environments. Inherent dative case is apparently not preserved in s-​passives/​unaccusatives, and ‘get’-​passives (Zaenen & Maling 1984). It is debated whether these constructions involve simple NP-​movement, or a more complex structure with a binding dependency between the nominative NP and the theta-​position. See Shimamura (2014a) and Wood (2014) for contrasting views.

 

114    Mark C. Baker and Jonathan David Bobaljik a NOM–​ACC alignment and an ERG–​ABS alignment under this view can be schematized as in (3). (3) 

ACC

NOM a.

[ [TP

T0

[ [vP

v0

EA

[VP

V

θ

IA

]]]

IA

]]]

θ

?NOM b.

[ [TP

T0

[ [vP

v0

EA θ,ERG

[VP

V θ

In the NOM–​ACC alignment (3a), v assigns a theta-​role to the external argument, but assigns structural accusative case to the internal argument. The external argument receives its case from (finite) T. By contrast, in the ERG–​ABS alignment, the EA receives case along with its theta-​role from v ((3b)). The IA is left to get case by other means: it may get nominative case from finite T, either at a distance via Agree or by moving to SpecTP; it may get (covert) accusative case from v (Legate 2008), or it could conceivably be left to get default case, as in the DCT. Proponents of such a view may point to ostensible differences between ergative subjects and nominative subjects. For example, they may claim (i) that ergative case is associated with a particular thematic role, such as agent, and (ii) that ergative does not alternate with other cases in ECM or raising-​type constructions (e.g. Woolford 2006).3 Subsumed under point (i) are two ways in which ergative case in certain languages may depart from the canonical ergative pattern of being associated with all and only the subjects of transitive verbs. On the one hand, there are languages, such as Georgian, Basque, and Hindi, in which the subjects of (some) agentive intransitive verbs (unergatives) bear ergative case. Examples of this sort suggest that ergative case is tied to the external (agent, actor) theta-​ role, not to transitivity as such. On the other hand, in some ergative languages, there are subjects of two-​argument verbs that fail to receive ergative case. This latter point can be formulated as the Ergative Case Generalization (ECG, Marantz 1991: 236): (4) Even when ergative case may go on the subject of an intransitive clause, ergative case will not appear on a derived subject. Legate (2012a:  183)  emphasizes this generalization, noting that:  “If the Ergative Case Generalization holds, it constitutes a powerful argument for the inherent analysis of ergative case—​since this analysis predicts the generalization to hold. Ergative is assigned by v to thematic subjects, and thus should not appear on derived subjects.” If derived 3 

Ergative and nominative subjects may also differ with respect to controlling agreement on the verb, or in their scope properties (Anand & Nevins 2006); these differences are not systematically correlated with ergative versus nominative alignment however. On the interaction of case and agreement, see Baker (2008, 2015) and Bobaljik (2008).

 

On inherent and dependent theories of ergative case    115 subjects move from a theta-​position to a non-​thematic position, then they will be ineligible to receive a theta-​related case in their surface position. We can now put the ICT and the DCT side by side. According to the ICT, the main factor in ergative case assignment should be what head a given NP gets its thematic role from. It should not matter (much) how many other NPs are in the clause. As a result, one would expect ergative case assignment to be relatively stable, little affected by properties of the larger structure, just like dative case on the complement of ‘help’ in Icelandic. In contrast, according to the DCT, the main factor in ergative case assignment should be how many NPs are in the same local domain. It should not matter (much) what the thematic roles of those NPs are. As a corollary to this, we might expect ergative case assignment to be relatively fluid, with the same NP getting different cases in different clauses depending on its context. Indeed, there are some constructions in some languages that look promising for the ICT, and have helped get it fixed in the beliefs of many in the field, as mentioned above. However, we want to show that, looking more broadly at other languages, there is much evidence of the sort that supports the DCT and looks problematic for the ICT. Indeed, our assessment is that the bulk of the data for the ergative languages we know about is on the side of the DCT. We look first with some care at cases in which nonagents (non-​external arguments) get ergative case despite having the wrong theta-​role, as long as there is another NP around, contrary to the ECG. Then we look more briefly at situations in which agents (theta-​ marked by v) fail to get ergative case even though they have the right theta-​role, because there is no other suitable NP in the vicinity. Finally, we take a broader typological view, pointing out that the ICT really predicts an active case pattern, rather than an ergative case pattern, but no paradigm instance of a dependent-​marking language with an active alignment pattern is known to typology. This gap is however expected on the DCT.

5.3  Ergative Case on Internal Arguments Our most complete instance of ergative case on internal arguments comes from Shipibo, a uniformly ergative language of the Panoan family, spoken in Eastern Peru; see Valenzuela (2003) (PV) for a thorough description; our synopsis here follows Baker (2014a). In this language, it is particularly clear that derived, nonagentive subjects can receive ergative case because Shipibo happens to have productive, morphologically overt applicative constructions. Legate (2012a: 183) points out explicitly why such applicative constructions are significant for theories of ergative case. She writes that: An additional way around the confound would be a two-​argument verb in which both arguments are internal, for example, the passive of a double object verb, or the applicative of an unaccusative verb. If the [ICT] holds, the subject of such verbs would not bear ergative case, despite the presence of two DP arguments.

 

116    Mark C. Baker and Jonathan David Bobaljik We accept this prediction, and the reasoning behind it, but observe that Shipibo shows it to be false, such that the ICT fails whereas the DCT succeeds for this language. In Shipibo, all dyadic verbs with agentive subjects get ergative case (-​n, -​nin, -​kan) (see (1a)), but no intransitive subjects do in simple clauses. This is shown in (5a) for unergatives and in (5b) for unaccusatives.4 (5) a. Joni-​bo-​ra teet-​ai; Rosa-​ra bewa-​ke. (see also PV: 336–​337) person-​PL-​PRT work-​IMPF Rosa-​PRT sing-​PRF ‘The people are working.’   ‘Rosa sang.’ b. Kokoti-​ra joshin-​ke; fruit-​PRT ripen-​PRF ‘The fruit ripened.’

Maria-​ra mawa-​ke. Maria-​PRT die-​PRF   ‘Maria died.’

Shipibo also has three applicative affixes (-​xon, -​anan, -​kin; see Valenzuela 2003:  ch. 17); we focus on -​xon, since it makes all the essential points. This affix can attach productively to transitive verbs and unergative verbs, adding an argument that is interpreted as affected by the event—​either a benefactive or a malefactive. This argument is structurally lower than the agent, but higher than the theme. Notice that the unergative subject in (6b) with the applicative bears ergative case, in contrast to (5a). (6) a. Jose-​kan-​ra Rosa atapa rete-​xon-​ke. Jose-​ERG-​PRT Rosa hen kill-​APPL-​PRF    ‘Jose killed a hen for Rosa.’ b. Papashoko-​n-​ra Rosa bewa-​xon-​ai. grandfather-​ERG-​PRT Rosa sing-​APPL-​IMPF ‘The grandfather is singing for Rosa.’

(also PV: 695–​699)

(also PV: 689–​690)

This is a high applicative in the sense of Pylkkänen (2008), and it fits well into the standard theory in which an applicative head theta-​marks the applied object, takes a VP potentially including the theme as its complement, and itself serves as the complement of the agent-​ assigning head v. Now crucially, -​xon can also attach to unaccusative verbs. Two examples are given in (7). Other attested examples have glosses like ‘spoil on,’ ‘get sick on,’ ‘grow up for,’ ‘turn sour (ferment) for,’ and ‘sink on’ (also PV: 691, 694).

4  For discussion of four possible exceptions to this (‘row,’ ‘pole,’ ‘breathe,’ and an alternative word for ‘die’ (rebes-​), mentioned by Valenzeula (2003), see Baker (2014a: 350–​352).

 

On inherent and dependent theories of ergative case    117 (7) a. Nokon   shino-​n-​ra      e-​a    mawa-​xon-​ke.   (*shino-​ra) my.GEN  monkey-​ERG-​PRT  me-​ABS  die-​APPL-​PRF  monkey.ABS-​PRT ‘My monkey died on me.’ b. Bimi-​n-​ra Rosa joshin-​xon-​ke. fruit-​ERG-​PRT Rosa ripen-​APPL-​PRF ‘The fruit ripened for Rosa.’

(*bimi-​ra) (*fruit-​PRT)

Note that the examples in (7) (and all similar examples) have theme arguments bearing ergative case; we do not, for example, get a double absolutive configuration in this construction. This goes against Legate’s (2012a) prediction, derived from the ICT. However, the DCT can account for this, since the examples in (7) have a second NP, not present in (5b). Hence it is not surprising that the theme argument is ergative in (7) but not (5b). More generally, comparing (5a) with (6b) and (5b) with (7) shows that what theta role an NP has (agent or theme), or what head it gets its theta role from (v or V), does not determine whether it is ergative or not, whereas how many NPs are in the clause (one or two) clearly does.5 Thus, we see that the ECG is spurious, although clear violations are seen only when various factors converge (see discussion of (13) below) Shipibo is special in that it has productive morphological applicative constructions, and we have independent evidence whether a verb is unaccusative or unergative (see Baker 2014a: 368–​371). Other ergative languages may not present quite as clear a picture, either because we know less about them, or because they lack productive applicatives of unaccusatives. Nevertheless, we do not think that Shipibo is unique in any important sense; rather, there is evidence from other ergative languages that, although more fragmentary, seems to go in the same direction. For example, Kalaallisut (West Greenlandic) is another ergative language that has a morphological applicative (underlyingly -​uti-​, Fortescue 1984: 89–​90). When this affix is added to an intransitive verb, the result is a transitive clause with ergative case on the argument that bears absolutive in the intransitive version. The affix is not fully regular and productive, but among the predicates that -​uti-​may attach to are non-​agentive predicates of emotion: (8) a. kamap-​p-​uq angry-​INDIC-​3SG.ABS ‘He/​she is angry.’ b. Arna-​p angut kama-​ap-​p-​aa woman-​ERG man.ABS be.angry-​APPL-​INDIC-​3SG>3SG ‘The woman is angry with the man.’ (Michael Fortescue, p.c.) 5  See Baker (2014a) for further details and refinements, including an explanation for why applicatives of unaccusatives are different from dyadic verbs with experiencer subjects, which do take two absolutive arguments and no ergative argument, in apparent support of the ECG (see also Baker, Chapter 31, in this volume, on analogous predicates in Burushaski).

 

118    Mark C. Baker and Jonathan David Bobaljik Note that in English, roots like ‘anger’ undergo a kind of transitivity alternation (Chris is angry vs. Pat angered Chris), suggesting that the experiencer can count as an internal argument. The corresponding experiencer is nevertheless marked for ergative case in the applicative version in (8b). Therefore, if predicates like ‘be angry’ are confirmed to be unaccusative in Kalaallisut (as claimed also by Spreng 2012), then (8b) replicates (7) from Shipibo—​additional evidence for the DCT over the ICT. Chukchi is an ergative language that does not, apparently, have a productive morphological applicative that combines with unaccusatives, but it does have inchoative and locative alternations that seem relevant, according to Nedjalkov’s (1976) survey of Chukchi’s many diathesis alternations. First, Chukchi has a morphologically unmarked causative–​inchoative alternation, which applies to verbs like ‘fill,’ among others: (9)

a. ətləg-​e jərʔen-​nin əʔtvʔet miml-​e father-​ERG fill-​3SG>3.SG   boat.ABS water-​INSTR ‘Father filled the boat with water.’ b. əʔtvʔet jərʔet-​gʔi miml-​e boat.ABS fill-​3SG water-​INSTR ‘The boat filled with water.’

Comparison with (9a) strongly suggests that both ‘boat’ and ‘water’ are internal arguments of ‘fill’ in (9b), and thus that ‘fill’ with these two arguments counts as an unaccusative verb with a derived subject. Further evidence for this is the fact that either ‘water’ or ‘boat’ can incorporate into the verb ‘fill’ (Nedjalkov 1976: 189, 208; see also Baker 1988). Chukchi also happens to have conative and locative-​type alternations, where a given argument can be projected as either a PP (realized as semantic case) or as a bare NP (Nedjalkov 1976: 193, 206, etc.). In particular, the locatum argument that is projected as an instrumental PP in (9a, b) can alternatively be projected as an NP. When this happens with the agentless version of ‘fill,’ it has two NPs as internal arguments—​and one of them (the locatum argument) is crucially marked with ergative case: (10) əʔtvʔet jərʔen-​nin boat.ABS fill-​3SG>3.SG ‘Water filled the boat.’

miml-​e water-​ERG

(Nedjalkov 1976: 195, 206)

Note that instrumental case and ergative are syncretic on inanimate NPs in Chukchi, but they are distinguished by agreement: the verb agrees with an ergative NP, but not an instrumental one. In particular, ‘fill’ agrees with ‘water’ in (10) but not in (9b), confirming that ‘water’ is ergative in (10), despite it being a theme-​type internal argument. (10) is thus another probable counterexample to the ICT (as noted in Bobaljik and Branigan 2006). In contrast, the DCT can explain why a theme subject gets ergative case if and only if there is another internal NP (not a PP) in the clause. ‘Cover’ is another verb that exhibits this diathesis pattern (Nedjalkov 1976: 195).

 

On inherent and dependent theories of ergative case    119 Yup’ik is yet another ergative language that seems relevant. It has a type of malefactive applicative in which the affix -​i-​is added to the verb along with an additional NP that expresses someone adversely affected by the event, as seen in (11b). Note that in this example the agent argument of ‘eat’ is ergative and the malefactee is absolutive, showing that the malefactee is a lower, internal argument. (11)

a. arnar neqa-​mek woman.ABS fish-​ABL ‘The woman ate a fish.’

ner’-​uq eat-​INDIC.3SG

(Mithun 2003: 565)

b. qimugte-​ m ner-​ i-​ a           angun  akuta-​ mek dog-​ERG  eat-​APPL-​3SG>3SG man.ABS mixture-​ABL ‘The dog ate some “akutaq” on the man (ate the man’s akutaq).’ (Mithun 2000: 97) Now (12a) is a typical unaccusative clause, and (12b) is a related malefactive construction derived from the same verb root. Crucially the malefactive argument bears ergative case in (12b). (12)

a. Maklagaq kit’e-​llru-​uq bearded.seal.ABS sink-​PAST-​INDIC.3SG ‘The bearded seal sank.’ b. Ing-​um maklagaq kic-​i-​lq-​aa that.one-​ERG bearded.seal.ABS sink-​APPL-​PAST-​INDIC.3SG>3SG ‘The bearded seal sank on that guy.’ (Yup’ik; Woodbury 1981: 332–​333)

We clearly cannot say that ‘that guy’ in (12b) is an external argument, or that it receives ergative case along with its malefactee theta-​role by (a generalization of) the ICT, because ‘man’ gets the same theta-​role from the same head (-​i) in (11b) but does not have the same case. The DCT, however, can work, because there is another NP in (12b) (but not in (11b)) which ‘that guy’ c-​commands, namely the theme ‘bearded seal.’ Hence the dependent-​case rule in (2) can apply correctly. While (12) in Yup’ik is similar to (7) in Shipibo in that ergative case is assigned in the applicative of an unaccusative, the languages differ in which of the two internal arguments (theme or affectee) moves to Spec,TP and serves as the subject of the clause. Baker (2014a) argues that applied arguments in Shipibo are NPs embedded in a null-​headed PP; this PP shell prevents the applied argument from moving to Spec,TP. Therefore, the theme argument must move, and it ends up c-​commanding the applied argument and getting ergative case.6 Evidently, in the Yup’ik example in (12) (and similar examples in

6 

Crucially the null P in this construction is not a phase head. That is why its complement is still visible to trigger ergative case on the theme after the theme moves to Spec,TP, whereas canonical PPs do not trigger ergative on the subject. Even overt Ps occasionally fail to be phase heads; see for example Baker & Vinokurova (2010: 623) on three such Ps in Sakha.

 

120    Mark C. Baker and Jonathan David Bobaljik distantly related Kalaallisut, Fortescue 1984: 269) the applied argument is truly an NP, so it can and does move to Spec,TP as the closer NP argument and gets ergative. Despite this difference, both languages provide instances of derived subjects receiving ergative case, contrary to the ECG. In this section, we have presented a series of examples that plausibly involve derived transitive subjects getting ergative case, which the ICT predicts to be impossible. Note that we are not claiming that applicatives of unaccusatives always yield an ERG–​ABS array in ergative languages. The DCT leaves room for arrays other than ERG–​ABS, thus permitting the analysis of a broader range of languages. Niuean provides a prominent example of an ergative language in which some applicatives yield an ABS–​ABS array; (13) is an example (from Massam 2006: 33, cf. Legate 2012a).7 (13) Ne faka-​kofu aki e PAST CAUS-​cover with ABS ‘The canoe was covered with leaves.’

vaka canoe

e ABS

tau PL

lauakau. leaf

Note that this example is thematically similar to (10) from Chukchi, where ergative is assigned. Similar ABS–​ABS patterns with nonagentive subjects are also found with reciprocals derived from ditransitive verbs and with a few psych verbs in Shipibo; these contrast with the applicatives of unaccusatives discussed above. Baker (2014a, 2015) develops one particular version of the DCT in which two arguments are ABS if three conditions hold: (i) the two arguments both start out inside the vP phase (hence are nonagentive), (ii) their initial c-​command relationship is not reversed by NP-​movement (as happens in Shipibo (7) but not Yup’ik (12b)), and (iii) ergative case is assigned at the spell out of TP but not VP. If these conditions do not hold, the higher argument will be ergative, even if it is nonagentive. Encouraging for a theory of this kind is the fact that the “location” argument appears to be the subject in Niuean (13) but the ‘locatum’ argument is the subject in (10) from Chukchi; this suggests that NP-​movement does reverse the arguments in Chukchi but not in Niuean. See Baker (Chapter 31, this volume) for a review of this proposal and some additional discussion, making connections

7 

This example is complicated by having a causative prefix in addition to the applicative aki, as Massam notes. This might suggest an alternative analysis in which it is a concealed ditransitive with a third, hidden (ergative) argument, hence an ERG ABS ABS array. Legate (2012a) also considers ERG–​ DAT and ABS–​DAT arrays in Warlpiri, arguing that the distribution of verbs selecting these frames supports the ECG. She notes, however, that the difference in the arrays may be due to differing properties of the DAT argument in the two verb classes (NP versus PP), an analysis pursued in Baker (2015, Chapter 31, this volume). Niuean also has locative applicative-​like transitive clauses formed from intransitive verbs that yield an ERG ABS array. Massam treats these as agentive intransitives, but some (‘sleep,’ ‘sit/​stay’) may actually be unaccusatives, given that they show reduplication for plural number of the NP that is the subject of the intransitive, and the ergative of the derived transitives (Seiter 1980: 64). This process of participant number marking seems to be otherwise limited to internal arguments; no non-​derived ergative triggers plural number marking on the verb.

 

On inherent and dependent theories of ergative case    121 to variation in ditransitive constructions. We can see, then, some room for variation in how derived subjects are case marked within a DCT, whereas in the ICT the ban on ergative case on derived subjects is expected to be quite rigid. Overall, then, the examples in this section fit with the DCT and go against a core prediction of the ICT. One could of course question whether these subjects are ‘derived’ in precisely the right sense picked out by the ECG. Since the advent of the VP-​internal subject hypothesis, there is a sense in which all subjects are derived. What our examples show, we claim, is that there is no type of theta-​role that is totally immune to ergative case: malefactives, locations, locatums, and even themes can all receive ergative case, if they end up as subjects with another NP lower in the same clause. Thus, what is relevant cross linguistically for ergative case assignment is transitivity within a local domain, not particular theta-​roles. If these configurations are somewhat rare, it is because the lower an NP is on the theta-​hierarchy, the more likely it is that, in a transitive context, the other NP will be the one that becomes the subject. If themes, for example, are quite low, then a configuration in which a theme becomes a transitive subject arises only when the higher argument fails to raise to subject position for some special reason, as in Shipibo.

5.4  Absolutive Case on External Arguments In section 5.3, we argued for the DCT over the ICT approach to ergative case by looking at themes and other internal arguments in noncanonical situations in which there is a second internal NP, showing that in many such situations they receive ergative case. In this section, we look more briefly at the converse situation: agents (external arguments) in ‘noncanonical’ situations in which an internal argument that would otherwise be there somehow becomes unavailable. According to the ICT, we would not expect this to matter much: the external argument presumably gets the same theta-​role from the same head (v) as it does in canonical transitive constructions; therefore, it should get the same inherent ergative case from v, all things being equal. In contrast, the DCT takes the presence of a suitable second argument to be essential to the subject receiving ergative case, so if something happens to that second argument, the subject should receive default absolutive case, by (2). In section 5.4.1, we show that again it is the DCT that makes the right prediction. In section 5.4.2, we then critically discuss proposals by Woolford and others to patch the ICT by supplementing it with a transitivity condition.

5.4.1 Ergative Lost under Detransitivization One relevant thing that can happen to the internal argument of a transitive verb in some ergative languages is that it can incorporate into the verb. This happens productively in

 

122    Mark C. Baker and Jonathan David Bobaljik Chukchi, for example. (14) gives a prototypical noun incorporation (NI) pair (Polinskaja and Nedjalkov 1987: 240). (14) a. ətləg-​e mətqəmət (kawkaw-​ək) Father-​ERG butter.ABS bread-​LOC ‘The father spread the butter (on the bread).’

kili-​nin spread.on-​3SG>3SG

b. ətləg-​ən (kawkaw-​ək) mətqə-​rkele-​nen Father-​ABS bread-​LOC butter-​spread.on-​3SG>3SG ‘The father spread butter (on the bread).’ When the object is not incorporated, the subject is ergative, as expected ((14a)), but when the object is incorporated, the subject is not ergative but absolutive ((14b)). We know of no reason to say that the subjects in these two examples are theta-​marked any differently, yet they differ in case. This is problematic for the ICT, since inherent case is supposed to remain constant when the surrounding syntactic context changes (as dative does in Icelandic). In contrast, this pattern follows from (2), as long as we assume that incorporated objects are inaccessible to the rule of dependent case assignment. This could be either because incorporation happens by compounding in the lexicon, so that there is no object present in the syntax at all, or (as we believe) because the trace of the syntactic movement that creates incorporation is invisible to rules of case and agreement.8 The same effect can be seen in Kalaallisut, with the slight complication that some verbal morphemes require NI in Kalaallisut and others forbid it, so one needs to compare different verbs with similar meanings in that language. A second thing that can happen to an internal argument in some ergative languages is that it can be projected not as an NP but as a PP (with P a phase head). We observed above that Chukchi is rich in these conative/​locative-​type alternations, and that is relevant here as well. In particular, the location argument which is expressed as a PP in (14b) can alternatively be expressed as a bare NP (compare English: I smeared butter on the bread vs. I smeared the bread with butter). When the locatum argument is incorporated but the location argument is projected as an NP, then the agent-​subject is marked ergative again, as shown in (15). (15) ətləg-​e kawkaw mətqə-​rkele-​nen Father-​ERG bread butter-​spread.on-​3SG>3SG ‘The father spread the bread with butter.’ (Polinskaja and Nedjalkov 1987: 240) It is very possible that (15) means something slightly different from (14b), with ‘bread’ counting as a location in (14) but as a theme in (15). But it is very doubtful that the 8  For example, Baker et al. (2005) argue that the phi-​features of the trace of head movement are deleted in Chukchi and some other languages, making it invisible to agreement. This deletion would bleed (2) if we understand ‘NP’ as ‘phrase bearing phi-​features.’

 

On inherent and dependent theories of ergative case    123 subject gets different theta-​roles in (14) and (15): it looks like a canonical agent in both. Comparing these three examples, then, it seems evident that the theta-​role of the subject does not determine whether it is ergative or not, but whether there is another NP in the same domain as the subject does. In (15), ‘bread’ is another NP in the same domain as the subject, but in (14b) it is not, given that PPs are usually separate domains (phases), the internal constituents of which are invisible to the outside world for purposes of case and agreement (see Baker, Chapter 31, this volume, but see note 7 for some exceptions). A third thing that can happen to a theme object in some ergative languages is that it can be removed as an object by antipassive. Descriptively speaking, antipassive is a morphological process that removes the object of a transitive verb from the core syntax, leaving it either unspecified or expressed as an oblique/​PP. Chukchi illustrates again: (16a) is a normal transitive; (16b) is the corresponding antipassive (Nedjalkov 1976: 201). (16) a. ətləg-​e qərir-​ə-​rkən-​en ekək. father-​ERG seek-​PRES-​3SG>3SG son.ABS ‘The father is seeking the son.’ b. ətləg-​ən ine-​lqərir-​ə-​rkən father-​ABS APASS-​seek-​PRES.3sS ‘The father is searching (for the son).’

(akka-​gtə). son-​DAT

Again, the subject is ergative in (16a) and absolutive in (16b), even though there is no detectable change in its theta-​role—​a problem for the ICT. However, whether a theme argument is syntactically present in the clause, and if so whether it is expressed as an NP or a PP, clearly does affect the case of the subject, exactly as expected under the DCT (although we do not commit to any particular view of the antipassive here). Similar facts can be given for Kalaallisut, and for various Australian languages.

5.4.2 Against Supplementing the ICT with  a Transitivity Condition Unlike the facts surrounding the Ergative Case Generalization, the facts outlined in this section are well known and not in dispute. Prima facie, they seem (to us) to provide a strong argument against the ICT, although the issue is oddly under-​discussed. Where the issue is addressed, proponents of the ICT contend that the thematic condition on ergative case assignment needs to be supplemented (in some languages) with a transitivity condition (Massam 2006: 32; Woolford 2006: 119–120; Legate 2012a: 182). However, the exact nature of this transitivity condition and its theoretical implications are usually left unexplored. (It is not obvious, for example, how v can ‘see’ whether there is an NP inside VP, and if so why that should affect its relationship with its specifier, as pointed

 

124    Mark C. Baker and Jonathan David Bobaljik out to us by Laura Kalin.)9 Canonical instances of inherent case assignment, such as dative in Icelandic, are patently not subject to a transitivity condition, since dative case in Icelandic is possible on the subjects of monadic predicates. Only Woolford (2006) makes a real effort to give independent motivation for a transitivity condition on inherent case, by arguing that a similar condition holds for dative case in Japanese and Basque. Much could be said about her particular examples. For example, in Japanese the putative transitivity condition is far from general, holding if at all in matrix clauses only (Shibatani 1977). Moreover, for languages in which it is true that dative arguments never occur as the sole NP in a clause, there are at least two ways in which this might be explained without recourse to a stipulated (and as yet unformulated) transitivity condition on inherent case assignment. First, it could well be that in some languages dative case is not an inherent case at all, but rather another instance of dependent case—​specifically dependent case assigned to the higher of two NPs generated inside the same VP (see for analyses of this type Baker 2015; Baker & Vinokurova 2010; Bobaljik & Branigan 2006). Second, there could be languages in which dative case is inherent, but in which inherent case nominals may not function as true subjects. Icelandic and German famously contrast in whether datives may (Icelandic) or may not (German) serve as grammatical subjects (Zaenen et al. 1985). Rezac (2008) (see also Davison 2004a) argues that the Basque datives pattern with German rather than Icelandic with regard to subjecthood diagnostics. If there is a requirement of a syntactic (EPP) or morphological (possibly default agreement) subject in every clause, then clauses with dative arguments will always appear to have an additional, possibly null, argument. This effect arises without a stipulated transitivity condition on inherent case. Importantly, since ergative NPs cross-​linguistically satisfy subjecthood tests, this reasoning about Basque datives cannot be extended to ergative. Thus, we conclude that there is no independent motivation for putting a transitivity restriction on the assignment of ergative case, as IC theorists need to do. Indeed, there are syntactic configurations in which the arguments that receive ergative in simple transitive clauses undergo case alternations even though there is no loss of transitivity. Rezac et al. (2014) present Basque perception verbs as part of an extended argument against the ICT. As in many languages, perception verbs may take a full CP complement ((17a)), or a reduced complement ((17b)): (17) a. Katu-​ek sagu-​ak harrapa-​tu cat-​PL.ERG mouse-​PL.ABS catch ‘I saw that the cats caught the mice.’

dituzte-​la ikusi dut. AUX-​that see AUX.1SG

9  Omer Preminger suggests one possible way of working this out. He points out that a v that assigns accusative case to an NP in its domain under Agree could be considered a distinct lexical item from a v that does not. Given this, the agent-​subject in a detransitivized structure (or unergative structure) is theta-​marked by a different v from one in a transitive structure, even though the thematic role is the same. This could account for their different cases, if the first v is not an assigner of inherent ergative case whereas the second v is.

 

On inherent and dependent theories of ergative case    125 b. Katu-​ak sagu-​ak harrapa-​tzen ikusi ditut. cat-​PL.ABS mouse-​PL.ABS catch-​ING seen AUX.1SG>3PL ‘I saw the cats catch mice.’ (Rezac et al. 2014: 1280) The configuration (17b) is a species of ECM environment. Rezac et al. show carefully that the NP katu-​ak ‘cats’ is thematically the external argument of the embedded clause, which is transitive, but it behaves syntactically as the object of the higher clause, and therefore bears absolutive case. On the DCT, (17b) is unremarkable; ‘cats’ does not get ergative case because it is in a case domain with the matrix subject (presumably because CP is missing, contrast (17a)) and it is not the highest NP in the domain.10 Yet on the ICT, the alternation is surprising: inherent case is supposed to be retained on arguments in ECM environments, as inherent dative case is in Icelandic, a model for the ICT. Stipulating that inherent ergative is only realized in a transitive clause—​even in the manner of n. 10—​does not avoid the problem in (17b), since both the matrix and embedded clauses are fully transitive. Causative constructions make a similar point. As in many languages, the causee (the embedded subject) in Basque is marked absolutive if the embedded predicate is intransitive but dative if the embedded predicate is transitive, as in (18). (18)

Jon-​ek Xabier-​ri Mikel hil-​arazi dio Jon-​ERG Xabier-​DAT Mikel.ABS kill-​CAUS AUX.3SG>3SG.3SG.DAT ‘John made Xavier kill Michael.’ (de Rijk 2007: 378; our gloss)

Here again, there is a case alternation (between ergative and dative) such that an agent argument gets different cases in different syntactic environments, contrary to the view that ergative case is an inherent case. Moreover, since ‘kill’ still has its theme argument in (18), no appeal to a transitivity restriction will explain why its subject ‘Xavier’ is not ergative in (18). Although ECM constructions like (17) may be rare across languages,

This approach in essence builds a version of the DCT (ergative is dependent on accusative in the same domain) into the lexical entries for v heads. It strikes us as providing no insight into why this condition might hold. For example, the same machinery would allow one to stipulate that only the v that does not assign accusative assigns inherent ergative to the NP it theta-​marks. This would yield a pattern where only unergative subjects are ergative, while transitive (and unaccusative) subjects are nominative. This pattern cannot be readily described on the DCT approach, and as far as we know it never occurs. Moreover, even if workable, the theoretical possibility sketched here addresses only the narrow issue of how a transitivity condition on ergative might be formulated within an ICT; it does not address the other evidence we have collected in favor of treating ergative as a structural (dependent) case. 10  Omer Preminger very rightly asks why, even given that (17b) is a single case domain, the lower subject ‘cats’ does not get ergative simply by virtue of c-​commanding the lower object ‘mice.’ For us, the ERG–​ABS–​ABS pattern in (17b) is exactly parallel to the ERG–​ABS–​ABS pattern found with simple ditransitive verbs in ergative languages ((almost) never ERG–​ERG–​ABS). However, there is indeed more to say about why this is so: Baker (2014a, 2015) attributes it to a cyclicity effect related to vP also being a phase; see Baker (Chapter 31, this volume) for a reprise.

 

126    Mark C. Baker and Jonathan David Bobaljik morphological causatives are fairly common; the same argument can be constructed in Inuit languages, for example.11 In sum, there is abundant evidence that the appearance of ergative case is tied to surface transitivity. On the ICT, this requires a special condition on the assignment of ergative case. We suggest that once the transitivity condition is brought fully into the light and elaborated properly, as it is in a DCT, there is no further need for a thematic condition that the ergative must be an agent—​at least for the pure ergative languages we are focusing on here, like Shipibo, Chukchi, and Kalaallisut. Moreover, there is evidence from embedded environments that transitivity alone is insufficient as a condition on ergative case, even for canonical agents, whereas the DCT seems to draw the right distinctions.

5.5  Typological Considerations: on the Rarity/​Absence of Active Case Systems 5.5.1 Preliminaries In fact, the challenges for the ICT mentioned in the previous section are arguably even more general. Following the ICT’s leading idea in its purest form naturally leads us to expect ergative case not only on the subjects of detransitivized transitive verbs, but also on the subjects of simple unergative verbs, given that they also receive the agent theta-​role from v on standard generative accounts. In other words, the widespread view that v assigns inherent case to the agent NP that it theta-​marks most naturally generates an active case pattern rather than a true ergative case pattern—​a pattern in which one case appears on the subjects of transitive verbs and unergative verbs, and a different case appears on the objects of transitive verbs and the subjects of unaccusative verbs. This is not the situation in languages like Shipibo, Inuit, and Chukchi, for which the ICT must invoke something like Woolford’s (2006) transitivity condition. But if the ICT is the right leading idea, then we might expect to see it working in purer form in some other languages, unalloyed with a transitivity condition. If this is not so—​if there are no genuinely active case marking languages—​we may begin to doubt not only whether the ICT should be the primary theory of ergative case, but whether it is even allowed by universal grammar. With this in mind, we argue (contra,

11  Julie Legate (p.c.) reminds us that the causee of a causative construction might be projected as an internal argument of the causative verb in a control-​like structure in some languages, rather than as the external argument of the lower verb in an ECM-​like structure. When that happens, it is not expected to get ergative case even within an ICT. However, one would still expect the transitive causee to be invariantly ergative in some subset of ergative languages with a morphological causative (namely those with an ECM-​style causative), whereas we know of no ergative language with this pattern.

 

On inherent and dependent theories of ergative case    127 e.g., Woolford 2015) that current typological knowledge does not offer any promising paradigm case of a dependent-​marking language with a true active case system. Space limitations (and some knowledge limitations) prevent us from discussing any putatively active language in depth, but we outline what we see as the major issues, as a spur to further work. Typological sources say that the active–​inactive/​stative alignment pattern (also called a split-​S or fluid-​S pattern) is attested in languages of the world: see, for example, Merlan (1985); Mithun (1991b); Dixon (1994: 70–83); and Comrie (2005). But our certainty that this is so is marred by the fact that these discussions generally conflate data from morphological case marking on nouns with data from agreement patterns on verbs. Mithun (1991b) is a typical example: of the five active languages that she discusses at some length, four are head-​marking languages (Lakhota, Guaraní, Caddo, Mohawk), and only one (Central Pomo) has overt case marking on NPs. Dixon’s discussion is similar, and he notes in passing (1994: 76) that “for most languages of this type morphological marking is achieved by cross-​referencing on the verb”; see also Merlan (1985: 353). The only languages Dixon mentions as having active case marking on NPs other than (Eastern) Pomo are two Caucasian languages: Laz and Tsova-​Tush. Data from the World Atlas of Language Structures confirms that there is a strong interaction between Nichols’s (1986) head-​marking/​dependent-​marking distinction and the distinction between active languages and true ergative languages. Siewierska (2005) lists 26 out of 380 languages as having an active agreement pattern, a respectable 6.8 percent. Indeed, in agreement-​oriented languages, an active system is slightly more common than a straight ergative system (19/​380, 5 percent). In contrast, Comrie (2005) lists only 4 out of 190 dependent marking languages as having an active case pattern (2.1 percent), and in this language type a straight ergative pattern is much more common than an active one (32/​190, 16.8 percent). For what it is worth, a rough chi-​squared contingency table test confirms that marking type and alignment type are not independent in this data (χ2=25.998, p < 0.00001), suggesting that one should not combine head-​marking and dependent-​marking languages in discussions of this topic. Once we refrain from doing so, we face the fact that active dependent-​marking languages are at best extremely rare. Comrie lists only Basque, Georgian, Imonda, and Drehu, to which we can add the Pomo languages discussed by Mithun and Dixon, Dixon’s Laz and Tsova-​Tush, and Lhasa Tibetan discussed by DeLancey (1984a, 2011). Of these languages, the first two are well known to generativists, and have (together with Hindi) encouraged the idea that ergative case is inherent case; in contrast, the others are little known to generativists. We briefly survey why we do not think that any of these languages provides a good prototype for the ICT to build on.

5.5.2 Split Active Languages Our first observation is that there are many languages such as Shipibo, Inuktitut, Tsez, and Chukchi, whose morphological ergativity shows a clean pattern, in the

 

128    Mark C. Baker and Jonathan David Bobaljik sense that transitivity alone is the determining factor for ergative case. By contrast, in many potential active languages, ‘ergative’ case is conditioned not only on the subject of the clause having an agent thematic role (as the ICT would lead one to expect), but also on a variety of other factors. In Hindi, for example, cited in support of the ICT, only a small subclass of unergative verbs (‘cough,’ ‘bark’) permit ergative subjects, and even with those verbs ergative is optional. Moreover, Hindi is famously a split ergative language, in which ergative only occurs in the perfect aspect (see Butt and King 2003), never in the imperfective. Similarly, in Georgian, the subjects of transitive verbs and some intransitive verbs are famously ergative only in aorist and perfect clauses (see Harris 1981). Similar remarks hold for Lhasa Tibetan, in which subjects of agentive intransitive verbs are sometimes marked with the ergative particle, but only in perfective clauses, never in imperfectives, as illustrated in (19): (19) a. ŋa-​s Seattle-​la I-​ERG Seattle-​to ‘I went to Seattle.’

phyin-​pa-​yin. went-​PRF-​VOL

b. ŋa-​(*s) Seattle-​la I.NOM(*ERG) Seattle-​to ‘I will go to Seattle.’

(DeLancey 1984a: 133)

’gro-​gi-​yin. go-​FUT-​VOL

Yet another example is Drehu, an Austronesian language spoken in New Caledonia (Tryon 1967; Moyse-​Faurie 1983), which actually draws a three-​way distinction among the subjects of present clauses (nominative), past clauses (marked nominative) and other tenses (canonical ergative, with rare exceptions). This interaction with tense–​aspect is well known, but its theoretical implications are not always kept clearly in mind. It strongly suggests that Tense and Aspect heads are heavily involved in the assignment of case—​or, for the DCT, in establishing the relevant case domains—​in these languages, not the theta-​role assigning head v. In none of these so-​called active languages is the subject marked ‘ergative’ if and only if it receives an agent theta-​role from v, as the ICT would expect.12 They are quite different in this respect from Icelandic, where verbs that take inherent dative case subjects do so in every tense–​aspect, just as we would expect on theoretical grounds. Since ergative

12  We admit that we do not have a fully worked out DCT analysis for these “split active” languages. But see Baker (2015, Chapter 31, this volume) for a proposal in which some aspect heads are extra phase heads, affecting when two NPs count as being in the same domain for (2). See also Nash (Chapter 8, this volume) for a new DCT-​style approach to the core facts of Georgian. Note also that in Drehu, noun incorporation of the object bleeds ergative case marking of the subject in relevant clause types (nonpast, nonpresent, Moyse-​Faurie 1983: 159). We took parallel facts in Chukchi to be evidence in favor of the DCT in section 5.4.1.

 

On inherent and dependent theories of ergative case    129 case varies across clauses in this way, it seems very doubtful that it should be thought of as an inherent case on the Icelandic model.13 Tsova-​Tush (aka Batsbi) also shows a split, but in terms of person, rather than tense–​ aspect. This language has been claimed to be a particularly good candidate for an active language, in that ergative case tracks the agentivity (volition and control) of the intransitive subject very closely (Holisky 1987; thanks to Omer Preminger for pointing this out). However, only first and second person pronouns may bear ergative in intransitives; third person pronouns and NPs are always nominative in this role, though they must be ergative as transitive subjects (Holisky 1987: 104–​105, 119). In addition, nuances of volitionality and control, which Holisky characterizes as components of meaning distinct from thematic roles, come into play in case-​selection only with intransitive subjects (1987: 122). Both the person split and the fluidity of interpretation are differences between transitive agents and intransitive agents that are entirely unexpected by the ICT, and they prevent Tsova-​Tush from counting as its missing prototype.

5.5.3 Active Languages with Unmarked ‘Ergative’ The next group of putatively active dependent marking languages to consider includes the Pomo languages (Hokan family, spoken in California; see especially O’Connor 1987) and Imonda (Papuan; Seiler 1985). These languages are all quite similar in that the putative ergative case on agentive subjects is actually morphologically unmarked (Ø); what is overtly marked is a kind of nonagentive case found on (some) direct objects and (some) nonagentive subjects of intransitive verbs: –​al on pronouns in Northern Pomo;  –​m in Imonda. This is already somewhat suspicious for the ICT, since it is extremely rare for ergative case to be morphologically unmarked in a canonical ergative language (Nias is perhaps the only attested example; see Baker 2015: ch. 3 for an analysis). Bittner and Hale (1996b) refer to these as “accusative active languages,” suggesting that it is an extension of the objective case that yields the pattern, rather than the existence of a thematically restricted but phonologically null ergative case. Indeed, O’Connor argues in some detail against the idea that Northern Pomo’s null-​marked case forms are connected to a particular theta-​role in the way that the ICT would hope, concluding that “the A case is semantically unmarked, it does not convey any information about volition, control, agentivity, etc.” (196). Most importantly, it is clear that the putative ergative marking found on intransitive subjects in these languages extends well beyond the canonical unergative class to include many typical unaccusative predicates: for example, ‘die,’ ‘sleep,’ and ‘misspeak’ in Northern Pomo, and ‘startle’ in Imonda. Indeed, Seiler (1985: 145–​148) reports

13  We set aside here the possibility of a hybrid model, in which inherent case-​marked NPs need abstract (structural) licensing in addition, as proposed for Icelandic by Cowper (1987); see Shimamura (2014b) for an extension of this to account for subject–​object extraction asymmetries in syntactically ergative languages.

 

130    Mark C. Baker and Jonathan David Bobaljik that only eight known verbs take the nonagentive case marker on their subject in Imonda. These languages, then, are not good prototypes for the ICT either. Instead, our tentative analysis of these languages is that they are really neutral languages, in which neither dependent ergative case nor dependent accusative case is consistently assigned. Rather -​al in Northern Pomo and -​m in Imonda are fundamentally dative case markers. As such, their core use is on the goal arguments of ditransitive verbs (indeed Seiler’s gloss for -​m is GL, short for ‘goal’). These case markers can also be used on the theme arguments of monotransitive verbs, but only as so-​called differential object markers (Bossong 1985; Aissen 2003b). Thus, overt affixes like -​al are used in Pomo only on pronouns, proper nouns, and (with a clitic or demonstrative) animate or human nouns used as direct objects; inanimate common nouns do not show overt case inflection. Similarly, in Imonda the case marker -​m marks only high-​animate objects (Seiler 1985: 163–​165). This is very reminiscent of markers such as Hindi -​ko, which serve both as differential markers for animate and/​or specific direct objects, and in dative functions such as marking recipients (the most common DOM pattern, Bossong 1985). Once we think of the overt cases in these languages as datives rather than accusatives, an easy hypothesis becomes available for why they are found on the subjects of a proper subset of the unaccusative predicates: these are simply predicates that select for quirky dative case on their subjects, like those known from Icelandic (cf. Marantz’s (1984) generalization that only direct internal arguments—​nonagents—​of a verb can receive lexical case from that verb in Icelandic). This fits well with the fact that only eight known verbs have case-​marked subjects in Imonda, not the whole class of unaccusatives. It also fits with O’Connor’s (1987) observation that having overt case-​marking on the subject in Northern Pomo sometimes expresses ‘empathy’ with the subject rather than agentivity; we interpret such subjects as being experiencers rather than themes, with lexical dative case being assigned to experiencer arguments only. If this is right, then Imonda and the Pomo languages are not the result of v assigning null ergative to agents, but the result of dative case being extended to some objects via DOM and to some nonagentive subjects as an instance of dative subject constructions—​two familiar developments that happen to come together in these languages. If this is on the right track, then these so-​called active languages are not directly relevant to comparing the ICT and the DCT as theories of ergativity.

5.5.4 Languages with Concealed Transitives This leaves only Basque and Laz to consider. Basque also proves problematic for the ICT (though it is in some ways also challenging for the DCT). In Basque, ergative case is indeed found on the subjects of some intransitive verbs as well as on the subjects of transitive verbs; an example is (20). (20) Euskara-​k noiz arte Basque-​ERG when until ‘How long will Basque last?’

iraungo last

du? AUX.3SG>3SG (de Rijk 2007: 265; our gloss)

 

On inherent and dependent theories of ergative case    131 However, the generative literature on this language has shown clearly that the case of the subject does not match up perfectly with whether the verb is unaccusative or unergative, as the ICT would hope. The verb ‘last’ in (20), for example, is a likely unaccusative verb; ‘boil’ is another that takes an ergative subject. Basque even has raising predicates like behar ‘need’ that take ergative subjects, as argued at length by Rezac et al. (2014) (but see Laka, Chapter 7, this volume for an opposing view). Moreover, as Preminger (2012) notes, examples where nearly synonymous verbs—​or even the same verb across different dialects—​ take subjects with different cases seem to speak against a view that ties case too closely to theta-​roles. Instead, some kind of lexical idiosyncrasy seems called for, as in Pomo and Imonda. However, it does not seem right simply to say that verbs like ‘last’ assign quirky/​lexical ergative case to their theme arguments. The reason is because ergative case on the subject of ‘last’ varies across syntactic structures, just as ‘regular’ ergative subjects do in Basque (see (17) and (18)). For example, (21) embeds ‘last’ in a causative construction: some varieties of Basque use dative case on the argument of ‘last,’ as in (21), others use absolutive case, but none preserve ergative case on this argument. (21)

Norbaite-​k eta zerbaite-​k iraun-​arazi dio hizkuntz-​ari. Someone-​ERG and something-​ERG last-​CAUS AUX.3SG>3SG.3SD language-​DAT ‘Someone and something has caused the language to last.’ (de Rijk 2007: 380; our gloss)

Similarly, intransitive verbs with ergative subjects can be embedded under a perception verb like ‘see’ to give sentences like ‘I saw the milk boiling.’ Then ‘milk’ has absolutive case (Karlos Arregi and Ikuska Ansola-​Badiola, p.c.), the result of ECM, not ergative case. The case of the subject in examples like (20) thus behaves like structural case, not inherent case. Given this, our DCT view leads us to locate the lexical idiosyncrasy of verbs like ‘last’ elsewhere: we suggest that they are concealed transitives, taking a second argument that is approximately meaningless and phonologically null, but nevertheless counts for triggering ergative case on its coargument by the dependent case rule in (2). For the more unergative verbs in this class, especially semantically monadic predicates that are syntactically expressed as a light (or ‘compound’) verb construction (e.g., lo egin ‘sleep do’ [=‘sleep’]), this is a conventional analysis (see Bobaljik 1993a; Hale & Keyser 1993; Laka 1993b).14 This view predicts that verbs like ‘dance’ and ‘last’ should also behave like transitive verbs for other syntactic diagnostics (where available). In this connection, we take it to be significant that the causative form of ‘last’ in (21) has a causee in dative case,

14  Preminger (2012) argues that not all clauses with unexpected ergatives in Basque should be analyzed as having a null absolutive nominal. However, his most compelling point is that dummy absolutive agreement on the auxiliary is not a positive argument for this view. It is still possible that these clauses have a null argument that is active for dependent case but inert for agreement—​indeed, so inert that it does not even count as a defective intervener, as dative NPs do in Basque. See Baker (2014a) for an analysis of this sort for the small class of verbs in Shipibo mentioned in n. 5.

 

132    Mark C. Baker and Jonathan David Bobaljik not absolutive case, in standard/​conservative/​Southern Basque dialects.15 This is crucially a property of transitive verbs, not intransitive ones (see (18)), supporting the claim that some kind of null object is indeed present with verbs of this type. If this hypothesis holds true in general, then a DCT approach to Basque may be tenable, whereas an ICT approach seems not to be. A concealed transitive approach may well be a promising analysis for the Kartvelian language Laz, as well. (22) shows an active-​style contrast between unergative and unaccusative verbs in this language: (22) a. Himu-​k i-​bgar-​s. s/​he-​ERG VAL-​cry-​PRES.3SG ‘S/​he is crying.’

(Öztürk & Pöchtrager 2011: 26)

b. Him ulu-​n s/​he.NOM go-​PRS.3SG ‘S/​he is going.’ In related Georgian, ergative marking is limited to the aorist (perfect) tenses, but in Laz, it has spread to both major tense/​aspect classes (Harris 1985). This lack of an aspect split leads Woolford (in Chapter 9, this volume) to cite Laz as the prime example of the pattern predicted by the ICT. Yet even without the split, there is an important point of comparison to Georgian: Nash (in Chapter 8, this volume) analyzes the prefix i-​in Georgian unergatives parallel to (22a) as a reflexive marker, occupying the internal argument position and providing a case competitor for the subject NP. Note that the cognate prefix (glossed VAL for valency) is present in (22a) but not (22b), and Öztürk & Pöchtrager (2011: 68) suggests that this difference is (reasonably) systematic in Laz.16 If correct, this suggests that in Laz, as in Basque, it is formal–​syntactic transitivity (possibly partially concealed), not theta-​roles per se, that determines case marking, as expected under the DCT.

5.5.5 Theoretical Implications Whatever the ultimate account of some of the putatively active languages surveyed here turns out to be, we have found no clear case of a uniformly active dependent marking language in the literature. This leads us to conjecture about why active alignment 15 

In innovative dialects where the causee is absolutive, we might tentatively say that ‘last-​CAUS’ has evolved into a lexical item in its own right, not composed syntactically from ‘last’ and ‘CAUS.’ As such, its case pattern is simply that of an ordinary transitive verb. 16  In contrast, Harris (1985) argues that the correspondence between i-​prefixation and ergative case-​ marking on unergatives is weaker in Laz than in Georgian. Harris provides evidence that even (some) non i-​marked unergatives are concealed transitives, though she ultimately does not accept this as a general synchronic account.

 

On inherent and dependent theories of ergative case    133 patterns are attested in head-​marking languages but not in dependent-​marking languages. This is, of course, quite a mysterious distribution from the point of view of standard Chomskyan theory, which holds that case and agreement are two sides of the same coin; from that perspective, one expects the very same alignment patterns to show up in languages that realize case overtly and languages that realize agreement overtly. But dependent case theory is not committed to there being a deep parallelism between case and agreement. On the contrary, dependent case crucially indicates a relationship between two NPs, with possibly agreeing functional heads playing no direct role. As a result, the transitivity of the clause is crucial, but the absolute position of a single NP within the clause (the unergative–​unaccusative distinction) will typically not be crucial. Therefore, we observe ergative case marking patterns, but few or no truly active case marking patterns. In contrast, agreement is crucially a relationship between an agreeing functional head and an NP. It is perfectly plausible, then, that the location of a single NP within a clause could make a difference for this: for example, the closest c-​ commanding head with agreeing features for a theme argument could well be different from the closest c-​commanding head with agreeing features for an agent argument (it could be v as opposed to T, for example; see Baker, Chapter 31, this volume for examples from Burushaski). Therefore, the DCT can contribute to a plausible theoretical explanation of why active–​inactive systems are not uncommon in languages with agreement systems, but are extremely rare or impossible in languages with overt case marking on nominals. In contrast, a straightforward ICT might expect languages with active case patterns to be at least as common as those with pure ergative patterns, contrary to fact.

5.6. Conclusion In this chapter, we have compared two contrasting theories of ergative case in some detail: a theory in which it is an inherent case assigned by v along with an external theta-​ role, on analogy with inherent dative case in Icelandic, and a dependent case theory in which it is assigned to the higher of two NPs in the same local domain. We have found many advantages for the second sort of theory: it can explain why nonagentive verbs with two NP arguments can have ergative case on one of those NPs when conditions are right, it can explain why agentive verbs that are detransitivized in one way or another typically lose ergative case on the subject, and it can explain why uniformly ergative languages are far more common than uniformly active case (as opposed to agreement) patterns throughout the world. Along the way, we have criticized the move of supplementing the ergative-​as-​inherent-​case theory with a transitivity condition, claiming that this sneaks the dependent case idea in through the back door and renders the thematic condition superfluous. We therefore submit that the inherent-​case view of ergative fails to find support over the dependent-​case view, at least for many canonical ergative languages. Indeed, the current typological record makes us question whether the inherent-​ergative case view is even allowed as an option by universal grammar.

 

134    Mark C. Baker and Jonathan David Bobaljik

Acknowledgments For discussion of the material presented here, we thank Jessica Coon, Laura Kalin, Julie Legate, Omer Preminger, Koji Shimamura, and Susi Wurmbrand, as well as audiences and seminar participants at CUNY, MIT, UConn, McGill, and Rutgers University.

Abbreviations We use the following abbreviations in glosses: ABL, ablative; ABS, absolutive; APASS, antipassive; APPL, applicative; AUX, auxiliary; CAUS, causative; DAT, dative; ERG, ergative; FUT, future; GEN, genitive; IMPF, imperfective; INDIC, indicative; ING, gerund(-like); INSTR, instrumental; LOC, locative; O+, objective/patientive case marking (Pomo); PL, plural; PRES, present; PRF, perfective; PRT (evidential), particle; SM, singular masculine; VAL, valence; VOL, volitionality; 1SG>PL indicates an unsegmented agreement morpheme for 1SG subject and 3PL object. Source abbreviation: PV = Valenzuela (2003).

 

Chapter 6

The l o cu s of ergative  c ase Julie Anne Legate

6.1 Introduction In this chapter, I demonstrate that the factors governing the assignment of ergative case vary significantly from language to language, are multifaceted, and are low in the clause, centered around vP, in a wide range of languages. I consider two languages for which the governing factors prima facie seem high in the clause, and find that instead these factors are in fact low. It is left open whether all apparent high-​ergative languages are similarly subject to reanalysis. The patterns discussed herein demonstrate that the assignment of ergative case cannot be reduced to a single factor and should not be oversimplified. In addition, the patterns are perhaps unexpected for two types of approaches to ergative case. One such approach treats ergative as a high case, based in the CP/​TP domain (for example, Levin and Massam 1985; Bobaljik 1993a; Chomsky 1993; Bittner and Hale 1996a; Bobaljik and Branigan 2006). (Note that care must be taken with such a high-​ergative analysis to ensure that we are not simply dealing with nomenclature—​a proposal for a language whereby “ergative” is a structural case assigned by T and “absolutive” is a structural case assigned by v is a proposal that the language in question is in fact nominative-​ accusative.) The second approach for which the patterns discussed here are perhaps unexpected are dependent case approaches, whereby the assignment of ergative is dependent on the presence of another DP in the same domain not already marked with a (lexical) case (for example, Marantz 1991; Bittner and Hale 1996a, 1996b; Baker 2014a). Before beginning, I should note that ergative as a term is used descriptively by authors from many different traditions, sometimes due to the unique patterning of the transitive subject in the relevant language, sometimes due to the unified patterning of the intransitive subject and the transitive object, sometimes for other reasons. We cannot expect that everything labelled “ergative” will turn out to be instances of a single phenomenon.

 

136   Julie Anne Legate In this chapter, I leave aside the unified patterning of the intransitive subject and the transitive object; see Legate (2008) for my thoughts on this matter. I focus instead on the unique patterning of the transitive subject, and include the extension of this patterning to unergative intransitive subjects in some languages (these forming a subset of split-​S languages). There are many ergative languages for which ergative case fails to be morphologically realized on a subset of nominals; I abstract away from this morphological realization here, but see Legate (2014a). However, I exclude languages that lack morphological realization of ergative case entirely—​agreement/​clitic patterns are not simply faithful representations of case patterns, and cannot be treated as such.

6.2  Low Ergative In this section, I review a number of unrelated languages in which ergative is assigned based on properties low in the clause, centered around vP. Indeed, this situation is well attested; I only provide a few representative examples here. The examples chosen also illustrate that the factors contributing to ergative case assignment differ across languages, and are often multifaceted within a language, and hence we cannot insist upon a simplistic, uniform analysis of ergative case assignment.

6.2.1 Tsova-​Tush Tsova-​Tush (aka Batsbi) (East Caucasian:  Georgia) (Holisky 1984, 1987; Holisky and Gagua 1994)1 exhibits a case pattern whereby transitive subjects bear ergative case, while intransitive subjects bear either ergative or nominative.2 The case found on the intransitive subject is based on the θ-​role borne by the DP, a property determined low, within the vP; to wit, subjects of unaccusatives bear nominative, whereas subjects of unergatives bear ergative. Holisky (1987) separates Tsova-​Tush intransitive predicates3 into classes, based on the propensity to use nominative or ergative case; she notes, however, that the classes are fluid, depending on what situation the speaker has in mind. The first class, consisting of approximately 31 verbs, allows nominative only on S, and only an unaccusative interpretation is possible. Examples include a=reva(d)dalar ‘be confused,’ and h”abdalar

1 

Thank you to Dee Ann Holisky for discussion of Tsova-​Tush. I use nominative here as the traditional term for this morphological form of the verb; it appears on intransitive subjects and transitive objects. I leave aside whether this form corresponds to nominative from T for both the subject and object, or nominative from T for the subject and accusative from v for the object. For related discussion, see for example, Aldridge (2008a); Legate (2008a); Coon et al. (2014). 3  Including those that appear with an oblique, rather than nominative, object; see Holisky 1984: 192, n.10. 2 

 

The locus of ergative case    137 ‘be mentioned, be remembered.’ Verbs in the second class, about 27 of them, preferentially use nominative, but allow ergative with a marked volitional interpretation (=unergative). For example of dah” dax:ar ‘drown, suffocate,’ a consultant suggested: “if a distraught rejected lover throws herself into a river and drowns, she could later, hypothetically, relate her death by using ergative marking” (Holisky 1987: 110). About 61 verbs appear either with nominative or ergative subjects; the corresponding meaning difference reflects an unaccusative versus unergative interpretation. For example, ʕopdalar with a nominative subject means ‘come to be hidden,’ in the context that something moves in front of you so that you end up hidden, whereas with an ergative subject means to hide oneself. Similarly, ‘fall’ can be understood as agentively with an ergative subject, or nonagentively with a nominative subject: (1)

a. (as) vuiž-​n-​as 1SG.ERG fell-​AOR-​1SG.ERG ‘I fell down, on purpose’ b. (so) vož-​en-​sO 1SG.NOM fell-​AOR-​1SG.NOM ‘I fell down, by accident.’

(Holisky 1987: 105)

For approximately 36 verbs, the subject is usually ergative, but may be nominative under a marked non-​volitional interpretation. ga=rek’a(d)dalar ‘run very fast’ falls into this class; a consultant suggested for the nominative interpretation “a person doesn’t want or intend to run, but starting down a hill, finds himself running because it is very steep” (Holisky 1987: 112). Finally, approximately 78 verbs appear only with ergative subjects under an unergative interpretation, including lavar/​levar ‘talk,’ and lap’c’ar ‘play.’ Furthermore, there is evidence of a structural difference between intransitives with nominative subjects and intransitives with ergative subjects. The intransitive marker -​ Dalar4 when added to a transitive eliminates the ergative subject, yielding an unaccusative. When added to an intransitive that normally takes an ergative subject, it yields an intransitive that takes a nominative subject, with an unaccusative “unintentional action” interpretation (Holisky and Gagua 1994). (2) a. ču Jiš-​n-​as PVB go.to.bed-​AOR-​1SG.ERG ‘I went to bed’ b. ču Jiš-​Jal-​in-​sŏ PVB go.to.bed-​INTR-​AOR-​1SG.NOM ‘I went to bed (unconsciously, without realizing it)’ (Holisky and Gagua 1994: [81b,c]) 4 

See Holisky and Gagua for the (morpho)phonological rules that yield the surface forms of this and other verbal morphology.

 

138   Julie Anne Legate Thus, the pattern of ergative case assignment in Tsova-​Tush shows significant sensitivity to the θ-​position of the DP—​external arguments receive ergative case, while subjects of unaccusatives receive nominative. This factor regulating ergative assignment is low in the clause, within the vP. However, the pattern is not uniquely determined by θ-​position; transitivity of the predicate and person features of the DP also play a role. To wit, the above pattern is limited to first-​/​second-​person DPs. Third-​person DPs remain nominative with intransitive verbs,5 regardless of interpretation, whereas third-​person DPs do bear ergative with a transitive verb.6 (3)  a. bader dah” dapx-​dalĩ child.NOM PVB undress-​AOR ‘The child got undressed’

(Holisky 1987: 104)

b. as      dah”   japx-​ jail-​ n-​ as 1SG.ERG  PVB    undress-​INTR-​AOR-​1SG.ERG ‘I got undressed’

(Holisky 1987: 105)

c. k’nat-​ ev    bader      dah” dapx-diẽ boy-​ERG  child.NOM  PVB  undress-​TR.AOR.3 ‘The boy undressed the child’

(Holisky 1987: 104)

The transitivity of the verb is again a property determined low in the structure. The person features of the DP are first ​present low in the structure, although higher structural heads could also be sensitive to these features. In addition, lexical selection, a relationship established within the vP, also seems to play a role, in that there are predicates with a nominative-​oblique case pattern, as well as predicates with an ergative-​oblique case pattern. The ergative-​oblique pattern is rare in the language, a “minor pattern” (Holisky and Gagua 1994, section 3.2.1.6), in contrast with the nominative-​oblique pattern, which is found on a “large group” of verbs (Holisky and Gagua 1994, section 3.2.1.2). This contrast indicates that the presence of a nominative (rather than oblique) object is also a factor in ergative assignment. Compare the following.

5 

Again, including verbs with an oblique object. Note that the contrast between (3b) versus (3c) is indeed due to the difference in person, not due to the difference in the status of the subject as a pronoun versus a full DP. Like full DPs, third-​person pronouns (derived from demonstratives) do show ergative in transitives; hence in the following example the subject is the third singular ergative oqus rather than the third singular nominative o. 6 

(i) oqus Jet: xen-​ex Bexk’-​in 3SG.ERG cow.NOM tree-​CON tie-​AOR.3SG ‘He tied the cow to the tree.’ (Holisky and Gagua 1994:[75a]) Thank you to an anonymous reviewer for raising this issue.

 

The locus of ergative case    139 (4) a. mastov mak qet-​in enemy.NOM PVB attack-​AOR ‘The enemy attacked us’

vai-​n 1PL-​DAT (Holisky and Gagua 1994:[76c])

b. tupliv        h”eč’q’O son shoe.PL.ERG  pinch     1SG.DAT ‘My shoes are pinching me’

(Holisky 1984: 188)

Lexical selection can also be found in intransitives like mušebadar “work” and gamarǰbadar “win, be victorious”, which obligatorily take ergative subjects even when the subject is third person. Holisky (1984:  189)  attributes this exceptional pattern to them being borrowed from Georgian. Overall, we find that there is not a single deciding factor in the assignment of ergative case in Tsova-​Tush, but rather a variety of factors play a role. These factors include the θ-​position of the DP, the person features of the DP, the presence of an object, the case borne by the object, and lexical selection by the predicate; these factors are overwhelmingly clustered low in the clause, in the vP domain.

6.2.2 Nez Perce Nez Perce (Sahaptin: North Idaho) (Rude 1985; Woolford 1997; Deal 2010a, 2010b) illustrates a different cluster of low properties governing ergative case assignment. The basic case pattern is tripartite, with ergative -​nim, accusative7 -​ne, and zero nominative/​elsewhere. These are illustrated in the following. (5) a. hi-​pním-​se 3SBJ-​sleep-​PFV ‘The cat is sleeping.’

pícpic. cat

b. ki-​nm      picpíc-​nim  pee-​p-​ú’        cu’yéem-​ne this-​ERG  cat-​ERG     3/​3-​eat-​PROSP fish-​ACC ‘This cat will eat the fish.’

(Deal 2010a: 77)

Ergative case fails to be assigned with pseudo noun incorporated objects (see Massam 2001 on pseudo noun incorporation, and Deal 2010b for discussion of the Nez Perce instance), as illustrated in (6).

7  It may be that this is better glossed as DAT. Note that in a double object construction, the goal must bear this case and the theme cannot. It is often referred to in the literature with the more neutral “objective.”

 

140   Julie Anne Legate (6) ’ipí hi-​qn’íi-​se 3SG 3SBJ-​dig-PFV ‘He digs qeqíit roots.’

qeqíit. edible.root

(Crook 1999: 238)

The lack of ergative case in such constructions may have several potential sources. It may be that these are syntactically treated as intransitive; indeed Rude (1985) argues for such an analysis for Nez Perce. Also, the object is unmarked for case; since ergative subjects do not occur with unmarked objects, ergative could be tied to assignment of accusative case. The object also fails to trigger object agreement, thus, the presence of object agreement may be a crucial factor in ergative case assignment. Additional data demonstrate that more than simple intransitivity is at issue. Deal (2013), building on Rude (1985), discusses transitive clauses in which the object is a possessed DP. If the possessor is disjoint in reference from the subject, it bears accusative case and triggers object agreement; the subject bears ergative case. Deal (2013) analyses this as an obligatory possessor raising construction. If the possessor is bound by the subject, on the other hand, the possessor bears genitive case and fails to trigger object agreement. Like in the possessor raising construction, however, the possessed DP also does not bear accusative or trigger object agreement.8 In the absence of accusative case and object agreement, ergative case is not assigned. (7) a. Pit’iin’-​imi paa-’yax̂-na’ny-∅-a ’ip-​nek picpic girl-​ERG 3SBJ-​find-​PR-​PFV-​REM.PST 3SG-​ACC cat ‘The girli found his/​herk/​∗i cat.’ b. Pit’iin’i  hi-’yaax̂-n-a          ’ip-​nimi          picpic girl   3SBJ-​find-​PFV-​REM.PST   3SG.​GEN  cat ‘The girli found heri/​∗k cat.’ (Deal 2013: 413) Causatives provide potential evidence disambiguating whether lack of accusative or lack of object agreement is the crucial factor. In the causative of a transitive in Nez Perce, the causee does not bear ergative case. It is important to recognize that causees in ergative languages are only expected to bear ergative case if they are introduced into the structure in a vP identical to the vP that introduces agents; see for example Ippolito (2000) and Legate (2014b) for arguments that causees are rather introduced into the structure more like (high) applicative objects. Specifically to Nez Perce, provisionally assuming the causee to be introduced in the specifier of the right type of v, as sketched in (8), there is accusative case associated with this v, but no object agreement, and no ergative case.

8  This requires additional explanation; Rude (1986) and Deal (2010a) point out that a genitive possessor in the subject blocks ergative as well, yielding a clause with an unmarked subject and accusative object. Deal (2010a) notes that the possessor must be treated as closer to a higher probe than the containing DP in possessor raising constructions; she treats the lack of agreement for bound possessors (which do not raise) as an anaphor agreement effect (see Rizzi 1990a; Woolford 1999).

 

The locus of ergative case    141 (8)

vP

v'

Agent.ERG

v

vP

[uAGR] v'

Causee.ACC v

VP V

Theme.ACC

Thus, the structure exhibits one set of object agreement, associated with the v that introduces the agent and agrees with the causee, but two accusative objects: the causee and the theme. Consider the examples in (9), which exhibit both an accusative causee and an accusative theme; the ergative agents are pro-​dropped. In (9a), the causee is plural and the theme is singular. The v that introduces the agent agrees with the causee, registering third-​person plural agreement, and assigns it accusative case; the v that introduces the causee registers no agreement with the singular theme, but does assign it accusative case. In (9b), the causee is singular and the theme is plural. The v that introduces the agent agrees with the causee, registering third-​person singular agreement, and assigns it accusative case; the v that introduces the causee registers no agreement with the plural theme, but does assign it accusative case. (9) a. ’e-​nee-​sepe-​cukwe-​n-​e ha-​’ayato-​na Bessie-​ne 3OBJ-​PL.OBJ-​CAUS-​know-​ASP-​REM.PST PL-​woman-​ACC Bessie-​ACC ‘I made the women know Bessie.’ ‘I introduced Bessie to the women.’  (Deal 2010a: 380) b. marsi-​ na   ’e-​ sepee-​ twik-​ ce         geyb    Marcie-​ACC  3OBJ-​CAUS-​accompany-​IPFV.PRS  Gabe  kaa   ceeki-​ne and  Jackie-​ACC ‘I make Marcie accompany Gabe and Jackie.’ (Deal 2010a: 380) These constructions then provide potential evidence for object agreement rather than accusative case assignment as a determining factor in ergative case assignment in Nez Perce. The v that introduces the causee assigns accusative case but does not agree with the object, and hence the causee does not bear ergative case. The v that introduces the

 

142   Julie Anne Legate agent assigns accusative case and agrees with the object, hence the agent bears ergative case. On the assumption that we have been making (following Deal (2010a, 2010b) for Nez Perce, and tracing back to Chomsky (1995) more generally) that object agreement is associated with v, this property is again based low in the clause. The issue remains underdetermined, however, in that the position of causees in the structure must be clarified.9 Object agreement is not the only factor in determining ergative case assignment in Nez Perce, however; person is also relevant: ergative case is found only on third-​person DPs. Deal (2016) uses the tests developed in Legate (2014a) to demonstrate that first-​ and second-​person DPs are not assigned ergative case in Nez Perce. The Nez Perce data thus stands in contrast with the widespread pattern whereby ergative case is assigned to all DPs, but is only realized morphologically on a subset of DPs; see Legate (2014a) for details.10 The following examples illustrate the Nez Perce pattern. (10) a. ’Ip-​ním pée-​’pewi-​se 3SG.-​ERG 3/​3-​look.for-​IPFV.PRS ‘She is looking for Mary.’

Méli-​ne. Mary-​ACC

b. ’Iin        ’ipéwi-​ se     Méli-​ ne. 1SG..NOM  look.for-​IPFV.PRS  Mary-​ACC ‘I am looking for Mary.’

(Deal 2016: [(2c)])

(Deal 2016: [(2a)])

In summary, the person features of the DP, and at least one of object agreement, accusative case, and the θ-​position of the DP are primary determinants of whether ergative case is assigned in Nez Perce. Again, while the role of person is potentially ambiguous in height, the other factors are clearly low in the clause, associated with vP.

6.2.3 Warlpiri Warlpiri (Pama-​Nyungan, South-​West:  Northern Territory, Australia) is often discussed in the literature, but the complex factors governing ergative case assignment are 9   Deal (2010a, 2010b) suggests an alternative explanation for the lack of ergative on causees, proposing that an ergative DP must enter an agreement relationship with T. In that this analysis is tantamount to a quirky case analysis of ergative, and given that ergative DPs do indeed undergo A-​movement, raising to become the grammatical subject, as expected of quirky case-​marked DPs, the analysis has plausibility. The necessity of agreement with T, however, cannot be established on the basis of causees alone, given that the lack of ergative on causees is subsumed under the requirement of object agreement for ergative assignment, and given that the θ-​position of the causee may be distinct from that of more standard external arguments. 10  Tsova-​Tush, discussed in the section 6.2.1, constitutes another language in which the person feature of the DP is a factor in syntactic case assignment, rather than in the morphological realization of case. Interestingly, in that instance it is the third-​person DPs that fail to be assigned ergative rather than the first and second-​person DPs.

 

The locus of ergative case    143 often glossed over. The presence of an object is indeed relevant. This is evidenced by the lack of a class of transitive verbs that have two absolutive arguments in their basic use (see for example, Swartz 1996), and by the fact that that many intransitive verbs take absolutive subjects.11 (11)

a. Mirni      ka-​lu        yapa       wangka-​mi-​lki thereabouts  PRS.IPFV-​3PL.SBJ  person.ABS  speak-​NPST-​now ‘There are people talking over there somewhere.’ b. Parnpa-​ngka, malkarri-​rla, ritual.type-​LOC ceremonial.shield.design-​LOC miirn-​nyina-​mi yapa       panu work-​sit-​NPST person.ABS many.ABS ‘Many people work at the parnpa ceremony.’

ka-​lu PRS.IPFV-​3PL.SBJ

c. Ngawininyi    ka         wararrkurra-​ parnka. snake.species.ABS    PRS.IPFV  slither-​run.NPST ‘The snake slithers away.’ However, an absolutive object is not required for ergative case assignment. For example, Warlpiri retains ergative with a dative unaffected object (see also for example, Djaru (Pama-​Nyungan, South-​West: Northern Territory, Australia) (Tsunoda 1981a), Gurindji (Pama-​Nyungan, South-​West: Northern Territory, Australia) (McConvell 1980)). (12)

a. Ngarrka-​ngku ka marlu man-​ERG PRS.IPFV kangaroo.ABS ‘The man is shooting the kangaroo.’

luwa-​rni shoot-​NPST

b. Ngarrka-​ngku  ka-​rla-​jinta           marlu-​ku     luwa-​rni man-​ERG     PRS.IPFV-​3DAT.OBJ-​3DAT.OBJ  kangaroo-​DAT  shoot-​NPST ‘The man is shooting at the kangaroo.’ (Hale et al. 1995: 1439)

Whether an object is required is less clear. It is difficult to unambiguously distinguish unergative verbs from transitive verbs in the language, given that the language has rampant pro-​drop, given that third-​person singular absolutive object agreement is null, and given that the language has productive applicative constructions (see for example, Legate 2003). It is worth noting that Swartz (1996) does not include for the language a class of verbs that are intransitive with an ergative subject. However, we do find interesting contrasts like the following, whereby yunparni ‘sing’ takes a subject in ergative case, even when apparently being used unergatively, whereas wirntimi ‘dance’12 takes an absolutive subject in the same context. 11  Warlpiri “absolutive” is nominative on S and accusative on O, see Legate (2008). Warlpiri data from Laughren et al. (2007) unless otherwise noted. 12  This is a particular style of dancing, typical of women.

 

144   Julie Anne Legate (13)

a. Yurapiti-​rli ka yunpa-​rni rabbit-​ERG PRS.IPFV sing-​NPST ‘The rabbit sings before sleep.’

jarda-​kungarnti-​rli. sleep-​PREP.COMP-​ERG

b. Wirnti-​ja-​lpa-​lu         karnta-​patu-​ju. dance-​PST-​PST.IPFV-​3PL.SBJ  women-​PAUC.ABS-​TOP ‘The women were dancing.’

Interestingly, Laughren et al. (2007) reports that there is dialectal variation on this point, whereby an ergative subject is used with wirntimi for some speakers from Lajamanu. It is clear that the presence of an object (absolutive or dative) is not the primary determinant of ergative case assignment, in that when an object is added to intransitives with an absolutive subject the subject remains absolutive. This is true whether the object is absolutive, (14a), or dative, as in (14b), which also illustrates that the dative passes objecthood tests in triggering object agreement and use of the object control complementizer -​kurra (see Hale 1983; Simpson and Bresnan 1983). (14c) provides an additional illustration of an absolutive subject with an agreeing object, this time a high applicative dative object (see Simpson 1991 and Legate 2001 on dative high applicatives in Warlpiri). (14) a. Warlpiri    ka-​ rna      ngajulu   wangka-mi. Warlpiri.ABS    IPFV-​1SG.SBJ  1SG.ABS  speak-​NPST ‘I am speaking Warlpiri.’ b. Karnta      ka-​ rla            wangka-​ mi  ngarrka-​ ku woman.ABS  PRS.IPFV-​3DAT.OBJ  speak-​NPST  man-​DAT [jarnti-​rninja-​kurra](-​ku). trim-​INF-​OBJ.COMP-​(DAT) ‘The woman is speaking to the man trimming it.’ (Simpson and Bresnan, 1983: 54) c. Karnta     ​ka-​rla            kurdu-ku   parnka-mi. woman.ABS ​ PRS.IPFV-3DAT.OBJ child-DAT run.NPST ‘The woman is running for the sake of the child.’       (Simpson, 1991: 381)

In Legate (2012a), I provide evidence that the θ-​position of the DP also plays a role in whether ergative case is assigned. I examine Warlpiri verbs with two arguments, one of which receives dative case. For those verbs that are ergative-​dative, the ergative is an external argument; examples are provided here for jinkami ‘support, help to walk’ and warrirni ‘seek.’ For those verbs that are absolutive-​dative, in contrast, many of the absolutives are internal arguments, especially themes/​ patients; examples are provided here for wiirr-​parntarrimi ‘be a white film over’ and rdipimi ‘come upon.’

 

The locus of ergative case    145 (15)

a. Ngati-​ nyanu-​ rlu    ka-​ rla            kurdu-​ ku mother-​ANAPH-​ERG  PRS.IPFV-​3DAT.OBJ  child-​DAT nyanungu-​nyangu-​ku    warru   jinka-​mi. 3-​POSS-​DAT        around  help.to.walk-​NPST ‘The mother is propping up her child as he walks around.’ b. Wati-​ ngki ka-​ rla        kurduku      warri-​ rni. man-​ERG  PRS.IPFV-​3DAT.OBJ  child-​DAT  seek-​NPST ‘The man is looking for the child.’

(16) a. Wiirr-​parntarri-​mi    ka-​rla          kurdu-​ku      yurlkurrpa white-​over.top-​NPST  PRES.IPFV-​3DAT  child-​DAT  soap.ABS palka-​juku. body-​still ‘The soap is still plastered over the child.’ b. Pardany-​ para-​ ja        ngula-​ ji      yangka chance.meeting-​follow-​PAST  that.ABS-​TOP  like kuja-​ ka-​ rla             yapa       —​ wati    marda, DECL.COMP-​ PRS.IPFV-​ 3DAT.OBJ person.ABS man.ABS maybe karnta       marda, kurdu     marda—​ rdipi-​ mi     warna-​ ku woman.ABS maybe  child.ABS maybe encounter-​ NPST    snake-​ DAT marda, wardap​ i-ki      marda, lungkarda-​ ku      marda, maybe  goanna-​ DAT maybe  blue.tongued.lizard-​ DAT maybe ngurrpa    marda, yangka marna-​ ngka-​ ku   marda, ignorant.ABS maybe  like     grass-​ LOC-​ DAT maybe ngulya-​ngka-​ku    marda. burrow-​LOC-​DAT  maybe ‘Pardany-​paraja is like when someone—​a man or a woman or a child—​comes across a snake or a goanna or a Blue Tongue lizard perhaps without knowing it was there—​like in the grass or in a burrow.’ Experiencer subjects may have either ergative or absolutive case, depending on the predicate; examples are provided for pulka-​pinyi ‘approve of, praise,’ which takes an ergative subject, and kapatimi ‘dislike,’ which takes an absolutive. (17)

a. Pulka-​ pi-​ nyi      ngula-​ ji       yangka approval-​hit-​NPST   that.ABS-​TOP  like kuja-​ ka-​ rla-​ jinta                  ngati-​ nyanu-​ rlu DECL.COMP-​PRS.IPFV-​3DAT.OBJ-​3DAT.OBJ  mother-​ANAPH-​ERG manu  kirda-​nyanu-​rlu       kulu-​parnta-​ku,     kurdu-​nyanu-​ku, or      father-​ANAPH-​ERG  anger-​having-​DAT  child-​ANAPH-​DAT ngula    yangka kuja-​ ka-​ jana              yapa-​kari    pi-​nyi that.ABS  like      DECL.COMP-​PRS.IPFV-​3PL.OBJ  person-​other  hit-​NPST

 

146   Julie Anne Legate kulu-​parnta-​rlu anger-​having-​ERG ‘Pulka-​pinyi is when a mother or father gives approval to their child who fights, like when he fights and beats up other people’ b. Warrki-​ki    ka-​rla           kapati-​mi     yangka    yapa. work-​DAT  PRS.IPFV-​3DAT.OBJ   dislike-​NPST  that.ABS  person.ABS ‘That person doesn’t like work.’ The role of lexical selection is also apparent in that certain agents appear with absolutive case in the presence of a dative object, including the subjects of jaka-​yirrarni ‘plan, plot’ and jurrurru-​yarnkami ‘seize.’ (18)

a. Napanangka-​rla       jaka-​yirra-​rnu     Napaljarri-​ki.          Kapu Napanangka.ABS-​3DAT.OBJ  plan-​put-​PST   Napaljarri-​DAT  FUT.COMP nganta   paka-​rni   kulu-ngku perhaps  hit-​NPST  anger-​ERG ‘Napanangka threatened Napaljarri that she would hit her in anger.’ b. Kulu      ka-​ rla         karnta fight.ABS  PRS.IPFV-​3DAT.OBJ  woman.ABS jinta-​kari-​ki-​rlangu-​ku        jurrurru-​yarnka-​mi    watiya-​ku one-​other-​DAT-​for.example-​DAT  grabbing-​grab-​NPST   wood-​DAT karlangu-​ku. digging.stick-​DAT ‘In a fight a woman grabs hold of the other woman’s stick.’

In summary, we see that at least the presence of an object and the θ-​position of the subject are relevant factors in the assignment of ergative case in Warlpiri. However, neither are determinative, and there is a significant role for lexical selection. These three factors are again low in the syntactic structure, within the vP, supporting an approach whereby ergative case assignment is determined within the vP.

6.2.4 Tshangla Tshangla (Tibeto-​Burman:  Bhutan) (Andvik 1999)13 contrasts the transitive subject marked with ergative -​gi,14 with the intransitive subject and transitive object, which are morphologically unmarked (and unglossed) for case.15 13 

In the Tshangla data, note that the morpheme glossed as a copula (COP) has several uses, including the imperfective; the morpheme glossed as a nominalizer (NMLZ) also has several uses, including marking past perfective in the finite affirmative; the stem extender (SE) is added to vowel-final verb roots. 14  The ergative also appears on instrumental and reason adjuncts; I leave this syncretism aside. 15  There is additional complexity involving information structure that I don’t consider here. Ergative is obligatory for focal transitive subjects, but optional for topical, and impossible for contrastively focused.

 

The locus of ergative case    147 (19)

a. Ji-​gi shing 1SG-​ERG tree ‘I cut the tree.’

chat-​pa cut-​NMLZ

b. Jang  yi-​pha 1SG   sleep-​NMLZ ‘I slept.’

(Andvik 1999: 200)

(Andvik 1999: 180)

Andvik (1999) examines the multiple factors involved in the appearance of ergative case, stating (1999: 193) “no single one of which is sufficient on its own to motivate agentive marking”. From the above examples we see that some notion of transitivity is relevant, but the details need to be determined. To begin, assignment of ergative is not dependent on an object that bears the unmarked case: verbs that select for locative/​dative16 objects also take ergative subjects. For example, ‘to cheat,’ ‘to rebel against,’ ‘to bother,’ and ‘to scold’ fall into this class; examples follow. (20) a. Tsongpen-​gi    a-​ha           tem    a-​wa. merchant-​ERG  1PL-​LOC/​DAT  cheat  do-​NMLZ ‘The merchant cheated us.’

(Andvik 1999: 221)

b. Ro-​ki     ro-​ka    apa-​ga        ngolok      a-​na. 3-​ERG  3-​GEN  father-​LOC/​DAT  rebellion  do-​COP ‘He is rebelling against his father.’ (Andvik 1999: 222) c. Kuchi,  ji-​ gi        nan-​ ga    trok-​ pa       na. excuse  1SG-​ERG  2SG-​LOC/​DAT  bother-​NMLZ  PRT ‘Excuse me, I have bothered you.’ (Andvik 1999: 243) d. Ro-​ki ja-​ga brang-​pa. 3-​ERG 1SG-​LOC/​DAT scold-​NMLZ ‘He scolded me.’

(Andvik 1999: 243)

Nor is the assignment of ergative dependent on a DP object. Verbs of cognition, including for example sele ‘to know,’ tsile ‘to reckon, consider,’ nale ‘to comply, agree,’ and verbs of utterance, including for example yekpe ‘to speak,’ and jime ‘to ask,’ take ergative whether used with a nominal complement, propositional complement, or a null complement. Two examples follow. (21)

a. Nyi nga-​ba-​ki “jang bu-​i jang bu-​i” yek-​pa-​la. PRT fish-​PL-​ERG   1SG take-​HOR 1SG take-​HOR speak-​NMLZ-​COP ‘And the fish said, “Take me! Take me!” ’ (Andvik 1999: 197)

Case morphology affected by information structure is found in non-ergative languages; see for example the well-discussed case of Japanese. 16  These are synthetic.

 

148   Julie Anne Legate b. Ji-​gi ma-​se-​la. Jang shi-​n nyok-​nyi-​sha ji-​gi se-​le. 1SG-​ERG NEG-​know-​PRT 1SG die-​SE receive-​NF-​PRT 1SG-​ERG know-​INF ‘I don’t know. Only if I had died would I know.’ (Andvik 1999: 196) The presence of a complement is relevant, though; predicates that normally appear as intransitive with an unmarked subject take ergative subjects when used with a complement (including null, as seen for the transitive in (21)). Andvik (1999: 215) characterizes this usage as “the action of the subject referent has consequences for another referent.” In the following examples, the same complex predicate (consisting of a light verb and noun), takes an unmarked subject with the unergative interpretation ‘pray,’ but an ergative subject with the transitive interpretation ‘entreat.’ (22) a. Nyi  shepa    phi-​n   chhum-​deke,  rokte  sewu   ta-​phe PRT  preach do-​ SE finish-​ NF     3PL    prayer make-​ INF ren-​pa-​kap-​nyi    ... prepare-​PTCP-​with-​NF ‘And when the preaching was finished, and the others were about to pray . . .’ (Andvik 1999: 221) b. Songo-​ba-​ki ro-​ka sewu ta-​pha-​la. person-​PL-​ERG 3-​LOC/​DAT prayer make-N ​ MLZ-C ​ OP ‘The people entreated him.’ (Andvik 1999: 221) Another example involves the verb ‘walk,’ which takes an unmarked subject in its unergative use, but in the following takes an ergative subject on the interpretation ‘walk ahead of.’ (23) Apa-​gi gum gum dang-​nyi, jang tshin-​ga lus chho-​wa father-​ERG ahead ahead walk-​NF 1SG after-​LOC leave stay-​NMLZ ‘Father walking ahead, I was left behind.’ (Andvik 1999: 216) Similarly, an experiencer subject may appear with ergative case in the presence of a complement. Thus, in the following, a complex predicate (consisting of a light verb and a noun) appears with an unmarked subject with the meaning ‘be happy’ but an ergative subject with the meaning ‘be pleased with.’ (24) a. Ro  kap-​nyi  chhas  phi-​nyi  jang-​ta   shonang  phi-​wa. 3SG  with-​NF   talk   do-​NF   1SG-​PRT  happy   feel-​NMLZ    ‘Talking with him I feel happy.’ (Andvik 1999: 223) b. Nyi  khaila onyen  shi-​ deke, semchen thamche-​ ki ribong-​ga   namesame PRT   tiger      DEM  die-​NF   animal    all-​ERG     rabbit-​LOC  very shonang phe-​nyi... happiness feel-​NF ‘And after the tiger died, all of the animals were very pleased (with the rabbit).’ (Andvik 1999: 224)

 

The locus of ergative case    149 The presence of a complement is only one of the factors determining ergative case assignment. The θ-​position of the DP is also relevant; thus, we find ergative only on thematic subjects, not two-​argument unaccusatives. (25) a.  Ama shi-​n chhum-​deke omchhang ata-​ga unyu mother die-​SE  finish-​NF another elder.brother-​LOC/​DAT DEM pruskin natsha-​rang     nyong-​pa similar disease-​EMPH  receive-​NMLZ ‘After mother died, the older brother also got the very same disease.’ (Andvik 1999: 212) b. Nyi unyu chhesung-​gi not-​dengai-​la songo thamchen yi PRT DEM spirit-​ERG harm-​NF-​PRT person all blood phros-​nyi   shile. vomit-​NF  die-​INF ‘And if this spirit makes them sick, everyone will vomit blood and die.’ (Andvik 1999: 204) Furthermore, verbs that are normally unaccusative do appear with ergative subjects when used in a marked agentive context, again demonstrating the relevance of the θ-​position of the DP. Compare the unaccusative use of ‘vomit’ in (25b), with the use with an ergative external argument in (26). (26) Shi-​wa songo-​gi bra songo-​ga not-​nyila, unyu die-​NOM  person-​ERG  other  person-​DAT/​LOC  make.sick-​COND  this shinang. ...   Shinang tshebang-​ gi  phros-​ pe. shinang   shinang   some-​ERG vomit-​NOM ‘If a dead person causes another person to become sick, this is “shinang”. ... Some shinangs vomit.’ (Andvik 1999: 204) Another example follows, in which the owner of a house is expecting a thief that night and thus is forcing himself to stay awake. (27) ro-​ki onya binang-​ga ma-​yi-​pha 3-​ERG  DEM  night-​LOC  NEG-​sleep-​PTCP ‘on that night he will not sleep.’

(Andvik 1999: 205)

The following use of ergative on the subject of a normally unaccusative predicate Andvik (1999: 205) treats as a separate phenomenon, as “contrary to expectations of what is normal.” In that violating expectations may require a volitional act, and in that the point of this utterance seems to be that the DP has agentive control over the situation, it is likely that this type of example may be assimilated to the two previous. (28) Apa nan  shuk chaka,  nan-​gi mar-​be hang-​rang mancha. father  2SG power  COP 2SG-​ERG  be.sick.INF  what-​EMPH  NEG.COP ‘Father, you are powerful; you are never going to get sick.’ (Andvik 1999: 205)

 

150   Julie Anne Legate Aspect is also relevant, and operates in the crosslinguistically expected direction: in the perfective, ergative on transitive subjects is obligatory, whereas in the imperfective, the ergative is optional. (29) a.  Ji-​gi/​*Jang shing  chat-​pa. 1SG-​ERG/​*1SG  tree cut-​NMLZ ‘I cut the tree.’  b. Ji-​gi/​Jang shing  chat-​cha. 1SG-​ERG/​1SG  tree cut-​COP ‘I am cutting the tree.’

(Andvik 1999: 200)

(Andvik 1999: 200)

Here I note simply that aspect falls under our generalization that ergative assignment is affected by properties low in the clause, in that Aspect is in a selectional relationship with vP. However, see for example Laka (2006a), Mateo-Toledo (2008), Mateo Pedro (2009), Coon (2010a) for approaches whereby the imperfective functions as an intransitive matrix predicate; such approaches are compatible with the current discussion. In sum, ergative case assignment in Tshangla is determined by a cluster of factors that are low in the clausal structure, including at least presence of a complement, the θ-​ position of the DP, and aspectual marking.

6.2.5 Hindi In this section we consider another well-​ studied language, Hindi/​ Urdu (Indo-​ Aryan: Pakistan, India) (Mahajan 1989; Mohanan 1994a; Butt and King 2004; Davison 2004a; among many others). The basic pattern in the perfective aspect is for the transitive subject to be marked ergative and the intransitive subject and transitive object to be unmarked.17 In other aspects, all of these core arguments are unmarked. (30) a.  gaaḍii muḍii. vehicle  turn.PFV ‘The vehicle turned.’

(Mohanan 1994a: 34) h

b.  raam-​ne darvaazaa  k olaa Ram-​ERG  door open.PFV ‘Ram opened the door.’

(Mohanan 1994a: 8)

17  The unmarked case is unglossed; see Legate (2008) for arguments that it corresponds to nominative on the intransitive subject and accusative on the transitive object.

 

The locus of ergative case    151 c.  raam darvaazaa kholegaa Ram door open.FUT ‘Ram will open the door.’

(Mohanan 1994a: 8)

The assignment of ergative is not dependent on the object bearing the unmarked case; dative objects are equally compatible with ergative subjects. (31)

raam-​ne ravii-​ko piiṭaa Ram-​ERG  Ravi-​DAT  beat.PFV ‘Ram beat Ravi.’ 

(Mohanan 1994a: 70)

Note that despite some inconsistency in the literature, ko on the object in Hindi/​Urdu is indeed dative rather than accusative. Synchronically, ko is used in contexts that are unambiguously dative, including experiencer subjects and the indirect object in a double object construction. These uses are also historically prior, appearing in Old Urdu in 1200AD for the indirect object in a double object construction and the object of ‘seek’ (Butt and Ahmed 2011)  (‘seek’ commonly appears with a dative object crosslinguistically, see for example the use in Warlpiri (section 6.2.3)), as well as for the goal of directed motion verbs (Butt, Ahmed, and Poudel 2008). Regarding the use of ko on objects in differential object marking (based on specificity/​animacy), such marking is commonly dative crosslinguistically. See also Legate (2008) for arguments that Hindi/​Urdu accusative is unmarked, morphologically syncretic with the nominative, and Bubenik (1998) for evidence that Old Indo-​Aryan nominative and accusative cases became syncretic in Middle Indo-​Aryan. While presence of a complement is relevant for ergative assignment, it is not determinative. A  few intransitive verbs allow ergative subjects, with a corresponding interpretive difference, including ‘cough,’ ‘sneeze,’ ‘smile,’ ‘spit,’ ‘cry,’ ‘laugh,’ ‘sleep.’ (See for example, Tuite et al. 1985, Hook et al. 1987, Mohanan 1994a, Davison 1999 for discussion.) (32) a.  (raam-​ko acaanak šer dikhaa.) (Ram-​DAT  suddenly  lion  appear.PFV)  (Ram suddenly saw a lion.) ‘He screamed.’ b. us-​ne /​*vah  jaan buujkhkar  cillaayaa. he-​ERG  he deliberately shout.PFV ‘He shouted deliberately.’

vah  /​*us-​ne cillaayaa. he he-​ERG  scream.PFV

(Mohanan 1994a: 72)

In addition, there are lexical effects, whereby particular lexical verbs unexpectedly appear without ergative, most notably ‘bring’ (Mahajan 1989), but also ‘speak’ and optionally ‘understand,’ among others.

 

152   Julie Anne Legate (33) raam(*-​ne)  šiišaa laayaa Ram-​ERG mirror  bring.PFV ‘Ram brought the mirror.’ 

(Mohanan 1994a: 72)

Again, this cluster of factors is low in the clause, within vP. The relationship between ergative case assignment and the vP is strongly supported by Hindi light verb constructions (see for example Butt 1995; Mahajan 2012). In such constructions, the presence/​absence of ergative case on the external argument is determined by the light verb. For example, when ‘bring’ as a lexical verb combines with the light verb ‘give,’ its subject does receive ergative case. (34) Kabir-​ne vo kitaab  laa dii Kabir-​ERG  that  book bring  give.PFV.F ‘Kabir brought that book.’ 

(Mahajan 2012: 208)

Conversely, when a lexical verb that normally takes an ergative subject combines with ‘bring’ as a light verb, its subject does not receive ergative case. The following illustrates with the lexical verb ‘cause to climb.’ (35) a.  us-​ne gaaRii  pǝhaaRii-​pǝr cǝRhaayii he-​ERG  car hill-​on climb.CAUS.PFV.F ‘He took the car up the hill.’ (=He caused the car to climb the hill) b. * us-​ne gaaRii  pǝhaaRii-​pǝr cǝRhaa laayii he-​ERG  car hill-​on climb.CAUS  bring.PFV.F ‘He took the car up the hill.’ (=He caused the car to climb the hill) c. vo gaaRii  pǝhaaRii-​pǝr cǝRhaa laayaa he  car hill-​on climb.CAUS bring.PFV.M ‘He took the car up the hill.’ (=He caused the car to climb the hill) (Mahajan 2012: 209) This is particularly clear evidence that the assignment of ergative case is determined low in the clause, within the vP. In summary, the assignment of ergative is dependent on perfective aspect, the presence of an object of the verb, the identity of the lexical verb (operating both ways, disallowing ergative with a transitive verb, and allowing ergative with an intransitive in a marked agentive interpretation), and the identity of the light verb. Again, this cluster of properties is low, centered around vP. To conclude this section, we have found substantial crosslinguistic variation in the distribution of ergative case, and multiple contributing factors within each language.

 

The locus of ergative case    153 However, in all examples the factors are low in the clause, centered around vP or the XPs in a selectional relationship with vP, including VP and AspectP.

6.3  High Ergative In this section, I discuss two languages for which the assignment of ergative has been described as dependent on properties high in the clausal structure. I demonstrate that for these two languages, at least, the descriptions should not lead us to posit a high source for ergative case. It may turn out that other more solid instances may be found of ergative dependent on a high source in the clause; if so, these would be a distinct phenomenon, and should be named differently in order to avoid confusion in the literature. “High ergative” may suffice.

6.3.1 Kurdish Past Tense Kurmanji Kurdish (Iranian:  Turkey, Iran) (Bynon 1979; Payne 1980; Haig 1998; Thackston 2006) exhibits a pattern that may be initially described as ergative dependent on past tense. Note that the ergative is marked with a generalized oblique case while the nominative is unmarked.18 In the past, the intransitive subject is unmarked, the transitive subject is oblique, i.e. ergative, and the transitive object is unmarked. (36) a.  ez ḑû-​m 1SG.DIR  go.PST-​1SG ‘I went.’ b. min nan xwar 1SG.OBL  bread  eat.PST.3SG ‘I ate the bread.’

(Haig 1998: 157)

(Haig 1998: 160)

In the present, in contrast, the subject is unmarked, while the object is oblique, in essentially an accusative pattern. (37) ez nên di-​xw-​im 1SG.DIR  bread.OBL  DUR-​eat.PRS-​1SG ‘I am eating the bread.’

(Haig 1998: 160)

18  Case marking is neutralized when the noun is modified. “Nominative” here is used as a traditional term for the unmarked case. In pronouns, the distinction between the oblique and the unmarked is suppletive. Unmarked nouns are unglossed; unmarked pronouns are glossed as direct (DIR). See, for example, Dorleijin 1996 for discussion of dialect variation in case marking, including extension of the marked ACC of the present into the past, yielding a transitive OBL-OBL pattern.

 

154   Julie Anne Legate We should not, however, conclude that ergative is assigned by past tense T, and hence dependent on a projection high in the clausal structure. The notion of “past” that is relevant to ergative case assignment here is not clausal tense associated with TP. Instead, the “past” is an allomorph of the verb stem, which evolved from an Old Iranian perfect participle (see for example Payne 1980; Haig 2008). The past versus present allomorph of the stem appears inside negation, aspect and agreement morphology, confirming that it is indeed based low in the clause. (38) a.  Goşt me ne-​ti-​xward. meat  1.PL.OBL  NEG-​DUR-​eat.PST.3SG ‘Meat, we didn’t eat.’ b. pe-​ǝt n-​ā-​le-​ǝm to-​2SG.DAT  NEG-​IPFV-​tell.PRS-​1SG.NOM  a-​ka-​ā IPFV-​do.PRS-​3SG.NOM ‘I shall not tell you what Hiwa is doing.’

(Haig 1998: 160) Hiwa chi Hiwa what (Karimi 2010: 697)

Furthermore, use of the past stem does not necessarily correspond to a clausal past tense interpretation. The present perfective combines the past stem with the perfective aspect, thus showing a dissociation between the verbal allomorph and clausal tense. Importantly, it is the verbal allomorph that determines case: ergative is indeed assigned in the present perfective, despite the clausal present tense.19 (39) a.  min nan xwar-​iye 1SG.OBL  bread eat.PST-​PFV.3SG ‘I’ve eaten the bread.’

(Haig 1998: 159)

b. Min heta niha çar kitêb çapkir-​ine. 1sg.OBL  until  now  four  books  publish.PST-​PFV.PL ‘Until now I have published four books.’ (Thackston 2006: 54) I conclude that the past allomorph of the verbal root is a factor in ergative assignment in Kurmanji Kurdish, not clausal past tense based in TP. The language thus confirms to the generalization that ergative is assigned based on properties low in the clausal structure.

6.3.2 Yukulta Irrealis Yukulta20 (Tangic: northwest Queensland, Australia) (McConvell 1981; Keen 1983) has been described as ergative dependent on the realis mood (see for example, Tsuonda 19 

20 

I have added glosses to the example from Thackston (2006). This language is also known as Gang(g)alidda.

 

The locus of ergative case    155 1981b). Lexical DPs in Yukulta have marked ergative and absolutive forms,21 pronouns show a single form for all of ergative, nominative, and accusative, while the clitic cluster has distinct agreement forms for each of ergative, nominative, and accusative. The following illustrate the basic pattern. (40) a.  ṭir-​iya -​ka-​nta snake-​ERG  -​TR-​PST.TR.R ‘The snake bit the boy.’ 

pa:tya maṇṭuwara bite.IND  boy.ABS (Keen 1983: 205)

b. ṭir-​a -​ŋka paritya waḷmat̪-​i kamar-​i snake-​ABS  -​PRS.INTR.3SG  crawl.IND  on.top-​LOC  stone-​LOC ‘The snake is crawling over the stone.’ (Keen 1983: 206) A nominative-​dative case frame is used for two-​argument predicates that are not canonically transitive, including predicates with an experiencer subject and predicates with a goal object. (41) a.  pulwitya -​ka-​ti ṭir-​inytya feel.fright.IND  -​1SG.NOM-​PRS.INTR.R  snake-​DAT ‘I’m frightened of snakes.’ (Keen 1983: 206) b. tyanitya -​ka-​ti look.IND  -​1SG.NOM-​PRS.INTR.R  ‘I’m looking for the boy.’

maṇṭuwara-​n̪ t ̪a boy-​DAT

(Keen 1983: 223)

The clitic cluster registers agreement, as well as information regarding transitivity, tense, and mood, under complex interactions. The past forms in the clitic cluster show a realis/​ irrealis distinction, but importantly, this distinction does not affect the assignment of ergative case. In the following, the first example is past realis and the second past irrealis; both have an ergative first-​person subject indicated in the clitic cluster and an absolutive object. (42) a. ṭat̪in-​ta -​ŋa-​nta warunta kuritya there-​ABS  1SG.ERG-​PST.TR.R  goanna.ABS  see.IND ‘I saw a goanna over there.’ (Keen 1983: 202) b. walira-​ŋa-​nti NEG-​1SG.ERG-​PST.TR.IRR  ‘I didn’t find your spear.’

kapa  ŋumpanta  miyaḷṭa find your.ABS spear.ABS (Keen 1983: 235)

This indicates that ergative assignment is not dependent on realis mood. 21  Ergative is syncretic with locative. Absolutive is used as a traditional term, referring to the morphological realization of nominative on the intransitive subject and accusative on the transitive object; see Legate (2008, 2014a) for discussion.

 

156   Julie Anne Legate Instead, the intransitive nominative-​dative frame is used in two contexts signaling reduced transitivity, in the sense of an unaffected/​goal object. The first context is negative non-​past clauses. In the first example of the following pair, the affirmative present appears with an ergative-​absolutive case frame; note that the object is a patient affected by the event. In the second example, the negative present appears with a nominative-​ dative case frame; note that the object cannot be affected by the non-​occurring event. The clitic cluster registers the first as a transitive verb and the second as an intransitive verb, but shows no realis versus irrealis distinction. (43) a. ṭaŋka-​ya -​ka-​ri ŋawu palat̪a man-​ERG  -​TR-​PRS.TR.R  dog.ABS  hit.IND ‘The man is hitting the dog.’ b. walira  -​ŋka ṭaŋka-​ṛa NEG -​PRS.INTR.3sg man-​ABS ‘The man isn’t hitting the dog.’

(Keen 1983: 206)

ŋawu-​n̪ t a̪ palat̪a dog-​DAT hit.IND (Keen 1983: 206)

The second context of reduced transitivity marked by use of the nominative-​dative case frame is desideratives. The use of nominative-​dative in the desiderative is optional, and “can suggest that the expectation of an action being completed or experienced is reduced due to outside factors” (Keen 1983: 239). Note that the desiderative itself is low in the clausal structure, appearing as a form of the lexical verb, below the higher inflectional information registered on the auxiliary. The following examples illustrate the desiderative nominative-​dative, in contrast with the ergative-​absolutive. Again, the distinction between the two is registered in the auxiliary as a difference in transitivity, not mood; both appear in the realis. (44) a.  kaṇata -​ka-​ti cook.DES  1SG.NOM-​PRS.INTR.R  ‘I’d like to cook some tucker.’ b. wuḷanta -​ŋa-​ri food.ABS  1SG.ERG-​PRS.TR.R  ‘I’m cooking tucker.’

wuḷan-​inytya food-​DAT

kaṇatya cook.IND

(Keen 1983: 221)

(Keen 1983: 221)

As an aside, note that this case frame is also employed when the object outranks the subject, according to a hierarchy whereby first-​person nonsingular pronouns outrank first-​person singular and second-​person pronouns, which in turn outrank third-​person pronouns and nominals. The following illustrates; note that the third-​person subject ‘mosquito’ is in the nominative/​absolutive, and the clitic cluster marks the clause as intransitive realis, and the first-​person agreement clitic is the oblique form as triggered by a dative.

 

The locus of ergative case    157 (45) kuŋul-​ta -​tu̪ -​yiŋka mosquito-​ABS  1SG.OBL-​PAST.INTR.R  ‘A mosquito bit me.’

pa:tya bite.IND

(Keen 1983: 234)

To summarize, Yukulta ergative is not assigned by realis mood. Instead, ergative is assigned to the thematic subject of transitive verbs. In selected contexts of reduced transitivity, and in contexts in which the thematic object outranks the thematic subject, the intransitive nominative-​dative case frame is used instead. This case frame is otherwise used for two-​argument predicates that are not canonically transitive, including experiencer subject predicates, and predicates with a goal object. The factors of transitivity, the θ-​role of the subject and object, and the case borne by the object, are all low in the clause. The language therefore is in fact consistent with the generalization that ergative assignment is dependent on factors low in the clause, within the vP.

6.4 Conclusion This chapter has had modest goals: to demonstrate that the assignment of ergative case is multifaceted, both within and across languages, and to demonstrate that ergative case in a range of unrelated languages is assigned based on properties low in the clause, centered around vP. We reexamined two languages, Kurmanji Kurdish and Yukulta, that have been described as exhibiting ergative dependent on factors high in the clausal structure, past tense, and realis mood respectively. We discovered that ergative case assignment in these languages in fact is not dependent on tense and mood, but rather on properties that are determined low in the vP: an allomorph of the lexical verb determined within vP, the θ-​role borne by the subject and object, the case of the object, the desiderative form of the lexical verb, and the relative ranking of the person features of the subject and object. We leave open whether true “high ergative” languages may be found, that is languages in which assignment of ergative is dependent on factors high in the clausal structure, in the TP/​CP domain. Note that only languages in which ergative can be clearly differentiated from nominative assigned by TP/​CP would the label “high ergative” be appropriate for that case; otherwise, the case would simply be nominative. In addition to Kurmanji Kurdish and Yukulta, we examined five typologically disparate languages in which the ergative is assigned based on a variety of factors centered around vP: Tsova-​Tush, Nez Perce, Warlpiri, Tshangla, and Hindi. Many of the factors identified may fall under the notion of transitivity broadly conceived, including the presence of a complement, the assignment of accusative case, the presence of object agreement, and the thematic interpretation of the subject and the object. Other

 

158   Julie Anne Legate low factors identified include the identity of the lexical predicate, the identity of the light verb, and the clausal aspect. These properties are clustered around vP, and are independent of higher projections in the TP/​CP domain. I conclude that for a wide range of “low ergative” languages, the locus of ergative case is vP.

Acknowledgements Thank you to the editors, Jessica Coon, Diane Massam, and Lisa Travis, to an anonymous reviewer, to the audience at the 2013 LSA Annual Meeting, and to all those who have discussed ergativity with me over the years.

Abbreviations Abbreviations in glosses follow the Leipzig conventions, with the following additions: ANAPH = anaphoric, AOR = aorist, ASP = aspect, CON = contact, DES = desiderative, DIR = direct, EMPH = emphatic, HOR = hortative, NF = nonfinal verbal suffix, PAUC = paucal, PR = possessor raising, PREP = preparatory, PROSP = prospective, PRT = particle, PVB = preverb, R = realis, REM = remote, SE = stem extender.

 

Chapter 7

Ergative n e e d n ot spl it :  An expl orat i on into the T otalE rg hy p oth e si s Itziar Laka

7.1 Introduction:  The TotalErg Hypothesis Ergativity has received significant attention during the last decades in generative grammar; the number of languages and phenomena under scrutiny has increased, and as a consequence our knowledge about the properties and the range of variation in the ergative class has deepened considerably. As a result of this intensive focus, there has been an increasing trend toward convergence in the hypotheses as to what constitutes the core of ergativity. This convergence involves two different but related hypotheses, the sum of which I will name the TotalErg hypothesis: (1)  The TotalErg hypothesis: (a) Ergative case is inherent. (b) Ergativity does not split. The inherent ergative hypothesis has been increasingly gaining empirical support, and has been defended for a large variety of ergative languages (Levin 1983; Mahajan 1989; Johns 1992; Oyharçabal 1992; Woolford 1997, 2001, 2006; Massam 1998, 2006; Holmer 1999; Legate 2002, 2008; Aldridge 2004, 2008a; Stepanov 2004; Anand and Nevins 2006; Laka 2006a, 2006b; Wiltschko 2008; among others, see Baker and Bobaljik (Chapter 5, this volume), Legate (Chapter 6, this volume), and Sheenan (Chapter 3,

 

160   Itziar Laka this volume) for an overview). Aldridge (2008a: 987), following Legate (2002, 2006), claims that ergative morphology is in fact this very property: “Morphological ergativity is defined by the uniform assignment of inherent case to the subject by transitive v, as proposed by Legate (2002, 2008).” One widespread implementation of this generalization that I will assume in this chapter is to say that ergative case is associated to the specifier of (a subset of) little vs (see Sheenan, Chapter 3, this volume); other mechanisms have also been proposed in the literature (see for instance the review by Polinsky and Preminger 2014). The inherent ergative hypothesis dissociates ergative case marking from structural Case licensing by Tense; ergative case is thus independent of Tense/​ finiteness, unlike nominative case. The hypothesis that ergative case is inherent predicts that internal arguments cannot bear ergative case, and therefore that raising to ergative cannot exist, a claim originally made by Marantz (1984). To my knowledge, the earliest proposal that ergative case is inherent was put forth by Levin (1983) in order to provide an account of ergativity in Basque. Working within the Government and Biding Theory, Levin (1983) observed there was no evidence for a dissociation between theta roles and case morphology in Basque, the type of dissociation commonly found in nominative languages that justified the proposal of Case Theory as independent from Theta Theory (Chomsky 1981). The implementation proposed by Levin (1983) involved case assignment at D-​structure, the level at which thematic relations were established in that model. Levin’s original proposal was not followed by subsequent generative studies on Basque in the 1980s and 1990s, probably because the assignment of case at D-​structure represented too radical a departure from the standard GB claim that case was an S-​structure licensing mechanism where Tense was crucially involved. Ortiz de Urbina (1989), for instance, argued that case in Basque was structural, and in particular that ergative was assigned to the specifier of Inflection by Tense. Many subsequent accounts have pursued the structural case hypothesis with variations (see, among many others, Laka 1993b; Fernández 1997; more recently see Rezac et al. 2014, and Berro and Etxepare, Chapter 32, this volume). The inherent ergative hypothesis for Basque has been defended by a minority of works in different forms, as an instance of lexical case by Oyharçabal (1992), who combines it with a structural licensing as well, or as inherent case related to AspP/​vP (Holmer 1999; Laka 2006b). A second converging line of research on ergativity seeks to show that split ergativity is a misnomer (Laka 2006a; Coon 2010a, 2013a, 2013b; Coon and Preminger, Chapter 10, this volume). Under this hypothesis, linguistic phenomena labeled splits, suggesting a change from an ergative pattern to a nominative one, follow naturally given the grammatical properties of the languages under study, within the boundaries of an ergative system as understood here. Ultimately, this approach should provide a parsimonious understanding of ergativity where apparent changes in case morphology are not due to pockets of lexical or grammatical exceptions, but rather, they result from unexceptional processes acting across different syntactic structures. As Berro and Etxepare discuss in detail and with great nuance in this volume (Chapter 32), and originally pointed out by Levin (1983), ergative marking in Basque is sensitive to the unaccusative/​unergative contrast. This type of ergative grammars have

 

Ergative need not split   161 been argued to instantiate a variety of split ergativity triggered by the lexical semantics of the predicates involved, and this is why they have been called Split-​S languages by Dixon (1979, 1994), for instance. The account defended in this chapter claims that Split-​S languages like Basque are in fact split-​less: only external arguments carry ergative case, and that suffices to account for the contrast without any need to resort to claiming that the case pattern splits from one mode (ergative) to another (nominative). This, in turn, entails that case morphology in grammars like Basque is determined within the vP domain, because differences within the vP domain entail differences in case marking. As Sheenan (Chapter 3, this volume) and Berro and Etxepare (Chapter 32, this volume) discuss at length, languages and language varieties make different cutting points as to what subset of arguments are treated as external, that is, as to what subset of little vs assign ergative case; in particular, the set of arguments need not be restricted to those entailing causation and can include experiencers and holders as well. Another conclusion to be drawn at the end of the chapter is that agreement-​based evidence is not necessarily valid to argue for case-​related phenomena. I will claim that case and agreement are not always morphological reflexes of one syntactic operation; in particular, if ergative is inherent and vP related, as several authors argue in this volume, it is doubtful that agreement data can bear directly on discussions on case and ergativity. This is particularly relevant in a language like Basque, where agreement morphology is only licensed in finite configurations, but case morphology is insensitive to finiteness. Therefore, I will argue, evidence based on agreement facts does not necessarily hinge on case. This chapter presents an account of behar ‘need’ and associated syntactic configurations and predicates in Basque under the TotalErg hypothesis, which I argue provides a parsimonious account of the phenomena at stake. The interest of exploring this area of Basque syntax is due to the fact that it has become the focus of the debate between structural vs. inherent approaches to Basque ergativity, given recent claims by Rezac et al. (2014) that the behavior of this predicate provides crucial evidence in favor of a structural, T-​dependent account of ergativity in this language. I will therefore consider the syntax of behar ‘need’ in Basque and argue that the changes in case assignment related to sentences headed by behar can be accounted for within the TotalErg hypothesis, without resort to a structurally assigned ergative case. In fact, I will argue that behar ‘need’ and similarly behaving predicates in Basque offer more empirical evidence supporting (a)  and (b), thus converging with several other cross-​linguistic accounts of ergativity. The chapter is structured as follows: in section 7.2, I briefly introduce the central phenomenon, seemingly involving a change in the case assigned to the subject of clauses when behar ‘need’ is introduced, and lay out the basics of the explanation of this contrast to be defended. In section 7.3, I  discuss Etxepare and Uribe-​Etxebarria (2012) and Harves and Kayne (2012), the two accounts upon which I base the derivation of the apparent split provoked by need. Although the accounts differ in certain details, the common ground they share suffices for the derivation I propose. In section 7.4, I discuss the categorical status of behar, and I also discuss the proposal in Rezac et al. (2014) that

 

162   Itziar Laka the presence of behar does indeed provoke a split and change the case assigning pattern of the clause. Finally, in section 7.5 I discuss the role that agreement-​based evidence has on discussions of morphological case; I argue that agreement and case are dissociated though related phenomena, and hence agreement is not direct evidence for case. Finally, in section 7.6 I sum up the proposal and its consequences for the TotalErg hypothesis.

7.2 Ergative Need in a Nutshell The central grammatical phenomenon under discussion in this chapter is illustrated by the pair of sentences in (2):1 (2)  a.  zu Bilbo-​ra joan zara you Bilbao-​to gone 2sg.be ‘You have gone to Bilbao.’ b. zu-​k Bilbo-​ra joan behar duzu you-​erg Bilbao-​to gone need have.2sg ‘You need to go to Bilbao.’ The sentence in (2a) is headed by the unaccusative verb joan ‘go’ and the DP argument is case-​marked absolutive (zero morphology). The introduction of behar ‘need’ in (2b) appears to change the case borne by the DP, which is now marked ergative (morpheme k). I will argue that the differences in case marking in (2) are the necessary consequence of the syntactic structure of each sentence, in a way very similar to what happens in the progressive, as discussed in Laka (2006a): while (2a) is a monoclausal structure, (2b) is biclausal, and the main clause predicate introduces its own external argument. To this end, I will follow the accounts in Etxepare and Uribe-​Etxebarria (2012) and Harves and Kayne (2012), who agree in claiming that the noun behar ‘need’ in sentence (2b) introduces an external argument of its own. If this is so, and if ergative case is inherently assigned to external arguments in Basque (Levin 1983; Oyharçabal 1992; Holmer 1999; Laka 2006b), then the external argument introduced by behar necessarily carries ergative case, and the contrast between (2a) and (2b) follows.2

1  The contrast illustrated in (2) and discussed throughout the chapter does not hold for behar in western varieties, although it does for other predicates, like nahi ‘want.’ Thus, in western Basque, both (2a) and (2b) have absolutive subjects and intransitive auxiliaries. Western Basque appears to have grammaticalized behar as an element in the functional structure of the clause, in a pattern reminiscent of the grammaticalized ari progressive discussed in Laka (2006a, section 4.1) that lacked the case-​changing phenomena generally induced by the progressive. 2  For the purposes of exposition, I focus on behar ‘need’ but it must be noted that nahi ‘want’ and a cohort of other predicates have the same nature and behavior. See Berro and Etxepare (this volume, section 32.5.1), for a fuller list and a description of the semantic classes they belong to.

 

Ergative need not split   163 If (2a) is monoclausal but (2b) is biclausal, then the absolutive DP in (2a) is the theme argument of joan ‘go,’ but the ergative DP in (2b) is the external argument introduced by behar ‘need.’ The key structural difference behind (2a, b) is schematized in (3a, b) respectively: (3) a. [theme go] b. [ext.arg.i need [themei go]] Given the structural differences in (2)–​(3), it must be concluded that there is no split ergativity at play, no dissociation between semantic class and case morphology, and no change in the case assignment pattern from an ergative to a nominative one or vice-​ versa: the pronoun zu ‘you’ in (2b) carries ergative marking (-​k) because it is an external argument, and it bears (zero) absolutive case in (2a) because it is an internal argument. The relation between the external and internal arguments in (2b) is one of control, as illustrated by the co-​indexing in (3b), but it is possible to have referentially autonomous arguments in each clause (see later in the discussion, in e­ xample 11c). In sum, the general pattern behind the contrast between (2a) and (2b) is the same that has been attested cross-​linguistically in so-​called TAM Splits, which have been derived by showing that the seemingly split structures involve different complementation configurations which are at the base of the morphological changes observed (Laka 2006a; Coon 2010a, 2013a, 2013b; Coon and Preminger, Chapter 10, this volume).

7.3  Two Ways to Derive Transitive Need from a Noun I will now discuss in greater detail the account in Etxepare and Uribe-​Etxebarria (2012) and compare it to that of Harves and Kayne (2012), showing that both accounts provide the necessary elements to derive the contrast in (2) within the tenets of TotalErg. Etxepare and Uribe-​Etxebarria (2012) put forth an account of behar and associated syntactic configurations under the hypothesis that there is only one lexical entry for behar: it is always a noun. Since I will follow the basic structure provided by their account and focus on its consequences for the distribution and nature of morphological case in Basque and ergative morphology at large, I will lay it out and compare it to the cross-​linguistic account of the syntax of transitive need defended by Harves and Kayne (2012),3 which also shares the initial hypothesis that transitive need originates as a noun. As we will see, the main difference between these two approaches involves the behavior of this noun during the derivation: whereas in Harves and Kayne (2012) the noun is an argument and incorporates into the verb have, in Etxepare and Uribe-​Etxebarria (2012) it is a predicate and it does not incorporate. This difference, in turn, does not directly 3 

Earlier versions of these proposals are found in Etxepare & Uribe-​Etxebarria (2010) and in Harves (2008).

 

164   Itziar Laka bear on the discussion in this chapter, as we will see, because both accounts correctly predict that constructions with behar will necessarily yield ergative case-​marked external arguments given the TotalErg hypothesis. Etxepare and Uribe-​Etxebarria (2012) specifically discuss the syntactic structure associated with behar ‘need’ in sentences like (2a, b); they argue that behar is a noun acting as a predicate in a small clause [SC [DP/​vP] behar] where it can combine either with a DP constituent (see ­example 11a for an illustration) or a vP constituent (see ­example 11b for an illustration). In this analysis, the ‘needer,’ which the authors conceive as an experiencer, and Berro and Etxepare (Chapter 32, this volume) categorize as a holder, is an external argument introduced in the specifier of an applicative head P that combines with the small clause, [PP experiencer P [SC]] and introduces the experiencer of the need, that is, the external argument, in its specifier position. (4)  vP v

PP DPEXP

P' SC

DP/vP

be P

N

[HAVE be + appl]

appl

behar This PP, in turn, combines with the copula yielding [vP be [PP]]. The incorporation of the applicative P onto be yields the verb have (following Freeze 1992; Kayne 1993). Hence, (2a, b) are transitive due to the incorporation of P, the head that introduces the experiencer argument in its specifier (see also the structural descriptions in 5c, d). The account in Etxepare and Uribe-​Etxebarria (2012) shares several basic features with the one in Harves and Kayne (2012), who put forth a cross-​linguistic account of transitive need and relate it to auxiliary selection: languages with transitive need are necessarily languages with transitive have. According to Harves and Kayne (2012), transitive need emerges from the incorporation of the noun need onto the verb have: [VP have+needi [NP ei [DP]]. As the authors note, this derivation is compatible with the decompositional approach to have resulting from incorporation of P onto be (Freeze 1992 and Kayne 1993). Harves and Kayne (2012) explicitly include Basque within the group of languages where noun incorporation onto have yields transitive need, though they do not enter into the specifics of its syntax. Both accounts take as their starting assumption that behar ‘need’ is a noun; the difference between the two proposals lies in the syntactic configuration behar enters into, which in turn bears on whether this element incorporates into have or not. In Harves and Kayne (2012), transitive need results from noun incorporation (Baker 1988), and the constituent where need and its internal argument combine is a NP (see

 

Ergative need not split   165 the structural description in 5a). The incorporated noun does not require case, and the accusative case available for assignment from have can license the complement DP of the incorporated noun (5b). (5) a.  [VP have [NP need [DP]] b. [VP have+needi [NP ei [DP]] c. be [PP experiencer P [SC[DP/​vP] behar]] d. [[havebe+Pi] [PP experiencer ei [SC[DP/​vP] behar]] In Etxepare and Uribe-​Etxebarria (2012), behar is also a noun, but it acts as a predicate in a small clause. Therefore, it is not an argument and it does not require case (4), (5c), which in turn makes noun incorporation unnecessary (5d). The common ground shared by Etxepare and Uribe-​Etxebarria (2012) and Harves and Kayne (2012) suffices to account for the transitive behavior of behar and its ergative marked external argument within the boundaries of TotalErg and without appeal to any exceptional or language-​particular grammatical mechanism. Both accounts converge in arguing that the predicate structure of behar/​need necessarily involves the introduction of an external argument independently of the structure of the complement (in the case of Harves and Kayne 2012) or sister constituent (in the case of Etxepare and Uribe-​Etxebarria 2012)  of the argument (in the case of Harves and Kayne 2012) or predicate (in the case of Etxepare and Uribe-​Etxebarria 2012) of behar/​ need. We compare the structural descriptions of a sentence like (2b) as argued for by each account: (6) a. Etxepare and Uribe-Etxebarria (2012)

b. Harves and Kayne (2012)

vP PP P'

DP zuk

SC

vP [PRO Bilbora joan]

vP v

DP

HAVE be + appl

zuk

P N

tappl

v' v

NP

vP [PRO Bilbora joan]

N

HAVE

behar

behar

As can be seen in the structural representations in (6), both accounts provide a specifier position within the vP domain where the external argument of behar/​need is introduced. Given the inherent ergative hypothesis, that configuration requires the assignment of ergative case. Therefore, the syntax of behar as conceived by these accounts is compatible with the hypothesis that ergative case is inherently assigned to external arguments at the vP level, in accordance with the TotalErg hypothesis.

 

166   Itziar Laka

7.4  Is Basque Behar a Noun or a Raising Modal? I will now discuss the categorical status of behar, which has been subject to discussion in the recent literature. Two main hypotheses have been defended regarding the syntactic status of this element: (a) it is a lexical category, either a noun or a verb; (b) it is a functional category, a raising modal belonging in the inflectional structure of the sentence. Each hypothesis signals a different departure point and yields very different accounts; this is why it is important to consider what independent evidence can be found in support of one or the other. I will argue there is substantial independent evidence for a noun behar in Basque (as hypothesized by Etxepare and Uribe-​Etxebarria 2012), and that the evidence in favor of the view that there is also a raising modal behar (as hypothesized by Rezac et al. 2014) is weaker. The more detailed traditional descriptions of Basque consider behar to be a lexical element, either a noun or a verb. De Rijk (2007: section 14.5) and Michelena (1990) extensively discuss the similarities and differences between behar the noun and behar the transitive verb, and the different types of constructions they can enter into. The generative literature on behar, for the most part, agrees with traditional descriptions of the language, and takes it to be a lexical category, a noun in some constructions and a verb in others, akin to English need. Proposals within generative grammar, like Goenaga (1985), and Ormazabal (1991) provide restructuring accounts of biclausal behar structures like (2b), and start from the assumption that in those structures behar is a verb. Goenaga (2006) discusses the categorical status of behar and concludes that it is underdetermined between a noun and an adjective. Ortiz de Urbina (2003: 300) refers to behar as a “semi auxiliary” verb, because “in some of their usages, seem to be ‘transparent’ with respect to their subordinate clausal complements, so that clause union effects may be perceived.” Haddican (2005) argues that behar is a ‘quasi-​functional’ verb in the sense of Cardinaletti and Shlonsky (2004): it is a verb because it drives the selection of the auxiliary, but it is like a functional element because it is transparent to agreement with embedded arguments (clitic climbing). When considering the full array of constructions behar can enter into, some accounts have to assume that there is more than one behar in the mental lexicon of Basque speakers, each belonging to a different syntactic category. I will first consider the case of behar as a noun. There is agreement in the literature regarding the existence of a noun behar ‘need’ in the Basque lexicon (Michelena, 1990). As a common noun, behar can head argument DPs and behave as a nominal argument, as shown in (7):4

4 

The change from a simple r in behar to a double rr in beharra is the orthographic reflex of the fact that in word final position the trill (grapheme rr) is neutralized (grapheme r).

 

Ergative need not split   167 (7) a. zu-​k [DP[NP[PP liburu bat-​en] beharr]-a] duzu you-​erg book one-​gen need-​det have.2sg ‘You have the need of one/​a book.’ b. zu-​ki [DP[NP[PP PROi liburu bat irakur-​tze-​ko] beharr]-a] duzu you-​ergi PROi book one reed-​nom-​gen need-​det have.2sg ‘You have the need to read one/​a book.’ The examples (7a) and (7b) only differ in the type of complement the noun behar ‘need’ takes. In (7a), the complement of behar is a genitive PP containing the DP liburu bat ‘one/​a book’; in (7b) the complement is also a genitive PP containing a nominalized nonfinite clause liburu bat irakurtzeko ‘to read one/​a book.’ As shown in (8), these DP arguments headed by the noun behar ‘need’ behave like any other DP argument; they carry a determiner (-​a in the examples), and they bear ergative (8a, b), or dative case (8c, d) when required: (8) a. [DP[NP[PP   liburu bat-​en] beharr]-a]-k ez du gure ikerketa geldituko       book one-​gen need-​det-​erg not have.3SG our research.Det stop.asp ‘The need of one/​a book will not stop our research.’ b. [DP[NP[PP PROarb liburua irakur-​tze-​ko] beharr]-a]-k    ez du book.det read-​nom-​gen need-​det-​erg   not have.3SG gure ikerketa geldituko our research.det stop.asp ‘The need to read one/​a book will not stop our research.’ c. egoera hau [DP[NP[PP liburu bat-​en] beharr]-a]-ri zor zaio situation this book one-​gen need-​det-​dat owe be.3sg.3sg ‘This situation is due [Lit: owes] to the need of a book.’ d. egoera hau [DP[NP[PP PROarb liburu bat irakur-​tze-​ko] beharr]-a]-ri zor zaio situation this        book  one read-​nom-​gen need-​det-​dat owe be.3sg.3sg ‘This situation is due [Lit: owes] to the need to read a book.’

Michelena (1990), Goenaga (2006), De Rijk (2007), and Etxepare and Uribe-​Etxebarria (2012) among others, offer a wide variety of examples and evidence supporting the existence of a noun behar, including instances like (9a), where behar combines with a locative postposition, (9b) where it is modified by an adjective, or (9c) where it bears the partitive determiner, as shown in these examples: (9) a. ni   [PP  zu-​re   laguntza-​ren beharr-​ean] nago I.abs    you-​gen  help-​gen need-​in be/​loc.1sg ‘I am in need of your help.’ b. ni-​k  [DP  zu-​re   laguntza-​ren behar gorria] dut I-​erg       you-​ gen  help-​gen need red have.1sg ‘Lit: I have a red need of your help.’ ‘I am in dire need of your help.’

 

168   Itziar Laka c. ni-​k ez dut [DP  zu-​re   laguntza-​ren  beharr-​ik] I-​erg not have.1sg   you-​ gen   help-​ gen   need-​ part ‘I do not have any need of your help.’ I therefore take it as an uncontroversial fact that there is a noun behar ‘need’ in the Basque lexicon. This noun can take either DPs or non-​finite clauses as complements, as shown in (7), (8), and (9). A full account of behar that postulates only one lexical entry is, all other things equal, preferable to an alternative account postulating various homophonous lexical entries. This is what Etxepare and Uribe-​Etxebarria (2012) set out to do: they provide a full account of behar and associated syntactic structures by postulating a single lexical entry beharN. In the context of advocating for a structural assignment of ergative case in Basque, Rezac et al. (2014) discuss cases like (2b), which they refer to as the ‘INF+behar construction.’ Their hypothesis is that, in this particular construction, behar is a raising modal akin to English must. They follow the account in Wurmbrand (1999) who argued that modal verbs like English must are raising predicates involving syntactic movement, and not control structures. Hence, according to Rezac et al. (2014), the contrast in (2) involves the same thematic array of arguments: the unaccusative verb joan ‘go’ is the only thematic predicate in both (2a) and (2b), and the argument zu ‘you’ is a theme in both sentences, as shown in the derivations (10a) and (10b) respectively (adapted from Rezac et al. 2014: 1291, ­example 22): (10)  a. [TP [DP zui] … [vP ti [PPBilbora] [vjoan] … zara] b. [TP [DP zuki] … [[vP ti [PPBilbora] [vjoan] …] behar] duzu] The presence of the modal element behar in (10b) triggers a change in the case assigned to the raised argument, which is now assigned ergative case by Tense. According to Rezac et al. (2014), this account of the contrast in (2) involving raising to ergative provides crucial evidence that ergative case is structural (Tense-​dependent) and not inherent (vP-​dependent) in Basque. I will argue that this raising account requires a more complex lexicon with different homophonous entries and the deployment of language-​particular or exceptional mechanisms that burden our theory of grammar, and therefore, that the proposal in Etxepare and Uribe-​Etxebarria (2012), combined with the TotalErg hypothesis, is both empirically and theoretically preferable. Rezac et  al. (2014) do not discuss the lexical (nominal or verbal) behar; unlike Etxepare and Uribe-​Etxebarria (2012), Rezac et al. (2014) consider only cases like (2b), restricting their account to that type of construction, which they distinctly label as INF+behar. However, they do not discuss other syntactic configurations behar enters into, like the ones shown and discussed in (7), (8), (9), or how their account of configurations like (2b) relate to them. Consider the following examples involving the predicate behar in (11):

 

Ergative need not split   169 (11)  a.  Zu-​k liburu-​a behar duzu you-​erg book-​det need have.2sg ‘You need the book.’ b. Zu-​k liburu-​a irakurri behar duzu You-​erg book-​det read need have.2sg ‘You need to read the book.’ c. Ni-​k zu-​k liburu-​a irakur-​tzea behar dut I-​erg you-​erg book-​det read-​inf need have.1sg ‘I need you to read the book.’ The proposal in Etxepare and Uribe-​Etxebarria (2012) with behar as a noun, offers a unified explanation for the sentences in (11), which differ only in the type of constituent the predicate noun behar combines with in the small clause it heads: in (11a) it combines with a DP, in (11b) it combines with a vP, and in (11c) it combines with an AspP, as schematically shown in (12): (12) a. be [PP DP P [SC [DP] behar]] b. be [PP DP P [SC [vP] behar]] c. be [PP DP P [SC [AspP] behar]] The raising analysis of INF+behar (the construction it sets out to account for is illustrated in ­example 11b) begs the question of how it relates to cases where behar acts as a transitive predicate (11a, 12a), or cases where behar takes non-​control infinitival complements whose external argument is not co-referential with the one in the finite main clause (11c, 12c). A raising account of behar for (11b) is also not straightforwardly compatible with the possibility of having non-​finite clauses headed by behar, which is possible in Basque, as shown in (13): (13) normala da [ni-​k liburu-​a berriz irakurri behar iza-​te]-a] normal is [I-​erg book-​det again read need be-​nom]-det] ‘It is normal for me to need to/​have to read the book again.’ (cf. English *it is normal for me to must read the book again) An account of (11b) in terms of a raising modal structure entails that (11a, c) and (13) involve at least one different lexical entry for behar which is homophonous with the raising modal deployed in INF+behar constructions as analyzed by Rezac et al. (2014). However, a nominal account of behar like the one put forth by Etxepare and Uribe-​Etxebarria (2012) or Harves and Kayne (2012) does not need to postulate multiple homophonous lexical entries for behar; it can provide a full account of its syntax taking it to be a single lexical element of category N that enters into various syntactic combinations yielding the large array of constructions this nominal element can enter into, either as a predicate or as an argument.

 

170   Itziar Laka Rezac et  al. (2014) argue that the behar+INF construction provides evidence that ergative is not inherently assigned in Basque: “Like English INF + must, Basque INF + behar proves to be a raising construction, but raising confers ergativity on the raisee even if it would otherwise be absolutive. Thus, ergativity is again dissociated from thematic relations” (Rezac et al. 2014: 1289). But if behar in (2b, 11b) is a raising modal with no external thematic role to assign, the assignment of ergative case to the subject is puzzling, because, as the authors acknowledge, other raising predicates like the copular verbs izan ‘be(individual level)’ egon ‘be(stage level)’ do not assign ergative to their subjects (as shown in 14b,d); they have absolutive subjects (as shown in 14a,c) (see Zabala 2003 for an extensive study of copular constructions in Basque). That is, if ergative case is structurally assigned in Basque by Tense, then all raising configurations should yield the same case for the raised subject. However, this is not so, as shown by the following examples of copular sentences: (14) a. Zu ikaslea zara you student be.2sg ‘You are a student.’ b. *zu-​k ikaslea zara you-​erg student be.2sg c. Gu Bilbo-n gaude we Bilbao-​in are.1pl ‘We are in Bilbao.’ d. *Gu-​k Bilbo-n gaude we-​erg Bilbao-​in are.1pl ‘We are in Bilbao.’ Rezac et al. (2014) do not discuss the reason why structures with behar assign ergative case to the raised argument. This issue is of significant relevance in a discussion on ergativity, because accepting that structures with behar + INF involve raising entails accepting that Basque stands out as a grammar involving ‘raising-​to-​ergative,’ a type of grammatical process that has been widely argued not to be possible in human language (Marantz 2000; Woolford 2006; Legate 2008, 2012a; Sheenan, Chapter 3, this volume, among others), and moreover, the account bears the burden of explaining why only some cases of raising in Basque yield this exceptional effect. Given the highly exceptional status of raising-​to-​ergative in grammatical theory, it is necessary to ascertain that other explanatory trails cannot be more successfully tread. The argument put forth in this chapter is that there is no need to postulate such an exceptional mechanism as raising to ergative to understand the syntax of behar and ergativity in Basque. By extension, a non-​raising account of behar like that put forth by Etxepare and Uribe-​Etxebarria (2012) is preferable both on empirical as well as theoretical grounds, because of the following reasons: (a) it does not need to postulate more than one lexical

 

Ergative need not split   171 entry for behar, revealed to be always a noun; (b) it does not need to postulate raising to ergative; (c) it does not need to postulate language particular exceptional processes to ensure that some raising structures will yield absolutive subjects while others yield ergative subjects; (d) it does not need to postulate invisible Tense projections in non-​finite clauses like (11c, 13) where ergative case is perfectly grammatical; (e) it is compatible with the TotalErg hypothesis: inherent ergative and no splits. In contrast, the account in Rezac et al. (2014) must postulate at least two different lexical/​categorical entries for behar: if there is a functional category behar akin to English ‘must,’ then there must also be at least a lexical noun behar in order to account for cases where behar is clearly akin to the English noun ‘need.’ It must conclude that raising to ergative is a possibility for Basque grammar, even though the reason why the raised argument is assigned ergative is not motivated, and by extension it yields a less constrained theory of grammar. The raising account relies on Tense as the sole assigner of ergative case, and in principle it begs the question of how to account for cases like (13), where ergative is possible in nonfinite and agreement-less clauses. Finally, the raising account of behar/​need yields a highly exceptional and very language-​specific picture of Basque that sets it apart from converging accounts of ergativity, and burdens our theory of grammar.

7.5  Agreement Issues: Little to  Do with Case In this section, I  briefly discuss the agreement morphology patterns generated by structures with behar. I argue that agreement morphology is dissociated from case morphology in Basque, a dissociation that is to be expected if ergative case is inherent and associated to vP configurations. In order to do that, I will first lay out some basic facts of agreement in Basque (see also Berro and Etxepare in this volume for a fuller view). As is well known, Basque has pluripersonal agreement (ergative, dative and absolutive arguments must obligatorily agree with verbal inflection), and agreement morphology, unlike case morphology, surfaces only in finite clauses, as shown in (15): (15) a. ni-​k zu-​ri liburua eman dizut I-​erg you-​dat book.abs given 3sg.have.2sg.1sg ‘I have given the book to you.’ b. [ni-​k zu-​ri liburua ema-​te-​a] ezinezkoa da I-​erg you-​dat book.abs give-​nom-​det impossible is ‘It is impossible for me to give you the book.’ (Lit: *it is impossible I you the book give)

 

172   Itziar Laka Both (15a) and (15b) contain case-​marked DPs, but only the finite clause (15a) has agreement morphology referencing the case-​marked DPs. In (15b), the clause containing the arguments is not finite and there is no agreement morphology in the nominalized verbal form ematea ‘to give’; the main clause contains an inflected auxiliary that does not agree with the DPs from the embedded clause. This contrast can straightforwardly be captured if case and agreement morphology are not manifestations of the same licensing condition, as claimed in Laka (2006b). Given a dissociation between case and agreement morphology, it is not surprising that children display different acquisition stages for case and agreement morphology (Ezeizabarrena 1996), that speakers suffering agrammatism produce many agreement errors but few case-​morphology errors (Laka and Erriondo Korostola 2001), or that proficient non-​native speakers of Basque who are native speakers of Spanish generate different electrophysiological components from native speakers when processing ergative case morphology, but not when processing ergative or absolutive agreement morphology (Zawiszewski et al. 2011; Zawiszewski, Chapter 28, this volume). Regarding the agreement patterns associated with behar, it has been widely noted in the literature that control configurations like (11b) above, repeated here as (16a) behave like clitic climbing structures in Romance, because the matrix auxiliary obligatorily agrees with the object inside the embedded clause (16b): (16) a. Zu-​k liburu-​a irakurri behar du-​zu you-​erg book-​det read need have-​2sg ‘You need to read the book.’ b. Zu-​k liburu-​ak irakurri behar d-​it-​u-​zu you-​erg book-​detpl read need 3pl-​have-​2sg ‘You need to read the books.’ c. *Zu-​k liburu-​ak irakurri behar d-​u-​zu you-​erg book-​detpl read need have-​2sg (You need to read the books) These agreement patterns have often been taken as direct evidence for case assignment, for instance by Etxepare and Uribe-​Etxebarria (2012), who argue as follows:  “Recall that at the beginning of this section we have shown that the matrix auxiliary has to display agreement with all the arguments of the embedded predicate. If this is correct, it suggests that, in contrast with other non-​finite constructions, the embedded arguments cannot check their Case and agreement features within the non-​finite clause and must look at the matrix domain to do it.” I depart from the assumption that case and agreement morphology necessarily reflect the externalization of a single grammatical process (Chomsky 2000), and follow instead the hypothesis that, at least in ergative languages, they are separate grammatical mechanisms. Note that this is a necessary consequence of the hypothesis that ergative is inherent, assigned within the vP domain.

 

Ergative need not split   173 Dissociating case from agreement provides a way of accounting for the numerous mismatches between case and agreement morphology attested cross-​linguistically, many of which involve ergative languages, like Warlpiri or Chukchi, with both ergative–​ absolutive case morphology and nominative–​ accusative agreement (Dixon 1994), the Spurious Antipassive in Chukchi (Bobaljik and Branigan 2006), or Ergative Displacement in Basque (Laka 1988, 1993a), where agreement markers are dissociated from case morphology. This dissociation also provides a straightforward path to explain data from nonfinite environments where ergative case occurs freely in the absence of agreement and finiteness, like the examples provided in (11c), (13), (15b).

7.6 Discussion: No Need to Split The proposal in this chapter attempts to advance our understanding of ergativity with minimal appeal to language-​particular exceptions, and assuming that ergativity is a uniform property manifested in some human grammars, a position that I  have named the TotalErg hypothesis, bringing together two distinct but related hypotheses: (a) ergative case is inherent: its morphology signals external arguments associated to the specifier of the vP domain; (b) ergativity is a uniform grammatical phenomenon and it does not split; there are no nominative ruptures within ergative grammars. The TotalErg hypothesis refers to morphological case, and not to abstract Case, the overt NP licensing condition postulated in Government and Binding and the Minimalist Program (Chomsky 1981, 1995, 2000). As discussed in Bobaljik and Wurmbrand (2008) and Polinsky and Preminger (2014), morphological and abstract case were originally thought to be directly related, but this relation has become increasingly more tenuous in contemporary studies. I subscribe the suggestion in Bobaljik and Wurmbrand (2008) that once the dissociation of abstract Case and morphological case is in place, “this further level of abstraction leaves open the possibility that ergativity is best described as a morphological phenomenon, lying squarely outside the domain of Case Theory, and masking a (more) uniform syntax.” Bobaljik and Wurmbrand (2008) also note, quoting Chomsky (2000), that the relevance of abstract/​structural case has been significantly reduced in Minimalism, given the feature-​matching view of probe–​goal relations. “The question arises still more sharply if matching is just identity, so that Case can never be attracted; operations are not induced by Case-​checking requirements” (Chomsky 2000: 127). Within this view of what triggers grammatical operations, agreement (a φ-feature identity match) can be more adequately thought of as the externalization of finiteness-​dependent operations than case. Morphological case, in turn, need not be a reflex of finiteness-​dependent operations, but it can signal vP related configurations, as it does in the case of ergative languages, if the approach defended here is on the right track.

 

174   Itziar Laka

Acknowledgments I want to thank Richard Kayne, Martina Wiltschko, and Adam Zawiszewski for very helpful comments and suggestions, and the editors Jessica Coon, Diane Massam, and Lisa Travis for their patience and encouragement. The name TotalErg is inspired by an Italian company whose logo prompted me to think about the issues discussed in this chapter. Research funding from the European Community (EC FP7/​SSH-​2013-​1 AThEME (613465)), the Basque government (IT665-​13), and the Spanish government (FFI2012-​31360) is gratefully acknowledged.

Abbreviations 1pl, first person plural; 1sg, first person singular; 2sg, second person singular; 3sg, third person singular; asp, aspect; dat, dative; Det, determiner; erg, ergative; gen, genitive; inf, infinitive; loc, locative; nom, nominalizer.

 

Chapter 8

T he structura l s ou rc e of split erg at i v i t y and ergati v e c ase in Georg ia n Léa Nash

8.1 Introduction The heterogeneous class of ergative languages shares one distinguishing property: the transitive subject bears a special marking, the ergative case, which does not appear on the intransitive subject. There is a large, albeit not unanimous, consensus in recent linguistic theorizing concerning the inherent nature of the ergative case, assigned by the thematic licenser, the little v (see Johns et al. 2006; Legate 2008; among others; and section 8.5.2). A question arises as to why the same category cannot assign the inherent ergative in all languages. Is this due to different parametric settings of case features of v, or is the ergative case epiphenomenal and conditioned by case-​independent configurational factors at work in ergative systems? One of the reasons to opt for the second line of inquiry is that ergative languages are never fully ergative and display properties typical of nominative systems under certain structural circumstances. Languages vary in the degree of manifesting this mixed behaviour known as split-​ergativity (Dixon 1994:2): Basque shifts to nominative only in the progressive aspect, and is ergative in all main tenses (Laka 2006a), while Hindi, Samoan, and Georgian shift in a wider range of constructions—​they are nominative in imperfective aspects and ergative in perfective aspects. The existence of split ergativity leads to the reformulation of our initial questions: why can v assign an inherent case in some constructions but not in others in the same language? Why in a given configuration (e.g. perfective), is ergative available in certain languages but not in others? Why are some languages mixed ergative–​nominative and others pure nominative?

 

176   Léa Nash In the present work, I attempt to single out the determining property that conditions the shift of an ergative language to nominative (or vice versa), on the basis of a close investigation of split ergativity in Georgian. This language, nominative in imperfective tenses and ergative in non-​imperfective tenses, is particularly interesting because other than the change in case marking of main arguments, the shift does not affect its other structural properties: absence of auxiliary support, word order, number/​person agreement, relativization strategies (Polinsky, in press), scope facts (Anand & Nevins 2006) are identical in ergative and nominative environments. My investigation of aspectual properties of imperfective clauses in Georgian that show nominative case alignment and their comparison to clauses underspecified for grammatical aspect that show ergative alignment leads me to the conclusion that ergative splits are conditioned by functional complexity in this language, and, by extension, universally. Concretely, I contend that the difference between a nominative and an ergative behaviour of the same language, and by extension between nominative and ergative languages, consists in the capacity of the transitive subject to be theta-​licensed, and by consequence case-​licensed, in a position outside vP only in the nominative type. An outcome of this difference is that the transitive subject in ergative languages is licensed in vP, which is also the minimal domain containing the direct object. As both arguments of the transitive verb stay in vP, they are case-​licensed by the same c-​commanding functional head, according to the mechanism of Dependent Case assignment as originally proposed by Marantz (1991) (see also Bobaljik 2008; Baker & Vinokurova 2010). (1)

If there are two distinct NPs in the same phase such that NP1 c-​commands NP2 , then value the case feature of NP1 as ergative unless NP2 has already been marked for case.1 [from Baker & Vinokurova 2010: 595]

The reason why (1)  holds in a language, which under the present analysis should be defined as a language with ergative (or at least with non-​nominative) case alignment, is due to the functional impoverishment between T and vP. Specifically, a minimal functional spine of a nominative clause includes an aspectual category, which I will call Event borrowing the term from Ramchand (2013) (see also Ramchand & Svenonius 2014), it expresses the viewpoint aspect that ties eventualities in vP with the reference time in T and theta-​licenses the subject of Event in its specifier, à la Kratzer’s (1994) Voice. The highest of vP arguments introduced in Spec,EventP, has its case valued by T, while the other argument, the object DP in vP checks case by Event. As this functional category is absent in ergative systems, T directly c-​commands vP and case checking proceeds as in (1) (2) [TP T [EvP DPcase Ev [vP v […DPcase ]]]] (3) [TP T [vP DPerg v […DPcase ]]]

1 

nominative system ergative system

It will be shown in section 8.5.2 that the notion of NP in the formulation should be understood in a larger sense, encompassing CPs and weak implicit arguments as in Landau (2010).

 

Split ergativity and ergative case in Georgian    177 While I contend that relations between arguments and the verbal predicate are identical in both systems, nominative systems provide a different thematic licensing position for the transitive subject, Spec,EventP. But how does the agent DP which is the argument of v, the highest of primitive verbal predicates in a decomposed transitive vP, get to Spec,EventP in nominative languages? Is this due to A-​movement, and, if yes, what motivates it? Building on Kratzer’s (1994) original insight that the theta-​licensing category, Voice, obtains information about the semantic role of the external argument via a semantic mechanism of Event Identification, I claim that the syntactic implementation of this idea is possible through the mechanism of semantic control, essential in restructuring contexts as proposed by Wurmbrand (2003) (see also Harley 2013). The Event category is then to be viewed as a functional head of the same type as an auxiliary verb: it constitutes a complex predicate with the lexical predicate and licenses the argument of the latter via restructuring. The unavailability of Event in ergative systems should have other repercussions than case-​licensing, namely this absence should affect the temporal interpretation of an eventuality expressed by v, which I show to be the case in Georgian. In section 8.2, I show how nominative tenses in this language are unambiguously imperfective, while ergative tenses are aspectually deficient—​neither perfective nor imperfective. In section 8.3, I  analyse this aspectual underspecification in syntactic terms; I propose that a viewpoint aspectual category, which I label as Event, is absent in ergative tenses. Section 8.4 is devoted to justifying the presence of the category Event in nominative tenses and its effect on thematic licensing of the transitive subject via Event-​ v restructuring in the sense of Wurmbrand (2003). In section 8.5, I analyse the impact of Event’s absence on case licensing in ergative tenses. I first expose properties of ergative DPs in Georgian focusing on the fact that these arguments behave as structurally case-​ marked arguments, and then argue that in T-​vP configurations, case assignment proceeds along the mechanism of dependent case as in Marantz (1991) whereby the highest of vP arguments is marked as ergative. In section 8.6, I compare the conclusions of this study to an alternative analysis of ergative split initiated by Laka (2006a) and show how both approaches hinge on one central idea pertaining to a richer structural articulation of temporal relations, by adding a clause or a functional category, in nominative systems but not in ergative systems.

8.2  Ergative Split in Georgian 8.2.1 Nominative and Ergative Tenses In the present, future, imperfective past tenses, referred henceforth as nominative tenses, Georgian shows nominative case alignment: the transitive subject is nominative, the direct object is accusative. The case affix on direct and indirect objects is identical in nominative tenses, but their licensing properties are not: unlike direct objects (theme

 

178   Léa Nash arguments), dative arguments (goals, benefactors, affected locations, experiencers) are licensed by applicatives and show constant case behaviour in nominative and ergative tenses. Therefore, I will continue to refer to the case on direct objects as accusative. In the aorist and subjunctive tenses, henceforth ergative tenses, Georgian behaves as a typical ergative language: the transitive subject is marked with the ergative case (also known in the Kartvelian grammatical tradition as narrative (see Tuite, Chapter  45, this volume) and the direct object is nominative. As the case borne by direct objects in ergative tenses is strictly conditioned by the verb’s finiteness (see section 8.5.1.2), I consider this morphologically absolutive case to correspond to the structural nominative checked by T rather than to the abstract “accusative” assigned by V (see Legate’s 2008 distinction of absolutive as abstract V or abstract T cases). nominative tenses: (4) a. vano-​ø xaT-​av-​s /​ xaT-​av-​d-​a mankana-​s Vano-​nom draw-​ts-​3s/​ draw-​ts-​past-​3s car-​acc ‘Vano is drawing/​was drawing a car’ b. vano-​ø Ċam-​ø-​s /​ Ċam-​ø-​d-​a kada-​s Vano-​nom eat-​ts-​3s /​ eat-​ts-​past-​3s cake-​acc ‘Vano is eating/​was eating a cake’ c. vano-​ø alag-​eb-​s /​ alag-​eb-​d-​a otax-​s Vano-​nom tidy-​ts-​3s /​ tidy-​ts-​past-​3s room-​acc ‘Vano is/​was tidying the room’ ergative tenses: (5) a. vano-​m xaT-​a /​ xaT-​o-​s mankana-​ø Vano-​erg draw-​aor3s/​ draw-​sbjn-​3s car-​nom ‘Vano drew/​draw a car’ b. vano-​m Ċam-​a /​ Ċam-​o-​s kada-​ø Vano-​erg eat-​aor3s/​ eat-​sbjn-​3s cake-​nom ‘Vano ate/​eat a cake c. vano-​m alag-​a /​ alag-​o-​s otax-​i Vano-​erg tidy-​aor3s/​ tidy-​sbjn-​3s room-​nom ‘Vano tidied/​tidy the room’

8.2.2 Meaning of Nominative and Ergative Tenses In Georgian, the present and the past imperfective tenses are used in progressive, generic, and habitual contexts. These tenses are imperfective as they successfully pass any test that shows that the event they convey contains the reference time of the utterance. In (6a), only

 

Split ergativity and ergative case in Georgian    179 simultaneity with the other temporal clause is implied, and in (6b), the sentence only has durative, but not inclusive reading. (6) a. roca Kar-​i ga=v-​aγ-​e nino-​ø Telepon-​ze laParaK-​ob-​d-​a when door prev=1-​open-​aor Nino-​nom phone-​on speak-​ts-​past-​3s ‘When I opened the door, Nino was speaking on the phone’ (–​> simultaneous talking and opening) b. nino-​ø Telepon-​ze ori-​dan sam-​amde laParaK-​ob-​d-​a Nino-​nom phone-​on two-​from three-​until speak-​ts-​past-​3s ‘Nino was speaking on the phone from two to three’ (–​> continued after three) Although both the past imperfective and the aorist are used to describe situations that happened before the reference time, the past imperfective is employed to report simple general facts about the past, while the aorist is not felicitous in these contexts. So (7b) can only describe a specific event in the past, e.g. a TV talk show, where people spoke five languages. (7) a. adre xalx-​i xut ena-​ze laParaK-​ob-​da before people-​nom five language-​on speak-​ts-​past3s

past imperfective

b. #adre xalx-​ma xut ena-​ze i-​laParaK-​a aorist before, people-​erg five language-​on riam2-​speak-​aor3s ‘Before, people spoke five languages’

8.2.3 The Aorist Although it is commonly stated that Georgian switches to ergative case system in perfective tenses, a close look at the aorist reveals a more complex situation. The Georgian aorist does describe events terminated before the reference time. But although they are completed and do not continue unlike (6b) beyond the reference time, events in the aorist are interpreted as non-​culminated in the sense that the natural endpoint inherent to the meaning of the predicate is not necessarily reached upon the termination of the event. In (8a), we tend to understand that the event has ended but the car is not drawn, the dress is not sewn in (8b), the table is not painted in (8c). (8) a. vano-​m xaT-​a mankana-​ø vano-​erg draw-​aor3s car-​nom ‘Vano drew a car’ 2 

RIAM stands for the reflexive implicit argument marker and its properties will be discussed in section 8.4.3.1.

 

180   Léa Nash b. vano-​m Ker-​a es Kaba-​ø vano-​erg sew-​aor3sg this dress-​nom ‘Vano sewed a dress’ c. vano-​m γeb-​a magida-​ø vano-​erg paint-​aor3s table-​nom ‘Vano painted the table’ Non-​culmination is not entailed but strongly implied, as both continuations of (9) are equally felicitous. (9)

vano-​m Vano-​erg

c’mind-​a kotan-​i, … clean-​aor3s pot-​nom a. … magram  ver  ga=c’mind-​a but   can’t prev=clean-​aor3s b.  … da Kargad  ga=c’mind-​a and well     prev=clean-​aor3s ‘Vano cleaned the pot … but (he) couldn’t clean it/​and cleaned it well’

In (9) the pot can be understood as absolutely not cleaned as a result of the event, as partially cleaned, or completely cleaned. The only meaning entailed in (9) is that Vano started and ended some action with the pot that can be defined as cleaning but the outcome of his doings is not entailed. If the direct object is a plural DP as in (10), the sentence entails that Vano did something to each of the pots, without a further entailment that any of them got cleaned. The situation is very different in imperfective contexts with plural direct objects as in (11), where the sentence is true even if Vano was in the process of cleaning only one pot and in the end he cleans just three3 (see Singh 1998 for similar effects in Hindi neutral perfectives). (10) gušin, vano-​m c’mind-​a xut-​i kotan-​i yesterday, vano-​erg clean-​aor3s five-​nom pot-​nom ‘Yesterday Vano cleaned five pots’ (entails: 5 pots got affected by Vano’s activity) (11)

Vano-​ø c’mind-​av-​d-​a vano-​nom clean-​ts-​past-​3s ‘Vano was cleaning five pots’

xut-​ø kotan-​s five-​acc pot-​acc (does not entail: 5 pots were being cleaned)

3  The contrast between (10) and (11) undermines the idea that non-​culminated accomplishments should be analysed in a similar fashion as imperfectives in terms of Dowty (1979)’s semantic notion of inertia worlds, as has been proposed by Bar-el et al. (2005) for St’át’imcets and Koenig and Muansuwan (2001) for Thai.

 

Split ergativity and ergative case in Georgian    181 As events without a culmination point are atelic, a speaker chooses to use the aorist form in order to convey the duration of a naturally telic accomplishment, as in (12). (12) a. vano-​m Kargad da=alag-​a otax-​i, ori saati alag-​a Vano-​erg well prev=tidy-​aor3s room-​nom, two hours tidy-​aor3s ‘Vano cleaned the room well, he cleaned it for two hours’ b. vano-​m ver ga=Tex-​a kotan-​i, ori saati Ki Tex-​a Vano-​erg can’t break-​aor3s vase-​nom, two hours yet break-​aor3s ‘Vano could not break the vase, yet he broke it for two hours’ From the left parts of assertions in (12) and in (13), we see that the culmination of terminated events is expressed by a perfectivizing morpheme (preverb) prefixed to the aorist forms. Georgian preverbs, which generally have a locative meaning and are in many aspects comparable to Slavic perfectivizing preverbs are not only dedicated to the aorist, they are added to imperfective verbs in the present and the past tense to shift their meaning to the future. They are also independent of finiteness and optionally surface in nominalizations, as in Slavic. (13) a. Vano-​m da=xaT-a mankana-​ ø Vano-​ERG prev=draw-​aor3s car-​nom ‘Vano drew (and finished) the car’ b. Vano-​m še=Ker-​a Kaba-​ø Vano-​ERG prev=sew-​aor3s dress-​nom ‘Vano sewed (and finished) a dress’ c. Vano-​m ga=c’mind-​a otax-​i Vano-​erg prev=clean-​aor3s room-​nom ‘Vano cleaned (and finished) the room’ On the basis of the exposed properties of the aorist, I conclude that this tense expresses events which took place and terminated before the reference time but it is not formally specified for the perfective aspect. The situation is different from the past and the present tenses where clauses are positively specified for the imperfective aspect.

8.3  Aspectual Deficiency as Structural Impoverishment 8.3.1 Neutral Aspect and Georgian Aorist The implication of non-​culmination of terminated events, typical for the Georgian aorist, has been reported to exist in other languages (Singh 1998; Bar-el et al. 2005;

 

182   Léa Nash Tatevosov 2008) and naturally fits the description of neutral viewpoint aspect put forth in Smith (1991). According to the author, sentences in the neutral aspect, which is morphologically default, describe situations that include the initial point and at least one stage of the situation. This aspect shares some properties with the perfective, and others with the imperfective. As witnessed above and summarized in what follows, the Georgian aorist perfectly manifests all the properties ascribed to this type of viewpoint aspect (see also Iatridou et al. 2001) (a) Neutral aspect makes reference only to the beginning of the internal temporal structure of an eventuality. Similarly to the imperfective (14a), and in contrast to the perfective (14b), (14c) does not assert that a table gets painted. (14)

a. vano-​ø γeb-​av-​d-​a magida-​s Vano-​nom paint-​ts-​past-​3s table-​acc ‘Vano was painting the table’

imperfective

b. vano-​m še=γeb-​a magida-​ø preverb+aorist vano-​erg prev=paint-aor3s table-​nom ‘Vano painted the table (and finished) c. vano-​m γeb-​a magida-​ø aorist Vano-​erg paint-​aor3s table-​nom ‘Vano painted the table’ (=was engaged in table painting) (b) Predicates in neutral aspect disallow completive adverbials (e.g. ‘in two hours’), (15)

a. *nino-​m    γeb-​a        magida-​ø     or   saat-​ši  [ok with še=γeba (prev+aor)] Nino-​erg  paint-​aor3s table-​ nom  two    hour-​in ‘Nino painted the table in two hours’ b. *nino-​ø γeb-​av-​d-​a magida-​s Nino-​nom paint-​ts-​past-​3s table-​acc ‘*Nino was painting the table in two hours’

or two

saat-​ši hour-​in

(c) Neutral aspect allows both durative and inclusive interpretation of time intervals (e.g. between 3 and 4 o’clock), like the perfective, and unlike the imperfective. (16) a. nino-​m γeb-​a magida-​ø sami-​dan otx-​amde Nino-​erg painted-​aor3s table-​nom three-​from four-​to ‘Nino painted the table from 3 to 4’ (entails: Nino did not paint after 4)

 

Split ergativity and ergative case in Georgian    183 b. nino-​ø γeb-​av-​d-​a magida-​s sami-​dan otx-​amde Nino-​nom paint-​ts-​past-​3s table-​acc three-​from four-​to ‘Nino was painting the table from 3 to 4’ (does not entail: Nino did not paint after three) (d) the neutral patterns in temporal sequencing like perfective eventualities, and unlike imperfectives which only allow simultaneous readings with temporal clauses. (17) a. nino-​m γeb-​a magida-​ø roca vano-​ø c’a=vid-​a Nino-​erg paint-​aor3s table-​nom when Vano-​nom prev=go-​aor3s ‘Nino painted the table when Vano left’ (sequencing) b. nino-​ø γeb-​av-​d-​a magida-​s roca vano-​ø c’a=vid-​a Nino-​nom paint-​ts-​past-​3s table-​acc when Vano-​nom prev=go-​aor3s ‘Nino was painting the table when Vano left’ (simultaneity)

8.3.2 No Category of Aspect in the Neutral Aspect Smith (1991) questions whether neutral aspect is a third type of viewpoint aspect, along with the perfective and the imperfective. There is an ongoing debate concerning its ontological validity, and the effects ascribed to this aspect can be claimed to be deductible from other semantic properties such as telicity and cumulativity, which are independently required (Artshuler 2013). I contend that the deficient perfectivity of sentences in Georgian aorist is not due to the presence of a formal neutral viewpoint Aspect category, but is rather a result of the absence of any Aspect category in the clausal functional structure. As opposed, past and present imperfective tenses have their aspectual properties encoded by such a category. My view is grounded in the conception of grammatical viewpoint aspect as a distinct notion from the aspectual organization of vP (Aktionsart) and constituting a distinct syntactic category with a binary value (perfective–​imperfective), situated in the clause between T and vP (see Smith 1991 for distinguishing the two notions of aspect; Depreatere 1995; see also Stowell 1996 for properties of ZeitP).

8.3.2.1 The Aorist and the Subjunctive Share Structural Deficiency: Imperative Clauses Additional support for my claim that sentences in the aorist do not contain a category expressing the viewpoint aspect comes from the properties that the aorist shares with the subjunctive in Georgian. In both “tenses”, the transitive subject is marked with ergative and the verb surfaces in its impoverished form to be discussed in section 8.4.1. Most significantly, aorist and subjunctive forms of the verb are used in imperative clauses as well.

 

184   Léa Nash As Georgian does not have infinitives subjunctive forms are used in modal contexts (after can/​must/​wish) and in counterfactuals, and are marked by a special irrealis mood affix -​o-​. As in the aorist, the perfectivizing preverb is added to the verb to convey the natural telicity of the event, otherwise the meaning of the verb is neutral. Example (18a) means that I wish Vano to engage in the event of car drawing, by drawing at least some lines that eventually, in normal circumstances, can yield a picture of a car. The sentence also serves to convey the duration of the event that I wish Vano to be engaged in. (18)

a. m-​i-​nda (rom) vano-​m xaT-​o-​s mankana-​ø (ori saati) 1o-​appl-​wish that Vano-​erg draw-​sbjn-​3s car-​nom (two hour) ‘I wish that Vano draw a car (for two hours)’ b. Kargi ikneba vano-​m (rom) xaT-​o-​s mankana-​ø good befut Vano-​erg that draw-​sbjn-​3s car-​nom ‘It would be great if Vano draw a car’

In imperative clauses, the subjunctive and the aorist forms share the labor: the subjunctive form is used in negative contexts (19a) and with 1st pl, 3rd person subjects (let us, him, them), (19b). The aorist form is used with 2nd person imperatives (19c): (19) a. ar ø-​xaT-​o mankana-​ø neg 2-​draw-​sbjn car-​nom ‘Don’t draw a car’ b. (man) xaT-​o-​s mankana-​ø (he-​erg) draw-​sbjn-​3s car-​nom ‘Let him draw a car’ c. ø-​xaT-​e mankana-​ø 2-​draw-​aor car-​nom ‘Draw a car!’ Aorist and subjunctive forms are used with the imperative mood and in irrealis, non-​referential temporal contexts with no temporal anchoring in the reference time. I conclude that these forms occur in temporally deficient contexts with impoverished functional structures (see Zanuttini 1997, among others, who analyses Romance imperative clauses as functionally deficient and lacking tense).

8.3.3 Interim Conclusion: Ergative–​Nominative Asymmetry We have seen that in Georgian aspectually underspecified events are expressed by ergative tenses, and imperfective events by nominative tenses. This missing aspectual specification, typical of ergative clauses, generally serves to make a time flow transition between

 

Split ergativity and ergative case in Georgian    185 the speech time and the eventuality itself. Syntactically, this ‘bridging’ task is carried out by a special functional category Event (or Aspect, or Zeit), generated in the clausal functional spine between T (time) and vP (eventuality) (see Ramchand & Svenonius 2014). (20) nominative systems: [TPT..-​..Ev-​vP ] ergative systems: [TP T..-​Ev-​vP] The interpretative cost of structural deficiency in ergative tenses is that events are interpreted as distal, not anchored with respect to the reference time—​the event is just named without referring to the specifics of its internal temporal organization as if it were “nominalized” in some sorts. The morphological consequence of (20) is a systematic absence of thematic suffixes, glossed as TS, in ergative tenses in Georgian. In section 8.4, I show that this suffix stands for imperfective aspect and as such formally fleshes out the category Event. We saw that its absence does not shift the verb into perfective in Georgian; what Georgian lacks, unlike English, is a zero morpheme positively marking the perfective aspect of the clause. This is the reason why an assertion in the perfective aspect such as John broke the window but could not break it sounds as a contradiction in English but not in Georgian.

8.4  The Category Event in Nominative Tenses 8.4.1 Morphological Evidence: TS as Markers of Imperfectivity In nominative tenses, transitive, unergative and unaccusative verbs show up with a suffix attached between the root and tense/​agreement markers. Following Georgian grammarian tradition, I refer to them as thematic suffixes, ts, and show that below they spell out the imperfective aspect (see Tuite, Chapter 45, this volume). While unaccusative verbs all have the same ts -​eb (see section 8.4.3.1), unergative and transitive verbs appear with a small set: -​av,-​eb, -​i, -​ob, -​ø. Each ts can coarsely be tied to a semantic class: creation/​destruction/​reconfiguration transitive verbs—​manner verbs in the sense of Rappaport Hovav & Levin (2010)—​have -​av, deadjectival causatives—​ Rappaport Hovav and Levin’s result verbs—​come with -​eb. As for unergatives, verbs of sound and noise emission take -​eb, manner of motion -​av, manner of behaviour -​ob.4 Thematic suffixes also appear on corresponding nominalizations but must be followed by the nominalizer -​a. 4 

Stative verbs never take a ts, never show up in the aorist or in the future where only bounded, i.e. terminated, events appear.

 

186   Léa Nash (21) verbs (present) Root+ts+t/​agr draw sow frighten whiten

xaT-​av-​s tes-​av-​s (a)šin-​eb-​s5 (a)tetr-​eb-​s

reign (act+king) act+nervous roll swim

mep-​ob-​s nerviul-​ob-​s gor-​av-​s cur-​av-​s

Nominalization transitive drawing sowing frightening whitening unergative reigning acting-​nervous rolling swimming

Root+ts+n xat-​av-​a tes-​av-​a šin-​eb-​a tetr-​eb-​a mep-​ob-​a nerviul-​ob-​a gor-​av-​a cur-​av-​a

As thematic affixes are incompatible with non-​imperfective ergative tenses, I take them to spell out the imperfective aspect of the eventuality in clauses.

8.4.2 Event as an Aspectual Head That Triggers Restructuring The imperfective aspect is formally represented as a syntactic category Event which has a double function cross-​linguistically (see Ramchand 2013 and section 8.6.3): it selects vP and theta-​licenses its highest argument via the mechanism of semantic control, as proposed by Wurmbrand (2003) for restructuring constructions. Semantic control results from Event’s binding an open variable in v, just as in restructuring contexts the higher of the two verbs binds a thematic variable of its V complement. (22) [EvP DPi Evi [vP vi [VP…]]]

complex predicate Ev-​v

This head-to-head binding is at the heart of complex predicate composition and argument sharing. The capacity to theta-​license an argument makes Event look as an argument taking predicate but it functions as an auxiliary because, unlike lexical verbs, it cannot have its own thematic properties but only inherit those of its complement. In Georgian, this privileged relation between Event and v affects the shape of ts, their form is sensitive to—​in a sense “agrees” with—​the Aktionsart properties of the selected verb. For example, it is noteworthy that all (anti)causative result verbs appear with -​eb (see section 8.4.3.1). 5  The morpheme a-​prefixed to the root in deadjectival causatives signals that the event is causative, it shows that in a bieventive verb, the upper causing event and the lower change event are not identical. It can be qualified, as a shorthand, as a causative marker, or as an applicative.

 

Split ergativity and ergative case in Georgian    187 While subjects of transitive verbs are theta-​licensed in nominative systems by a functional category Event, as in (22), all thematic licensing proceeds at the level of vP in ergative tenses. Ergative case signals that agents are base-​generated in vP, as specifiers of their lexical predicate. In nominative systems, the agent is external to its selecting predicate as its licensing happens after restructuring.6 A similar conception of agent licensing that implies that the highest lexical verbal predicate, labelled as v-​caus, or as v, or as Cause, does not project its argument locally goes back to Hale and Keyser’s (1993, 2002) syntactic decomposition of verbs and has been repeatedly argued for by many (see Pylkkänen 1999; Schäfer 2012b; Harley 2013, among others). My analysis differs from these authors’ in not taking this property of v as a primitive but deriving it from structural requirements as in (30) which crucially hinge on the existence of the restructuring category Event.

8.4.3 Event Does Not Express Voice Distinctions Nash (1995) analyses Georgian thematic suffixes as realizations of Kratzer’s (1994) Voice, as their presence coincides with that of accusative objects, following Burzio’s generalization that the category that theta-​licenses the external argument also checks accusative. As Voice is absent in ergative tenses, case marking of main arguments shifts and the transitive subject is marked with the inherent verb case. At that period of syntactic theorizing there was much confusion concerning the distinction between Kratzer’s Voice and Chomsky’s (1995) little v, the transitivizing head absent in unaccusative VPs. Nash (1995) does not provide a clear answer whether the absent head is a transitivizer in which case Georgian clauses in ergative tenses are predicted to have a radically different VP structure than in nominative systems—​transitive VPs would be mono-​level unaccusative structures with an adjoined agent marked with the inherent ergative. In nominative tenses, with Voice projected, Georgian transitive predicates are predicted to have a more articulated substructure and clearly show voice distinctions. Drawing on the conclusions of the present analysis about aspectual properties of the aorist in Georgian, I can maintain the same insight as in Nash (1995) concerning functional impoverishment of ergative tenses and show that the absent Event category, which only expresses temporal properties of the clause, does not affect the expression of voice distinctions which involve operations lower in the structure, at the level of vP. 6  Collins’s (2005) analysis of ergative–​nominative case asymmetry in terms of movement from Spec,vP to Spec,VoiceP constitutes an alternative to the restructuring operation. I do not adopt it here as its motivation remains problematic—​if DPs are lexically marked for case, movement for case in the process of which the lexical marker is absorbed by the selecting head is unwarranted. If the ergative case-​ affix is analysed as a preposition (see Polinsky in press), such movement operation reduces to preposition incorporation, comparable in spirit to Kayne’s (1993) treatment of auxiliary selection and Mahajan’s (1994) analysis of ergativity in Hindi. As shown, Georgian ergative nominals cannot be analysed as PPs and preposition incorporation in one tense but not in the other would require additional elaboration of the link between Voice and vP.

 

188   Léa Nash

8.4.3.1 Voice Distinctions in Georgian The principal reason why ergativity cannot be tied to voice distinctions in Georgian is that verbal predicates in non-​active voice have the same shape across nominative and ergative tenses. There are three ways to express non-​active voice: (i) non-​active counterparts of agentive transitives are marked with the prefix i-​, which also serves to mark reflexive verbs; by its polysemy, i-​ resembles the Romance SE/​SI clitic (and is glossed as reflexive-​implicit argument marker RIAM; see also section 8.5.1.3), (ii) anti-causative deadjectival verbs are formed by means of the suffix -​d, the fientive primitive predicate become, (iii) some unaccusatives are unmarked, their semantic class is heterogenous containing both underived unaccusatives such as rč-​eb-​a (stay) and change of state verbs such as tb-​eb-​a (warm). (23) a. vano-​ø xe-​s γeb-​av-​s, a-tetr-​eb-​s, a-lP-​ob-​s Vano-​nom tree-​acc paint-​ts-​3s, appl-whiten-​ts-​3s, APPL-rot-​ts-​3s ‘Vano is painting/​whitening/​rotting the tree’ b. xe-​ø i-​γeb-​eb-​a, tetr-​d-​eb-​a, lP-​eb-​a tree-​nom riam-​paint-​ts-​3s, white-​become-​ts-​3s, rot-​ts-​3s ‘the tree is being painted, is whitening, is rotting c. xe-​ø i-​γeb-​a, ga=tetr-​d-​a, da=lP-​a tree-​nom riam-​paint-​aor3s, prev=white-​become-​aor3s, prev=rot-​aor3s7 ‘the tree was painted (but not finished)’, whitened, rotted’ (23b) shows that in nominative tenses non-​active morphemes coexist with ts -​eb, so they cannot spell out the same functional head. There is robust morphological evidence that non-​active morphemes attach to the verb before ts. According to Baker’s (1985) Mirror Principle, the order of morphemes reflects the structural hierarchy of corresponding syntactic categories, therefore, Event scopes over the fientive predicate as the non-​active -​d attaches to the root before ts in (23b). The same cannot be as easily asserted for riam, a prefix. Yet, there is some indirect evidence that ts scopes over riam as well. As mentioned in section 8.4.1, all non-​active verbs take ts -​eb, while transitive verbs combine with different ts. Normally, transitive agentive manner verbs take -​av but their corresponding non-​active forms still show up with -​eb, (24). If riam were affixed after the composition of V with ts, it is unclear how that would result in the change of -​av to -​eb. (24) a. vano-​ø γeb-​av-​s /​ c’mind-​av-​s kotan-​s Vano-​nom paint-​ts-​3s /​ clean-​ts-​3s pot-​acc ‘Vano is painting/​cleaning the pot’ 7  Perfective preverbs are obligatory on unaccusatives but not on non-​active forms which are derived from agentive transitives and carry the riam prefix. As unaccusatives are achievements, the expression of telicity becomes mandatory for this aspectual class.

 

Split ergativity and ergative case in Georgian    189 b. *kotan-​i i-​γeb-​av-​a /    ​i-​c’mind-​av-​a b’. kotan-​i i-​γeb-​eb-​a /    i-​c’mind-​eb-​a pot-​nom riam-​paint-​ts-​3s /​ riam-​clean-​ts-​3s ‘The pot is being painted/​cleaned’ The morphological evidence presented above suggests that voice related operations take place at the level of vP below Event, when this category is instantiated. The little v is active when it selects a referential DP argument, Agent or Causer, and it is non-​active when its argument is not realized, or non-​referential, in which case the predicate bears the riam i-​(see Embick 1998; Alexiadou et al. 2006; Schäffer 2008). As for unaccusatives built from the fientive predicate -​d become, I conjecture that they are monoeventive predicates just as their English translation suggests and are not to be viewed as truncated causatives. Below I summarize crucial morphological pieces which serve to build non-​active and unaccusative predicates in Georgian. (25) a. i-​ —​[reflexive implicit argument marking on] non-​active v b. -​d —​fientive V become c. -​eb —​imperfective aspect realized in Event and absent in the aorist/​subjunctive Active and non-​active v are thematic predicates, while the category that guarantees the nominative behaviour of Georgian is functional, it provides aspectual information and theta-​licenses the higher argument of the active vP. The transitive subject and the direct object in nominative tenses belong to two different domains, EventP and vP. The subject is case-​licensed by the c-​commanding T via standard structural case-​Agreement mechanisms as in Chomsky (2000), while the direct object checks its case against the closest functional c-​commanding category Event. (26) [TP T [EvP DPi case Evi [vP vi [VP DPcase V]]]] In the absence of Event in ergative tenses, T directly takes a vP as its complement. If the verbal domain contains one argument, its case is checked against T and is nominative. However, in transitive contexts where two arguments need to have their case checked, it is the lower of the two, the direct object, that case-​Agrees with T. The higher agent argument is case-​marked with the ergative case, and accounting for the source of this marking has been a notorious conundrum of linguistic theorizing. In a nutshell, the problem may be tackled from three angles: ergative can be analysed as (i) an inherent case (to be discussed in section 8.5.3), (ii) a structural case checked by a specific feature-​content of some functional category (see section 8.5.4), or (iii) a dependent case, assigned when some specific structural configurational requirements are met. I will adopt the third path in section 8.5, first showing why it is unwarranted to view the ergative case in Georgian as inherent, contra Nash (1995).

 

190   Léa Nash

8.5  Case Assignment in Ergative Tenses 8.5.1 Properties of Ergative Arguments in Georgian In this section, I  expose properties of arguments marked with the ergative case in Georgian which reveal that these arguments behave like their homologues in nominative tenses: both are sensitive to finiteness of the clause, both trigger number agreement with the verb and both case-​concord with secondary predicates.

8.5.1.1 Ergative Nominals are DPs The ergative case morpheme, the suffix -​ma/​-​m, has no other functions in Georgian, unlike other ergative languages where the affix or the adposition which marks the transitive subject is also used to mark locative arguments (e.g. Avar), instrumentals (e.g. Udi), genitives (e.g. Lak, Inuit) (see Tuite, Chapter 45, this volume). The ergative case marking in Georgian cannot qualify as a postposition because unlike other postpositions and like nominative and accusative/​dative case affixes it is subject to a noun–​adjective case concord within a DP (see Polinsky in press on analysis of ergative as a prepositional marker). (27) a. ert-​ma lamaz-​ma Kac-​ma

oneerg handsomeerg manerg

b. ert-​s lamaz-​s Kac-​s

oneacc handsomeacc manacc

c. ert-​i lamaz-​i Kac-​i

onenom handsomenom mannom

d. ert(*-​ze) lamaz(*-​ze) Kac-​ze one-​on  handsome-​on man-​on ‘on one handsome man’

8.5.1.2 Ergative DPs Occur in Finite Contexts Ergative, like nominative, dative and accusative, is unavailable in non-​finite contexts, which in Georgian involve nominalizations. Nor does the ergative affix mark optional agents in passive constructions that are formed in Georgian as projections of a postposition mier specifically designated to mark passivized agents, (28a). In nominalizations, genitive is assigned to the sole argument of a nominalized predicate, and in case of nominalizations of transitive predicates genitive marks the theme and the agent optionally surfaces as a mier-​PP of passive constructions (28b–​c). (28) a. vano-​ø iQo [nino-​s mier] nakebi Vano-​nom was Nino-​gen by praise.past.p ‘Vano was praised by Nino’

 

Split ergativity and ergative case in Georgian    191 b. ga=v-​igon-​e [nino-s laParaK]-​i /​ *[nino-m laParaK]-​i prev=1-​hear-​aor Nino-​gen talking-​nom /​ Nino-​erg talking-​nom ‘I heard Nino(’s) talking’ c. ga=v-​igon-​e [[nino-​s mier] vano-​s keba]-​ø prev=1-​hear-​aor Nino-​gen by Vano-​gen praising-​nom ‘I heard Nino(’s) praising (of) Vano’

8.5.1.3 Ergative DPs Are Subjects of “Transitivized” Unergatives Subjects of unergative verbs are uniformly marked as ergative, which can be taken as an indication that this case is semantically tied to the expression of agentivity. Yet, unergative constructions are not syntactically monoargumental as clearly witnessed by the morphological shape of corresponding verbs which obligatorily contain the RIAM i-​, which as we saw in section 8.4.3.1 is also employed in Georgian to signal the implicit argument in non-​active contexts. The presence of this affix is a non-​ambiguous indication that the internal structure of unergative predicates is complex in ergative tenses. Nash (2016) analyses them as transitive bieventive internally caused accomplishments where the causer is coreferential with the argument of the process whose structural presence is signalled by riam. This reanalysis of the aspectual class of activities into accomplishments is not directly tied to ergativity, as the same riam also appears on unergative verbs in the nominative future tense. It is rather forced by the mismatch between interpretative requirements to express completed events (the future is perfective in Georgian) and formal syntactic means to execute them: in the absence of a functional category that expresses perfectivity, this information must be read off the internal structure of vP:  the structure of accomplishments with two subevents, activity and result, provides the required meaning of the completion of eventuality.8 (29) nino-​m i-​laParaK-​a, i-​varǰiš-​a, Nino-​erg riam-​talk-​aor3s riam-​exercise-​aor3s, ‘Nino talked, exercised, hissed’ literally, ‘Nino caused her talking, exercising, hissing’ ‘Nino had herself talk, exercise, hiss’

i-​sisin-​a riam-​hiss-​aor3s

8.5.1.4 Ergative DPs Trigger Number Agreement A good indication that ergative DPs are structurally case-​marked is that they trigger number agreement with the verb in the same fashion as nominative subjects.9 8 

Subjects of stative verbs in Georgian such as love, fear, hate are never marked with ergative; this class of verbs has a radically different behaviour than dynamic verbs and will not be treated in this work. It is sufficient to point out that the dynamic–​stative split is an indication that ergative is tied to complex internal vP structure, namely to the availability of bieventive decomposition of the eventuality expressed by the verb. 9  Examples in (30–​31) show that 3rd person nominative or accusative plural direct objects never agree with the verb. Number agreement in Georgian is subject-​oriented, triggered by the highest argument (see also n. 12)

 

192   Léa Nash (30) a. Kac-​i xaT-​av-​s kal-​eb-​s man-​nom draw-​ts-​3s woman-​pl-​acc ‘a man draws women’ b. Kac-​eb-​i xaT-​av-​en man-​pl-​nom draw-​ts-​3pl ‘men draw a woman’ (31)

kal-​s woman-​acc

a. Kac-​ma xaT-​a kal-​eb-​i man-​erg draw-​aor3s woman-​pl-​nom ‘a man drew women’ b. Kac-​eb-​ma xat-​es kal-​i man-​pl-​erg draw-​aor3pl woman-​nom ‘men drew a woman’

8.5.1.5 Ergative and 1st/​2nd Person Arguments Only 3rd person arguments can appear with the ergative case marking. First-​and second-​person pronouns do not bear any case endings in Georgian and are coreferenced on the verb by person prefixes which have been analysed as clitics by Nash-​ Haran 1992; Halle & Marantz 1993; Nash 1997. I conjecture that 1st/​2nd arguments, due to their special status as discourse participants, are uniformly represented as referential pro in vP (or in EventP in nominative tenses) and are doubled by a corresponding clitic in T, similarly to the situation attested in many Romance languages (Manzini & Savoia 2005). The optional “caseless” full pronouns are henceforth best viewed as adjuncts to case-​marked pro. This is particularly clear when we consider agreement patterns of depictive nominal and adjectival secondary predicates. The secondary predicate agrees with the silent pro marked as ergative in (32a, 33a) and as nominative in (32b, 33b), while the optional “caseless” full 1sg pronoun me does not change its form across tenses.

(32) a. bavš-​ma   (me) proerg da=v-​Tov-​e tbilis-​i child-​erg I prev=1-​leave-​aor Tbilisi-​nom ‘I left Tbilisi (while) a child’ b. bavšv-​i   (me) pronom da=v-​obl-​di child-​nom I prev=1-​orphan-​become-​aor ‘I became an orphan, (while) a child’ (33) a. (me) proerg v-​nax-​e I 1-​see-​aor ‘Ii saw Vano drunki’

vano-​ø mtvral-​ma Vano-​nom drunk-​erg

 

Split ergativity and ergative case in Georgian    193 b. (me) pronom v-​naxul-​ob-​di vano-​s mtvral-​i I 1-​see-​ts-​past Vano-​acc drunk-​nom ‘Ii used to see Vano drunki’ Having illustrated that ergative arguments behave as structurally case-​marked DPs with respect to subject–​verb agreement and case concord, I turn in section 8.5.2 to laying out an account of ergative case assignment in functionally impoverished environments where one functional category, instead of two in nominative contexts, checks structural case of main arguments of transitive clauses.

8.5.2 Ergative as a Dependent Case According to the present analysis, Event is absent in the functional spine of clauses in ergative tenses, and Tense, the locus of reference time information, directly selects vP. If the verbal domain contains one argument, its case is naturally checked by the closest c-​commanding functional head, T, and is nominative. But when the vP contains a transitive predicate, two arguments must have their cases checked against the same functional head. I propose that case checking in these circumstances proceeds along the algorithm of Dependent Case, as proposed in Marantz (1991) and further elaborated in Bobaljik (2008), Baker & Vinokurova (2010), Baker (2014a). The central idea of this principle is that a “dependent” structural case may be assigned to one argument if there is a second argument in the same domain that needs its case checked.10 In other words, two nominals in the same domain that have no other structural/​lexical means to have their case licensed compete for the same case-​checking source (see Bittner & Hale 1996a). The dependent case assigned to the higher of two is ergative. The notion of nominals is to be taken here in a larger sense and should apply to implicit and clausal arguments. The first reason is that in Georgian, the subject of the transitive clause is marked ergative even when the direct object is CP (34). Therefore, the dependent case mechanism should apply not only to DPs per se but to argument positions that nominal may occupy. (34) vano-​m tkv-​a [CProm bebia-​ø ak vano-​erg say-​aor3s   that grandma-​nom here ‘Vano said that Grandmother is here’

ari-​s] be-​pres3sg

Second, some predicates take ergative and dative arguments, so the ergative subject does not “see” a case competitor, as dative is assigned by the applicative head. 10  Specifically, according to Marantz’s (1991) formulation, dependent case is assigned by V+I to a position governed by V+I when a distinct position governed by V+I is: a. not “marked” (does not have lexically governed case) b. distinct from the chain being assigned dependent case. Dependent case assigned up to subject ergative Dependent case assigned down to object accusative

 

194   Léa Nash (35)

a. vano-​m u-​Kbin-​a nino-​s Vano-​erg appl-​bite-​aor3s Nino-​dat ‘Vano bit (at) Nino’ b. vano-​m u-​daraj-​a Vano-​erg appl-​guard-​aor3s ‘Vano guarded the house’

nino-​s Nino-​dat

c. vano-​m u-​Qvir-​a nino-​s Vano-​erg appl-​shout-​aor3s Nino-​dat ‘Vano shouted at Nino’ These constructions seem problematic if we want to maintain the idea that ergative is dependent on the presence of another element as no other argument is required in (35), as their translations indicate. Yet, the presence of applicative u-​implies that there is more than meets the eye in these configurations. u-​generally adds possessive/​benefactive arguments and can be regarded as a low applicative head that relates two individuals in the sense of Pylkkänen (2008). The sentences in (35) should then be interpreted not as their English translations suggest but rather as events of attribution (causing to have) of the result of the activity to another entity. So (35a) means Vano gave Nino a bite; (35b) Vano gave the house a guard(ing), (35c) Vano gave Nino a shout. I propose that in these constructions the dative possessor is related via the applicative u-​to a weak implicit argument, with a cognate object meaning, as in Landau (2010). This implicit argument is syntactically represented as pro and counts as a case-​competitor for the DP agent.11 (36) vano-​m u-​Kbin-​a [Nino-​s Vano-​erg appl-​bite-​aor3s Nino-​dat Lit.: ‘Vano caused Nino to have a bite

u-​ pro] appl

An independent proof that there is a hidden nominal in (37b–​c) is that these predicates are compatible with the adjectival intensifier iseti ‘such … that’ that only modifies nominals, as in (37a). (37) a. nino-​m   [iset-​i      Kaba-​ø]    iQid-​a         [rom  Qvela-​ø  ga=giž-​d-​a] N-​erg  such-​nom  dress-​nom  buy-​aor3s that   all-​ nom  prev=crazy​become-​aor3s ‘Nino bought such a dress that everyone went crazy’ b. suares-​ma [iset-​i pro] u-​Kbin-​a vano-​s [rom Qvela-​ø Suarez-​erg such-​nom    appl-​bite-​aor3s vano-​dat that all-​nom ga=giž-​d-​a] prev=crazy-​​become-​aor3s ‘Suarez bit Vano “such” that everyone went crazy’ 11 

Referential pronominal argument drop is common in Georgian, in which case the missing argument is represented as pro. These silent pronouns count as visible case-​competitors, obviously.

 

Split ergativity and ergative case in Georgian    195 c. nino-​m [iset-​i pro] u-​Qvir-​a vano-​s [rom Qvela-​ø N-​erg such-​nom      appl-​shout-​aor3s Vano-​dat that all-​nom ga=giž-​d-​a] prev=crazy-​​become-​aor3s ‘Nino shouted “such” at Vano that everyone went crazy’

I therefore conclude that in spite of some apparently problematic configurations where the direct object is not expressed as a “visible” nominal, ergative is always assigned in configurations where T sees two arguments with unvalued case features in the same vP domain and marks the higher of the two with the dependent case. I now turn to the discussion of why accounting for ergative as an inherent case, at least in Georgian and desirably more generally, is less warranted than the present analysis.

8.5.3 Ergative Is Not an Inherent Case The literature on ergativity abounds with analyses that treat ergative as an inherent case assigned within VP confines, by V or little v (Mahajan 1989; Jelinek 1993; Johns 1993; Nash 1995; Butt 1995; Massam 2002; Woolford 2006; Legate 2008). According to Woolford (2006: 12), structural cases are assigned via Agreement that a DP establishes with the closest c-​commanding functional head (Chomsky 2000), while inherent cases are associated with certain θ-​positions and are typically assigned by a lexical head to the DP base-​generated in its specifier. There is good cross-​linguistic evidence that ergative can qualify as an inherent case: (a) it is not assigned to derived subjects (Marantz 1991; Legate 2012a); (b) it is assigned to agents, causers, and instruments, i.e. to Proto-​Agents in Dowty’s (1991) sense. The link between ergative and agentive participant is stronger than that between nominative and agent (Polinsky in press); (c)  morphologically, ergative cross-​reference markers are closer to the verb root than absolutive markers (Jelinek 1993 on Lummi). These properties of ergative can also be easily accounted for if it is analysed as a structural dependent case assigned to the highest of vP arguments, along Marantz’s algorithm. Properties (a) and (b) follow from the fact that dependent case is assigned within vP where each structural position is thematically motivated—​there is no raising into theta-​positions. Property (c) is a natural consequence of ergative as the first case to be assigned when two nominals compete for the case valuation from T. We can predict that in languages where ergative is a dependent case, the second DP also checks its case by T and is sensitive to its finiteness, as attested in Georgian (see section 8.5.1.2). On a more general level, the principal weakness of theories that treat ergative as inherent assigned by v pertains to the identification of the relevant property which enables the verb to assign ergative. If v universally assigns the agent theta role why is it a case assigner only in a subpart of languages, namely ergative languages? Is the reason to be sought in different thematic capacities or in different case-​assigning capacities of

 

196   Léa Nash v in two case systems? Most accounts just state that v in ergative languages is endowed with the special [Case] feature. But if parametric variation hinges on this feature—​two systems differ because there is a special case feature available in one system but not in the other system—​I conclude that this amounts to a simple restatement of the parameter rather than to its explanation.12

8.5.4 Are Ergative and Nominative Checked by Different Functional Categories? Another line of thought that argues against the inherent nature of ergative and considers both nominative and ergative as structural cases checked by different functional categories has been developed in Murasugi (1992) and Bobaljik (1993b):  languages obey Obligatory Case Principle which fixes which case must be obligatorily assigned in the structure—​nominative systems set the subject case to be obligatory, while ergative systems identify the object case to be obligatory. This principle is subject to the same questioning; why do languages set the values as they do? If case is not a syntactic primitive but rather a derived notion which serves as a shorthand for a privileged structural relation between a (functional) category and a DP in the clause, OCP can be viewed as a restatement of the problem according to which different functional categories in each language type have certain case features without attributing this asymmetry to independent factors of their respective grammars. The present analysis attempts to pinpoint the triggering factor in one language, Georgian, which is responsible for the difference. This factor pertains to the structural means that a language disposes to link an eventuality to the reference time. Languages may dispose a Tense–​Event–​vP chain where a dedicated functional category, Event, expresses how the event described by the verb is articulated with respect to the reference time: whether it attains its natural end (perfective), or whether it still goes on (imperfective). In ergative systems, as witnessed by Georgian, this category may be absent and the eventuality is not explicitly located with respect to the reference time by functional means and is therefore understood as some dynamic process that happened (began and ended) without further entailments about attaining the natural telos or about its 12 

A reviewer suggests that my account of ergativity in terms of functional deficiency can be made compatible with an analysis of the ergative case as an inherent case assigned by v. Under such an analysis, v universally assigns the inherent ergative case locally, to its specifier. But as in nominative systems Event-​ v reanalysis takes place, the agent is generated too far from v, in Spec,EventP, which blocks its marking as ergative. Although a theory along these lines is conceptually appealing (because no extra mechanism of dependent case is needed) it fails to explain, without additional assumptions, why the ergative is contingent on finiteness in Georgian. What would prevent an agent in nominalisations to surface with the inherent ergative instead of the genitive? More importantly, such an analysis fails to account for the number agreement triggered by ergative DPs. A way out cannot be sought in Multiple Agree mechanism (see Hiraiwa 2001) that allows not only a lower structurally marked DP to be visible for feature valuation by T, but an intervening inherently case-​marked DP too. As shown in (31) and in n. 9, transitive ergative subjects trigger number agreement but nominative objects never do.

 

Split ergativity and ergative case in Georgian    197 continuation. A direct consequence of this structural deficiency leads to dependent case assignment in a language, as one functional category, T, has to check two unvalued cases of vP arguments. It ensues from this analysis that thematic properties of verbal predicates are identical in both case systems. Yet in nominative languages, thematic licensing of the agent is relegated to the specifier of the category Event via the mechanism of semantic control, as in (22).

8.6  An Alternative Account of Ergative Splits 8.6.1 Nominative Systems as Biclausal Structures On the basis of the analysis of progressive constructions in Basque, Laka (2006a) proposes that an ergative language can shift to a nominative system when the syntactic expression of the tense requires a biclausal configuration. In Basque, a transitive VP in the progressive tense surfaces as a nominalization embedded under a locative PP (38). The main verb ari (be engaged) is intransitive and its only argument is naturally marked with nominative. The nominative argument of ari syntactically controls PRO in the embedded nominalized VP. In a nutshell, Basque clauses in the progressive tense (39) are analysed as their English archaic homologues as in Johni is a(t) [PROi writing the letter], while other tenses, with the ergative transitive subject, are expressed by monoclausal structures. (38) [TPDPi T [VPV ariintrans [PP P [NP N [VPPROi DP Vtrans ] (39) emakume-​ai [PROi ogi-​a ja-​te-​n] ari da woman-​det bread-​det eat-​nom-​loc engage is ‘the woman is (engaged in) eating the bread.’ [Laka 2006a: 183, ex. 16] Building on Laka’s proposal, Coon (2010b), Coon and Preminger (2012) argue that the shift from ergative to nominative systems should not viewed as a shift of case-​assigning features of functional heads but rather as a result of adding structure to clauses with ergative subjects: ergative clauses are monoclausal and nominative ones biclausal. The question may be raised as to whether a biclausal structure is always needed to tie a simple eventuality to the reference time. The authors answer positively to this question and propose that the progressive tense, which syntactically is a manifestation of imperfective aspect, may universally require the nominalization of the eventuality expressed by vP and its subsequent embedding under a locative P: X is in [NPV-​ing]. As a consequence of this syntactic representation of imperfectivity, the external argument of the eventuality is realized in the nominalized VP as a PRO and the main verb is intransitive.

 

198   Léa Nash

8.6.2 Temporal Relations: Biclausal Configurations or Restructuring Importantly, the Laka–​Coon line of thought and the present analysis hinge on one central idea: when a language shifts from ergative to nominative, the added structure—​ another clause or an extra functional category—​serves to syntactically control as in (38) or semantically control, as in (22), the highest argument of the transitive predicate. There is no linguistic evidence to analyse Georgian imperfective tenses as biclausal: no auxiliaries, no nominalizations and no locative P are involved in building nominative clauses. The imperfective marker ts, which minimally distinguishes the verbal form in imperfective and non-​imperfective aspects, cannot be further split into the nominalizing part and the re-​verbalizing affix functioning as the main verb. And although ts appears in nominalizations as shown in (21), it is not a nominalizer by itself—​an additional nominalizing affix is required to categorize the verb as a noun. Under a larger perspective, one may question whether an analysis of progressives as biclausal control structures, viable for Basque, can be easily extended to other languages. Does the periphrastic progressive tense in English also involve syntactic control or is the surface subject the argument of -​ing verbs rather than the argument of be? Ramchand (2013) provides a convincing support for the second option and shows that -​ing forms and main verbs surface in the same position.

8.6.3 There Is a Phase Head above vP in English: Ramchand (2013) Recent work by Ramchand (2013) and Ramchand and Svenonius (2014) on the semantics of functional inventory sheds new light on the embedding of functional categories in English. They argue that several syntactic tests, such as VP-​fronting, pseudo-​cleft constructions, British non-​finite do substitution, show that English progressive -​ing forms, -​en passives, and main verbs behave differently than perfect participles or modal infinitives in English. This is an indication that these verbal forms originate low in the VP structure. I reproduce here just one test showing that do can substitute for an infinitive modal complement or a perfect participle (40a–​b), but not for a progressive or passive participle. (40) a. b. c. d.

John might leave, and Mary might do also. John has left, and Mary has done also. John is leaving, and Mary is (*doing) also. John was arrested, and Mary was (*done) also

 

Split ergativity and ergative case in Georgian    199 Other tests cited by the authors leave little doubt that the syntactic size of the simple VP and of the progressive VP is identical, which goes against the idea, at least synchronically, that progressive clauses are biclausal configurations, unlike present perfect clauses in (40b). Ramchand (2013) introduces a frontier (signalled in (41) by #) in the functional spine of English clauses, where the perfect and T are separated from the progressive arguing that this cut corresponds to the phase boundary and tries to identify the category that demarcates the phase. (41) [TPT[have] ASPperf[ -​en] #ASP[-​ing] V[PASS-​en] V…] Ramchand labels it as Event, the term that I have borrowed in the present work, and contends that this category “guarantees the sortal shift from event descriptions to temporally enriched situational descriptions and it also determines a choice of subject of predication for the whole complex event, the Topic of Event”. The author claims that the existence of Event does not impose the rejection of the little v, which must thematically license the agent (the initiator) and semantically delineate an eventuality. In this work, I attempt to show that the meaning Ramchand ascribes to this phasal head Event corresponds to the category spelled out by ts in Georgian. In nominative tenses, Georgian, just like English, has topics (syntactically, specifiers) of Events, while in ergative tenses, Georgian has topics of eventualities.

8.7 Conclusion On the basis of the analysis of ergative split in Georgian I have attempted to show that nominative and ergative tenses within the same language, and, by extension, in nominative and ergative languages in general, differ with respect to functional inventories that are available to express temporal relations syntactically in each system. Nominative systems involve a functional category Event, a phasal head above vP, which combines two functions. On the one hand, Event has the properties commonly attributed to Kratzer’s Voice as it thematically licenses the external argument of vP via semantic control with v. On the other hand, it also has the properties commonly ascribed to the Aspect category as it expresses the localization of the eventuality with respect to the reference time in T. The argument in Spec,EventP is marked as nominative by the c-​commanding functional category T in nominative languages. Ergative systems dispose a smaller functional clausal spine as they lack the Event head. As a result, in these systems (i) the eventuality stays temporally opaque with respect to the time of utterance, and (ii) the transitive subject is generated in a different position and lower than in nominative tenses: it is thematically licensed by v in Spec,vP and is marked as ergative via the mechanism of dependent case assignment. To conclude, the ergative case on a DP is taken in the system I have elaborated in the present analysis to signal the absence of an intervening

 

200   Léa Nash functional category between T and v that serves as a thematic licenser of subjects in a nominative language.

Abbreviations 1st, first-person singular; 1o, first-person singular object; 2nd, second-person singular; 3s, third-person singular; ACC, accusative; AOR, aorist; APPL, applicative; ASP, aspect; DAT, dative; ERG, ergative; GEN, genitive; LOC, locative; NEG, negation; NOM, nominative; PAST.P, past participle; PAST, past tense; PL, plural; PRES, present tense; PREV, aspectual preverb; RIAM, reflexive implicit; TS, thematic suffix.

 

PA RT  I I

C HA R AC T E R I ST IC S A N D E X T E N SION S

 

 

Characteristics

 

 

Chapter 9

Split ergat i v i t y i n sy ntax a nd at morphol o gica l spe l l ou t Ellen Woolford

9.1 Introduction A split ergative case pattern is one in which not all of the subjects that could be marked with ergative case actually are. Two well-​known types are person splits where only third person subjects are marked with ergative case, and aspect splits where ergative case is only used in the perfective aspect.1 A language with a split ergative case pattern qualifies as a split ergative language under anyone’s definition, but there is a wide range of opinion among linguists as to what else would qualify a language to be called a split ergative language. Under the narrowest definition, which I favor, a split ergative language necessarily has a split ergative case pattern.2 In contrast, Aldridge (2008a) considers Warlpiri to be a split ergative language, even though its ergative case pattern is not split, nor is its agreement (which follows a purely nominative–​accusative pattern). Warlpiri qualifies as a split ergative language under a broader definition wherein an ergative language is classified as split ergative unless both its case and agreement systems are fully ergative. Deal (2015) uses an even broader definition, under which even a language with a fully ergative case and agreement pattern would not qualify as fully ergative unless the language

1 

There are many discussions/​surveys of types of split ergativity in the literature. Some of the most well-​known include Silverstein (1976); Comrie (1978); Dixon (1979, 1994); DeLancey (1981); and Song (2001). 2  This includes languages with a split ergative agreement pattern if the case system (overt or covert) is ergative.

 

206   Ellen Woolford also had syntactic ergativity.3 Dixon (1994: 161) does not lump morphological and syntactic ergativity together, but instead classifies the morphological and syntactic pattern of a language separately, describing Dyirbal as split with respect to case (because it has a person split), but fully ergative at the syntactic level. Dixon (1994) and McGregor (2009) classify languages as split ergative even if all main clauses are fully ergative, if embedded/​dependent clauses are not. Another point of disagreement concerns whether active (Split S) languages count as split ergative.4 Dahlstrom (1983) maintains that active languages are not split ergative, but Dixon (1994) maintains that they are, although of a different type. Each view stems from a different definition of split ergativity. I will add my view here, that there are actually two different types of active languages, which I refer to as active and active ergative, but neither is split ergative. In fact, in my view, active ergative languages are the only ergative languages that are not split ergative. Perhaps the narrowest definition of split ergativity is that of Coon (2013a) who maintains that in a technical sense there is no split ergativity; instead what we describe as a split ergative pattern always involves two different syntactic constructions, only one of which allows/​licenses ergative case. The question of how split ergativity is defined becomes important in evaluating the validity of typological claims such as the common statement in the literature that (almost) every ergative language is split ergative (e.g. Silverstein 1976; Moravcsik 1978b; Dixon 1979). It is also theoretically important in that there is a tendency to assume that full ergativity is the basic pattern for an ergative language whereas split ergativity needs to be explained. Sections 9.2 and 9.3 of this chapter focus on person and aspect splits and the literature and controversies concerning the proper theoretical analysis of these two well-​known types of ergative splits. Two lesser known types of split ergativity are described in sections 9.4 and 9.5, one in Nepali based on Kratzer’s (1995) distinction between stage and individual level predicates and one in two unrelated languages, Folopa and Mongsen Ao, which is partially determined by social conventions. Section 9.6 attempts to clarify the controversy over whether active languages are split ergative, and the various definitions of split ergativity upon which this controversy is based. Section 9.7 turns a little-​known type of ergative split mentioned in Silverstein (1976), but seldom included in subsequent surveys. This type of ergative case split is based on the features of the object. I agree with Silverstein that this is a type of split ergativity, even though the literature generally does not describe the relevant languages as split ergative. Languages with this type of ergative case split include Inuit (Bittner and Hale 1996b), Niuean (Massam 2000), and Nez Perce (Rude 1988; Deal 2015). For Inuit and Niuean, the object feature that matters is specificity, while in Nez Perce it is described as topicality (Rude 1988). Finally, section 9.8 3 

A language is said to be syntactically ergative if syntactic rules (appear to) make reference to ergative case. 4  Active languages, also called Split S languages, have two classes of verbs. In one verb class, all subjects, transitive and intransitive are marked alike; in the other class, intransitive subjects are marked like objects.

 

Split in syntax and at morphological spellout    207 describes an ergative case split based on the proximate/​obviate distinction in Sahaptin (Rigsby and Rude 1996).

9.2  Person/​Animacy/​NP Splits Person splits are often referred to as animacy splits or NP splits since they are not limited to person but can involve other features of NPs such as animacy. Person/​animacy/​ NP splits are “almost universal in Australian languages and well-​attested elsewhere (in North America, and in some Siberian and Tibeto-​Burman languages)” according to DeLancey (1981: 628). In the most common type of person split, ergative case marks only third person subjects, never first or second. We can illustrate this person split with examples from Marathi. There is a contrast between example (1)  where the third person subject is marked with ergative case and example (2) where the first person subject is not marked with ergative case:5 (1) Ram-ne     acɘvlɘ. Ram-erg washed.hands.neut.3sg ‘Ram washed his hands.’ (2) Mi  acɘvlɘ.  I washed.hands.neut.3sg ‘I washed my hands.’

[Marathi] (Comrie 1984: 862 (15)) [Marathi] (Comrie 1984: 862 (16))

9.2.1 The Silverstein Hierarchy and Markedness Discussions of person/​animacy/​NP splits always include Silverstein’s (1976) classic work on this topic, and his proposal that in such splits, ergative case marks subjects whose features are low the person/​animacy hierarchy: (3) Person/​Animacy Hierarchy (simplified) 1st > 2nd > animate > inanimate For example, one can characterize the Marathi person split shown in (1) and (2) by drawing a line between 2nd person and human, indicating that ergative case marks the 5  Note that the agreement in the Marathi example in (1) is the default 3rd person neuter which occurs when there is no nominative in the clause, and this default agreement also occurs in (2), indicating that the first person subject in (2) has ergative case in syntax, where agreement features are determined. This is an example of a purely morphological ergative split. (See the discussion in section 9.2.2.)

 

208   Ellen Woolford lower ranked third person subjects, but not the higher ranked first or second person subjects: (4) Marathi pattern 1st >  2nd │> animate > inanimate nominative ergative -​-​-​-​-​-​-​-​-​-​-​-​→ However, not all person splits can be characterized by dividing this standard hierarchy into two neat parts. Silverstein discusses several more complex patterns. Dhirari marks all subjects ergative except 1st and 2nd person plural subjects. This is unexpected under the simple version of the hierarchy as outlined. Describing the Dhirari pattern requires a different hierarchy, with 1st and 2nd person plurals at the top/​left. Using this (language specific) hierarchy, it is possible to characterize the pattern as divided into two neat parts, where ergative case marks only the lower elements: (5) Dhirari pattern 1st pl > 2nd pl │> 1st sg > 2nd sg > animate > inanimate nominative ergative -​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​→ Silverstein describes a slight variant of the Dhirari pattern which occurs in Bandjalang, where all subjects except 1st plural are marked with ergative case. This pattern can be described using the same hierarchy as in Dhirari, but moving the line leftward: (6) Bandjalang pattern 1st pl │> 2nd pl > 1st sg > 2nd sg > animate > inanimate nom ergative -​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​-​→ Silverstein (1976:  127)  describes a more complex in Aranda, where only 1st singular pronouns and 3rd person inanimates are marked with ergative case (although Legate (2014a) points out based on the original data source that 3rd animates are also marked ergative in this language). Marking the lower inanimates with ergative case (or all third persons), but nothing higher would be an expected pattern under the animacy hierarchy approach; the problem is the fact that the very high 1st singular is also marked with ergative case: (7) Aranda pattern (with the correction noted in Legate (2014a)) 1st sg > │ 1st pl > 2nd sg > 2nd pl │> animate > inanimate ergative nominative ergative -​-​-​-​-​-​-​-​-​-​-​-​-​→ Silverstein describes Aranda as having two separate split systems of ergative marking. That is, the pattern of this language can be described using two separate hierarchies, one separating local persons )1st and 2nd from 3rd persons, and one involving only local persons, which places 1st singular at the bottom:

 

Split in syntax and at morphological spellout    209 (8) Two hierarchy Approach to Aranda local │> 3rd nom ergative 2nd pl  > 1st pl > 2nd sg │>  1st sg nominative ergative It is clear from Silverstein’s discussion that he does not claim to account for all person/​ animacy/​NP splits in all languages, but rather he proposes an approach to this kind of split as the basis for further work: “There are numerous ‘holes’ in the pattern and these mean that we have the opportunity for further constraint of the system” (1976: 125).6 Two important points that Silverstein makes are that ergative splits are not random and that the patterns of NP splits have to do with markedness.

9.2.2 Person/​Animacy/​NP Splits Occur at Morphological Spellout In the introduction to this section, I pointed out that the person split in Marathi is not present in syntax, based on the fact that it has no effect on agreement. Instead, this person split, and perhaps all person splits are only splits at the morphological level. Goddard (1982) argues that there is no split in ergative case in Australian languages, but only a split in ergative case marking, and he views these as instances of “case homonymy.” Baerman et al. (2005) argue more generally that all ergative splits of the person/​ animacy/​NP type are instances of case syncretism, which they define broadly to include any situation in which a case feature present in syntax is not morphologically realized/​ distinguished. Legate (2014a:  183)  presents detailed arguments for this position, that “split ergativity based on nominal type is a morphological phenomenon, not a syntactic one.” Although Legate notes that the formal analysis of this type of splits depends on one’s theory of morphology, she gives a formal account of this type of ergative split in the framework of Distributed Morphology. In that framework, there is a level between syntax and morpheme insertion where certain operations on features can take place. Legate proposes that at this level, the ergative case feature on the subject in syntax is deleted in certain contexts, such as in the presence of a local person feature. As a result, the ergative case feature is not morphologically spelled out/​distinguished, in that context. Under this view, person/​animacy/​NP splits involving ergative case are much like what one might call a ‘gender split’ in English pronouns, where gender is morphologically marked in third person pronouns (he, she, it), but not in first or second person 6  Comrie (1981a) notes that although the Saibai dialect of Kalaw Lagaw Ya comes close to matching Silverstein’s hierarchy predictions, there are exceptions. One exception is that proper nouns behave as if they are higher ranked than pronouns. Additional exceptions to the strict predictions of the animacy hierarchy are pointed out in Filimonova (2005) and Legate (2014a).

 

210   Ellen Woolford pronouns.7 A more direct parallel to ergative person splits is the dative person split in Italian wherein pronominal clitics are only distinctly marked for dative case in the third person: (9)

Dative split in Italian pronominal clitics 1st mi accusative/​dative 2nd ti accusative/​dative 3rd lo accusative.masculine 3rd gli dative.masculine

9.3  Aspect/​TAM Splits We can illustrate an aspect split with the Marathi examples in (10) and (11), where ergative case is used in the perfective aspect, but not in the imperfective. (10) Ravi-​ni kavitaa vaac-​l-​i. Ravi-​erg poem read-​perf-​3sg.fem ‘Ravi read the poem.’ (11)

Ravi kavitaa vaac-​t-​o. Ravi.nom poem read-​impf-​3sg.masc ‘Ravi reads a poem.’

[Marathi] (Gair and Wali 1988: 96 (18b)) [Marathi] (Gair and Wali 1988: 96 (18a))

Splits of this type are often called TAM splits because they are said to involve tense, aspect, or mood (e.g. Dixon 1994). However, some recent literature questions whether tense or mood splits actually exist. Coon (2013a: 189) cites literature suggesting that purported examples of tense splits are better analyzed as aspect splits. Salanova (2007: 47) in particular states this conclusion directly: “so-​called tense-​aspect-​mood splits essentially boil down to aspectually conditioned splits.” Salanova also doubts that mood splits exist, as does Coon who notes that mood splits often involve imperatives and argues that all mood splits are “reducible to a clause type split” (2013a: 189). Of course, some would count clause type splits as ergative splits, but Coon’s position is that the syntactic structure of the two clauses in such splits differs such that one would not expect ergative case to be assigned in one of them. In Marathi and other languages which have agreement, it is clear that the aspect split is present in syntax, given that it affects the agreement. Note that the agreement in (10) is 3rd feminine, agreeing with the feminine nominative object, poem, while the agreement in (11) is 3rd masculine, agreeing with the nominative subject, Ravi. 7  A reviewer notes that in other languages with this gender split in pronouns, such as French and Russian, there is evidence in the form of predicate adjective agreement and participle agreement that gender features are present in syntax on first and second person pronouns.

 

Split in syntax and at morphological spellout    211 Coon (2013b) has an excellent recent survey of examples of ergative aspect splits, including some less often cited examples such as Samoan. Coon (2013b: 184) gives the following example from Milner (1973): (12) Na va’  ai-​a e le tama le   i’a. past look.at-​perf erg the boy the fish ‘The boy spotted the fish.’ (13) Na va’ai le tama i le i’a. past look.at  the boy at the fish ‘The boy looked at the fish.’

[Samoan] (perfective aspect)

(imperfective aspect)

While many examples of ergative aspect splits are like that of Hindi, shown in (10) and (11), which manifests a split in the case marking of transitive clauses depending on the aspect, some of the Mayan languages have an aspect split which is limited to intransitive clauses, as we see in the Chol examples from Coon (2013a) ((14)–​(15)). The transitive pair in (14) and (15), which differ in aspect, both show the same agreement pattern: the set A prefix cross-​references the subject and the set B suffix cross-​references the object: (14) Tyi i-​jats’-​ä-​yoñ. perf A3-​hit-​tr-​B1 ‘She hit me.’

[Chol] (Coon 2013a: 11 (15a))

(15) Choñkol i-​jats’-​oñ. prog A3-​hit-​​B1 ‘She’s hitting me.’

(Coon 2013a: 11 (16a))

In contrast, the intransitive pair in (16) and (17) are different in agreement marking. The subject in the perfective example in (16) is marked like the transitive object, while the subject in the imperfective example in (17) is marked like the transitive subject: (16) Tyi majl-​i-​yoñ. perf go-​intr-​B1 ‘I went.’

[Chol] (Coon 2013a: 11 (15b))

(17) Choñkol i-​majl-​el. prog A3-​go-​nml ‘She’s going.’

(Coon 2013a: 11 (16b))

There is currently no standard view as to how aspect splits should be analyzed. Existing approaches differ as to whether they situate some or all aspects in syntax or at morphological spellout. Davison (2004b) proposes a differential licensing approach for Hindi, postulating that the features/​conditions required to license ergative case in syntax

 

212   Ellen Woolford are different in different aspects, and Ura (2006) pursues this type of approach as well. Woolford (2013) explores a differential faithfulness approach, drawing a parallel between aspect splits in syntax and positional/​contextual faithfulness effects in phonology, e.g. Beckman (1998). Coon (2013a, 2013b) argues for a differential structures approach, under which ergative case cannot be licensed in imperfectives in some languages because the structure/​verb is one that does not take an external argument (Coon 2013a, 2013b). Coon’s approach is partially motivated by Laka’s (2006) account of the lack of ergative case in the Basque progressive, as in (18), in contrast to the presence of ergative case on the subject with the same verb in other contexts such as the simple present in (19): (18) Emakume-​a ogi-​a jaten woman-​det bread-​det eating ‘The woman is eating (the) bread.’ (19) Emakume-​a-​k ogi-​a woman-​det-​erg bread-​det ‘The woman eats (the) bread.’

ari da. prog is jaten du. eating has

[Basque] (Laka 2006a: 173 (1b))

(Laka 2006a: 173 (1a))

Although this is not the familiar ergative aspect split between perfective and imperfective aspect, it does involve an aspect, the progressive. Laka argues that morpheme, ari, glossed as progressive aspect in the example in (18), is really a main verb, with a meaning something like ‘engaged in,’ and that this progressive example actually has a biclausal construction, as shown in (20): (20) Emakume-​a ogi-​a ja-​te-​n ari da. woman-​det bread-​det eat-​nom-​loc engaged is ‘The woman is (engaged in) eating the bread.’

[Basque] (Laka 2006a: 174 (2c))

Supporting evidence that verbs in the higher clause does not take an external argument, and cannot license ergative case, is the presence of the be auxiliary, which occurs in unaccusative constructions in Basque, as in (21), rather than the have auxiliary, which occurs with verbs that do take an external argument, as in the example with an ergative subject in (19). (21)

Emakume-​a hurbil-​tzen woman-​det get/​near-​impf ‘The woman gets closer.’

da. is

[Basque] (Laka 2006a: 179 (7))

Coon (2013a) analyzes the aspect split in the agreement pattern of the Chol examples as involving very different syntactic structures in the two aspects: the perfective examples have an ordinary clausal structure, but the proposal is that the imperfective examples involve a possessive construction, so that a better translation of the progressive example in (17), repeated as (22) would be something like ‘Her going continues’:

 

Split in syntax and at morphological spellout    213 (22) Choñkol i-​majl-​el. prog A3-​go-​nml ‘She is going.’

[Chol] (Coon 2013a: 11 (16b))

In contrast, this aspect split, which occurs in some Mayan languages, but not others, is treated as a purely morphological phenomenon under the approach in Woolford (2013), involving a choice of which of the two cross-​referencing series, A (true agreement) or B (a pronominal clitic), to use for an intransitive subject. An important generalization that the correct formal approach to aspect splits should capture is the correlation between ergative case and the perfective aspect in such splits: (23) “If a split is conditioned by tense or aspect, the ergative marking is always found either in past tense or in perfective aspect (Dixon 1994: 99).” Coon’s approach will capture this generalization if the hypothesis holds cross-​linguistically that imperfectives in such splits always involve the verb ‘be,’ or some other verb or construction does not take an external argument, and can thus not license ergative case. As Coon notes, this approach is similar to that of Tsunoda (1981b) who tries to unite aspect splits with active–​stative languages, arguing that both involve alternations between types of verbs which do and do not take agentive/​external arguments.

9.4  Nepali: A Split Based on Stage versus Individual Level Predicates A new type of ergative split is described by Butt and Poudel (2007) in Nepali. The split involves the distinction between individual level and stage level predicates described in Kratzer (1995). The following minimal pair illustrates this ergative split:8 (24) Raam-​le angreji jaan-​da-​cha. [individual level predicate, Nepali] Ram-​erg English know-​impf-​nonpast.masc.3sg ‘Ram knows English.’ (Butt and Poudel 2007 (16a)) (25) Raam (aajaa) angriji bol-​da-​cha. [stage level predicate, Nepali] Ram.nom (today) English speak-​impf-​nonpast.masc.3sg ‘Ram will speak English (today).’ (Butt and Poudel 2007 (16b)) An individual level predicate “predicates a lasting/​inherent property of a referent” (Butt and Poudel 2007: 4). In contrast, a stage level predicate “says something about a property 8 

There is also a complex pattern of optionality of ergative marking in Nepali described in Li (2007).

 

214   Ellen Woolford of a referent that holds for a slice of that referent’s spatio-​temporal existence” (Butt and Poudel 2007: 4). Additional evidence that Nepali makes a distinction between stage and individual level predicates comes from copular clauses, where are different copulas for the two types of predicates: (26) Saru Bhakta kabi hun. [individual level predicate, Nepali] Saru Bhakta poet be.nonpast.3masc.hon ‘Saru Bhakta is a poet.’ (Butt and Poudel 2007 (13)) (27) Saru Bjalta aaja khusi Saru Bhakta today happy Saru Bhakta is happy today.

chan. [stage level predicate, Nepali] be.nonpast.3masc.hon (Butt and Poudel 2007 (14))

In other types of clauses, lacking a copula to mark the type of predicate, ergative case functions to mark the predicate as individual level, as in (24), while the absence of ergative case indicates a stage level predicate, as in (25). However, this pattern is present only in the imperfective aspect/​nonpast tense. Nepali is like Marathi in that in the perfective aspect/​past tense, all the verbs that can take an ergative subject do so, regardless of whether the predicate is stage or individual level. The fact that Nepali has two cross-​ cutting ergative splits, one based on aspect and one based on predicate type, may present an additional challenge (or perhaps an additional clue) for those who attempt a formal account of this pattern.

9.5  Socially Controlled Ergative Case Splits Dixon (1994) discusses an ergative case split which occurs in two unrelated languages, Folopa (Anderson and Wade 1988) and Mongsen Ao (Coupe 2008). This type of ergative case split appears to be determined by social factors; using the ergative morpheme is interpreted in certain contexts as asserting or emphasizing the subject’s control/​ownership of the object, which can be rude in some social situations, and indicates theft in others.

9.5.1 Folopa Anderson and Wade (1988) describe a language of Papua New Guinea, Folopa, which is a counterexample to the following claim in Dixon (1979): while languages can be Fluid-​S with respect to intransitives (using either ergative or nominative case for subjects), no language is Fluid-​S with respect to transitive clauses. Dixon (1994) modifies his typology and classifies as Folopa as Fluid-​S in both intransitive and transitive clauses.

 

Split in syntax and at morphological spellout    215 Although Folopa does allow a choice of an ergative or a nominative subject for most transitive verbs, the primary usage pattern according to Anderson and Wade (1988) is to use ergative case. Using/​spelling out the ergative case morpheme is the most frequent choice in texts and in judgments of sentences out of context. Anderson and Wade point out that ergative is the normal choice even if there is no possibility of ambiguity as to which argument is the subject, as in the following example: (28) U so-​né ama wa teo doko-​ta-​pό. [Folopa] that woman-​erg her string.bag unfinished weave-​pres-​ind ‘That woman is weaving her unfinished string bag.’ (Anderson and Wade 1988: 6 (16)) However, in situations involving food, using an ergative subject carries an implication of stinginess, because marking the subject with an ergative morpheme adds an interpretation an overt assertion of control or ownership. With respect to food, the norm is to share. “If a person uses the ergative and thus states his control, he will be considered stingy” (Anderson and Wade 1988: 11). Thus the socially correct version of the statement in (29) uses a nominative subject. (29) ę o faa-​ta. I.nom sago break.open-​pres ‘I am breaking this sago.’

[Folopa] (Anderson and Wade 1988: 11 (23))

In contrast, using I.erg yąlo instead of I.nom ę in the example (29) would be asserting control/​ownership over the food and would be considered stingy. In situations involving things the speaker does not own, using ergative case on the subject also asserts control/​ownership. In the version of the statement in (30), the speaker uses a nominative subject to indicate that he is not asserting control/​ownership of his brother’s sago palm, i.e. he mistakenly cut it down. In the second version in (31), the speaker uses an ergative subject to indicate that he is asserting control/​ownership, in that he intentionally cut down his brother’s sago palm. (30) No-​ό kale naaǫ o make ę di-​ale-​pό. [Folopa] brother-​voc the your sago young I.nom cut.down-​past-​ind ‘Brother, I (mistakenly) cut down your young sago tree.’ (Anderson and Wade 1988: 7) (31) No-​ό naaǫ o make yąlo di-​ale-​pό. brother-​voc your sago young I.erg cut.down-​past-​ind ‘Brother, I (intentionally) cut down your young sago tree.’ (Anderson and Wade 1988: 7) In section 9.5.2 we see this same split in an unrelated language, Mongsen Ao.

 

216   Ellen Woolford

9.5.2 Mongsen Ao The Mongsen dialect of Ao, a Tibeto-​Burman language described in Coupe (2008), has an ergative split that looks very much like what we saw in Folopa. As in Folopa, using ergative case in certain contexts, involving food and other people’s belongings, is taken to indicate that the speaker claims or emphasizes ownership, which is rude in certain contexts, and indicates theft in other contexts. The following minimal pair is given by Coupe (2008) to illustrate a situation in which using/​pronouncing ergative case on the subject indicates that the subject is willfully stealing. The first version of the sentence lacks an ergative case morpheme and lacks this interpretation; in contrast the second version has an overt ergative case morpheme and has the interpretation of willful theft: (32) A-​hɘn a-​tʃak tʃàɁ-​ɘ̀ɹ-​ùɁ. nrl-​chicken nrl-​paddy consume-​pres-​dec9 ‘The chickens are eating paddy.’

[Mongsen Ao] (Coupe 2008: 157 (5.22))

(33) A-​hɘn nɘ a-​tʃak tʃàɁ-​ɘ̀ɹ-​ùɁ.  [Mongsen Ao] nrl-​chicken erg nrl-​paddy consume-​pres-​dec ‘The chickens are eating paddy.’ (implying that they are stealing it) (Coupe 2008: 157 (5.23)) Mongsen Ao is related to the Tibetan languages where ergative case marking is said to be optional and often used to disambiguate sentences. It seems likely that the morphological spellout of the ergative case feature in Folopa and Mongsen Ao is optional with respect to the grammar, and thus available for uses determined by the social context. Mongsen Ao appears to have another ergative split as well, one of the type described earlier for Nepali involving stage versus individual level predicates. We see this is the following minimal pair from Coupe (2008: 156). The first example, in (34), has a stage level predicate, one that refers to something that the subject is doing at the moment (chopping wood). The second example, in (35), has an individual level predicate that refers to a property of the subject, his occupation. Paralleling the pattern we saw in Nepali, only the example with an individual level predicate has ergative on the subject: (34) Nì a-​sɘ́ŋ sɘ-​ɘ̀ɹ. 1sg nrl-​wood chop-​pres ‘I’m chopping wood.’

9 

[Mongsen Ao] (Coupe 2008: 156 (5.20))

The gloss nrl abbreviates ‘non-​relational prefix,’ and the gloss dec abbreviates ‘declarative mood clitic.’

 

Split in syntax and at morphological spellout    217 (35) Nì nɘ a-​sɘ́ŋ sɘ-​ɘ̀ɹ. 1sg erg nrl-​wood chop-​pres ‘I chop wood.’ (i.e. habitually, as an occupation)

(Coupe 2008: 156 (5.21))

9.6  Are Active Languages Split Ergative? Active languages (often called active–​stative languages) have two classes of intransitive verbs whose case or agreement marking patterns differ. The active pattern is usually described as follows: verbs in one class case mark or cross-​reference intransitive subjects like transitive subjects, while verbs in the other class case mark or cross-​reference intransitive subjects like transitive objects. An example of an active language is Choctaw. Intransitive verbs in the more active class (verb that take an external argument/​initial 1/​initial subject) cross-​reference their subjects with what is labeled Series I in the Choctaw literature, as in (36). Intransitive verbs in the more stative class (which do not take an external argument/​initial 1/​initial subject) cross-​reference their subjects with Series II, as in (37): (36) Hilha-​li-​tok. dance-​1sg.seriesI-​past ‘I danced.’ (37) Sa-​niya-​h. 1sg.seriesII-​fat-​tns ‘I am fat.’

[Choctaw] (Davies 1986: 14 (1a))

(Broadwell 2006: 33 (9))

In most transitive verbs, Series I  cross-​references subjects, while Series II cross-​ references objects: (38) Chi-​pí-​li-​h. 2sg.seriesII-​see-​1sg.seriesI-​tense ‘I see you.’

[Choctaw] (Broadwell 2006: 45 (71))

However, there are a few transitive verbs which do not take an external argument, and these verbs cross-​reference both subject and object with Series II: (39) Chi-​sa-​yimmi-​h. 2sg.seriesII-​1sg.seriesII-​believe-​pred ‘I believe you.’

[Choctaw] (Davies 1986: 77 (30))

 

218   Ellen Woolford The generalization is that external arguments are cross-​referenced with Series I  and internal arguments are cross-​referenced with Series II. The case marking pattern in Laz is also classified typologically as active (e.g. Dixon 1994; Song 2001), although here the pattern involves case rather than cross-​referencing. In Laz, one class of verbs marks all subjects with ergative case, even intransitives as in (40), while in the other verb class, intransitive subjects get nominative case, as in (41): (40) Bere-​k imgars. child-​erg 3sg.cry ‘The child cries.’

[Laz] (Harris 1985: 52 (38))

(41) Bere oxori-​s doskidu. child.nom  house-​dat  3sg.stay ‘The child stayed in the house.’

(Harris 1985: 53 (49))

There is disagreement in the literature as to whether active languages qualify as split ergative. As we will see in this section, this disagreement is due to a difference of opinion concerning the definition of split ergativity.

9.6.1 Dahlstrom 1983: Active Languages Are Not Split Ergative Dahlstrom (1983) argues that active languages are not split ergative. This conclusion is based on the view that a case or agreement pattern qualifies as split ergative only if the same verb manifests a nominative–​accusative pattern in one context, but an ergative pattern in another context. This disqualifies active languages because each verb class always marks subjects in the same way, regardless of the context.10

9.6.2 Dixon 1994: Active (Split S) Languages Are Split Ergative In contrast, Dixon (1994) argues that active languages are split ergative. To understand why Dixon disagrees with Dahlstom, we need to understand the details of the pattern-​ based definition of split ergative that Dixon is using: if a language has both nominative–​ accusative case patterns and ergative case patterns, it qualifies as split ergative in Dixon’s view. It is important here to note that Dixon’s strictly pattern-​based definition is 10  Some verbs appear to be members of both verb classes, but there is typically a difference in meaning involving control or intentionality. Languages with many such verbs are referred to as Fluid-​S in the typological literature, e.g. Dixon (1994).

 

Split in syntax and at morphological spellout    219 independent of the actual identity of the cases involved. Under that definition, an active language such as Laz does qualify as split ergative because the verb class that uses ergative case for both intransitive and transitive subjects has a nominative–​accusative case pattern in Dixon’s terms (because S and A are marked alike), while the verb class which use ergative case only for transitive subjects, has as an ergative case pattern (because S and O are marked alike):11 (42) Active ergative case patterns under Dixon’s definition of split ergative: Verb class 1:  intransitive: transitive: Verb class 2: intransitive:  transitive:

S-​ergative A-​ergative S-​nom/​abs  A-​ergative

[pattern: nominative–​accusative] O-​nom/​abs [pattern: ergative–​absolutive] O-​nom/​abs

However, Dixon still recognizes a difference between active languages and other split ergative languages: he considers active languages to be a distinct type of split ergative language which he calls Split S. Dixon’s motivation for using a purely pattern-​based definition of ergativity and split ergative is sound from a typological perspective, especially given that at the time this approach was developed, in the 1970s, syntactic theory was in its infancy and provided little or no guidance in determining the identity of different cases, especially ergative. Moreover, there was also considerable disagreement in the descriptive literature as to how to label cases. In that situation, it was entirely reasonable to conclude that the only reliable way to type case systems would be in terms of their patterns. Moreover, that kind of typology is relevant to a question frequently explored in the typological literature: how do different languages mark/​ differentiate subjects and objects? However, to answer the question of what syntactic and morphological conditions contribute to the distribution of ergative case, a question that I wish to explore, it is helpful if patterns described as ergative actually have ergative case. It is confusing if verbs or languages that mark all subjects with ergative case are coded as having a nominative–​accusative case pattern. It would be better if we had separate terms for patterns versus cases. For example, since a nominative–​accusative case pattern is simply one that marks all subjects alike, we could refer to this simply as a subject pattern or a subject–​ object pattern. Since ergative is a case, but absolutive is not (Legate 2008), we could refer to the ergative–​absolutive pattern as simply an absolutive pattern.

9.6.3 Two Types of Active Languages: Neither Is Split Ergative I agree with Dahlstrom (1983) that active languages do not qualify as split ergative. However, I argue that Choctaw and Laz are really two different types of languages. I reserve the label 11  These two verb classes are sometimes unfortunately labeled unergative and ergative/​unaccusative, which is very confusing since in active languages with ergative case, it is the unergative class that takes an ergative subject.

 

220   Ellen Woolford ‘active’ for languages such as Choctaw, and label languages such as Laz as active ergative (or simply ergative) for reasons which will become clear in the following.

9.6.3.1 Active Languages Under a strictly case-​based definition, a language cannot be split ergative unless it is ergative, and it is not ergative unless it uses ergative case. I argue that active languages such as Choctaw are not split ergative because they are not ergative. They do not use ergative case. The case pattern of Choctaw is nominative–​accusative (Broadwell 2006):12 (43) John-​at tákkon(-​a) chopa-​h. John-​nom  peach(-​acc)  bought-​tense ‘John bought a peach.’

[Choctaw] (Broadwell 2006: 39)

(44) Hattak-​at alla-​yã towa(-​yã) ĩ-​pila-​tok. man-​nom child-​acc ball(-​acc) appl-​throw-​past ‘The man threw the child the ball.’ 

(Davies 1986: 7)

I argue in Woolford (2010) that the difference between the Series I and Series II cross-​ referencing forms is that Series I is ordinary agreement, while Series II consists of nominative and accusative pronominal clitics, which are not morphologically distinguished by case. In Choctaw, ordinary agreement cross-​references external arguments, while a nominative or accusative clitic is used to cross-​reference internal arguments. Thus there is no disconnect between the case of arguments and the case of the cross-​referencing elements in Choctaw.

9.6.3.2 Active–​Ergative Languages: Fully Ergative In contrast to Choctaw, Laz uses ergative case and thus is an ergative language. Nevertheless, in my view, the Laz case pattern is not a split ergative pattern. Rather, Laz has the most fully ergative case pattern possible, where every verb that can license ergative case does so, regardless of the context, and regardless of whether or not there is an object in the clause. Another active ergative language is Warlpiri. Verbs in the verb class that can license ergative case always mark their subjects ergative, while verbs in the other verb class which cannot license ergative case never mark their subjects with ergative case. In (45) and (46), we see an example of each verb type, in roughly the same transitivity context, where an oblique object is present in the clause: (45) Ngajulu-​rlu ka-​ma-​rla jurlarda-​ku  me-​erg pres-​1sg-​3sg  honey-​dat ‘I am looking for honey.’ 12 

warni-​mi. [Warlpiri] look.for-​nonpast (Bittner and Hale 1996b: 561 (49a))

The accusative morpheme need not be spelled out in Choctaw when the object is adjacent to the verb (Broadwell 2006).

 

Split in syntax and at morphological spellout    221 (46) Ngarrka ka-​rla wangka-​mi kurdu-​ku. man.nom  pres.3sg-​3sg  speak-​nonpast  child-​dat ‘The man is speaking to the child.’ (Bittner and Hale 1996b: 561 (49b)) However, not even an oblique object is required in order for the subject to take ergative case in Warlpiri; verbs in the ergative case licensing verb class take an ergative subject even when no object is present: (47) Ngarrka-​ngku  ka yunpa-​rni. man-​erg pres  sing-​nonpast ‘The man is singing.’

[Warlpiri] (Hale 1982: 237 (30a))

Verbs that take an ergative subject have an external argument (called an initial 1 or an initial subject in other frameworks), and such verbs are classified as transitive under the Hale and Keyser (1993) definition, where all verbs that take an external argument are classified as transitive (whether or not an object is present in syntax, under their assumption that such verbs always have an object at the argument structure level). The connection between external arguments and ergative case is expected under the view that ergative is an inherent case like dative, licensed to external arguments in connection with theta marking (see Woolford 2006 and references cited therein). However, it is important to note that the class of verbs that takes an ergative subject/​external argument varies from language to language, and is not entirely predictable from the semantics of the verb.

9.7  Object Shift Split Ergative In this section, we turn to a type of ergative split which is mentioned in Silverstein (1976: 125) but seldom found in other surveys of types of split ergativity. In this type, the case marking of the subject depends on the features of the object. Bittner and Hale (1996b) contrast Warlpiri, where all verbs with an external argument mark their subjects with ergative case, with Inuit, where the subject is ergative only when a [+specific] direct object is present in the clause. Based on evidence in Bittner (1994), they argue that specific objects move out of the VP in Inuit, but non-​specific objects remain in situ inside the VP. Constructions with a specific object have an ergative subject, as in (48), while constructions with the same verb, but a non-​specific object have a nominative subject, as in (49): (48) Juuna-​p miiqqa-​t paar(i-​v)-​a-​i. [Inuit] Juuna-​erg  child-​pl.nom  look.after-​ind-​[+tr]-​3sg.3pl ‘Juuna is looking after the children.’ (specific object)(Bittner and Hale 1996b: 544 (22b))

 

222   Ellen Woolford (49) Juuna atuakka-​nik marlun-​nik.  pi-​si-​v-​u-​q Juuna.nom  book-​pl.ins  two-​pl.ins paid.for-​get-​ind-​[-​tr]-​3sg ‘Juuna bought two books.’ (non-​specific object) (Bittner 1994: 72 (45b)) This type of movement where the object moves out of the VP to a position below the subject is often called object shift; it is well known from work on Icelandic and other Germanic languages (e.g. Holmberg 1986; Vikner 1991; and Diesing 1996). Other languages where the subject gets ergative case only if object shift occurs include Nez Perce (Woolford 1997; Deal 2013)  and Niuean (Massam 2000). These two languages differ from Inuit in that both the shifted and unshifted objects are clearly direct objects. We see this in the Nez Perce example: (50) Háama-​nm  pée-​'wi-​ye wewúkiye-​ne. man-​erg 3.3-​shoot-​asp  elk-​obj ‘The man shot an elk.’ (topical object) (51)

Háama  hi-​'wí-​ye wewúkiye. man 3-​shoot-​asp  elk ‘The man shot an elk.’ (non-​topical object)

[Nez Perce] (Rude 1988: 552 (30))

(Rude 1988: 552 (31))

Niuean provides word order evidence that the object has moved out of the VP in ergative subject constructions. Both constructions have VP fronting, but in the VSO version in (52), the specific object moves before the VP fronts, while in the VOS version in (53), the non-​specific object remains inside the VP and fronts with it (Massam 2000): (52) Ne inu e Sione e kofe. past  drink  erg  Sione  nom  coffee ‘Sione drank the coffee.’ (specific object) (53)

Ne inu kofe a Sione. past  drink  coffee  nom  Sione ‘Sione drank coffee.’ (nonspecific object)

[VSO Niuean]13   (Massam 2000: 98 (2a)) [VOS [Niuean]]   (Massam 2000: 98 (2b))

Although it is true that the presence of an ergative subject correlates with a feature of the object in this kind of ergative split, as Silverstein (1976) states, the relationship between the object features and the subject features is indirect: the features of the object determine whether or not it will undergo object shift, and in turn, moving the object to a position outside the VP results in the use of ergative case on the subject. There are currently 13 

The ergative case morpheme for proper nouns in Niuean, e, happens to look just like the nominative/​ absolutive case morpheme for common nouns, e.

 

Split in syntax and at morphological spellout    223 three hypotheses in the literature as to why ergative case is only used when the object moves out of the VP in this type of language: Bittner and Hale (1996b) propose that in this type of language, VP is a barrier, and that the object has to move across that barrier in order to be in the same domain as the subject, so as to serve as a case competitor for the subject, so as to allow/​force ergative case to be licensed on the subject. Deal (2013) proposes that in Nez Perce, moving the object out of the VP causes the agreement to be portmanteau, which in turn allows/​causes the subject to be ergative. Woolford (2015) proposes that moving the object above the base position of the subject/​external argument interferes with the ability of Infl/​T to probe past it to get to the base position of the external argument to assign it nominative case. Using ergative case on the subject instead, as a last resort, avoids this locality violation. Since ergative case is used only when object shift occurs in this type of ergative split, I refer to this as an object shift ergative split.

9.8  Proximate Obviate Ergative Case Split In this last section, we look at another type of ergative case split that is not often included in typological surveys of split ergativity. This type of split is based on the proximate/​ obviate distinction.14 This proximate/​obviate ergative split occurs in Sahaptin. Sahaptin clearly qualifies as a split ergative language in that subjects of some transitive clauses have nominative case, while subjects of other transitive clauses have ergative case. This ergative split is governed by whether the subject is proximate or obviate: proximate subjects are nominative as in (54), while obviate subjects are ergative as in (55): (54) Ɨwínš i-​q̓ínun-​a wapaanƚá-​n. man.nom  3-​see-​past  grizzly-​obj ‘The man (proximate) saw a grizzly (obviate).’

[Sahaptin] (Rigsby and Rude 1996: 677 (21))

(55) Ɨwínš-​in pá-​tuχnana  yáamaš-​na. man-​obv.erg  3.3-​shot mule.deer-​obj ‘The man (obviate) shot a mule deer (proximate).’ (Rigsby and Rude 1996: 676 (13)) Now, languages generally only mark the proximate/​obviate distinction in clauses with two third person arguments. However, ergative case can mark the subject when the 14  The proximate argument is an established referent that is essentially the main character/​focus of empathy in a narrative or span of discourse. An obviate argument is some other character.

 

224   Ellen Woolford object is first or second person. In this situation, Sahaptin uses different forms of the ergative case morpheme, one with and one without the feature [+obviate]: (56) Ɨwínš-​nɨm=nam  i-​q̓ínu-​ša. man-​erg=2sgcl  3-​see-​imperf ‘The man sees you.’

[Sahaptin] (Rigsby and Rude 1996: 677 (22))

Thus one might say that Sahaptin qualifies as split ergative in a second, unusual sense as well: its ergative case morpheme is split into two.

9.9 Conclusion This chapter has described a range of types of ergative splits, including the well-​ known person and aspect splits, as well as several less well-​known types. I  have sided here with Dahlstrom’s (1993) view that active languages are not split ergative; however, I have divided active languages into two very different types: active languages, which are not ergative at all, and active ergative languages such as Laz and Warlpiri that are ergative. I have taken the unusual position of defining the ergative pattern in active ergative languages as the only fully ergative case pattern there is, based on the following definition of a fully ergative case pattern: the case pattern of a language is fully ergative if all verbs that can license ergative case do so in all contexts and regardless of transitivity. I support Silverstein’s (1976) view that there is a type of split ergative language wherein the ergativity of the subject depends on the features of the object, and illustrate this type with Inuit, Nez Perce, and Niuean. Following Massam (2000) and Deal (2013), I argue that this type of split is actually based on whether or not object shift occurs, which is in turn determined by the features of the object. This chapter has included discussion of several controversies as to whether particular types of patterns should count as instances of split ergativity, and has pointed out differences in the definition of split ergativity that are the source of these differences of opinion. The chapter has also included a discussion of various theoretical proposals in the literature to account for different types of split ergativity. A major theme of this chapter is determining whether the various types of ergative splits are present in syntax or are purely morphological. Most of the splits discussed here appear to be present in syntax, with the notable exception of person/​animacy/​NP splits, which are occur after syntax, at the stage of morphological spellout of the features from syntax.

 

Split in syntax and at morphological spellout    225

Abbreviations ASP, aspect; DAT, dative; DEC, declarative; DET, determiner; ERG, ergative; FEM, feminine; HON, honorific; IMPF, imperfective; IND, indicative; INTR, intransitive; LOC, locative; MASC, masculine; NEUT, neuter; NML, nominal; NOM, nominative; NRL, non-relational; OBJ, object; OBV, obviate; PAST, past tense; PERF, perfective; PL, plural; PRED, predicate; PROG, progressive; SG, singular; TR, transitive; VOC, vocative.

 

Chapter 10

Split ergati v i t y i s not a b ou t erg at i v i t y Jessica Coon and Omer Preminger

10.1 Introduction It has been frequently noted in the literature on ergativity that few—​if any—​ergative systems are purely ergative. Rather, many ergative languages exhibit a phenomenon known as “split ergativity” in which the ergative pattern is lost in certain parts of the grammar. No ergative language is fully consistent in carrying through the ergative principle throughout its entire morphology, syntax, and lexicon:  all languages that exhibit ergative patterning in their commonest case-​marking system also exhibit some accusative pattern somewhere in the rest of their grammar. [Moravcsik 1978b: 237]

The two most commonly described types of split ergativity are (i) aspectual splits, and (ii) person splits. (See also the contributions by Laka, Chapter 7, Nash, Chapter 8, Woolford, Chapter 9, this volume.) In the former, the ergative pattern is lost in some subset of non-​perfective aspects (or possibly certain tenses; see Coon 2013b for discussion); and in the latter, the ergative pattern is lost with some particular combination of “highly-​ranked” nominal arguments (we return to the details of such “rankings”). The central argument put forth in this chapter is that split ergativity is epiphenomenal, and that the factors which trigger the appearance of such splits are not limited to ergative systems in the first place. In both aspectual and person splits, we argue, the split is the result of a bifurcation of the clause into two distinct case/​agreement domains; this bifurcation results in the subject being, in structural terms, an intransitive subject.1 Since intransitive subjects do not appear with ergative marking, this straightforwardly accounts 1  For other structural accounts of ergative splits, see, among others, Laka (2006a, Chapter 7, this volume) and Nash (Chapter 8, this volume).

 

Split ergativity is not about ergativity    227 for the absence of ergative morphology in those cases. But crucially, such bifurcation is not specific to ergative-​patterning languages; rather, it is obfuscated in nominative-​ accusative environments because—​by definition—​transitive and intransitive subjects pattern alike in those environments, and the terminology in question (‘ergative’ vs. ‘non-​ ergative’) specifically tracks the behavior of subjects. Thus, Moravcsik’s generalization, as quoted, does not reflect any deep instability of ergative systems, nor a real asymmetry between ergativity and accusativity (contra Visser 2006, for example). In an ergative system that exhibits this type of split, ergative-​absolutive alignment is always associated with a fixed set of substantive values (e.g. perfective for aspectual splits, 3rd person for person splits). The account we will present derives this universal directionality of splits by connecting the addition of extra structure to independently attested facts:  the use of locative constructions in progressive and non-​perfective aspects (Bybee, Perkins, & Pagliuca 1994; Laka 2006a; Coon 2013a), and the requirement that 1st and 2nd person arguments be structurally licensed (Béjar & Rezac 2003, Preminger 2014). The remainder of this chapter is organized as follows. Section 10.2 briefly reviews ergative and split ergative patterns. Section 10.3 focuses on aspectual splits, summarizing the different types of split patterns that emerge, and presenting a structural motivation for the loss of ergativity in certain aspects. In section 10.4 we turn to person splits, and present a structural account of this phenomenon, as well. Section 10.5 concludes.

10.2  Ergativity and Split Ergativity Typical ergative-​ absolutive and nominative-​ accusative alignment systems are represented in (1)  and (2), where we follow Dixon (1979) in using the following labels: A = transitive subject; P = transitive object; and S = intransitive subject. (1) ERGATIVE– ABSOLUTIVE SYSTEM

transitive:

intransitive:

(2) NOMINATIVE– ACCUSATIVE SYSTEM

A

P

A

P

ERG

ABS

NOM

ACC

S

S

ABS

NOM

 

228    Jessica Coon and Omer Preminger As evidenced by the range of work in the present volume, accounts of ergative case and agreement patterns within the generative tradition have been numerous, and we do not aim to adjudicate among them here. In light of this wide array of approaches, we choose not to commit to any one particular analysis of ergativity itself, and instead aim to show how a structure-​based account of split ergativity can account for non-​ergative patterning in an otherwise ergative system, regardless of the specific theory of ergative case assignment adopted. Throughout this chapter, we will speak of ergative patterns, rather than ergative languages. As noted at the outset, a given language rarely shows a consistent ergative pattern of alignment throughout its entire grammar (see Silverstein 1976; Comrie 1978; Moravcsik 1978b; DeLancey 1981; Tsunoda 1981b for discussion; as well as numerous examples). As observed by Dixon (1994), the two main factors that one finds as triggers of split ergativity cross-​linguistically are aspectual splits (discussed in section 10.3), and person splits (discussed in section 10.4). We take it to be a relevant and crucial fact that both in aspect-​based and in person-​based splits, one finds a universally fixed directionality (Dixon 1994, inter alia). What this means is that for each split-​triggering substantive category, the ergative-​absolutive alignment is always associated with a fixed value (or set of values) of that substantive category (e.g. perfective for aspectual splits, 3rd person for person splits). This fixed directionality is schematized for aspectual splits in (3) and for person splits in (4):2 (3) fixed directionality of aspectual splits ←  ergative  non-​ergative      → ≫

perfective 

imperfective 

(4) fixed directionality of person splits ←  ergative common nous

≫

≫      progressive

non-​ergative  

proper ≫ demonstratives, nouns 3rd person pronouns

≫

→ 1st/​2nd person pronouns

In this chapter, developing the proposal in Coon & Preminger 2012, we offer a unified account of aspect-​based and person-​based split ergativity, which captures the universal directionality across these two substantive categories. Crucially, the proposed account reduces both types of splits to structural factors that are not specific to ergative languages in particular—​while explaining why, in a nominative-​accusative language, these same factors would not result in what one would characterize as an “alignment split.” 2  While person splits are commonly described as involving this type of scale, we demonstrate in section 10.4 that the relevant cut may in fact always be between 1st/​2nd person pronouns and all other nouns. We discuss factors that have obscured this distinction, particularly in the language Dyirbal.

 

Split ergativity is not about ergativity    229

10.3 Aspectual Splits This section provides a brief synopsis of aspectual splits, discussed in much greater detail in Coon 2013a,b.3 As (3)  illustrates, different languages may make the split in different places along the aspectual scale, but ergativity crucially remains consistently anchored to the left side of this scale (i.e., to the perfective).4 Coon (2013b) argues that aspectual splits have two underlying causes: (i) the introduction of complex syntactic structure associated with non-​perfective aspects (discussed in section 10.3.1); and (ii) reduced transitivity, manifested as demotion of the direct object, triggered in non-​ perfective aspects (discussed in section 10.3.2). In both scenarios, the transitive subject of a “split”-​patterning non-​perfective construction will not receive ergative marking because it is, structurally speaking, no longer a transitive subject (where being a transitive subject means being the higher of two non-​oblique noun phrases in a single, non-​ bifurcated clause). Furthermore, as we will show, the factors responsible for these splits are not limited to ergative-​patterning languages. We therefore need not think of split ergativity as a property that is particular to “ergative languages” to the exclusion of other languages. Rather, when the structural conditions for a split arise, the result is that transitive subjects pattern with intransitive subjects; this goes unnoticed in a non-​ergative system because, in these systems, all subjects receive the same marking (i.e. nominative).

10.3.1 Added Structure In Basque, one finds an aspectual split that is demonstrably triggered by the addition of syntactic structure, in this case in the progressive aspect (Laka 2006a, Chapter  7, this volume). In the perfective and imperfective aspects, Basque shows an ergative-​ absolutive alignment in the forms of the suffixal article. Taking singular noun phrases as an illustration, A arguments take the article -​ak (ergative sg.), while P and S arguments take the article -​a (absolutive sg.): (5) Basque —​perfective a.  [A Ehiztari-​ak ] [P otso-​a ] harrapatu  d-​ϕ-​u-​ϕ.   hunter-​artsg.erg     wolf-​artsg(abs)  caught 3.abs-​sg.abs-​​ ‘The hunter has caught the wolf.’

AUX -3sg.erg

3  Though these are sometimes called TAM (i.e. tense, aspect, mood) splits, Salanova (2007) and Coon (2013b) call into question the existence of tense-​or mood-​based splits that do not also involve aspect or clause-​type. In what follows, we assume that aspect is the only true trigger of this type of split—​though in practice, of course, perfective aspect frequently overlaps with past tense. 4  Note that since progressive aspect is a subtype of imperfective aspect, the hierarchy in (3) is predictable in another way: the split can either target just the progressive, or also the imperfective, which properly contains the progressive; see Coon 2013a for discussion.

 

230    Jessica Coon and Omer Preminger b. [S Otso-​a ] etorrid-​a.  wolf-​ artsg(abs)    arrived 3.abs‘The wolf has arrived.’

AUX

(3sg.abs)​

[Laka 1996]

In the progressive, however, the ergative marking is lost. Here all singular arguments take the -​a suffix, as shown in (6). Furthermore, while the auxiliary agrees with both the subject and the object in the ergative-​patterning perfective (as in (5)), only the subject is agreed with in the progressive—​even if the lexical verb is transitive, as in (6a):5 (6)

Basque —​progressive a.  [A emakume-​a ] [P ogi-​a ] ja-​te-​n ari  d-​a.     woman-​artsg(abs)     bread-​artsg(abs)  eat-​ nml-​loc prog 3.abs- AUX (3sg.abs) ‘The woman is eating the bread.’

b. [S emakume-​a ] dantza-​n ari d-​a.        woman-​artsg(abs)   dance-​ loc  prog  3.abs‘The woman is dancing.’

AUX ​(3sg.abs)

[Laka 1996]

Note that the split in the Basque progressive crucially does not involve the language switching from an ergative-​absolutive to a nominative-​accusative pattern, though this is frequently how such splits are informally described. While it is the case that both subjects pattern alike in (6), the pattern seen here is more accurately described as “neutral”: all core arguments are in the unmarked absolutive form, including the transitive object (ogi-​a “bread-​artsg(abs)”). Similar patterns are found in Nakh-​Daghestanian and Indo-​Aryan languages (see Coon 2013b). Laka (2006a) argues that these “split” forms in the Basque progressive are in fact bi-​clausal, involving a progressive auxiliary (ari; see (6a–​b)) which embeds a locative-​ marked subordinate clause containing the lexical verb and its object. This proposal accounts elegantly for the properties shown above: the A argument is not marked ergative, because it is not a transitive subject; it is the subject of an auxiliary whose complement is something other than a non-​oblique nominal—​and therefore, this auxiliary is formally intransitive (in Basque, the complement of this subordinator is oblique, as reflected in the glosses of (6a–​b)). The P argument does not trigger agreement because it is in a separate, lower clause. Similar proposals have been made for Nakh-​Daghestanian languages; see Kazenin 1998, 2001c; Kazenin & Testelec 1999; Forker 2010; Gagliardi et al. 2014.6 This analysis is advanced further in Coon 2010b, 2013a, where it is argued 5  This is somewhat of a simplification. As shown by Arregi & Nevins (2008, 2012) and Preminger (2009), so-​called “subject agreement” (or “ergative agreement”) and “indirect object agreement” (or “dative agreement”) in Basque are actually instances of obligatory clitic doubling. 6  The proposal that non-​perfective aspects may involve added structure also provides an explanation for the behavior of aspectual splits in the Mayan family, for example in Yucatec (Bricker 1981), in Chol (Vázquez Álvarez 2002; Coon 2010b), and in Q’anjob’al (Mateo-Toledo 2003; Mateo Pedro 2009); see

 

Split ergativity is not about ergativity    231 that such clausal bifurcation is always found in non-​perfective aspects rather than in the perfective, because it is precisely non-​perfective aspects whose structure is constructed using spatial/​locative building blocks (see also Bybee, Perkins & Pagliuca 1994; Demirdache & Uribe-​Etxebarria 2000). There are two important consequences of this analysis. First, it provides a natural explanation for why no such “splits” have been observed in nominative-​accusative languages. In many nominative-​accusative languages, the progressive, and sometimes the imperfective, are built on locative constructions—​just as they are in Basque. An example is given in (7). (7) Ik ben  het  huis aan  het I am the house at the ‘I am building the house.’

bouwen.  (Dutch) build [Demirdache & Uribe-​Etxebarria 2000: 178]

The difference is that in a nominative-​accusative system, this insulation of the subject for case purposes has no effect on the marking of subjects: A and S arguments in a nominative-​accusative system receive the same marking (nominative) regardless of whether or not a direct object is syntactically accessible. Second, examples like (7), as well as their myriad cross-​linguistic counterparts (Bybee, Perkins & Pagliuca 1994), illustrate that independently of split ergativity—​or even ergativity more generally—​it is typically the non-​perfective aspects that are built using elements of locative morphosyntax not found in the perfective. In conjunction with the clausal bifurcation analysis presented here, this provides an explanation for the cross-​ linguistically fixed directionality of aspectual splits (recall (3)): given a language that is underlyingly ergative, the perfective—​in which no structure is added to the basic clausal skeleton—​will reveal the underlying ergative alignment. But the structure added in a non-​perfective aspect could, if it bifurcates the clause into two separate case/​agreement domains, result in what looks like a “shift” out of this underlyingly ergative pattern. Importantly, nothing said so far dictates that the structure added in a non-​perfective aspect must bifurcate the clause in this manner. It is conceivable that the relevant locative elements used to construct non-​perfective aspects would be syntactically opaque (e.g. phasal) in one language, but syntactically transparent in another. The parametric choice between these two options would yield, accordingly, the distinction between “split-​ergative” and “consistently ergative”—​now conceived of as an epiphenomenal, descriptive distinction.

10.3.2 Reduced Transitivity Like the languages surveyed in section 10.3.1, Samoan (Polynesian) has been described as a language with aspect-​based split ergativity. The pattern in Samoan, however, looks Larsen & Norman 1979; and Dayley 1981 for overviews. Splits in these languages involve an “extended ergative” pattern, not discussed here for reasons of space, but see Coon 2013b for an overview.

 

232    Jessica Coon and Omer Preminger slightly different than what we have seen so far. The basic ergative pattern is shown in (8a–​b), where the A argument in (8a) appears with the ergative case marker e; absolutive arguments, like the subject in (8b), are unmarked. (8) Samoan (Polynesian) a. Na fasi [A e le tama]  [P Sina]. pst  hit    erg  det  boy  Sina ‘The boy hit Sina.’ b. ’ olo ’ o moe [S le tama]. pres.prog  sleep      det  boy ‘The boy is sleeping.’

[Ochs 1988: 89]

Samoan exhibits a split between the perfective and imperfective aspects, as illustrated by the pair of transitives in (9a–​b). Just as we saw for Basque in section 10.3.1, the ergative marking on the A argument in (9b) is lost in the “split” pattern, and replaced with (unmarked) absolutive. However, while in the languages from section 10.3.1 we saw absolutive marking retained on the P argument (resulting in what we described as a ‘neutral’ pattern, with all core arguments marked absolutive), here the P argument takes an oblique suffix. We will refer to this as an ‘abs-​obl’ pattern. (9) a.  perfective na va ’ ai-​a [A e le tama]  [P le i’a] pst  look.at-​prfv      erg the  boy     the Fish ‘The boy spotted the fish.’ b. imperfective na va’ai [A le tama] [P i le i’a] pst look.at  the boy     obl the fish ‘The boy looked at the fish.’

[Milner 1973]

In the linguistic literature on Polynesian, forms like those in (9a) are known as “ergative,” while those in (9b) are labeled “objective” (Milner 1973) or “middle” (Chung 1978).7 But despite the term “objective,” the object in the imperfective is marked with i, whose function throughout Polynesian is that of an oblique marker (Chung 1978: 26). The same essential pattern is found in Warrungu (Pama-​Nyungan), and a related pattern is found in Adyghe (NW Caucasian), also discussed in Tsunoda 1981b

7 

Though the distinction between forms like (9a) and (9b) has previously been treated as a voice contrast, Milner (1973) argues that the distinction is aspectual, and that the English translations of some pairs are often best captured by using distinct lexical items—​e.g. spotted vs. looked at—​one which emphasizes the “totality” of the action (i.e., perfective), and the other which focuses on “the action itself ” (i.e., imperfective); see Milner (1973: 631).

 

Split ergativity is not about ergativity    233 (citing Anderson 1976). Here, however, the variation (shown in (10a) and (10b)) is not per-​aspect, but per-​lexeme: -š’a- (“lead”) follows an ergative pattern, while -ža(“wait”) follows an abs-​obl pattern (certain verbs in Adyghe can appear in both constructions). (10) Adyghe (NW Caucasian) a. [A a-š’ ] [P a-χe-r ] ϕ -ә-š’a-ɣe-χ   he-​ erg       he-pl-abs-​   o3sg-s3sg-lead-prfv-pl ‘He led them.’ b.  [A a-χe-​r ] [P a-š’ ] ϕ -je-ža-ɣe-χ  he-pl-abs    he-obl   s3-io3sg-wait-prfv-pl  ‘They waited for him.’ [Kumakhov & Vamling 2009: 94–96] In order to account for patterns like this, Tsunoda (1981b) proposes an Effectiveness Condition (‘EF-​CON’)—​which appeals to notions like “effectiveness,” “conclusiveness,” “definiteness,” “actualness,” and several others—​meant to account both for aspectual splits like the one in (9), and lexeme-​based splits as in the Adyghe data in (10a–​b). (A similar proposal was developed independently by Hopper & Thompson 1980; see Malchukov 2005 for a recent survey.) The idea behind EF-​CON is that failures to meet some portion of the factors associated with high transitivity have a common morpho-​syntactic consequence—​namely, the P argument surfacing as oblique—​regardless of whether these factors are related to grammatical aspect, or alternatively, tied to the specific meaning of a given verb. Tsunoda concludes: “Verb-​split and TAM-​split are fundamentally no different from each other, their semantics and case-​marking mechanisms involving common principles” (Tsunoda 1981b: 391). In non-​perfective aspects, objects are generally more likely to be indefinite, non-​referential, and less affected; in the perfective the focus is on the culmination of the event, and objects are more likely to be affected. We might also add to this group “antipassive” constructions, described in a number of ergative languages (including Iñupiaq, Chukchi, Salish, and Dyirbal; see for example Heath 1976 and Spreng 2010).8 Generally speaking, antipassive constructions have the effect of demoting the notional P argument, by relegating it to an oblique, by incorporating it into the verb, or by omitting it altogether. Crucially, this triggers a concomitant change in marking of the A argument from ergative to absolutive (see Polinsky, Chapter 13, this volume, for a review). In the domain of grammatical aspect, antipassives are typically found in imperfective or “unbounded” aspects; the internal argument of an antipassive is often described as non-​specific and/​or indefinite. Formally, then, these constructions appear strikingly similar to the aspectual and “verb-​type” splits reviewed above. Indeed, Spreng (2010: 563) notes: “If we review the triggers for ergativity splits across languages, we find that the Antipassive occurs under some of the same conditions.” 8  While far more commonly described in languages that exhibit an ergative argument alignment, the antipassive construction is not exclusive to such languages (see, e.g., Dryer & Haspelmath 2013).

 

234    Jessica Coon and Omer Preminger While the abs-​obl aspectual splits, “verb-​type” splits, and antipassive constructions discussed in this section are most commonly noted in so-​called “ergative languages,” Tsunoda (1981b) notes that the general phenomenon of object demotion in less affected environments is not limited to languages with ergative alignments. Consider the English conative alternation: in (11a), the bear is assigned a direct case (presumably, accusative/​ objective), and is clearly an affected argument; in (11b), the Patient is expressed as a PP, and there is no requirement that the act of shooting be successful—​that is, the bear may be totally unaffected by the event. In English, of course, subjects carry unmarked nominative regardless of whether they are transitive subjects or intransitive subjects; but if English were ergative, we would expect to find ergative on the (transitive) subject in (11a), but absolutive on the (intransitive) subject in (11b). See Kiparsky 1998a for a related discussion of partitive alternations that arise with Finnish objects. (11)

a. Sam shot [P the bear]. b.  Sam shot [P at the bear].

A survey of the various proposals put forth to account for conative and related alternations is beyond the scope of this chapter (see, e.g., Levin 1993; Kiparsky 1998a; Borer 2005b). Let us assume that some account of these alternation is in place. That means we have a way of predicting that loss of affectedness (or some closely related notion) on the part of the P argument will correlate with oblique marking on that argument, as it does in (11a–​b). Crucially, such oblique marking creates a very similar state of affairs, syntactically, to what we saw in section 10.3.1: in the Basque progressive, for example, the A argument was the subject of a syntactically intransitive aspectual auxiliary, because that auxiliary selected a PP as its complement. Here, in cases of reduced affectedness, it is the main lexical verb that selects a PP complement. The absence of ergative marking on the subject of these split constructions, then, simply reflects the fact that the A argument is no longer a transitive subject (in the sense that it no longer has a non-​oblique clausemate object)—​as was the case in the “split” patterns surveyed in section 10.3.1. In other words, the pattern shown here is triggered by the very same syntactic factor that triggered the aspectual splits: a PP layer separating the subject from the object.9

10.3.3 Summary In this section, we have examined several types of aspectual splits. Importantly, it is not clear that any of these patterns really instantiate a split between ergative-absolutive and 9 

In recent work, Kalin & van Urk (2015) argue that a complex pattern of “agreement reversal” in certain Neo-​Aramaic languages can also be captured through the addition of structure precisely in non-​perfective aspects. Crucially, neither of the Neo-​Aramaic patterns in question is characterized as “ergative”—​lending further support to the proposal here that the structural factors which trigger the appearance of “split ergativity” are independent of ergativity or an ergative argument alignment.

 

Split ergativity is not about ergativity    235 nominative-​accusative per se; that verdict might ultimately be a matter of terminology (e.g. whether one wants to call the marking on the P argument in non-​ergative alignments ‘accusative’ even though it is identical to a marker that serves as oblique elsewhere in the language). The splits share the following two properties. First, the split is not the result of special rules of case assignment or agreement, active only in certain aspects and not others; case-​assigning functional heads in different aspects need not bear different case/​agreement features (cf. Anand & Nevins’ 2006 analysis of Hindi, as well Ura 2006). Second, the underlying mechanisms responsible for these splits are not a special property of ergative-​patterning languages. The phenomena above are also found in predominantly nominative-​accusative languages (for example, the progressive construction in Dutch and the English conative alternation). The difference between, e.g., Samoan on the one hand and, e.g., English on the other, is that—​by definition—​transitive and intransitive subjects are marked alike in a nominative-​accusative system, making it impossible to see what would otherwise be a split in subject marking.

10.4 Person Splits In this section, we turn to person-​or NP-​based split ergativity. This refers to systems that exhibit a non-​ergative alignment when certain types of arguments are involved (e.g. 1st/​2nd person pronouns), but exhibit an ergative alignment otherwise. There are important parallels between person-​based split ergativity and Differential Argument-​ Marking—​which we use, for now, as a cover-​term for both Differential Subject Marking (DSM) and Differential Object Marking (DOM); see Malchukov, Chapter 11, this volume. This refers to the relatively well-​studied phenomenon whereby a certain class of arguments (e.g. definites, proper nouns), when they occur in a particular structural position (e.g. direct object), bear a case-​marking that is atypical for arguments in that position (e.g. dative, instead of accusative). We believe that these parallels are significant and should be considered carefully. We therefore turn first to Differential Argument Marking, and outline a particular approach to this phenomenon and a proposal for how it arises. This proposal will then be related back to the structure-​based account of split ergativity put forth in section 10.3, above.

10.4.1 Differential Argument Marking as a Configurational Phenomenon We start with a basic observation drawn from the work of Baker & Vinokurova (2010) on Sakha (Turkic): that DOM—​at least in this language—​is determined configurationally. In Sakha, the presence of overt accusative marking on the direct object covaries with specificity: specific objects bear overt accusative marking, non-​specific ones

 

236    Jessica Coon and Omer Preminger do not.10 So far, this characterization would fit nicely within the foregoing description of ‘Differential Argument Marking’ (in particular, DOM). But Sakha also exhibits a correlation—​familiar from other languages in the Turkic family and beyond—​between specificity and the structural position of the object. In effect, what one finds, at least in simple cases, is a three-​way correlation between position, specificity, and (overt) case-​ marking: the specific object in (12a) bears accusative case marking and appears outside of the VP (as demarcated by the adverb türgennik “quickly”), while the non-​specific object in (12b) has no accusative marking and remains VP-​internal. (12)

a.  Masha  salamaat1-​*(y) [VP türgennik  t1 sie-​te ]. Masha porridge-​*(acc)      quickly    eat-​past.3sg.subj ‘Masha ate the porridge quickly.’

(Sakha)

b. Masha   [VP türgennik salamaat-​(#y) sie-​te ]. Masha quickly porridge-​(#acc)  eat-​past.3sg.subj ‘Masha ate porridge quickly.’ [B&V: 602; annotations added] This three-​ way correlation between specificity, structural position, and case-​ marking begs the question of cause and effect: is one of these three factors the underlying cause, from which the other two stem? We take short object movement of the kind shown in (12a) to be an operation that obligatorily applies to all specific noun phrases within the VP; and whose successful culmination depends on whether the noun phrase in question is in a position from which vacating the VP is possible (see Preminger 2011a, 2014 for discussion, building on Diesing 1992, 1997, Vikner 1997, and others). In this sense, the morpho-​semantic feature of specificity is the underlying cause for object movement. Turning to the morphological side, the reason DOM manifests itself as case morphology is because case is assigned configurationally—​and therefore, case depends on the positions of different noun phrases in the clause; see also Baker (Chapter 31, this volume) and Baker & Bobaljik (Chapter 5, this volume).11 When a (specific) object moves out of VP, it is in a local enough configuration with the subject to receive dependent case (viz. accusative), as shown in (13a). When the object remains inside VP, it receives a non-​specific interpretation, and is not in a local enough configuration with the subject to receive dependent case, as shown in (13b).

10 

An exception to this involves accusative-​marked objects that are contrastively focused; we will not discuss this here (see Baker & Vinokurova 2010: 602). 11  Baker & Vinokurova’s (2010) actual claim is that the facts of case in Sakha require a ‘hybrid’ theory of case, consisting of a configurational component and a probe-​goal component. However, Levin & Preminger (2015) have shown that this is incorrect, and that the facts regarding case in Sakha can be derived in a completely configurational manner, given certain independently motivated changes to Baker & Vinokurova’s theory of agreement.

 

Split ergativity is not about ergativity    237 (13)  a. Masha

porridge1 - ACC

[VP . . . t1 . . . ]

dependent acc b. Masha . . .  

[VP . . . porridge . . . ] dependent acc

The picture that emerges, concerning the causal relations between specificity, structural position, and case-​marking, is therefore the following:12 (14) Causal Relations in Differential Argument Marking morpho-semantic property (e.g.non-specific/specific)

structural configuration (e.g. object inside/outside VP)

case-marking (e.g.bare/ACC)

Extending this approach to ergative systems, let us suppose that just like accusative on Sakha objects, ergative on subjects (at least in those languages that show differential ergative marking) depends on a second DP occupying a position in the same case-​ domain (following Marantz 1991; see also Baker 2015). If something were to disrupt this sufficiently local configuration of two DPs, it would bleed the assignment of ergative—​ just like the assignment of accusative is bled in the Sakha (12b/​13b). In fact, it appears that precisely such a disruption is attested. As shown by Woolford (2008), Massam (2013), and Baker (2015), an object that remains inside VP can bleed the assignment of ergative to the subject of said VP. This is illustrated for Eastern Ostyak (Finno-​Ugric, Siberia) in (15). In (15a), the specific object appears outside of the VP and the subject is marked with ergative case. In (15b), in contrast, the object remains VP-​internal, and the subject appears in its nominative form. (15) a.  mə-​ŋən ləɣə1 [VP əllə  juɣ kanŋa t1 aməɣaloɣ       ].  (Eastern Ostyak) we-​erg  them       large tree  beside  put.past.3pl.obj/​1pl.subj ‘We put them (pots of berries) beside a big tree.’ b. mä [VP t’əkäjəɣlämnä ula mənɣäləm ]. we.dual(nom)     younger.sister.com berry pick.past.1pl.subj ‘I went to pick berries with my younger sister.’ [Gulya 1966, via Baker 2015; annotations added]

12 

See Merchant (2009) for a different (and predating) proposal—​building on Aissen (1999a, 2003) and implemented in a ‘cartographic’ fashion—​that nevertheless preserves the fundamental insight that position is the independent variable in Differential Argument Marking, whereas changes in interpretation and case-​marking are derivative.

 

238    Jessica Coon and Omer Preminger Here, failure of the object to move into a sufficiently local configuration with the subject results in a failure to assign dependent case to the latter—​since this is an ergative language. Crucially, though, this state of affairs is completely parallel to what we saw in Sakha, aside from the directionality of dependent case assignment; compare (13) with (16): (16)  a. We-ERG them1 [VP . . . t1 . . . ] dependent erg b. We . . . [VP . . . berry . . . ]

 

dependent erg

In a similar vein, Massam (2013) shows that in Niuean, specificity/​non-​specificity of the object has the same effect on the case-​marking of the subject (i.e. ergative vs. bare), even when the object in question is null—​all but ruling out an account of this subject case alternation in terms of (pseudo-​)noun-​incorporation of the object. Overall, these patterns are a rather powerful demonstration that Differential Argument Marking is fundamentally a configurational phenomenon. Changes in the case-​marking of a particular argument are not a response to changes in the semantic properties of that argument per se, except insofar as those semantic changes affect the argument’s structural position. Indeed, in the limiting case, changes in the semantic properties of a given argument can trigger changes in the case-​marking on a different argument (if the case-​marking on the latter depends on the structural relation between the two, as in (16)). A theory of Differential Argument Marking that attempts to derive the morphological marking of an argument directly from that argument’s own properties (e.g. Legate 2014a) thus fails to generalize to cases like (15a–​b). But a configurational theory, where the independent variable is not semantic type (specific/​non-​specific) but relative position, does generalize to such cases. Before concluding this subsection, it might be instructive to consider a question of terminology: would the difference between the Sakha (12a) (with a specific, accusative-​ marked object) and (12b) (with a non-​specific, unmarked object) have led anyone to characterize Sakha as a “split accusative” language? That is, to assert that in the presence of a non-​specific object, the language “shifts” out of its normal accusative alignment, and into a neutral alignment (where all core arguments are unmarked)? Note that similar alternations in subject marking routinely result in a language being classified as “split ergative”; this terminological bias toward focusing on the marking of subjects, we contend, is behind the impression that ergative languages—​but not nominative–​accusative ones—​are where “splits” tend to arise (see the discussion in section 10.1). We return to this terminological bias in section 10.4.7

 

Split ergativity is not about ergativity    239

10.4.2 “Binary Split” Systems and the Prospects of Unification with Differential Argument Marking In light of the results presented in section 10.4.1, it is obviously intriguing to explore the possibility of unifying person-​based split ergativity with Differential Argument Marking, under a configurational approach to both. However, there is one type of person-​based split ergativity, known as a “binary split” system, which seems to pose a serious challenge to such a unification. Dyirbal is an example of such a system. Focusing on clauses with lexical (i.e., non-​pronominal) arguments, we find an ergative alignment, as shown in (17): the A argument is marked with the suffix -​ŋgu, while P and S arguments are morphologically unmarked. (17) a.  transitive: lexical subject, lexical object [P ŋuma]  [A yabu-​ŋgu ] bura-​n     father   mother-​ erg  see-​nonfut ‘Mother saw father.’ b. intransitive: lexical subject [S ŋuma] miyanda-​nyu  father laugh-​nonfut ‘Father laughed.’

(Dyirbal)

[Dixon 1994: 161]

Turning to clauses which involve 1st and 2nd person pronominals, as in (18), we find the opposite pattern. Here, only the P argument receives morphological case, resulting in a nominative-​accusative alignment pattern. (18) a. transitive: 1st/​2nd person subject, 1st/​2nd person object [A ŋana] [P nyurra-​na] bura-​n  we   y’all-​ acc see-​nonfut ‘We saw y’all.’ b. intransitive: 1st/​2nd person subject [S ŋana] miyanda-​nyu  we laugh-​nonfut ‘We laughed.’

[Dixon 1994: 161]

Importantly, and in contrast with the dependent case configurations examined in section 10.4.1, the choice of marking on a given nominal appears independent of properties of other nominals in the clause. As shown in (19), unmarked subjects may coincide with unmarked object as in (19a); and marked subjects may coincide with marked objects, as in (19b).

 

240    Jessica Coon and Omer Preminger (19) transitive: mixed clauses a. 1st/​2nd person subject, lexical object [A ŋana] [P ŋuma] bura-​n  we    father see-​nonfut ‘We saw father.’ b. lexical subject, 1st/​2nd person object [P ŋana-​na] [A ŋuma-​ŋgu] bura-​n  we-​ acc     father-​erg see-​nonfut ‘Father saw us.’

[Dixon 1994: 130]

These facts are summarized in (20): (20) Dyirbal argument marking —​a summary 1st/​2nd person pronouns

other nominals

A

∅

-​ŋgu (‘erg’)

S

∅

∅

P

-​na (‘acc’)

∅

The crucial observations are as follows (Dixon 1994): ‘accusative’ and ‘ergative’ can co-​occur, as in (19b); but neither depends on the presence of the other, as shown in (17a) and (18a). It is therefore tempting to view this as two separate systems—​a subject system and an object system—​both of which are sensitive to interpretation (1st/​2nd person vs. 3rd person), but each of which operates independently of the other. This type of view is schematized in (21) (we note that there have been several attempts to derive (21), or something like it, from factors such as “prototypicality” and “iconicity”; see, e.g., Wierzbicka 1981; and see Silverstein 1981 for a critical discussion). (21)

a. subjects: • 1st/​2nd person → unmarked • 3rd person → marked (‘erg’) b. objects: • 3rd person → unmarked • 1st/​2nd person → marked (‘acc’)

Recall, however, that such a treatment would not generalize to the Eastern Ostyak and Niuean patterns discussed in section 10.4.1; ideally, we would want a configurational account of these Dyirbal facts, as well. In what follows, we propose such an account, based on the premise that Dyirbal has the kind of very short object movement proposed by Johnson (1991), inter alia:

 

Split ergativity is not about ergativity    241 (22)

··· µP

···

µ'

µ0

νP ν'

SUBJ ν0

VP V0

  We follow Baker & Vinokurova (2010) in assuming that dependent case is assigned as soon as possible; and that it can be relativized to a particular cyclic domain (in Baker & Vinokurova’s analysis of Sakha, dative is assigned to the higher of two DPs within the VP domain, whereas accusative is assigned to the lower of two DPs within the CP domain). With this assumption in place, we can begin to derive the distribution of ‘accusative’ and ‘ergative’ in Dyirbal in purely configurational terms. Consider a first attempt, given in (23):13 (23) case in Dyirbal a. ‘accusative’ in Dyirbal: dependent case assigned to the higher of two DPs outside of VP. b. ‘ergative’ in Dyirbal: dependent case assigned to the higher of two DPs inside of vP. Given (23a), the object will only be ‘accusative’ if it undergoes the kind of short object movement that, in Sakha, was associated with specificity. Only then will there be two noun phrases outside of VP; and the higher of these two will be the object, at least prior to any subsequent movement of the subject (and recall that dependent case is assigned as soon as the conditions on its assignment are met). The only difference is that in Dyirbal, the relevant semantic property associated with short object movement is not specificity per se, but some other property that groups 1st and 2nd person pronouns together to the exclusion of all other nominals; we return to the precise nature of this property in section 10.4.5, below. 13  Varying the directionality of dependent case in (23a–​b) creates, in both instances, non-​differential assignment of case. Changing the direction of (23b) to lower results in invariable assignment of dependent case to the object (which, terminologically, would then be called ‘accusative’ rather than ‘ergative’). Changing the direction of (23a) to lower results in invariable assignment of dependent case to the transitive subject (‘ergative’, rather than ‘accusative’). Importantly, then, these parametric options do not seem to overgenerate.

 

242    Jessica Coon and Omer Preminger The problem with (22–​23) as an account of case in Dyirbal pertains to (23b): as it stands, the subject will always be marked with ‘ergative’ case, since the subject and object satisfy (23b) in their base positions. The solution to this problem, we argue, lies in recognizing the unique agreement requirements that apply to 1st/​2nd person arguments, which we turn to now.

10.4.3 The Formal Agreement Requirement on 1st/​2nd Person Pronouns The Person Case Constraint (PCC), also known is the *Me-​Lui Constraint, is a prohibition against certain combinations of arguments, usually affecting the two internal arguments of a ditransitive. There are at least four attested varieties of the PCC (see Nevins 2007 for a recent review); of interest here is the so-​called “strong” PCC. The strong PCC prohibits 1st/​2nd person direct objects from co-​occurring with an indirect object. In other words, it amounts to the requirement that direct objects of ditransitives be 3rd person. (See Bonet 1991, Anagnostopoulou 2005, Nevins 2007, inter alia, for further discussion.) But when one looks at formal accounts of the (strong) PCC, what one finds is that they do not rule out 1st/​2nd person direct objects per se—​what they rule out is 1st/​ 2nd person object agreement, or 1st/​2nd person object clitics. This is for good reason: the PCC arises precisely where object-​agreement/​object-​clitics are found (e.g. the Basque (24)), and its effects disappear when no such object marking is found (e.g. the embedded clause in (25), which is the infinitival counterpart of (24b)). (24) a. Zuk niri liburu-​a saldu d-​i-​ϕ-​da-​zu. (Basque) you.erg me.dat book-​artsg(abs) sold 3.abs- -​sg.abs-​1sg.dat-​2sg.erg ‘You have the book to me.’ b. * Zuk harakin-​ari ni saldu  n-​(a)i-​ϕ-​o-​zu.   you.erg butcher-​artsg.dat me(abs) sold 1.abs- -​sg.abs-​3sg.dat-​2sg.erg Intended: ‘You have sold me to the butcher.’ (25) Gaizki iruditzen ϕ-​zai-​ϕ-​t [zuk wrong look-​impf 3.abs-​ -​sg.abs-​1sg.dat   you.erg harakin-​ ari    saltzea]. butcher-​artsg.dat  sold-​nmz-​artsg(abs) ‘It seems wrong to me for you to sell me to the butcher.’

ni me(abs)

Thus, what the PCC rules out is not a given combination of (internal) arguments, but rather a given combination of object-​agreement markers (or object-​clitics). Without further provisions, however, the expectation arises that a 1st/​2nd person strong pronoun in direct object position of a ditransitive would be just fine, just as long as the finite verbal element carried agreement morphology expressing 3rd person features (rather than the 1st/​2nd person features of the actual object). But this expectation is of course false:

 

Split ergativity is not about ergativity    243 (26) * Zuk harakin-​ari ni saldu d-​i-​ϕ-​o-​zu.  you.erg butcher-​artsg.dat me(abs) sold 3.abs-​ -​sg.abs-​3sg.dat-​2sg.erg Intended: ‘You have sold me to the butcher.’ To handle this, accounts of the PCC are commonly supplemented with something along the lines of (27)—​or, on Preminger’s (2014) formulation, (28).14 (27) Person Licensing Condition (PLC) Interpretable 1st/​2nd person features must be licensed by entering into an Agree relation with an appropriate functional category.  [Béjar & Rezac 2003] (28) Person Licensing Condition (PLC) —​alternative formulation The feature [participant] on a pronoun must participate in a valuation relation. NB: ‘[participant]’ is the feature that distinguishes 1st/​2nd person nominals from 3rd person ones; see Harley & Ritter (2002) and McGinnis (2005) for further discussion. As noted by Preminger (2011b), this requirement appears to be a sui generis requirement on marked person features, which does which does not extend to other φ-​feature classes (number and gender; pace Baker 2008).

10.4.4 The Person Licensing Condition Meets Differential Subject Marking Recall the analysis of the Dyirbal “binary split” system put forth in section 10.4.2, repeated here: ···

(29) 

[=(22)] µP

···

µ'

µ0

νP ν'

SUBJ ν0

VP V0

14  A principle like (27) or (28) will correctly rule out cases like (26), but it leaves open the question of how the local direct object pronoun in the infinitival clause in (25) satisfies its licensing requirements. See Preminger (2011b) for an explanation based on locality.

 

244    Jessica Coon and Omer Preminger (30)   Case in Dyirbal a. ‘accusative’ in Dyirbal: [=(23a–​b)] dependent case assigned to the higher of two DPs outside of VP. b.  ‘ergative’ in Dyirbal: dependent case assigned to the higher of two DPs inside of vP. The problem with this version was that it predicted invariable ergative case on the subject, contrary to fact (see section 10.4.2 for details). This can be remedied, we propose, if—​in light of the Person Licensing Condition—​we add the following provision:15 (31)

a. If the External Argument bears [participant], agreement with v0 (upon first merge in [Spec,vP]) will satisfy its PLC requirement. b. [participant]-​bearing v0 is phasal: yes/​no. ← per-​language parameterization

In a language where (31b) is set to ‘yes’, the assignment of ergative will be bled precisely when the subject bears [participant] features (i.e. when the subject is 1st/​2nd person). That is because the phasehood of v0 will render the object, located inside the complement of v0, inaccessible for the purposes of (30b).16 And, crucially, (30a) will be unaffected by phasehood (or lack thereof) of vP, because the subject is already at the edge of vP. Thus, the assignment of ‘accusative’ will not be sensitive to the features of the subject, even if (31b) is set to ‘yes.’ This produces the attested behavior of the “binary split” system of Dyirbal. The object will be assigned ‘accusative’ only if it undergoes short object movement—​which, by hypothesis, applies only to 1st/​2nd person pronouns in Dyirbal. The subject will be assigned ‘ergative’ unless it is 1st/​2nd person, in which case the phasehood of vP will prevent (30b) from applying successfully.

10.4.5 DSM vs. DOM There is a residual issue with the account of the “binary split” system that was put forth in section 10.4.4. If Differential Argument Marking is fundamentally a configurational phenomenon, as argued here, then it comes out as a coincidence that in Dyirbal, the same 15  One might wonder whether, and to what degree, the parameterization in (31b) conflicts with the idea that vP is a category whose phasal status alternates based on the transitivity of the verb (Chomsky 2001, et seq.). In the Baker & Vinokurova treatment of DOM—​which we build upon—​it is crucially VP (and not vP) that is the ‘phase’ for the purposes of case computation. We take this to mean that the “classic” vP phase is simply a poor fit for the kind of locality domain required in an adequate case-​assignment algorithm. We therefore set aside the traditional, transitivity-​based definition of vP phasehood for the present purposes. 16  Since case assignment rules apply as soon as possible (see section 10.4.1), it is conceivable that later movement of the subject out of vP will once again bring the subject and object into a single case domain, resulting in the assignment of ‘ergative’ to the subject (or even to the object, depending on their relative hierarchical configuration). To avoid such spurious case assignment configurations, we tentatively assume that the rule in (30a) does not apply outside of µP.

 

Split ergativity is not about ergativity    245 factors that regulate movement of the object out of VP also regulate the phasehood of vP—​ namely, the presence or absence of a [participant] feature. We contend, however, that this is actually a desirable state of affairs: we want this to be a Dyirbal-​specific coincidence. To see why, we must first attend to the featural bases for DSM and for DOM, cross-​linguistically. Both DSM and DOM have historically often been claimed to adhere to a scale like (32): (32) ← subj marked erg common proper ≫ ≫ nouns nouns

obj marked acc → demonstratives, 1st/​2nd person ≫ 3rd person pronouns pronouns       [Dixon 1994, Silverstein 1976, inter alia]

While this has proven to be an extremely useful (and influential) first approximation, it belies the finer typological differences one finds between DSM and DOM. To wit: Intuitively we could expect [the DSM] split to be found between humans and non-​humans, or between animates and inanimates. Actually, no language places the split in such positions; most of them indeed single out 1/​2 pronouns from the rest. [Cocchi 1999: 112] If [a unified approach to DOM and DSM is correct], we expect to find an equal diversity of types of subject and object splits in the world’s languages; however, that prediction is not borne out. Instead, there are very few kinds of subject splits, in contrast to an enormous diversity of object splits. For example, Comrie (1981b‌:  123)  notes that while definiteness is frequently the basis of object splits, there is an embarrassing absence of clear attestations of the predicted marked indefinite subject. [Woolford 2001: 535]

And with respect to Australian languages (like Dyirbal) in particular: My key point is simply that [Hopper & Thompson (1980)] clearly establish that special accusative marking tends to occur with proper nouns, human and animate nouns, and definite, referential usages in many language families. There is no reason to link this phenomenon when it occurs in Australian languages with the incidental fact that Australian languages have ergative case. [Goddard 1982: 191]

Overall, while differential ergative marking on subjects (DSM) typically correlates with the 1st/​2nd person vs. 3rd person distinction, differential case marking on objects (DOM) is typically sensitive to features like animacy, specificity, and definiteness.17

17 

One potentially problematic case involves patterns of apparent DOM that are sensitive to person features per se, rather than the canonical DOM features discussed here. Roberta D’Alessandro (p.c.) points out that in the Italian dialect Ariellese, an a marker is added to 1st/​2nd person object

 

246    Jessica Coon and Omer Preminger Consequently, a language like, e.g., Balochi (NW Iranian) might be a far better representative for how “binary split” systems usually work than Dyirbal is (we return to Dyirbal itself in section 10.4.6). In Balochi, DSM distinguishes 1st/​2nd person from 3rd person, but DOM is sensitive to definiteness. This pattern is summarized in (33); see Farrell 1995: 224 for details.18 (33)

Balochi Argument Marking 1st/​2nd person pronouns

3rd person pronouns, and other definites

indefinites

A

∅

‘erg’

‘erg’

S

∅

∅

∅

P

‘dat’

‘dat’

∅

To capture these crosslinguistic generalizations, which Balochi exemplifies, we propose the following generalization: (34) Differential Argument Marking Generalization a. DSM (i.e. of A) is based on the presence or absence of the feature [participant] b. DOM (i.e. of P) is governed by definiteness, specificity, and/​or animacy Returning to our proposal regarding binary split systems, the basic structure of which is repeated again in (35), the generalizations in (34a–​b) can be captured as follows. DOM (34b) is regulated by movement to [Spec,µP], and can therefore depend on one of a set of possible features—​definiteness, specificity, and/​or animacy—​ depending on what it is that µ0 probes for. DSM (34a), on the other hand, is regulated by the phasehood of vP, which on our proposal is determined by a single, fixed feature: [participant].

pronouns, but not to the ‘unmarked’ 3rd person pronoun esse (Ariellese also has a system of proximate/​medial/​distal 3rd person demonstratives, all of which apparently marginally tolerate the a marker). A similar though non-​identical system, involving optionality of a with 3rd person pronouns (but obligatoriness of it with 1st/​2nd person ones) is reported by Manzini & Savoia (2005) for the Canosa Sannita dialect. While nothing in our account specifically rules out person-​sensitive DOM, the overwhelming tendency of DOM to be based on other features (in contrast to DSM, which is frequently—​perhaps always—​based on person distinctions) is still a very robust cross-​linguistic generalization. We therefore tentatively set aside the pattern in Ariellese and Canosa Sannita for future work. 18  An essentially identical pattern—​DSM based on 1st/​2nd person vs. 3rd person, with DOM based on definiteness vs. indefiniteness—​is found in Kham (Tibeto-​Burman; see DeLancey 1981, Watters 2002, Merchant 2009).

 

Split ergativity is not about ergativity    247 ···

(35)  ···

[≈(29)] µP µ'

µ0

νP ν'

SUBJ

VP

ν0 V0

10.4.6  A New (or Old) Take on Dyirbal Objects While the proposal in (29–​30) successfully derives systems like the one exemplified by Balochi, DOM in Dyirbal now appears to be an exception to the rule—​since it is described as tracking the 1st/​2nd person vs. 3rd person distinction. In this subsection, we demonstrate that this may be based on a mischaracterization of what the relevant distinction really is. The crucial observation is that Dyirbal doesn’t actually have 3rd person pronouns. This means that we can recast (36) (repeated from earlier) as (37), without altering the empirical coverage whatsoever: (36) Dyirbal Argument Marking—​Summary 1st/​2nd person pronouns

[=(20)]

other nominals

A

∅

-​ŋgu (‘erg’)

S

∅

∅

P

-​na (‘acc’)

∅

(37) Dyirbal Argument Marking —​Expanded version 1st/​2nd person pronouns

3rd person pronouns

other nominals

A

∅

-​ŋgu (‘erg’)

-​ŋgu (‘erg’)

S

∅

∅

∅

P

-​na (‘acc’)

-​na (‘acc’)

∅

 

248    Jessica Coon and Omer Preminger To be clear, we are not saying that there is a Dyirbal-​internal argument in favor of (37); only that given the lack of 3rd person pronouns in the language, there is no Dyirbal-​ internal argument against (37). It is worth noting, in this respect, that the kind of pattern we are asserting to exist in Dyirbal is attested—​in a language with 3rd person pronominals, in Cashinawa (Panoan; Dixon 1994). As the table in (38) shows, the split in the marking of A arguments in Cashinawa divides 1st/​2nd person from 3rd person, while for objects the distinction is between pronouns (regardless of person) and all other nominals. (38) Cashinawa (Dixon 1994: 86) 1st/​2nd-​person 3rd-​person pronouns pronouns other nouns A

-​∅

habũ

nasalization

S

-​∅

habu

-​∅

P

-​a

haa

-​∅

In fact, this view of Dyirbal itself is not without precedent, either (see also Legate 2008): Proper and some common nouns (usually just those referring to humans) can take the suffix -​na, but only when they are in transitive object function. [Dixon 1972: 43]

We can therefore maintain the view that while DSM in Dyirbal is about [participant] vs. lack thereof, DOM in Dyirbal is about pronominality—​or perhaps, given this quote from Dixon (1972), animacy. This brings Dyirbal into accordance with the Differential Argument Marking generalization given in (34) and, in turn, allows those mechanisms we have proposed to derive (34) to derive the behavior of Dyirbal, as well.

10.4.7 Summary We have argued that Differential Argument Marking, including DSM, is a configurational phenomenon—​and crucially, that DSM encompasses what in an ergative system is characterized as “split ergativity”. First, morpho-​semantic distinctions (such as specificity, definiteness, or animacy) may give rise to changes in the relative positions of core arguments, via the mechanism of Object Shift (or other very short object movement). This, in turn, may affect the way case is assigned in the clause—​normally resulting in the morpho-​semantic properties of an argument co-​varying with its own case marking; but in certain cases, via the mechanism of dependent case, resulting in properties of one argument co-​varying with the morphological marking of another (as in Eastern Ostyak and Niuean). Second, even subject marking alternations based on whether the subject is 1st/​2nd person or 3rd person can be recast configurationally, given the independently motivated formal agreement requirement on 1st/​2nd person

 

Split ergativity is not about ergativity    249 arguments (Béjar & Rezac 2003; Anagnostopoulou 2005). Finally, even a language like Dyirbal does not counterexemplify these generalizations, since the lack of 3rd person pronouns in the language does not actually allow one to distinguish between a pronominality/​animacy distinction and a distinction based on 1st/​2nd person vs. 3rd person. We now turn to a note on terminology. There is a common impression that DSM (also known as ‘split ergativity’) is more common in ergative languages than in non-​ ergative ones, an impression that might arise even from surveying the preceding subsections. Why would this be? Our answer—​echoing the discussion of aspectual splits summarized in section 10.3.3—​is that this is nothing more than a terminological bias towards properties of the subject (and not of the object) as the defining property of a morphological marking system. Recall that, on the current analysis, DSM is about the disruption (or non-​disruption) of the dependent case relation that the subject participates in. In a language with an ergative-​absolutive alignment, disrupting this relation alters the case-​marking of the subject itself (namely, ‘ergative’). But in a nominative-​accusative system, disrupting this relation does not affect the subject’s marking—​instead, it affects the assignment of ‘accusative’ to the object. So, by fixating on the subject, we would not notice that anything “noteworthy” has changed (recall the discussion in section 10.4.1 of Sakha as a ‘split accusative’ language).19 If we look carefully, we should be able to find something that looks like the mirror image of the Eastern Ostyak and Niuean pattern, discussed in section 10.4.1: a scenario where the [participant] features on a 1st/​2nd person subject—​and the concomitant phasehood of v0—​affect a morphosyntactic property other than the subject’s own morphological marking. As discussed in Coon & Preminger 2012, this might be exactly what is happening in the Abruzzese dialect of Italian: (39) a. Ji  so’      magnate. I   eaten.sg  am(be) ‘I have eaten.’ b. Esse  a        magnate she    has(have) eaten.sg ‘She has eaten.’

(Abruzzese)

[D’Alessandro & Roberts 2010: 54–​55]

On the assumption that have is be + X0 (where X0 is some clausal particle of category D or P; see Freeze 1992, Kayne 1993), a phasal v0—​caused by a [participant]-​bearing

19 

There is one noteworthy scenario where disrupting a dependent case relation in a nominative-​ accusative system would affect the case-​marking on a subject—​when the relation in question is between the subject and an even higher noun phrase. Accusative subjects in Turkic (see George & Kornfilt 1981; Kornfilt 1984, 2003, 2006 on Turkish; as well as Baker & Vinokurova 2010 on Sakha) may constitute precisely such a scenario.

 

250    Jessica Coon and Omer Preminger subject, in same manner proposed in section 10.4.4—​would block the head-​movement/​ incorporation of X0 into be. This seems to be exactly the state of affairs in Abruzzese (see D’Alessandro & Roberts 2010).20

10.5 Conclusion We argued that split ergativity, of both the aspectual sort and the person-​based sort, is epiphenomenal. The absence of ergative marking on (what appears to be) a transitive subject is the result of structural properties of the clauses in question, which affect the configurational basis upon which case and agreement depend. In the domain of aspectual splits (section 10.3), we showed that non-​perfective aspects involve additional structure, either in the form of an aspectual auxiliary (e.g. Basque), or demotion of the P argument (e.g. Samoan). As for person splits (section 10.4), we showed that general properties of Differential Argument Marking are implicated in so-​called person-​based split ergativity. Specifically, while the marking of objects (DOM) is sensitive, in most if not all cases, to features such as animacy, definiteness, and specificity—​the usual features responsible for Object Shift—​the marking of subjects (DSM) is sensitive, in most if not all cases, to the distinction between 1st/​2nd person and 3rd person. As we have shown, there is good reason to treat these patterns, too, as configurationally triggered. Object shift would bring the object into the same case domain as the subject, facilitating the assignment of dependent case. 1st/​2nd person arguments are known to have unique licensing requirements, and if the projection satisfying these requirements on a 1st/​2nd person subject is phasal, it will disrupt otherwise available dependent case relations. We have presented some specific case studies in how the addition of structure may arise; but note that these are not intended to be exhaustive. Our general proposal is not that every language with split ergativity is necessarily a variation on one of these specific cases. Rather, our proposal is that splits, in general, arise as the result of structural changes to the relevant clauses, and not as the result of special mechanisms or featural specifications not found in nominative-​accusative languages. An illustrative example is provided by Nash (Chapter 8, this volume). She argues that while split ergativity in Georgian (Kartvelian) does not involve the addition of morphosyntactic structure of the type seen in section 10.3 (contra the suggestion in Coon 2013b), it does nonetheless fit the general proposal advanced here: ergative-​patterning aorist tenses are in fact an example of neutral aspect, and lack a phasal EventP present in split-​patterning non-​aorist tenses. 20 

What we have not yet been able to find is a language where [participant] features on the subject affect the case marking of the object per se—​i.e. a language where the presence of a 1st/​2nd person subject bleeds the assignment of accusative to the object. We are not sure, at this juncture, whether this is a true typological gap, given that its mirror image—​properties of the object bleeding the assignment of ergative to the subject—​was only found in two languages in our sample (Eastern Ostyak and Niuean).

 

Split ergativity is not about ergativity    251 We have also not examined in detail here proposals in which 1st and 2nd person arguments are base-​generated in a higher clausal position (Nash, Chapter 8, this volume, on Georgian; Wiltschko 2006 on Halkomelem (Salish)), though this again fits the general pattern of splits being the result of different structures, and of special licensing requirements applying only to 1st/​2nd person arguments. We have argued further that the structural differences that cause splits are not limited to ergative systems in the first place. Rather, the issue boils down to the defining characteristic used by linguists to distinguish ‘ergative systems’ and ‘accusative systems’ from one another. The simple fact that in a nominative-​accusative system, both transitive and intransitive subjects receive the same marking (nominative), obscures the fact that some of these so-​called transitive subjects may actually be intransitive subjects in the morphosyntactic sense (e.g. in the presence of a complex progressive aspectual construction, a demoted object, or phasal vP). In other words, the difference between a transitive and intransitive subject (as in the English John shot the bear vs. John shot at the bear, for example) does not catch our attention as readily in a nominative-​accusative system as it does in an ergative one (see also Polinsky, Chapter 13, this volume). As noted above, this analysis gives us a handle on the seeming paucity of consistently ergative languages, compared to consistently accusative languages—​an asymmetry which on the current view is rather superficial, and has more to do with our readiness to apply the terminology of ‘split’ than it has to do with any deep grammatical properties. Furthermore, the proposed analyses of both aspect-​based and person-​based split ergativity are able to account for the universal directionality of these splits:  the fact that ergative-​absolutive alignment and nominative-​accusative alignment each remain anchored, cross-​linguistically, to fixed ends of the relevant scale (be it an aspectual scale or an NP “prominence” scale). That is because in both domains (aspect and NP-​type), there is independent evidence that the values on one end of the relevant ‘scale’—​but not the other—​are cross-​linguistically associated with additional syntactic structure (in ergative languages and accusative ones alike). It is this additional structure, on the current account, that is responsible for the relevant bifurcation of the clause, in those languages where the added structure happens to be syntactically opaque; and it is this bifurcation that results, in a language that is ergative to begin with, in the appearance of a split, due to altering the configurational properties of the clause. Proposals which seek to account for split ergativity by making direct reference to the scales in (3) and (4)  (e.g. DeLancey 1981)  overgenerate:  in the domain of person splits, for example, it was shown that most of the splits predicted by such a scale are in fact unattested. Overwhelmingly, the only category on this scale that is relevant to subjects is 1st/​2nd person—​a category which is independently known to require special licensing mechanisms (cf. PCC effects). While we make no new proposals about why some languages display an ergative-​ absolutive pattern and others display a nominative-​accusative one, the resulting picture is one in which a given language consistently exhibits one or the other pattern—​at least as far as the core principles of case assignment are concerned (see also Laka, Chapter 7, this volume). There is no ‘split’ in the set of case-​assignment rules or principles operative

 

252    Jessica Coon and Omer Preminger in the language as a whole; different constructions may result in detransitivization, which as discussed, is not a phenomenon that is specific to ergative systems in the first place. This “demystification” of split ergativity contributes to a growing body of work that suggests there is no ergative macro-​parameter, which would group together ergative and split-​ergative languages under a single setting, in opposition to other, fully accusative languages. Furthermore, the same results cast doubt on the idea that there is anything especially marked about ergativity (cf. Visser 2006).

Acknowledgements The authors’ names appear in alphabetical order. Many thanks to Peter Arkadiev, Roberta D’Alessandro, Amy Rose Deal, Itziar Laka, Anoop Mahajan, Jason Merchant, David Pesetsky, Masha Polinsky, Milan Rezac, Norvin Richards, Ian Roberts, Andrés Pablo Salanova, and Thomas Weir, as well as to audiences at the University of Cambridge, Reed College, CLS 47, WCCFL 29, and SLE 2013, for feedback and comments at various stages of this project.

Abbreviations 1, 2, 3, 1st-, 2nd-, 3rd-person; ABS, absolutive; AOR, aorist; ASP, aspect; AUX, auxiliary; DET, determiner; ERG, ergative; FEM, feminine; FUT, future; GEN, genitive; HON, honorific; IMPF, imperfective; LOC, locative; MASC, masculine; NML, nominal; NONFUT, non-future; OBL, oblique; PL, plural; PRES, present; PRFV, perfective; PROG, progressive; PREP, preposition; PST, past; POSS, possessive; PTCP, participle; SG, singular. In some cases, glosses have been simplified or modified from those of the original authors for consistency. We follow the original authors’ choices regarding capitalization and punctuation. If no citation is provided, data is from the authors’ field notes.

 

Chapter 11

Ergativit y a nd differentia l c ase marki ng Andrej Malchukov

11.1 Introduction The issue of differential subject marking (DSM) in ergative languages has been primarily discussed by Silverstein (1976) and others in connection to patterns of split ergativity found in Australian languages. A typical Australian pattern of split ergativity makes a split between pronouns (sometimes, just 1st and 2nd person pronouns) vs. nouns: nouns show overwhelmingly an ergative pattern, while pronouns an accusative pattern. The study of differential object marking (DOM) has a somewhat different source. DOM has been introduced into general literature in the work by Bossong, Comrie, and Lazard among others and has been extensively documented by Bossong (1985) for Iranian languages. Later on, however, the two strands of research have started to converge, as the Australian pattern of split ergativity with pronouns showing the accusative pattern has been identified as one of classical cases of DOM. In accordance with the focus of the volume on ergativity, I will have more to say in this chapter about the DSM, which is more prominent in ergative languages as compared to DOM (see section 11.9 for discussion and explanation), yet for proper understanding of differential case marking a comparison of DSM and DOM phenomena is important. In what follows I discuss the phenomenon of differential case marking in ergative languages and propose an explanation for the attested patterns in terms of competing motivations as practiced in functional typology, which can be also formalized in Optimality Theory. Section 11.2 introduces DOM, and discusses a popular explanation of the attested patterns in terms of markedness. Section 11.3 extends the discussion of differential case marking to DSM in ergative languages, showing that markedness cannot provide a complete explanation for DSM patterns. In particular, it will be claimed

 

254   Andrej Malchukov that while in the domain of DOM we encounter just one pattern, complying with the predictions of the Silverstein hierarchy, in the domain of DSM we find two opposite patterns; a “Silverstein-​pattern” and an “anti-​Silverstein pattern.” Section 11.4 introduces a two-​factor approach, in terms of indexing and distinguishability constraints, which is able to explain asymmetries between DOM and DSM. Section 11.5 extends the discussion to other features implicated in differential case marking involving definiteness and discourse factors. The following sections deal with varieties of differential case marking:  they address symmetric and asymmetric case marking (section 11.6), split vs. fluid marking (section 11.7), and show how this taxonomy relates to the two case-​ marking strategies driven by indexing or distinguishability constraints (section 11.8). The next section (section 11.9) explores a relation between differential case marking and voice alternations such as passive and antipassive. Section 11.10 provides a discussion of diachronic scenarios behind differential case marking, and section 11.11 is a brief conclusion.

11.2  DOM: The Markedness Approach Among the phenomena known under the rubric of differential case marking, DOM is definitely better researched (Bossong 1985; Lazard 1998; Aissen 2003b; among others). These studies have shown that in many languages objects higher on the animacy/​definiteness hierarchy tend to be (case) marked while those which are lower on the hierarchy need not be. It has also been observed that in some languages DOM is more sensitive to the definiteness dimension (e.g. in Persian, where the postposition -​râ occurs mostly with definite NPs), while in some other languages DOM is more sensitive to animacy proper (e.g. in Guaraní, the postposition pe is normally found with animate NPs). In yet other languages DOM is sensitive to both features. One of the best studied cases is Hindi, where animate NPs should be marked by accusative/​dative case, while inanimates are marked only if definite (Mohanan 1990): (1)

Hindi (Mohanan 1990: 104) Ilaa-​ne bacce-​ko /​ (*baccaa)  uTaayaa. Ila-​ERG  child-​ACC /​ (*child) lift.PERF ‘Ila lifted a/​the child.’

(2) Ilaa-​ne haar uTaayaa. Ila-​ERG  necklace  lift.PERF ‘Ila lifted a/​the necklace.’ (3) Ilaa-​ne haar-​ko uTaayaa. Ila-​ERG necklace-​ACC lift.PERF ‘Ila lifted the necklace.’

 

Ergativity and differential case marking    255 The usual explanation for DOM relies on the concept of markedness. This explanation, originally due to Silverstein (1976), is formulated by Comrie (1989: 128) as follows: “In a standard transitive scenario, A is animate and definite while P is inanimate /​indefinite (or at least less animate and definite than A), so any deviation from this scenario should be marked.” This account correctly predicts that animate/​definite objects, which manifest a marked (less natural) combination of role and semantic features, should be case-​ marked, while inanimate/​indefinite objects, which manifest an unmarked combination, need not be marked. This markedness explanation of DOM patterns also lies at the heart of Aissen’s (2003) optimality–​theoretic account of DOM. Aissen proposes to capture the markedness pattern through a harmonic alignment between a scale of grammatical functions (subject and object) and prominence scales incorporating the Animacy Hierarchy. On Aissen’s account the insight that animate arguments are less natural qua objects (than qua subjects), is captured by a constraint ranking, where a constraint against having unmarked animate objects is stronger than a constraint against having unmarked inanimate objects. Thus, the pattern in Hindi where only humans are obligatorily case-​marked is captured in Aissen’s account in the following fashion (a simplified fragment of the constraint hierarchy is represented in (4)): (4) …*Oj/​Hum & Øc >> *Case >>….>> *Oj/​Inan & Øc… In this representation, the constraint against having zero marked human objects (*Oj/​ Hum & Øc) dominates the economy constraint prohibiting any case marking (*Case), which in turn dominates a constraint against having zero marked inanimate objects (*Oj/​Inan & Øc). As a result, human objects are obligatorily marked in Hindi, while inanimates may be left unmarked (see Aissen 2003b for a full account). Aissen’s approach has been highly influential and was followed up and refined in a number of respects in subsequent literature. de Swart (2007), de Hoop and Malchukov (2008), and Malchukov (2008a) proposed to subsume Aissen’s markedness constraints under Distinguishability (as instantiation of the discriminating function of cases). In this way they extended this analysis to cases where case assignment is ‘global’ (in terms of Silverstein 1976), that is, sensitive to properties of both arguments in a clause. Thus, de Swart (2007) cites an example form Malayalam, where object marking is usually lacking on inanimate arguments, but may appear on the P if the A is inanimate as well (in examples like ‘The waves destroyed the ship’). Malchukov (2008a) discusses a pattern of global DOM in the Papuan language Awtuw (Feldman 1986). In Awtuw, ACC is obligatorily used if a P equals or is higher than an A on the Animacy Hierarchy (as in (5)), and is left unmarked otherwise (as in (6)): Awtuw (Feldman 1986: 110) (5) Tey tale-​re yaw d-​æl-​i. 3FS woman-​ACC pig FA-​bite-​P ‘The pig bit the woman.’

 

256   Andrej Malchukov (6) Tey tale yaw d-​æl-​i. 3FS woman pig FA-​bite-​P ‘The woman bit the pig.’ This pattern is clearly due to the differentiating strategy; the ACC marks untypical Ps (animate, human), which may be confused with the subject otherwise. The distinction between local and global disambiguation is not clear-​cut, though. Thus, while in Hindi markedness is local, in a genetically related Kashmiri it is (at least partially) global in that the P takes an object (ACC/​DAT) case if the A is lower than the P on the Animacy/​Person Hierarchy: Kashmiri (Wali and Koul 1997: 155) (7) Bı chus-​ath tsı parına:va:an. 1SG.NOM  am.2SG.ACC  2SG.NOM  teaching ‘I am teaching you.’ (8) Su chu-​y tse parına:va:an. 3SG.NOM  is-​2SG  2SG.ACC  teaching ‘He is teaching you.’ In (7), the P is lower than the A on the Animacy/​Person Hierarchy and remains in the NOM case; in (8), the P is higher than the A and therefore takes the object case. Given that the distinction between local and global distinguishability is not clear-​cut, I consider both local and global distinguishability as manifestations of basically the same strategy. The relation between global and local distinguishability can also be conceived in diachronic terms: what starts as a pattern of global distinguishability, where use of a marker is optional (dependent on the context, which can always make its use dispensable), is eventually conventionalized as a pattern of local distinguishability (where, say, animate objects are always marked irrespective of the context). This is consistent with what we know about the extension of DOM in individual languages (see, e.g., Aissen 2003b for a diachronic discussion of DOM in Spanish). A related development occurred in Persian where an animacy based DOM (attested in early Judaeo-​Persian texts) developed into definiteness based pattern (Stilo 2009). This conclusion is also compatible with analyses which suggest that ‘pragmatic’ DOM can eventually evolve into ‘semantic’ DOM (Zeevat and Jäger 2002). Since distinguishability relates to ambiguity avoidance, a natural move would conceive it as a comprehension constraint in OT semantics, rather than a production constraints in OT syntax. Indeed, Zeevat and Jäger (2002) proposed to reconstruct Aissen’s analysis in OT semantics (Hendriks and de Hoop 2001), and Malchukov and de Hoop (2008) use distinguishability as a comprehension constraint in a bidirectional OT model. Finally, from a psycholinguistic perspective, de Hoop and Lamers (2006)

 

Ergativity and differential case marking    257 explicitly argue for incremental optimization of case interpretation making use of distinguishability constraints. So, it appears that by now we have a complete story for differential case marking (DCM) based on the concept of markedness/​distinguishability. Yet, as we will see in section 11.3, extending this analysis to DSM meets with mixed success.

11.3  DSM: A Problematic Pattern As aptly demonstrated in the literature (Moravcsik 1978a; Bossong 1985; Comrie 1989; Lazard 1998; Aissen 2003b; among others), DOM is both a pervasive and a cross-​ linguistically consistent phenomenon, even though languages differ in extensions of object marking along the animacy/​definiteness hierarchy. Less studied is the phenomenon of DSM, which may be largely due to the fact that ergative languages, which provide stronger cases of DSM, became the subject of a systematic study only recently.1 Does DSM conform to the markedness pattern, as does DOM? The predictions here would be that inanimate and/​or indefinite As which deviate from the prototype are preferably marked (Comrie 1989). Sometimes one indeed finds such a pattern. Consider the case of Qiang (Lapolla 2003: 125), where an A in a transitive causative clause does not take the “Agentive” case unless inanimate: (9)

Qiang (Lapolla 2003: 125) Moᴚu-​wu qa da-​tuə-​ᶎ. wind-​AGT  1SG  DIR-​fall.over-​CAUS ‘The wind knocked me down.’

Yet, such a pattern, with only inanimate As case-​marked, is rare. In most other split-​ ergative languages of Australia (as well as some Tibetan and Caucasian languages), markedness effects are manifested in a noun/​pronoun split, where pronouns (or just the first and second person pronouns, as in Dyirbal; Dixon 1972), which are highest on the Animacy Hierarchy, lack the ergative case (Silverstein 1976; Comrie 1989). This pattern is consistent with Silverstein’s generalization, which predicts that accusative marking spreads from the nominals on the top of the Animacy Hierarchy to lower nominals, while ergative marking spreads in the opposite direction starting from the nominals

1 

Note that in most accusative languages the syntactic status of “non-​canonical subjects” is controversial (see the chapters in Bhaskararao and Subbārāo 2004 for a recent discussion). For example, Icelandic seems to be quite exceptional among Germanic languages in having oblique subjects, involved in a DSM pattern. The corresponding constructions in other Germanic languages (such as German Mir gefällt das Buch [me.DAT please.PRES.3SG the.N book.NOM.N] ‘I like the book’) are usually considered to involve object experiencers, rather than subject experiencers, since the experiencer fails the standard tests for subjecthood (see, e.g., Zaenen et al. 1985; Sigurðsson 2002 for discussion of Icelandic/​German contrasts; see, however, Barðdal 2002 for a different view).

 

258   Andrej Malchukov on the bottom of the Animacy Hierarchy. This pattern can be straightforwardly captured through the Aissen-​style constraints, as illustrated above. Consider the pattern in Dyirbal, where only 1st and 2nd person pronouns, highest on the Animacy/​Person Hierarchy are left unmarked, while all other nominals feature ergative marking. This pattern can be captured by the constraint hierarchy in (10).2 (10) …*A/​Inan & Øc >>*A/​An & Øc >>…*A/​3rd & Øc >> *Case >>….>> *A/​1,2 & Øc… As in case of DOM, this discriminating account of case marking works well for cases when case-​marking pattern is global, that is when availability of ERG case on the A depends on properties of the P argument. Consider a case of Fore (Scott 1978), another language with a global pattern of case marking, which is in a way a mirror image of the Awtuw pattern illustrated in (5)–​(6) above. In Fore, the ergative marker (Scott’s “delineator”) is used only if the P is higher than the A on the Animacy Hierarchy (as in (12)), otherwise the A remains unmarked (as in (11)): Fore (Scott 1978: 116)

(11)

Yagaa wá aegúye. pig man 3SG.hit.3SG ‘The man hits (or kills) the pig.’

(12)

Yagaa-​wama wá aegúye. pig-​ERG man 3SG.hit.3SG ‘The pig hits the man.’

Also here the use of the case marker is motivated by the need to disambiguate: the ERG case appears on non-​prominent (non-​human) As which are more likely to be construed as objects than as subjects. Note also that ERG marking becomes dispensable if the arguments are disambiguated through the use of verbal agreement; on Foley’s account (1986:  173), this means that verbal (agreement) morphology takes precedence to the nominal (case) morphology as a disambiguation strategy. More discussion of the Fore pattern from the optimality–​theoretic perspective can be found in Donohue (1999) and de Swart (2007). While the cases above provide evidence for the markedness/​discriminating approach, there are also cases which directly contradict the predictions of the markedness approach. This is particularly clear when the DSM is attested for intransitive subjects, as also restrictedly attested in Hindi. As is well known, DSM in Hindi is primarily conditioned by aspect: ergative marking of transitive subjects is attested only in perfective

2  The representation in (10) does not appear in this form in Aissen (2003), since she relies on a general notion of subject, but appears in other studies that make a distinction between transitive (A) and intransitive (S) subjects (e.g. Stiebels 2002; Arkadiev 2008a).

 

Ergativity and differential case marking    259 tenses. There is also a group of intransitive verbs (Mohanan’s “class 2 verbs”) which allow a case alternation on subjects in the perfective, and where the choice between ERG and NOM depends on volitionality (see also Lee 2003 and de Hoop and Narasimhan 2005 for discussion): Hindi: Mohanan (1990: 94) (13) Vah cillaaya. he.NOM shout/​scream.PERF ‘He screamed.’ (14) Us-​ne cillaaya. He-​ERG shout/​scream.PERF ‘He shouted (deliberately).’ Note that ergative marking here is restricted to volitional and hence animate subjects, contrary to what is predicted under the markedness approach. This is not an isolated case. It has long been noted (see, e.g., DeLancey 1981), that DSM in many ergative languages (both consistently ergative languages and ergative languages with a tense/​aspect based split) does not follow the markedness/​economy predictions. For these languages, the presence of an ergative case on the subject is rather related to its agentivity (properties of volitionality/​control). Similar counterexamples to the expected markedness pattern are attested elsewhere. Consider the following examples from Samoan, where animate As are consistently encoded by ergative case, while inanimate As (e.g. those referring to a natural force) allow for an alternation between ergative and oblique (locative) cases: Samoan (Mosel and Hovdhaugen 1992: 424–​5) (15) Na tapuni e     le matagi  le     faitoto’a. PAST close    ERG ART wind   ART  door ‘The wind closed the door.’ (16) Na tapuni i le matagi  le      faitoto’a. PAST close   LOC ART wind    ART  door ‘The wind closed the door.’ This case is also incompatible with the markedness account, since it is a more animate A that is obligatorily marked by the ergative. Thus, it seems that markedness can account for the DOM pattern, but fails to provide a general explanation for DSM (see also Woolford 2001). Indeed, in case of DOM, one finds just one consistent pattern, with more prominent Ps preferentially marked, while in the domain of DSM, we obtain two opposite patterns:  a “Silverstein-​pattern” with less prominent As preferentially marked (as in Dyirbal and Qiang), and an “anti-​Silverstein pattern” with more prominent agentive As marked with an ergative case (as in Hindi and Samoan). In section 11.4

 

260   Andrej Malchukov I present an explanation for the puzzling asymmetry between DOM and DSM in that respect.

11.4  Explaining Asymmetries in Differential Case Marking: The Two Factor Approach In the functional–​typological literature, two main functions of case marking have been identified: the indexing function, that is, cases are used to encode semantic roles, and the discriminating function, that is, the need to distinguish between the core arguments (subjects and objects) (Hopper and Thompson 1980; Mallinson and Blake 1981; Kibrik 1985; Comrie 1989). At first, recognition of these functions has given rise to two analyses of case marking that were viewed as alternative and even competing. Nowadays, however, it is generally acknowledged that both are indispensable to account for the cross-​linguistic variation in case patterns (Mallinson and Blake 1981: 91–6; Song 2001: 156–67). For example, the indexing approach provides a better account of case marking of oblique arguments, and of the marking of core arguments in languages where case reflects semantic roles of arguments (‘role-​dominated’ languages in terms of Van Valin and Lapolla 1997). On the other hand, the indexing approach on its own fails to account for a well-​known tendency, attested both in accusative and ergative languages, to leave the single argument of an intransitive verb (S) as well as one of the core arguments of the transitive verb (A or P) unmarked. To account for this pattern one usually invokes other functional factors, such as the need to differentiate between the two arguments (that is, the discriminating function mentioned above), as well as economy that disfavors overt marking of (core) arguments (see Arkadiev 2008a for further discussion). Once these two motivations for case marking in general are recognized, we can consider what predictions follow for the domain of differential case marking. From a functional–​typological perspective, indexing, and differentiating can be viewed as two (potentially) competing motivations (see Du Bois 1985; Croft 1990), or, in optimality–​ theoretic terms, two conflicting constraints (Prince and Smolensky 2004). I assume that ‘differentiating’ (Diff) and ‘indexing’ (Index) are two general constraints (or rather constraint families) in the domain of case marking (see Malchukov 2008a; de Hoop and Malchukov 2008). • Diff: The arguments (A and P) must be distinguishable. • Index: Encode semantic roles (A and P). Index may be conceived as a family of role-​related faithfulness constraints, as commonly assumed in the literature (e.g. Woolford 2001; Lee 2003; Grimm 2011). Diff is similar to the markedness constraints introduced by Aissen (2003), yet its scope is broader as Diff

 

Ergativity and differential case marking    261 can be satisfied by other means than case such as agreement or word order (de Hoop and Lamers 2006; de Swart 2007). The question to be addressed is what consequences for the domain of case marking follow from recognition of these two general constraint families. Importantly, for the case of DOM, marking a prominent (animate) P is consistent with both constraints. Indeed, from a distinguishing perspective, preferential marking of animate Ps is understandable, as they are more likely to be confused with the subject. From an indexing perspective, marking of animate Ps is also favored, although for a different reason: Ps that are higher on the Animacy Hierarchy are more salient in discourse (are ‘prominent’ arguments in terms of Legendre et al. (1993), or ‘strong’ arguments, in terms of de Hoop 1996) and thus make a “better” patient (Hopper and Thompson 1980; Næss 2007). Table 11.1, as familiar from OT literature, shows cross-​linguistic preferences in the marking of prominent (P) and less prominent (p) objects, given these two constraints. Table 11.1  Constraint interaction in DOM Index

Diff

 P-​marking

Diff

Index

*!

*

 P-​marking *!

  p-​marking

*

   p-​marking

As is clear from Table 11.1, both Index and Diff constraints penalize marking of less prominent objects in preference to prominent ones, so marking of prominent Ps is preferred irrespective of the ranking of the two constraints. Indeed, the absence of case marking of low-​prominent (in particular, inanimate) patients can be attributed either to Index (under the constraint ranking Index-​P >> Econ >> {Index-​p, Diff-​p, Diff-​P}) or to Diff (under the constraint ranking Diff-​P >> Econ >> {Diff-​p, Index-​P, Index-​p}). As both constraints favour the same pattern, the cross-​linguistic consistency of the DOM pattern is (correctly) predicted. With regard to DSM, the predictions are rather different. Diff will disfavor ergative marking of ‘strong’ (prominent, animate) As; since the latter are identified as agents more easily, ergative marking is dispensable. By contrast, Index penalizes ERG on ‘weak’ (inanimate/​non-​sentient) As, as they are less typical as agents. Thus, in case of DSM, these two constraints give conflicting predictions with respect to the marking of prominent/​strong (A) or less prominent/​weak (a) subjects, as shown in Table 11.2. Table 11.2 Constraint interaction in DSM Index

*

A-m ​ arking     a-​marking

Diff

*!

Diff    A-​marking  a-​marking

Index

*! *

 

262   Andrej Malchukov Given that the two constraints are in conflict, and on a further assumption, that there is no universal fixed ranking between the two, this analysis correctly predicts a variation in DSM patterns across languages. Thus, it explains why DSM is less consistent cross-​ linguistically as compared to DOM. As predicted, two different DSM patterns are found. In some languages, more prominent nominals (e.g. pronouns, as in Dyirbal, or animates as in Qiang) will lack ERG case, in accordance with the differentiation/​markedness approach. Other languages restrict the use of ERG case to more agentive (and hence animate) nominals (cf. the pattern in Hindi and Samoan), following the predictions of the indexing approach. Thus, the Dyirbal pattern results from a constraint hierarchy where Diff outranks Index (Diff-​a >> Econ >> {Diff-​A, Index-​A, Index-​a}), while the Samoan pattern results from the opposite ranking of the two constraints (Index-​A >> Econ >> {Index-​a, Diff-​a, Diff-​A}). (See also de Hoop and Malchukov 2008 for a bidirectional OT analysis of these patterns.) Thus, it seems futile to try to reduce both factors (constraints) to one: the fact remains that in the domain of DSM we find two opposite patterns, a “Silverstein-​ pattern” and an “anti-​Silverstein” pattern, even if the latter has been largely disregarded in the literature on alignment splits. For the case of DOM, the two motivations converge, so some patterns (like in Hindi in (2)–​(3)) are amenable to explanation either in terms of Distinguishability or Indexing. In other cases, the effects of the two constraints can be discerned. Thus, the pattern of the global case marking, as illustrated for Awtuw ((5)–​(6)), is clearly due to Distinguishability. On the other hand, some other DOM patterns are arguably due to the indexing function. In particular, the role of Indexing becomes obvious when animacy effects in DOM are extended to DSM, as is the case in Central Pomo (Mithun 1991b). In Pomo the object (patient) case is only found on human patients, but also carries over to animate subjects of unaccusative intransitives (see Malchukov 2008a for discussion and illustration). This corroborates a connection between animacy, prominence, and affectedness, and implicates that some cases of DOM can be better explained under the indexing approach. Indeed, when the pattern of case marking in a transitive clause is extended to intransitive subject, the issues of distinguishability do not arise. Thus, animacy effects in differential case marking are epiphenomenal: in some cases, they are related to disambiguating function (and therefore can be compensated by agreement and/​or word order), in other cases are implicated by some role-​related features such as volitionality. Indeed, it is redundant to mark animacy per se.3

3  In a recent article dealing with DSM, Fauconnier and Verstraete (2014) argue that ‘motivity’ rather than animacy is a contributing factor to DSM (along with the features of Instigation and Directed Transmission). ‘Motivity’ is a broader concept than animacy as it relates to agents that possess an internal source of energy, including natural forces (such as fire, earthquakes, wind, or rain).

 

Ergativity and differential case marking    263

11.5  Extending the Two Factor Approach: Definiteness and Discourse Properties As noted above, Comrie (1981b) relied on markedness to explain why not only animate Ps, but also definite Ps tend to be marked in languages displaying a DOM pattern. We further observed that the role of animacy is different for DOM and DSM patterns, as in the latter case the indexing and differentiating strategies make different predictions. The question arises then whether we find a similar asymmetry between DOM and DSM with respect to the definiteness dimension of the nominal hierarchy, as well. In fact, Comrie (1981b: 129) notes an “embarrassing” absence of evidence for markedness effects in cases of differential encoding of definite and indefinite subjects. That is, one does not readily find cases where only indefinite As appear in the ergative case, as expected under the markedness approach. The same point has been repeated in subsequent literature, both functional–​typological and optimality–​theoretic (e.g., Woolford 2001; Næss 2004). This claim, however, is not entirely correct. Consider the case of Ika (Frank 1985), where we find exactly this pattern: new, indefinite As take the ergative case (see (17)), while given/​definite As do not, unless they are emphatic/​ contrastive (as in (18)): Ika (Frank 1985: 150) (17) Ikı gä-​ža kua ikä-​se’ gä-​ža? man eat-​MED or man-​ERG eat-​MED ‘They eat people or people eat them?’ (18) José-​se’ eigui keiwı tšei-​na. Jose-​ERG also right.away shoot-​DIST ‘Also José shot it.’ Frank (1985: 150) notes that A arguments that are newly introduced or contrastive are regularly marked by ERG in Ika, and explicitly relates this fact to the markedness pattern observed by Silverstein and Comrie. Ika is not exceptional in this respect. A similar pattern is found in many other languages displaying a phenomenon termed here ‘focal ergativity.’ Although the fact that in some languages ERG marking appears on emphatic, new, or contrastive As has not gone unnoticed in the literature (Plank 1979: 34; Van Valin 1992), its significance for the theory of case marking has not been sufficiently appreciated until recently. Yet, more recent work of McGregor and others (see, e.g., McGregor 2009; McGregor and Verstraete 2010)  demonstrated a connection between optional ergativity and discourse structure beyond any doubt. For example, in Gooniyandi (McGregor 1992), given/​predictable agents, when overt, need not be marked, but ERG

 

264   Andrej Malchukov regularly marks those transitive subjects that are unexpected as agents, as in the following example: (19) Gooniyandi (McGregor 1992: 289) Ned Colin-​ngga ridim-​ngangadda Ned Colin-​ERG ride-​3SG.A+3SG.P+1SG.D ‘Ned Colin rode my horse for me.’

yawarda. horse

As shown by McGregor (2009 and passim) patterns of focal ergativity are commonplace in Australian and Oceanic languages. Malchukov (2008a) discusses a number of examples from other linguistic areas; in particular, he cites Newari, where As take the ergative marker when rhematic (Givón 1984: 154). Thus, (20) would be appropriate as an answer to “Who is breaking the window?,” while (21) as an answer to “What is the man doing?” (Givón 1984: 154): Newari (Givón 1984: 154) (20) Wō ma̧nu-​nã jhya tajya-​na̧ co-​na̧. the man-​ERG window break-​AUX be-​AUX ‘The man is breaking the window.’ (21)

Wō ma̧nu jhya tajya-​na̧ the man window break-​AUX ‘The man is breaking the window.’

co-​na̧. be-​AUX

Importantly, both Malchukov (2008a) and McGregor (2009) explicitly relate this pattern to a violation of Du Bois’ “Given-​A constraint.” According to Du Bois (1987b), A  arguments are overwhelmingly given/​definite, hence when they are new and/​or indefinite they are expected to be eligible for special marking. Thus, markedness reversal effects in DSM are well attested with respect to definiteness, even though the data available indicate that the relevant distinction may be between topical/​given and focal/​new NPs, rather than between definite and indefinite ones. It should be noted though that the same conclusion emerges from recent studies of DOM: many DOM patterns which have been attributed to definiteness in earlier studies, have been reinterpreted in terms of information structure (topicality) in more recent work (Iemmolo 2010; Darlymple and Nikolaeva 2011).4 4 

It is interesting to observe that this literature, while agreeing that more topical Ps are preferentially marked, offers very different explanations for the attested patterns. On the one hand, Darlymple and Nikolaeva (2011) attribute preferential marking of more topical objects to the fact that objects are ‘secondary topics’ (following Givón who associates subjects with primary topics and objects with secondary topics). This is more in line with the approach of Hopper and Thompson (1980) who attribute preferential marking of more prominent (in particular, animate) objects to their high(er) individuation. On the other hand, Iemmolo (2010) explains preferential marking of topical objects rather by the need to signal non-​ prototypicality of topical objects, which is obviously more in line with the markedness approach.

 

Ergativity and differential case marking    265 However, the opposite pattern where the ergative case is missing on low-​prominent/​ non-​referential As is attested as well. Thus, Dixon (1994: 90) observes that some ergative languages restrict ergative marking to definite subjects. In Semelai (Kruspe 1999), specific As take the ergative case and are cross-​referenced by the ergative proclitic, while generic As are neither marked nor cross-​referenced: Semelai (Kruspe 1999: 253) (22) Cɔ jəl jkɔs. dog(s).DIR bark.at porcupine.DIR ‘Dogs bark at porcupines.’ (23) Jkɔs ki-​jəl la-​cɔ. porcupine.DIR 3SG-​bark.at ERG-​dog(s) ‘The dog(s) bark(s) at the porcupine.’ A similar pattern has been reported for Alsea (Mithun 1999), where ergative is restricted to referential As. While in the examples (22)–​(23), availability of ergative marking correlates with semantic referentiality (specificity), in other languages it is rather related to discourse referentiality, that is, communicative importance or salience of an argument in discourse (see Givón 1984: 423ff. on a distinction between semantic and pragmatic referentiality). Thus, in Timbe (Givón 2001b: 215, citing Foster) discourse prominent As (persistent topics) are more likely to take the ergative case. And in Gooniyandi, one of the functions of the ergative case is “stressing referentiality of an actor” (McGregor 1992). Thus, in the domain of definiteness we find asymmetries between DOM and DSM patterns similar to those we observed in the domain of animacy. Again, the DOM pattern is cross-​linguistically consistent: high-​prominent patients, which are specific/​referential and given/​topical are preferentially marked. Preferential marking of specific patients can be naturally explained in terms of prominence, and therefore attributed to Index (given the constraint ranking: Index-​P >> Econ >> {Index-​p, Diff-​p, Diff-​P}). The role of prominence constraints is probably most obvious for incorporating languages where low-​prominent (in particular, non-​specific) objects tend to be incorporated (Mithun 1984). On the other hand, preferential marking of topical and/​or given (hence, usually definite) objects can be attributed to Diff, as given/​topical arguments are likely to be construed as otherwise (given the constraint ranking: Diff-​P >> Econ >> {Diff-​p, Index-​P, Index-​p}). Yet given a natural correlation between topicality/​ givenness, definiteness, and specificity, the effects of the two strategies are often difficult to discern in the domain of DOM. In DSM we find less consistency with respect to definiteness effects. Thus, we find languages, like Semelai and Alsea, where ERG case is reserved only for referential subjects (strong As), which can be attributed to Index (Index-​A >> Econ >> {Index-​a, Diff-​a, Diff-​A}). In other languages (Ika, Newari), ERG marking is confined to non-​topical, new, indefinite subjects. This pattern is readily explained by Diff (given the constraint ranking: Diff-​a >> Econ >> {Diff-​A, Index-​A,

 

266   Andrej Malchukov Index-​a}), since non-​topical/​new/​indefinite arguments are more likely to be construed as objects than as (transitive) subjects, which are topics par excellence. Thus, asymmetries between DOM and DSM, in the domain of definiteness/​topicality are parallel to those observed in the domain of animacy, and can be also accounted for in terms of the same two constraints.5

11.6  Case-​M arking Strategies and Formal Classes of DCM We started with a discussion of different strategies of case marking—​indexing and differentiating—​and proceeded to a discussion of animacy effects peculiar to these strategies. An interesting question to address is whether the different strategies correspond to different patterns of differential case marking. Let me first introduce the fundamental distinction between two formal types of differential case marking, which are frequently confused in the literature, ‘asymmetric’ and ‘symmetric’ (de Hoop and Malchukov 2008; Malchukov 2008a). The ‘asymmetric’ type of differential marking involves an alternation of an overt case with a zero case (e.g. a nominative, in an accusative system, or absolutive, in an ergative system). The second type dubbed here ‘symmetric’ involves an alternation between two overt cases (such as an ergative/​oblique alternation for DSM and an accusative/​oblique alternation for DOM). Importantly, only the asymmetric type may be due to the differentiating strategy (and economy). It is hardly accidental that Aissen’s (2003) markedness approach was specifically designed to capture variation in asymmetric DOM patterns.6 On the other hand, considerations of disambiguation and economy hardly play a role in the symmetric type, as the opposition of the two markers does not manifest a clear markedness pattern. Indeed, both overt cases involved in a symmetric alternation satisfy the differentiating constraint, and neither of them, being overt, satisfies economy. The two formal types of differential case marking are independent of each other and may co-​occur in a single language. By way of illustration, consider a three-​way DOM in Finnish. As is well known, ACC in Finnish (formally identical to GEN on nouns) is involved in two different case-​alternations (see, e.g., Sulkala and Karjalainen 1992: 212–​19). On the one hand, the partitive case replaces the accusative in order to signal a less 5 

In their cross-​linguistic study of factors underlying DSM, Fauconnier and Verstraete (2014) confirm pervasiveness of patterns of focal ergativity, but claim that preferential marking of focal As can’t be viewed as a reversal of the DOM pattern with marking of topical Ps, because the focus cannot be regarded as the opposite of topicality. Yet what matters here is that A arguments are preferentially associated with (primary) topics/​given information, while P arguments are rather associated with focus/​ new information. Therefore, any deviation from this scenario (e.g. violation of the ‘Given A constraint’) may require extra marking on the markedness approach. 6  See, however, Keine and Müller (2015) for a proposal as to how this account (reconstructed in Distributed Morphology) can be extended to other cases, when the forms in variation reveal a formal asymmetry but both forms are overt.

 

Ergativity and differential case marking    267 affected/​indefinite P or imperfective aspect (see, Kiparsky 1998a for discussion). On the other hand, ACC is replaced by the unmarked (nominative) form when an A argument is missing, as in impersonal and imperative constructions (Comrie 1975). As predicted, the symmetric alternation is due to the indexing strategy, while the asymmetric alternation is due to differentiating strategy. Note also that in the latter case, case-​marking is “global” (e.g. case on P is dependent on the presence of the A argument), a frequent corollary to the differentiating strategy. Similar patterns observed above for two formal types of DOM, are also found in DSM. Here again it is important to distinguish between an asymmetric DSM type (an alternation between ergative and a zero case) and a symmetric alternation (between ergative and an oblique). Again, as in the previous case, only the former pattern can be readily explained in terms of markedness/​differentiation. The asymmetric DSM pattern is found in many split-​ergative languages of the Australian type, where, in accordance with Silverstein’s generalization, ERG case is not available for nominals high on the animacy hierarchy (Silverstein 1976; Blake 1977). The second—​symmetric—​DSM type can be exemplified by the “involuntary agent construction” in Lezgian (Haspelmath 1993). Note that ergative switches to an oblique (“adelative”) case to indicate lack of control volitionality on the part of the agent argument: Lezgian (Haspelmath 1993: 292) (24) Ajal-​di get’e xa-​na. child-​ERG pot.ABS break-​AOR ‘The child broke the pot.’ (25) Zamiira.di-​waj get’e xa-​na. Zamira-​AdEl pot.ABS break-​AOR ‘Zamira broke the pot (accidentally/​involuntarily).’ Clearly this DSM pattern is related to the indexing function, as it pertains to encoding of volitionality/​control. Similar patterns of symmetric DSM manifesting “involuntary agent construction” are attested elsewhere (see Kittilä 2002 for an overview). Thus, also in case of DSM we observe a correlation between a case-​marking strategy and a formal pattern: an indexing strategy yields a symmetric DSM, while a differentiating strategy yields an asymmetric DSM.

11.7  Case-​M arking Strategies and Distributional Types of DCM Above we have noted that differential case marking of the symmetric type shows correlation with the indexing strategy, while the asymmetric type with the differentiating strategy. Earlier a somewhat similar observation has been made by DeLancey (1981)

 

268   Andrej Malchukov who argued that one should distinguish between alternations of the ergative with an overt (oblique) case (as in many consistently ergative languages), and alternations of the ergative with zero (as in split ergative languages of the Australian type). According to DeLancey (1981), the former type is related to agentivity, while the latter type is not. Yet, as noted by Næss (2007), this is an oversimplification since alternations of the latter type (i.e. an asymmetric DSM) may be conditioned by agentivity as well. For example, in Tibetan (Denwood 1999) ERG marking is lost when A is non-​agentive ( Kittilä 2002: 239; Næss 2007): Tibetan (Denwood 1999: 137) (26) Sang.nyin nga-​s las.ka ‘di byed.kyi.yin. tomorrow I-​ERG work this do.LINK.AUX.VOL ‘I shall do this work tomorrow (of my own free will).’ (27) Sang.nyin nga las.ka ‘di byed.kyi.yin. tomorrow I work this do.LINK.AUX.INVOL I shall do this work tomorrow (whether I like it or not).’ Note that even though the DSM in Tibetan is asymmetric, it is driven by agentivity distinctions, not unlike those which we observed for symmetric DSM in Lezgian. A similar point of course can be made for Hindi, where the choice between ergative and nominative case on intransitive subjects is sensitive to volitionality (as in (13)–​(14)). Here, however, an important distinction between asymmetric alternations in split ergative Australian languages such as Dyirbal, on the one hand, and in Tibetan and Hindi, on the other hand, should be noted. The asymmetric marking in the Australian languages is of a “split” type, where different types of nominals select for different cases, while the alternation observed for other languages is of the “fluid” type, where the same nominal can alternatively take both case markers (the terms “split” vs. “fluid” have been adopted by de Hoop and Malchukov (2007) from Dixon’s (1994) classification of split-​ intransitive languages). Importantly, the “fluid” type of transitivity alternation correlates with the indexing strategy, even in situations where case alternation is asymmetric. The reason for this seems to be a general constraint against existence of two synonymous forms:7 if there are two alternative forms in variation there should also be some concomitant meaning differences associated with them. (We abstract away from cases where competition of two forms is indicative of language change.) In case two forms are in a complementary distribution they of course do not violate this constraint. This means, in effect, that whenever some two case forms are in alternation they will correlate with meaning differences, thus falling under the domain of the indexing strategy.8 7  This constraint is well known from the research in the field of lexical semantics/​pragmatics, where it is held responsible for the phenomenon of (lexical) blocking (see Kiparsky’s 1982a constraint Avoid Synonymy). 8  See de Hoop and Malchukov (2007) for an argument that fluid alternations can be adequately described by bidirectional OT, which computes (super)optimal candidate meaning–​form pairs. This is

 

Ergativity and differential case marking    269

11.8  Case-​M arking Strategies and Typology of Differential Case Marking: Recapitulation Above we have observed certain correlations between the type of differential case-​ marking, on the one hand, and case-​marking strategies, on the other. Thus, it was shown that only the asymmetric type can be due to the differentiating strategy. For symmetric DCM, issues of distinguishability and economy are irrelevant, as it does not manifest a markedness pattern. On the other hand, DCM of the fluid type shows a correlation with the indexing strategy. The latter correlation has been attributed to the fact that availability of paradigmatic opposition will always invite a semantic contrast, while there need not be such a contrast in DCM of the split type where two forms are in complementary distribution. Provided that the fluid DCM correlates with the indexing strategy, and the asymmetric DCM correlates with distinguishability, the connections between case-​marking strategies and types of DCM—​both formal and distributional—​can be represented as in Table 11.3. Table 11.3 Differential case-​marking types and case-​marking strategies Symmetric

Asymmetric

Fluid

Indexing

Indexing/​ Differentiating

Split

?

Differentiating

As is clear from Table 11.3, the two types of DCM which show diametrically opposite correlations in terms of strategies involved are the symmetric DCM of the fluid type, on the one hand, and the asymmetric split type, on the other hand. The former illustrated above by the subject alternation in Lezgian (as in (24)–​(25)) is always due to the indexing strategy, while the latter as found in many split ergative languages of the Australian type is due to the differentiating strategy. The other two possible types (the asymmetric fluid type, and the symmetric split type) can’t be unambiguously associated with either

so because the bidirectional OT, as advocated by Blutner (2000) and others, is specifically designed to capture iconicity effects, when a marked (more complex) expression is associated with a more marked (nonstandard) meaning, and the unmarked expression is associated with a standard meaning, whenever the two expressions appear in direct competition.

 

270   Andrej Malchukov of the two strategies. As noted above, the asymmetric fluid DSM, can be either due to the indexing strategy (see DSM in Tibetan, as in (26)–​(27)), or to the “global” differentiating strategy (see DSM in Fore, as in (11)–​(12)). Furthermore, it can be arguably due to both the indexing and the differentiating strategy as observed in the canonical cases of DOM, explained by Silverstein and Aissen in terms of markedness (see, e.g., the DOM pattern in Hindi in (1)–​(3), where ACC marking targets only animate and/​or definite objects). An opposite combination of features, the symmetric split pattern appears to be outside the scope of either indexing or differentiating strategy. Since it is symmetric it cannot be due to the differentiating strategy. On the other hand, since the case forms do not alternate with each other, the choice between the two forms does not invite a semantic contrast, and is more likely to be a morphological matter. Thus, alternative forms for the object case with different classes of nominals, as familiar from Indo-​European languages, are usually taken to represent different declensional types (Baerman 2009: 224), and the same seems to hold for the split realization of the A-​case in ergative languages (e.g. according to Spencer 2006, the two different ergative markers used for animates and inanimates in Chukchi should be regarded as different morphological realizations of a single syntactic case).

11.9  Ergativity, DCM, and Diathetic Alternations It has been observed in the literature, that differential case marking is also dependent on language typology: in particular, DOM is more common in nominative–​accusative languages, while DSM is more common in ergative languages (Bossong 1985; Drossard 1991: 452, 479). A correlation between accusative alignment and availability of DOM alternations, on the one hand, and ergative alignment and availability of DSM alternations, on the other hand, is also evident in the case of split ergative languages with a tense/​aspect based split. Thus, in Indo-​Aryan languages showing a tense-​based split, DSM of the type illustrated for Hindi in (13)–​(14) is only found in the ergative domain (perfective tenses), and is excluded from the accusative domain (imperfective tenses). Although DOM is found in many languages (including Hindi) in both domains, in some languages, like Kashmiri, it is restricted to the accusative domain (see Klaiman 1987: 77). To account for the skewed typology of DCM with respect to ergative and accusative languages, Malchukov (2006) introduces the economy constraints PAIP, which was originally an abbreviation of “Primary Actant Immunity Principle.” • PAIP: Avoid (case) marking of the primary argument. The primary argument (or primary term in Palmer’s 1994 terminology) refers to the argument of a transitive clause which is encoded like the intransitive subject (S). In

 

Ergativity and differential case marking    271 effect, PAIP penalizes case marking of an (otherwise) unmarked argument. Thus, in general, PAIP penalizes morphological case marking of the absolutive argument in ergative languages and of the nominative argument in nominative–​accusative languages. Of course, this can be seen as a reformulation of a constraint that states that one argument should always bear the unmarked case, or to put it differently, that the unmarked case (either nominative or absolutive) is obligatorily present in every sentence of a language. This constraint thus resembles Tsunoda’s (1981b) “Unmarked Case Constraint” as well as Bobaljik’s (1993b) “Obligatory Case Parameter.” On Malchukov’s (2006) account the economy constraint PAIP competes with the iconically motivated Relevance Principle, which is formulated in reliance to Hopper & Thomson’s (1980) transitivity parameters: • Relevance Principle: mark a transitivity parameter on the ‘relevant’ constituent (i.e. on the constituent to which a parameter pertains) Relevance Principle may be seen as a manifestation of Indexing constraints (belonging to Identify family in de Hoop and Malchukov 2008), but in addition carries a locality requirement, which (trivially) assumes that an argument should be marked for its own theta-​role. Relevance Principle would predict that an A-​related parameter (such as subject volitionality, or sentience) will be preferentially encoded on the A argument, while a P-​related parameter (such as P-​affectedness, or P-​individuation) would be preferentially marked on the P. This is, of course, the most usual case: manipulation in P-​related parameters is most often manifested in DOM rather than DSM, while manipulation in A-​related parameters is manifested in DSM rather than DOM (this is also clear from Kittilä’s cross-​linguistic survey of transitivity alternations; Kittilä 2002). Yet, occasional counterexamples to this principle are also found (Næss 2004; Malchukov 2006). Consider the following examples from Shipibo-​Conibo (Valenzuela 1997), also discussed by Kittilä (2002) and Næss (2004). In this Panoan language, an A must be marked by the ergative case only when a P is referential: Shipibo-​Conibo (Valenzuela 1997; Kittilä 2002: 212) (28) e-​n-​ra yapa-​ø pi-​kas-​ai. I-​ERG-​AS fish-​ABS eat-​DES-​INCOMPL ‘I want to eat fish (referential only).’ (29) ea-​ø-​ra yapa-​ø pi-​kas-​ai. I-​ABS-​AS fish-​ABS eat-​DES-​INCOMPL ‘I want to eat fish (referential or non-​referential).’ This case-​marking pattern is thus ‘global’ in Shipibo-​Conibo: as in the Fore examples (11)–​(12) above, case marking of one argument depends on the feature of another clausal argument. Yet, differently from Fore, it is referentiality rather than animacy that is at issue here; therefore this pattern is less naturally explained by distinguishability. On the

 

272   Andrej Malchukov other hand, capturing such global patterns from an indexing perspective is not straightforward either.9 A more general solution would be to attribute this case alternation to economy. On this approach, a markedness/​economy principle like PAIP (Malchukov 2006) would be held responsible for the fact that in an ergative language a P-​feature could affect encoding of A rather than the primary absolutive argument. In the example above, Relevance is violated in order to satisfy PAIP, resulting in P-​ features being A-​marked in an ergative language or A features being marked on P in an accusative language (Malchukov 2006). Similarly, PAIP can be violated in order to satisfy Relevance. An example of a violation of PAIP comes from Warlpiri (Hale 1973a): Warlpiri (Hale 1973a) (30) Njuntulu-​lu npa-​tju ŋatju. 2SG-​ERG 2SG-​1SG speared.1SG ‘You speared me.’ (31)

Njuntulu-​lu npa-​tju-​la ŋatju-​ku. 2SG-​ERG 2SG-​1SG-​la speared.1SG-​DAT ‘You speared at me.’/​‘You tried to spear me.’

Note that violation of PAIP in the ergative-​dative pattern in (31) is due to Relevance: the need to signal decreased affectedness of the P argument. This pattern is, however, rather exceptional for ergative languages which mostly adhere to PAIP (or more generally, to the Unmarked Case Constraint; Tsunoda 1981b). Indeed, in most ergative languages the ERG-​OBL pattern is disallowed, and demotion of a P in a would-​be DOM pattern triggers a shift to an antipassive construction. Warrungu (Tsunoda 1985) (32) Pama-​ngku yuri nyaka-​n. man-​ERG kangaroo-​ABS see-​NF ‘A man saw (found) a kangaroo.’ (33) Pama yuri-​wa nyaka-​kali-​n. man.ABS kangaroo-​DAT see-​APASS-​NF ‘A man was (is) looking for a kangaroo.’

9  One approach to account for such cases has been developed by Næss (2004, 2007). Næss, who generally adopts an indexing approach, suggests that a semantically transitive clause should additionally satisfy the condition of Maximal Semantic Distinguishability of its arguments: if they are not maximally distinguished (e.g. the subject is non-​volitional, or object is non-​affected), the transitive pattern may shift to intransitive. On this approach, a correlation between ergativity and DSM, on the one hand, and accusativity and DOM, on the other hand, is not expected.

 

Ergativity and differential case marking    273 Thus an antipassive construction is regularly used in ergative languages to signal decreased affectedness or referentiality of an object (Cooreman 1994). By using the antipassive construction, the subject becomes the unmarked argument (in the absolutive case), which means that PAIP is fulfilled. Similarly, we see evidence for PAIP in the nominative–​accusative languages as well. While in ergative languages a change in the strength/​prominence of the subject can affect the form of the subject exclusively (resulting in DSM), in nominative–​accusative languages, on the other hand, manipulating the strength/​prominence of an agentive argument regularly leads to passivization. As noted by Shibatani (1985), a passive is primarily used for ‘agent-​defocusing,’ when an agent is indefinite, non-​specific or not important in the discourse. Similarly, in some languages passive forms are used to indicate non-​volitionality of the subject (see Masica 1991 on Sinhala and Dhivehi). In ergative languages, on the other hand, lack of volitionality on the part of the A argument is more likely to be signaled through DSM (see (25) from Lezgian). Thus, features that trigger DSM in ergative languages may cause the use of a passive construction in nominative–​ accusative languages. Again, this can be straightforwardly explained by PAIP. To sum up, we have seen harmonic cases, where Relevance (Indexing) and PAIP reinforce each other. That is, both constraints can be satisfied in case of DOM in nominative–​ accusative languages and DSM in ergative languages. When the two constraints are in conflict, we commonly see that a voice alternation is a way to resolve the conflict. As predicted, passivization is applied when there is need to encode a subject alternation in Table 11.4 The relation between case and voice alternations Input

Output in a nominative–​accusative language

Output in an ergative–​absolutive language

A/​a prominence

Active/​passive voice alternation

DSM

P/​p prominence

DOM

Active/​antipassive voice alternation

a nominative–​accusative language, while antipassivization applies when an object alternation must be encoded in an ergative language (Malchukov 2006). This is illustrated in Table 11.4 (adopted from de Hoop and Malchukov 2008).10

10  This is in line with Legendre et al. (1993) who also argue that passives apply when the input is aP (with a non-​prominent subject), while antipassives apply when the input structure is Ap (with a non-​ prominent object). However, Legendre et al. do not account for the fact that passives are found more often in nominative–​accusative languages, while antipassives are found more often in ergative languages.

 

274   Andrej Malchukov

11.10 Diachronic Issues In conclusion let us briefly address diachronic scenarios behind evolution of DSM in ergative languages. As is clear from the discussion above, there are different types of DSM available; in particular, the distinction should be drawn between alternations of the asymmetric and symmetric type, on the one hand, and alternations of the split and fluid type, on the other hand. We will not attempt a comprehensive coverage of all scenarios leading to these systems. This is because the explanation of the scenarios leading to an asymmetric type amounts to explanation of the developments of the ergative system, which is covered in the volume elsewhere (see McGregor, Chapter 19, on grammaticalization of the ergative case). Suffice it to note that one well attested path of development of an ergative pattern is through reanalysis of an intransitive (resultative or passive) pattern, with the ergative originating in an oblique. More to the point is the development of the symmetric system with an oblique alternating with the ergative case. Also in this case, it is clear that the pattern with an oblique case was originally intransitive. In fact, some of the cases described above in terms of DSM remain synchronically controversial. Thus, the Lezgian pattern with an adelative case in (25) has been described as basically intransitive (Haspelmath 1993), since it is not found with canonical transitives, but rather with verbs that are either intransitive or labile (like ‘break’ in (24)–​(25)). While this analysis may indeed be appropriate for Lezgian, it should be noted that for some related languages, reanalysis of an intransitive to a transitive pattern proceeded further. Thus, in Agul (another Daghestanian language), the involuntary subject construction is also primarily found with intransitive and labile verbs, yet the adelative argument does not show any decrease of syntactic subject properties in comparison to the ergative subject (Ganenkov et al. 2008). At this stage, the non-​volitional agent can be adequately analyzed as a non-​canonical subject, that is, a subject showing exceptional marking, but displaying syntactic characteristics of a regular subject. Finally, at a still later stage, the variation between alternative case patterns may be ‘morphologisized’ and become associated with certain types of nominals, yielding an alternation of the split symmetric type. Thus, while the status of individual languages with respect to the cline may be controversial, the general stages of this development may be represented in the following way (see also Seržant 2013): • 1st stage: the initial construction is intransitive, with an oblique Cause/​Agent • 2nd stage: the agent acquires syntactic subject properties (is reanalyzed as a non-​ canonical subject) • 3rd stage: the alternation is conventionalized as DSM (of the split symmetric type) • 4th (final) stage: the construction is realigned to standard coding This diachronic perspective can also throw light on a controversial question of the taxonomy of differential case-​marking phenomena. It has been claimed by a number of authors, that differential case marking subsumes two different phenomena, some of them

 

Ergativity and differential case marking    275 being syntactic in nature, some other morphological (see, e.g., Woolford 2008; Legate 2014a; see also Spencer 2006 for a distinction between m-​case and s-​case). Other authors, however, treat differential case marking in a unified way, in syntax (e.g. Coon and Preminger, Chapter 10, this volume), or in morphology (e.g. Keine and Müller 2015). The same controversy is also found in the typological literature (see Goddard 1982), in particular, in the discussion of the relation between split ergativity and case syncretism (see also two contributions in the Oxford Handbook of Case by Iggesen (2009) and Baerman (2009) with rather different takes on this issue). From the discussion above, it should be clear that we view these distinctions as a part of a diachronic cline, which indisputably starts from different syntactic structures (possibly differing in transitivity value), which develops through a stage of noncanonical argument marking into a patterns of case syncretism. As also acknowledged by Baerman (2009: 229), who is generally a proponent of the syncretism approach, “the most widespread type of case syncretism, that of the core cases, may in many instances represent the outcome of desyntacticization, that is, the morphologized relic of what was once an active syntactic rule.” Another diachronic scenario leading to ergativity, or, rather, split intransitivity, involves transitive-​to-​intransitive reanalysis. This scenario involves reanalysis of transitive impersonal constructions (‘transimpersonals’; Malchukov 2008b) into a patient subject construction in a split intransitive language. (While split intransitivity might be viewed as a distinct phenomenon, it does yield a restricted ergative S=P alignment). As shown by Malchukov (2008b) this is a common grammaticalization path. This path starts from a transitive construction with a ‘weak’ (non-​agentive, non-​prominent, non-​ referential) A and involves reanalysis of a prominent P argument into a non-​canonical subject (and later eventually into a canonical subject). Koasati (Kimball 1991), like many other Muskogean languages, may serve as an illustration. In Koasati, a number of intransitive (unaccusative) verbs pattern with transitives insofar as the experiencer argument is indexed through object agreement (rather than subject agreement used elsewhere). In fact, if not for the case marking (which operates on the nominative–​accusative basis and independently of agreement), the patient subject pattern in (34) could be confused with a transitive pattern in (35): Koasati (Kimball 1991: 251) (34) (Anó-​k) ca-​libatli-​t. (I-​NOM) 1sg.obj-​burn-​past ‘I got burned.’ (35) Nihahci ikba-​k ca-​libatli-​t. grease hot-​NOM 1sg.obj-​burn-​past ‘The hot grease burned me.’ Given similarity of the two constructions, Kimball (1991) suggested that diachronically the patient subject pattern originated from reanalysis of transitive verbs with a third person (singular) agent. As independently observed by Mithun (2008) and Malchukov

 

276   Andrej Malchukov (2008b), the features which facilitated this reanalysis are:  (a)  the presence of object agreement on the verb; (b) the zero marking of a 3rd person subject. Now, as shown in both publications for Amerindian and other languages, these two features are typical of split intransitive languages, suggesting that transimpersonal constructions constitute an important historical source for split intransitivity (Malchukov 2008b; see Malchukov and Ogawa 2011 for further discussion). As noted by Malchukov (2008b), the functional motivation behind reanalysis of transimpersonal constructions into the patient subject construction is the functional pressure for realignment of the most prominent argument with the most prominent grammatical relation. The same explanation pertains to other cases of reanalysis, such as reanalysis of object experiencers into subject experiencers (Haspelmath 2001). Such cases of “please-​to-​like reanalysis” are well known from English (Allen 1995) and are also attested in other European languages (see Seržant 2013 for some discussion). Interestingly, the transimpersonal scenario above is arguably responsible for one of famous generalizations in generative literature, Burzio’s generalization. (36) Burzio’s generalization: All and only the verbs that can assign an external theta role to the subject can assign accusative Case to an object. (Burzio 1986: 178) The synchronic validity of this generalization may be contested; in fact, the transimpersonal constructions of the type illustrated above may be viewed as counterexamples, if taken at face value. Also, the theoretical status of this generalization remains controversial; in spite of its popularity there does not seem to be a consensus, how it can be derived from a more general principle (see Reuland 2000 for a variety of proposals). Now, I would like to suggest that this generalization makes a perfect sense diachronically, insofar as transimpersonal constructions lacking a referential A tend to be reanalyzed into intransitive structures. Thus the rationale behind this generalization should be stated in diachronic terms.11

11.11 Conclusions The present chapter argues that patterns of (differential) case marking can be explained in reliance to two general constraints related to (role)-​indexing, on the one hand, and

11 

In a recent article, Faarlund (2013) makes a somewhat similar suggestion relating reanalysis of object experiencers into subject experiencers to the loss of the pro-​subject in mainland Scandinavian languages. However, Faarlund does not explicitly relate reanalysis of transimpersonal constructions in Scandinavian to Burzio’s generalization.

 

Ergativity and differential case marking    277 distinguishability (or markedness) on the other hand. This approach correctly predicts asymmetries between DOM and DSM with regard to animacy, definiteness, as well as discourse features. With regard to all of these features, DSM has been found less consistent than DOM, in particular, an ergative case may be associated with either less prominent arguments (the “Silverstein pattern”), or more prominent ones (the “anti-​ Silverstein pattern”).12 The same approach can be also used to explain functional import of different varieties of DCM, symmetric, and asymmetric, split and fluid. I also show how this approach can be extended to capture a relation between case and voice alternation, as well as briefly outline diachronic scenarios leading to different types of differential case marking. A number of questions have been left unaddressed in this chapter for space reasons. Thus, I have not discussed TAM-​based split ergativity, which shows conceptual overlap with differential case marking. Yet, the approach outlined above can be naturally extended to cover such cases. Indeed, Malchukov (2015) and Malchukov and de Hoop (2011) present a comprehensive hierarchy of TAM-​based ergativity splits, and show how these splits can be represented on the basis of the constraints assumed above (indexing, distinguishability, economy). In order to capture a distinction between languages with morphological and syntactic ergativity one needs to invoke additional factors, yet also this distinction can also be fruitfully explored in a typological model based on competing motivations (which can be recast as OT constraints). In particular, Malchukov (2014) explains how patterns of syntactic ergativity/​accusativity can be captured through interaction of Bias constraints (embodying functional preferences) with Harmony constraints (embodying the analogical influence from coding to behavior). Other contributions to the same volume (MacWhinney, Malchukov, and Moravcsik 2014), provide further evidence for the power of the competing motivations approach.

Acknowledgments The research reported in this chapter has been carried out partially in collaboration with Helen de Hoop; I am grateful to Helen for useful feedback and discussions in the course of

12 

There are still few large scale typological statistical studies testing these predictions. One of such studies addressing both DSM and DOM (Bickel et al. 2015) found that DSM is indeed less consistent than DOM, which is in line with the present approach. Yet, Bickel and coauthors further showed that even for DOM evidence for the hierarchy effects is weaker than expected and furthermore that DOM patterns are prone to areal diffusion. On the other hand, the study by Sinnemäki (2014), which is confined to DOM, confirmed that DOM is a typologically predominant pattern; in fact, it was found to be more common than a consistent object marking. However, also this study could not identify either animacy or definiteness as the main contributing factors behind DOM. This finding, if confirmed, is consistent with approaches viewing animacy effects as epiphenomenal (see Malchukov 2008a), as well as with approaches treating definiteness effects as derivative from discourse factors (see Iemmolo 2010; Darlymple and Nikolaeva 2011).

 

278   Andrej Malchukov our collaborative work. I am also grateful to the editors for editorial feedback as well as to the anonymous reviewer for the useful comments on an earlier version of this chapter. The usual disclaimers apply.

Abbreviations ABS, absolutive; ACC, accusative; AdEl, adelative; AGT, agentive; AOR, aorist; APASS, antipassive; ART, article; AS, ascertained (direct) evidential; AUX, auxiliary; CAUS, Causative; D, dative; DAT, dative; DES, desiderative; DIR, directional; DIST, distal deictic aspect; ERG, ergative; FA, factive aspect; FS, feminine singular; INVOL, involuntary; INCOMPL, incompletive; LINK, linker; LOC, locative; MED, medial deictic aspect; NF, non-future; NOM nominative; OBJ, object; OT Optimality Theory; P, patient; PAST, past; PERF, perfective; SG, singular; VOL, voluntary.

 

Chapter 12

Three-way syst e ms d o not ex i st Gereon Müller and Daniela Thomas

12.1 Background There are three main types of encoding systems (via case or agreement) for core arguments in the world’s languages, where core arguments are external (DPext) and internal arguments (DPint) of transitive (Vt) and intransitive verbs (Vi) that are not lexically case-​ marked. In ergative systems, DPext of Vt is marked in some special way, as ergative, that is different from the marking of DPint of Vt, DPext of Vi (unergative intransitive context), and DPint of Vi (unaccusative intransitive context), all of which bear absolutive; see (1-​a). In contrast, in accusative encoding systems, the DPint of Vt is singled out (by bearing accusative), and the remaining core arguments are grouped together (nominative); see (1-​b). Finally, in active systems, argument encoding is oblivious to transitivity: DPext is encoded in one way, and DPint is encoded in some other way, with both Vt and Vi; see (1-​c). Such an active system of argument encoding may then in principle qualify either as an ergative system at its core (where the distribution of the ergative is extended to DPext of Vi; see Dixon (1994)), or as a modified accusative system (where the distribution of the accusative is extended to DPint of Vi; see Bittner & Hale (1996a,b)). (1)  a. Ergative system

b. Accusative system

c. Active system

DPext-Vi DPint-Vi

DPext-Vi DPint-Vi

DPext-Vi

DPint-Vi

DPext-Vt DPint-Vt

DPext-Vt DPint-Vt

DPext-Vt

DPint-Vt

erg/nom

abs/acc

erg

abs

nom

acc

These basic systems of argument encoding have been given simple accounts in Principles-​and-​Parameters-​type approaches. A common basic idea that has often been

 

280    Gereon Müller and Daniela Thomas pursued is that there are uniformly two functional heads (which for present purposes we may identify with T and v) that assign one structural case each in transitive contexts; and only one of the two case assigners remains active in intransitive contexts. On this view, of the four cases currently under consideration—​ergative, absolutive, nominative, accusative—​there are actually only two: one assigned by T, and one assigned by v. Throughout this article, we adopt a version of this approach according to which ergative is identified with accusative, and nominative with absolutive; see Murasugi (1992), Jelinek (1993), Ura (2000; 2006), Müller (2009), and Assmann et al. (2015).1 More specifically, given a TP-​vP analysis of clause structure, ergative/​accusative is a structural case assigned by v to either Specv or CompV in transitive contexts, and nominative/​absolutive is uniformly the case assigned by T. As shown in (2), on this view ergative and accusative systems differ in transitive contexts. Here v assigns case to DPext in ergative systems, and to DPint in accusative systems. T assigns case to the remaining argument (i.e. DPint in ergative systems, and DPext in accusative systems).2 (2)  Transitive context TP T'

T

nom abs

vP v'

DPext v acc

1 

erg V

VP DPint

Alternatively, it has been suggested that ergative is to be identified with nominative, and accusative with absolutive; see Levin & Massam (1985), Bobaljik (1993a), Laka (1993b), Chomsky (1995, ch. 3), Rezac (2003), and Bobaljik & Branigan (2006), among others. Most of what follows could also easily be made compatible with this view. 2  Here and in what follows, accusative encoding is represented by dashed lines in syntactic trees, ergative encoding by full lines. Note also that in our reconstructions of existing approaches, we adopt an Agree-​based approach (as in Chomsky (2001; 2008)) throughout, and disregard the possibility of case assignment being tied to movement of DPs.

 

Three-way systems do not exist    281 On this view, the parameter distinguishing ergative and accusative systems exclusively concerns v: Both upward and downward case assignment must be possible in principle, but there is a preference for upward case assignment in ergative systems, and a preference for downward case assignment in accusative systems.3 In contrast, ergative and accusative systems work in exactly the same way in intransitive contexts: Only T remains as a case-​assigning head here, with both DPint (see (3-​a)) and DPext (see (3-​b)). This corresponds directly to tendencies of morphological marking: The case associated with T is typically morphologically less marked than verbal case assigned by v. (3)  Intransitive unaccusative and unergative contexts a.

TP

b.

TP

T' vP

T nom

T'

v'

–

vP

T

v abs V

abs

VP DPint

v'

DPext

nom

v

VP V

–

This general approach can then be extended to active systems in various ways. For instance, one can postulate that whether T or v is the sole case assigner in intransitive contexts may be governed by different conditions, among which languages can choose. To sum up so far, this kind of analysis provides a simple account of the basic patterns of argument encoding in the world’s languages. Given the identification of two case-​ assigning heads with two structural cases, the question arises of how this approach fares with three-​way (tripartite) systems of argument encoding, as they are schematically depicted in (4) (see Dixon (1994), among others). 3 

Assmann et al. (2015) explicitly postulate a “Specifier-​Head Bias” which, if accompanied by an opposite “Complement-​Head Bias,” would capture the parametrization.

 

282    Gereon Müller and Daniela Thomas (4)  Three-way system: DPext-Vi

DPint-Vi nom/abs

DPext-Vt

DPint-Vt

erg

acc

In a three-​way system, there are three structural cases for core arguments, i.e. nominative/​absolutive for the sole DP of Vi, ergative for DPext of Vt, and accusative for DPint of Vt. As described in Bittner & Hale (1996b), the Central Australian language Antekerrepenhe instantiates such a three-​way system; see (5). (5) a. Arengke-​le aye-​nhe dog-​erg me-​acc ‘The dog bit me.’

ke-​ke bite-​pst

b. Apwerte-​le  athe    arengke-​nhe  we-​ke stones-​ins   I-​erg  dog-​acc    pelt-​pst ‘I pelted the dog with stones.’ c. Arengke     nterre-​ke dog-​nom  run-​pst ‘The dog ran.’ Three-​way systems are cross-​linguistically rare.4 However, they are potentially problematic for the type of analysis sketched above, where two case assigners (T, v) are responsible for two structural cases and each of {erg, acc} is identified with exactly one case of the other system.5 What can be done in view of this state of affairs? There are basically two kinds of solutions to the problem posed by three-​way systems. One possible way out is to enrich the

4  They qualify as non-​canonical from a typological perspective (see Corbett (2005); Corbett & Fedden (2014)). Note that this assessment presupposes that the voice systems of Austronesian languages do not instantiate three-​way patterns (with an accusative in actor topic contexts and an ergative in theme topic contexts); see Aldridge (2004), Paul & Travis (2006) for discussion. 5  This conclusion does not necessarily hold for approaches where structural case assignment in transitive contexts is relational; see Marantz (1991), Bittner & Hale (1996b), Wunderlich (1997; 2006), Kiparsky (1999), Stiebels (2002), McFadden (2004), Schäfer (2012a), and Baker (2015). Here, there are two separate sources for the ergative and for the accusative; and nominative/​absolutive is a default case for DPs in intransitive contexts. However, it is not clear whether these analyses have anything interesting to say about the cross-​linguistic rarity of three-​way systems. For this reason, we will disregard relational approaches in what follows.

 

Three-way systems do not exist    283 syntactic analysis with ad hoc assumptions about additional case features. Such a strategy will technically work (see Müller (2009)), but it does not strike us as particularly insightful. A second option is to leave the syntactic analysis as is, thus exclusively deriving two-​way systems, and relocate the phenomenon to morphology. On this view, three-​ way case systems do not exist as a syntactic phenomenon. It is this latter view that we will pursue in what follows. The challenge for a morphological reanalysis of seemingly tripartite systems is then twofold. First, it must be shown that a simple morphological analysis can indeed be given, one that does not have to resort to unnatural classes (see Bierkandt (2006) for this objection, based on evidence from Diyari), and that is ideally independently corroborated. And, second, it must be shown that the hypothesis that syntactically there are only two cases where traditionally three have been postulated, can be maintained in view of the classes of DP arguments that syntactic operations access in the relevant languages. The main bulk of the present chapter addresses the first question in sections 12.2 and 12.3; we turn to the second question in section 12.4. A first indication that a morphological reanalysis of three-​way systems might be worth pursuing is that independent evidence for distinguishing between case as a syntactic category and case as a morphological exponent has recently come to the fore. On the one hand, one and the same morphological case exponent may correspond to two different syntactic cases. Thus, as has been argued in Legate (2008), a zero exponent may either indicate a syntactic nominative, or it may act as the default realization of some other case (like accusative or ergative), depending on the language. On the other hand, one and the same syntactic case may correspond to two different morphological case exponents in a given language. In particular, it is argued in Keine & Müller (2011; 2015) that scale-​based differential object marking should be realized as a morphological phenomenon (i.e. as scale-​driven allomorphy associated with a single syntactic case). Crucially, three-​way systems typically also involve scale effects, such that, e.g., only non-​prototypical DPint arguments receive what at first sight looks like an accusative, or only non-​prototypical DPext arguments bear what at first sight looks like an ergative. This situation obtains in Nez Perce (see Rude (1985), Woolford (1997)): There is (what has been called) an accusative exponent for DPint of Vt; there is (what has been called) an ergative exponent for DPext-​3rd-​person of Vt (but not for other, more typical DPext types with Vt, i.e. for 1st/​2nd-​person); and there is a nominative (zero) exponent for DPint, DPext of Vi, and for DPext-​1st/​2nd-​person of Vt. All this is shown by the data in (6). (6) a. Kaa wéet’u’ núun-​ e    ká’la hinéesqicxne and  not     1pl-​acc  just    3nom.pldo.take.care.of.perf ‘And he just didn’t take care of us.’   (Rude (1985, 93)) b. ‘Iceyéeye-​ nm xáxaasna   hináaswapci’yawna coyote-​erg    grizzly-​acc  3nom.pldo.kill.perf ‘Coyote killed the grizzlies.’

(Rude (1985, 88))

 

284    Gereon Müller and Daniela Thomas c. (i) Núun  ∅-​papáayna we    1/​ 2nom-​pl.nom.arrive.perf ‘We arrived.’ (ii) Núun ’epe’wíye we    1/​ 2trANS.shoot.perf ‘We shot him.’

 (Rude (1985, 85))

Given this state of affairs, our main goal in what follows is to show that the morphological approach to differential object marking in terms of scale-​driven impoverishment developed in Keine & Müller (2011; 2015) (on the basis of Aissen (1999b; 2003)) can be extended to three-​way systems without major problems. It turns out that only one important new assumption is required: In addition to the standard prominence scales related to person, animacy, and definiteness (going back to Hale (1972) and Silverstein (1976)), we postulate that there is also a transitivity scale which participates in harmonic alignment processes that eventually bring about post-​syntactic impoverishment. We will argue that, as a consequence of scale-​driven impoverishment, case features are deleted in certain contexts (intransitive and “prototypical” transitive contexts in particular), and this leads to zero exponence with certain arguments bearing structural ergative or accusative case. Thus, we contend that what seem to be three-​way systems on the surface are underlyingly common ergative or accusative systems with overt markers for each of the two cases that disappear in intransitive (and, typically, other) contexts.

12.2  Theoretical Assumptions The optimality-​theoretic approach to prominence scale-​based differential argument encoding developed in Aissen (1999b; 2003) in terms of harmonic alignment and local conjunction does not distinguish between case as a morphological category and case as a syntactic category; it predicts that variation in, say, differential object marking can only be between an overt case exponent (i.e. presence of case) and no exponent (i.e. absence of case). In Keine & Müller (2015), it is shown that variation in argument encoding that is governed by exactly the same prominence scales can also be between two different overt case exponents; so there are non-​zero/​non-​zero alternations just as there are non-​zero/​zero alternations. To capture both kinds of effects, Aissen’s approach is reconstructed as an optimization procedure that applies at the interface between syntax and (post-​syntactic) morphology, and that deletes certain (but not necessarily all) subfeatures of syntactic cases that must independently be assumed to capture instances of syncretism (see Jakobson (1936), Bierwisch (1967)). For instance, assuming that the accusative is composed of the abstract case features [+gov,–​obl] in syntax, post-​syntactic optimization may result in the deletion of [–​obl]

 

Three-way systems do not exist    285 but not [+gov] in certain (prototypical) contexts, and subsequent vocabulary insertion (Halle & Marantz (1993)) may then choose a vocabulary item /​α/​bearing only the feature [+gov] rather than the otherwise expected, more specific vocabulary item /​β/​ characterized by the features [+gov,–​obl] because the latter is not compatible with the insertion site anymore (it does not realize a subset of the features of the syntactic context). Thus, these optimization procedures can be viewed as principled versions of impoverishment rules as they have widely been adopted in Distributed Morphology (see Halle & Marantz (1993)). Of course, in many instances the less specific morphological exponent that must be chosen after feature deletion will be a zero exponent; but it does not have to be, and this provides an argument for a morphological (rather than syntactic) approach. All that said, whereas we will presuppose the morphological version of Aissen’s harmonic alignment approach to differential argument encoding developed in Keine & Müller (2011; 2015) in what follows, we will not actually consider data where there is a non-​zero/​non-​ zero alternation; rather, all alternations discussed below will be between non-​zero and zero exponents. To begin with, suppose that the core structural cases are defined by the features in (7). (7)

Feature decomposition of cases a. ergative/​accusative: [+gov–​obl] b. absolutive/​nominative: [–​gov,–​obl]

(assigned by v) (assigned by T)

Ergative and accusative are [+governed] cases assigned by v, and absolutive and nominative are [–​governed] cases assigned by T. Both cases are structural (i.e. [–​oblique]), which separates them from lexical and oblique cases; however, since the latter do not play a role for the data under consideration in this chapter, we will generally ignore the feature [±obl] in what follows. Next, consider the prominence scales in (8). The scales in (8-​abc) go back to Hale (1972), Silverstein (1976), and Aissen (1999b; 2003). We take them to be ontological primitives in grammatical theory but will remain neutral as to their ultimate source (part of the language faculty or grounded in some extralinguistic domain). (8)

Scales: a. Person scale: Local Pers. (1,2) > 3. Pers. b. Animacy scale: Hum(an) > Anim(ate) > Inan(imate) c. Definiteness scale: Pro(noun) > Name (PN) > Def(inite) > Indefinite Specific (Spec) > NonSpecific (NSpec) d. Transitivity scale: vt(rans) > vi(ntrans)

 

286    Gereon Müller and Daniela Thomas The transitivity scale in (8-​d) is a new scale that we postulate; it will play an important role in deriving three-​way systems in a syntactic approach that only envisages two cases. The transitivity scale presupposes that transitive and intransitive v can be distinguished, in both ergative and accusative languages. This is unproblematic under the present syntactic analysis, where v is uniformly the inactive head in intransitive contexts. In addition to these scales (which can in principle have arbitrarily many members), a basic binary scale is needed in the theory of harmonic alignment. We assume that this is the DP case scale in (9), which simply states that DPs with the marked value for the feature [±gov] (which captures DPs bearing ergative/​accusative) are more prominent than DPs with the unmarked value (which captures DPs with nominative/​absolutive). This scale replaces the grammatical function scale in Aissen (1999b; 2003) and Keine & Müller (2011; 2015).6 (9)

DP case scale: DP[+gov] > DP[−gov]

Harmonic alignment (Prince & Smolensky (2004)) then takes a basic scale X > Y and another scale a > b > … > z, and produces a pair of complex markedness scales X/​a ≻ X/​b ≻ ... ≻ X/​z and Y/​z ≻ … ≻ Y/​b ≻ Y/​a; these scale are then transferred to fixed hierarchies of constraints with reversed order: *X/​z ≫… ≫ *X/​b ≫ *X/​a; *Y/​a ≫ *Y/​b ≫ … ≫ *Y/​z. Harmonic alignment of the binary DP case scale with the transitivity scale yields the two constraint hierarchies with invariant internal order in (10). (10)  a. *DP[+gov]/​vi ≫ *DP[+gov]/​vt b. *DP[−gov]/​vt ≫ *DP[−gov]/​vi As it stands, constraints like *DP[−gov]/​vt and *DP[−gov]/​vi in (10-​b) indiscriminately block configurations where a nominative/​absolutive DP co-​occurs with vt or vi in a clause. This is not yet adequate because it is not the DPs themselves that need to be deleted in certain (prototypical) contexts, but rather their case features. For this reason, following Aissen (1999b; 2003), a faithfulness constraint Max(case) demanding case feature preservation is locally conjoined with the members of the hierarchies derived by harmonic alignment, which is then counteracted by a general markedness constraint demanding case feature deletion.7 Local conjunction is a mechanism introduced by Smolensky (1995) (also see Legendre et al. (1998), Smolensky (2006)). 6 

This then qualifies as a second, though minor, difference, in addition to the introduction of the transitivity scale. The underlying reason why the grammatical function scale (consisting of “subject,” i.e., Specv, and “object,” i.e., CompV) that Keine & Müller (2011; 2015) adopt from Aissen (1999b; 2003) needs to be replaced is that the notion of subject becomes unclear when ergative systems and intransitive contexts are taken into consideration (which the analysis in Keine & Müller (2011; 2015) is not concerned with). 7  Strictly speaking, the role of the faithfulness constraint Max(case) is played by the markedness constraint *Øcase in Aissen’s original approach, where case is not yet present in an input.

 

Three-way systems do not exist    287 Under local conjunction, two constraints A, B are combined to form a new constraint A&B which inherently outranks both A and B. A&B is violated if both conjoined constraints are violated (in a certain local domain). Importantly, local conjunction of a constraint A with members of a fixed constraint hierarchy B1 ≫ B2 ≫ … ≫ Bn derived by harmonic alignment preserves order. For the case feature [–​gov], the two additional constraints that play a role are given in (11) (analogous constraints exist for [+gov]; see below). (11)

a. Max(case): Preserve case features. b. *[–​gov]: Avoid the feature [–​gov].

Max(case) is a faithfulness constraint that demands preservation of a case feature in a syntactic structure in the post-​syntactic morphological component; this constraint can be conjoined with a constraint hierarchy derived from harmonic alignment. On the other hand, *[–​gov] is a markedness constraint that forces deletion of [–​gov] before vocabulary insertion—​i.e. it brings about impoverishment; this constraint cannot be conjoined with a constraint hierarchy (see Aissen (1999b; 2003), Keine & Müller (2011; 2015)). The result of local conjunction of Max(case) with the fixed constraint hierarchy in (10-​b) is given in (12). (12)

*DP[−gov]/​vt & Max(case) ≫ *DP[−gov]/​vi & Max(case)

The constraint *DP[−gov]/​vt & Max(case) is violated by a post-​syntactic (pre-​vocabulary insertion) representation if there is a nominative/​absolutive DP in a transitive clause that has its [–​gov] feature deleted; similarly, *DP[−gov]/​vi & Max(case) is violated if a nominative/​absolutive DP in an intransitive clause gets its [–​gov] feature deleted.8 If both these constraints outrank *[–​gov], all syntactic [–​gov] case features will be preserved 8 

A technical question arises regarding the dual use of case features in a constraint like *DP[−gov]/​ vt & Max(case): It must be ensured that a case feature like [–​gov] that is deleted (thereby violating MAX(case)) can still be accessed so as to determine the violation (i.e., [–​gov] is needed to characterize the class of DPs that are subject to the constraint). In principle, there would seem to be two possibilities. First, one could distinguish between deletion and erasure, as it is suggested in a structurally similar context in Chomsky (1995): On this view, deleted material would be inaccessible for morphological realization, but still accessible for constraint evaluation. Second, one can postulate that constraints like *DP[−gov]/​vt & Max(case) are not only output-​sensitive, but also input-​ sensitive (see Trommer (2006)). Thus, [–​gov] in “*DP[−gov]/​vt” refers to the input (i.e., the syntactic representation where feature deletion is not yet an issue), whereas [–​gov] in “Max(case)” refers to the output (i.e., the post-​syntactic representation in which feature deletion may or may not have applied). In view of the fact that the relevant constraints arise as a result of combining two separate constraints by local conjunction, and given that we are dealing with syntax/​morphology interface optimization procedures, this second solution strikes us as vastly superior, and we will adopt it in the remainder of this chapter.

 

288    Gereon Müller and Daniela Thomas post-​syntactically, and the regular exponents for nominative/​absolutive will show up as a result of vocabulary insertion. If *[–​gov] outranks both these constraints, all syntactic [–​gov] case features will be deleted post-​syntactically, and zero marking will often arise—​or at least a “retreat to the general case” (see Halle & Marantz (1993)), depending on whether or not there are other, more underspecified non-​zero exponents that can realize what is left (e.g. [–​obl]). The most interesting situation is where *[–​gov] is interspersed between *DP[−gov]/​vt & Max(case) and *DP[−gov]/​vi & Max(case) in a language, as in the ranking in (13). (Recall that the order of the two relativized faithfulness constraints themselves is fixed once and for all, as a consequence of harmonic alignment.) (13) *DP[−gov]/​vt & Max(case) ≫ *[–​gov] ≫ *DP[−gov]/​vi & Max(case) Now [–​gov] will be preserved post-​syntactically in transitive contexts but deleted in intransitive contexts. Subsequent vocabulary insertion can then lead to a [–​gov]-​marked exponent as a case marker for DP in transitive contexts, but given that vocabulary insertion obeys the Subset Principle (Halle & Marantz (1993), Halle (1997)), it will have to resort to an underspecified (typically zero) exponent not bearing [–​gov] in intransitive contexts. We would like to suggest that this represents one basic situation with apparent three-​way systems:9 What at first sight looks like a separate occurrence of an accusative and an absolutive marker emerges as an allomorphic realization of an absolutive marker; there is no accusative present at any point in the derivation. At this point, one may think that a viable alternative to feature deletion via optimization based on harmonic alignment and local conjunction might be to postulate an appropriate impoverishment rule like the one in (14) (with the contextual information interpreted loosely, not necessarily requiring either adjacency or the linearization indicated here). (14) [–​gov] → Ø /​ DP_​_​[vi] However, whereas (14) simply stipulates the context in which deletion takes place, (13) derives this context. In addition, (13) (again in contrast to (14)) predicts that there can be no language where deletion of [–​gov] takes place in transitive but not in intransitive contexts; and we take this to be a correct generalization. Another difference that strikes us as even more important will become clear below: As noted above, three-​way systems typically also involve (other) scale effects; so it remains to be shown how harmonic alignment and local conjunction with the other scales can be brought into the picture. It will turn out that the optimization approach captures these multidimensional scale effects in a fairly straightforward way whereas a standard, rule-​based impoverishment approach will face what look like insurmountable obstacles because the deletion contexts do not form natural classes.

9 

The other possible situation is completely analogous, with [–​gov] replaced by [+gov]; see below.

 

Three-way systems do not exist    289 Still, before addressing this issue by carrying out some case studies, a general question needs to be clarified that concerns locality. In order to evaluate a constraint like *DP[−gov]/​vt & Max(case) or *DP[−gov]/​vi & Max(case), both the properties of the DP (either DPext or DPint) and the properties of v must be taken into account. This suggests that the local domain for constraint evaluation at the interface might be the phase (see Chomsky (2001)), with feature deletion and vocabulary insertion applying cyclically.10 Against this background, let us now turn to some case studies. We will consider putative three-​way systems in Kham, Djapu, Nez Perce, Upriver Halkomelem, and Dyirbal, and we will show how they can all be given simple accounts in terms of syntax/​morphology interface optimizations via harmonic alignment and local conjunction.

12.3 Case Studies 12.3.1  Kham 12.3.1.1 Data The Tibeto-​Burman language Kham has been argued to rely on a three-​way system of argument encoding by case. The distribution of the case markers is shown in (15) (cf. Watters (2002, 66f.)). (15)

Distribution of case markers 1st

2nd 3rd, definite

3rd, indefinite

DPext-​Vt

-​∅

-​∅

-​e/​-​ye

-​e/​-​ye

DPext/​int-​Vi

-​∅

-​∅

-​∅

-​∅

DPint-​Vt

-​lai -​lai

-​lai

-​∅

This can be taken to mean that there is a nominative/​absolutive case for sole DPext/​int-​ Vi arguments, for prototypical (local person) DPext-​Vt arguments, and for prototypical (3rd person indefinite) DPint-​Vt arguments; an ergative case for marked 3rd person DPext-​Vt arguments; and finally, an accusative case for marked (local person or 3rd person definite) DPint-​Vt arguments. Clearly, a system of rules for syntactic case assignment designed to capture this distribution would necessarily be much more complex than is standardly assumed for structural case. In contrast, as indicated in (15), we argue that Kham basically exhibits a standard ergative system in the syntax, with -​e/​-​ye as the 10 Postulating cyclic phase-​based spell-​out will also ensure that *DP [−gov]/​vt & Max(case) cannot be

violated by deletion of some other case feature (e.g. of *DP[+gov]) in the presence of DP[−gov]/​vt in the same phase: Spell-​out of the complement XP of a phase head Y can only delete features in XP (but is sensitive to the properties of Y); and subsequent spell-​out of the phase edge (as part of the next higher phase complement) cannot access previously spelled-​out material again.

 

290    Gereon Müller and Daniela Thomas canonical ergative marker and -​lai as the canonical absolutive marker. On this view, the simple person-​based split in ergative contexts, and the more complex transitivity-​/​ definiteness-​based split in absolutive contexts, are instances of allomorphic variation reducible to scale-​driven optimization.11 Let us address the two cases in turn, beginning with the more complex situation with absolutive realization.

12.3.1.2 Absolutive Marking In addition to the binary DP case scale as the basic scale (see (9)), (15) illustrates that two scales are relevant, namely, the transitivity scale (see (8-​d)) and the definiteness scale (see (8-​c)). Harmonic alignment of the case scale with the transitivity scale and the definiteness scale yields the two constraint hierarchies for absolutive ([–​gov]) DPs in (16). (16) a. *DP[−gov]/​vt ≫ *DP[−gov]/​vi b. *DP[−gov]/​Pro ≫ *DP[−gov]/​PN ≫ *DP[−gov]/​Def ≫ *DP[−gov]/​Spec ≫ *DP[−gov]/​NSpec Next, the two hierarchies with fixed internal rankings thus derived are locally conjoined with one another, giving rise to two-​dimensional local conjunction (see Aissen (2003)). Here, each constraint of one hierarchy is locally conjoined with each constraint of the other hierarchy, preserving original orders, as before. The new hierarchies that result are given in (17-​ab).12 (17) a. *DP[−gov]/​Pro/​vt ≫ *DP[−gov]/​PN/​vt ≫ *DP[−gov]/​Def/​vt ≫ *DP[−gov]/​Spec/​vt ≫ *DP[−gov]/​NSpec/​vt b. *DP[−gov]/​Pro/​vi ≫ *DP[−gov]/​PN/​vi ≫ *DP[−gov]/​Def/​vi ≫ *DP[−gov]/​Spec/​vi ≫ *DP[−gov]/​NSpec/​vi c. *DP[−gov]/​Pro/​vt ≫ *DP[−gov]/​Pro/​vi d. *DP[−gov]/​PN/​vt ≫ *DP[−gov]/​PN/​vi e. *DP[−gov]/​Def/​vt ≫ *DP[−gov]/​Def/​vi  f. *DP[−gov]/​Spec/​vt ≫ *DP[−gov]/​Spec/​vi g. *DP[−gov]/​NSpec/​vt ≫ *DP[−gov]/​NSpec/​vi Finally, the hierarchies in (17) are locally conjoined with Max(case), again preserving original orders. Thus, we end up with constraints like *DP[−gov]/​Pro/​vt & Max(case); this constraint is violated if there is an absolutive DP that is a pronoun, and the absolutive DP shows up with a transitive verb, and its [–​gov] case feature is deleted—​i.e. if the case feature of an absolutive pronoun is deleted in transitive contexts. 11 

This implies that the absolutive can indeed be a non-​zero case throughout in the languages that we consider here, notwithstanding its cross-​linguistic tendency to be less marked segmentally. See Handschuh (2014). 12  Following Aissen (2003), we adopt a simplified notational variant where, e.g. “*DP [−gov]/​Pro/​vt” stands for “*DP[−gov]/​Pro & DP[−gov]/​vt” (where linear order of the two conjuncts is irrelevant).

 

Three-way systems do not exist    291 As a consequence, a two-​dimensional system of argument encoding arises where some constraint pairs exhibit a fixed ranking, and others do not (such that languages simply can choose how they rank the constraints with respect to one another). Following Aissen’s (2003) conventions, fixed and variable rankings among the constraints generated by successive harmonic alignment and local conjunction with Max(case) can be represented as in (18). In this graph, constraints that stand in a domination relation invariantly have a fixed ranking (as a consequence of the mechanics of harmonic alignment and local conjunction), whereas constraints that do not stand in a domination relation are freely ordered with respect to each other. (18)

Absolutive allomorphy in Kham *DP[−gov]/Pro/vt I: /lai/

& Max(case)

*DP[−gov]/PN/vt

*DP[−gov]/Pro/vi

& Max(case)

& Max(case)

*DP[−gov]/Def/vt

*DP[−gov]/PN/vi

& Max(case)

& Max(case)

*DP[−gov]/Spec/vt

*DP[−gov]/Def/vi

& Max(case)

II: / /

& Max(case)

*DP[−gov]/NSpec/vt

*DP[−gov]/Spec/vi

& Max(case)

& Max(case)

*DP[−gov]/NSpec/vi & Max(case)

All the constraints in (18) demand case feature preservation. At this point, the ranking of the conflicting constraint demanding case feature deletion becomes relevant: *[–​gov] leads to zero-​marking for DPs with the feature combinations identified by the constraints that are ranked below it. In Kham, this constraint must be ranked above *DP[−gov]/​Spec/​vt & Max(case) and *DP[−gov]/​Pro/​vi & Max(case), and below *DP[−gov]/​ Def/​vt & Max(case) and *DP[−gov]/​Pro/​vt & Max(case), thereby separating the system in (18) into two discrete areas (given transitivity of ranking relations and the fixed rankings

 

292    Gereon Müller and Daniela Thomas established under harmonic alignment and local conjunction), which are here referred to as I and II. The absolutive case feature [–​gov] is preserved in area I and removed in area II, which leads to the fully specified exponent /​lai/​in I configurations and to the elsewhere exponent /​Ø/​in II configurations. We can now ask whether this pattern could also be captured in a similar way by adopting a standard impoverishment rule as the source of case feature deletion. It turns out that this is not the case: One would have to postulate two separate impoverishment rules, as in (19), since the contexts in which [–​gov] deletion takes place (intransitive clause and indefinite interpretation of DP) cannot be referred to as a natural class. Furthermore, (19) would give rise to redundancies with indefinite (specific or non-​specific) DPs in intransitive contexts. (19) a. [–​gov] → ∅ /​ DP_​_​[vi] b. [–​gov] → ∅ /​ DP[indef]_​_​ Thus, in contrast to an approach in terms of genuine impoverishment rules, an optimality-​ theoretic approach employing scale-​driven deletion makes it possible to refer to the diverse contexts where case feature deletion takes place as a natural class (defined by the relative ranking of the constraint demanding case feature deletion with respect to the ranking of the constraints demanding case-​feature preservation in the various contexts).

12.3.1.3 Ergative Marking Turning next to allomorphy in ergative realization, the account is simple. It does not involve two-​dimensional argument encoding because only the person scale is relevant for harmonic alignment with the binary basic case scale, of which the relevant member now is DP[+gov]. Harmonic alignment of the case scale and the person scale plus local conjunction with Max(case) yields the invariant hierarchy in (20). (20) *DP[+gov]/​3 & Max(case) ≫ *DP[+gov]/​loc & Max(case) Interleaving of *[+gov] between the two constraints of this hierarchy then produces zero exponence in cases where there is a prototypical external argument DP; see (21). (21) Ergative allomorphy in Kham *DP[+gov]/3 & Max(case)

I: /(y)e/

*[+gov] *DP[+gov]/loc & Max(case)

II: /Ø/

 

Three-way systems do not exist    293 Thus, overt ergative marking is blocked with transitive 1st or 2nd person subjects (because [+gov] is deleted here, and a retreat to the general elsewhere marker must take place), but available with 3rd person subjects (because [+gov] is preserved here, so that the vocabulary item /​(y)e/​that is characterized by the feature [+gov] can be inserted without violating the Subset Principle). To conclude, Kham has an ergative encoding system where both ergative and absolutive can be non-​zero cases, and both ergative and absolutive can be zero-​marked, as a consequence of case feature deletion in prototypical configurations.

12.3.2  Djapu 12.3.2.1 Data The Pama-​Nyungan language Djapu has also been analyzed in terms of a three-​way system comprising nominative/​absolutive, ergative, and accusative. The distribution of case exponents is shown in (22) (see Morphy (1983, 34–​35)). (22) Distribution of case markers Pron

+hu

–​hu

DPext-​Vt

-​∅

-​DHu

-​DHu

DPext/​int-​Vi

-​∅

-​∅

-​∅

DPint-​Vt

-​NHA

-​NHA

-​∅

Again, we assume that Djapu underlyingly exhibits an ergative system, with -​DHu as the ergative marker and -​NHA as the absolutive marker. Thus, overt absolutive marking is suspended in intransitive contexts and for non-​human objects; overt ergative marking does not show up on pronominal transitive subjects.

12.3.2.2 Absolutive Marking The relevant scales determining the distribution of morphological case exponents are the transitivity scale and the animacy scale. Both are harmonically aligned with the basic case scale, yielding (23-​a) and (23-​b) for absolutive contexts. (23)

a. *DP[−gov]/​vt ≫ *DP[−gov]/​vi b. *DP[−gov]/​Hum ≫ *DP[−gov]/​Anim ≫ *DP[−gov]/​Inan

Local conjunction among the members of these constraint hierarchies with fixed internal order produces the strict rankings in (24). (24) a. *DP[−gov]/​Hum/​vt ≫ *DP[−gov]/​Anim/​vt ≫ *DP[−gov]/​Inan/​vt b. *DP[−gov]/​Hum/​vi ≫ *DP[−gov]/​Anim/​vi ≫ *DP[−gov]/​Inan/​vi c. *DP[−gov]/​Hum/​vt ≫ *DP[−gov]/​Hum/​vi

 

294    Gereon Müller and Daniela Thomas d. *DP[−gov]/​Anim/​vt ≫ *DP[−gov]/​Anim/​vi e. *DP[−gov]/​Inan/​vt ≫ *DP[−gov]/​Inan/​vi Finally, order-​preserving local conjunction with Max(case) gives rise to the two-​ dimensional system in (25). (25) Absolutive allomorphy in Djapu I: /NHA/

*DP[−gov]/Hum/vt & Max(case)

*DP[−gov]/Anim/vt

*DP[−gov]/Hum/vi

& Max(case) *DP[−gov]/Inan/vt

& Max(case) *DP[−gov]/Anim/vi

& Max(case)

& Max(case) *DP[−gov]/Inan/vi

II: /Ø/

& Max(case) Since all absolutive arguments except for human objects remain without an overt marker, the conflicting markedness constraint *[–​gov] demanding case feature deletion must be located below *DP[−gov]/​Hum/​vt & Max(case), and directly above both *DP[−gov]/​Anim/​vt & Max(case) on the one hand (in transitive contexts), and *DP[−gov]/​Hum/​vi & Max(case) on the other (in intransitive contexts). Given that the absolutive marker /​NHA/​is specified for the feature [–​gov], it can only show up in transitive contexts with a human referent interpretation of the internal argument DP. As before, an impoverishment account would be inferior because the contexts in which case feature deletion takes place do not form a natural class definable in terms of shared morpho-​syntactic features.

12.3.2.3  Ergative Marking Again, allomorphic variation in the ergative system is somewhat simpler. The relevant scales are the case scale and the definiteness scale; see (26). (26) *DP[+gov]/​Nspec ≫ *DP[+gov]/​Spec ≫ *DP[+gov]/​Def ≫ *DP[+gov]/​PN ≫ *DP[+gov]/​Pron Local conjunction with Max(case) and interleaving of *[+gov] between *DP[+gov]/​PN & Max(case) and *DP[+gov]/​Pron & Max(case) yields a distribution of the overt ergative exponent /​DHu/​that involves all DPext arguments of transitive contexts except for pronouns. This is shown in (27).

 

Three-way systems do not exist    295 (27)

Ergative allomorphy in Djapu: *DP[+gov]/Nspec & Max(case) *DP[+gov]/Spec & Max(case) *DP[+gov]/Def & Max(case)

I: /DHu/

*DP[+gov]/PN & Max(case) *[+gov] *DP[+gov]/Pron & Max(case)

II: /Ø/

12.3.3  Nez Perce 12.3.3.1 Data In Nez Perce (Penutian), sole arguments of intransitive verbs are unmarked, as are local person external arguments of transitive verbs; 3rd person external arguments of transitive verbs receive a special marker, and the same goes for all internal arguments of transitive verbs; see (28) (Rude (1985, 82f.)). As noted above, this is usually taken to imply a three-​way system based on a general accusative, a general nominative/​absolutive, and a scale-​dependent ergative. (28) Distribution of case markers 1/​2 pronouns

3 pronouns

proper names

common nouns

DPext-​Vt

-​∅

-​(n(i))m

-​(n(i))m

-​(n(i))m

DPext/​int-​Vi

-​∅

-​∅

-​∅

-​∅

DPint-​Vt

-​ne

-​ne

-​ne

-​ne

As before, we suggest that this pattern be reanalyzed as a canonical ergative system, with scale-​driven allomorphy affecting both ergative and absolutive contexts.

12.3.3.2 Absolutive Marking This time, the absolutive alternation pattern is very simple:  It results from a simple local conjunction of the transitivity and case scales, which yields the fixed order in (29).

 

296    Gereon Müller and Daniela Thomas (29) *DP[−gov]/​vt & Max(case) ≫ *DP[−gov]/​vi & Max(case) The conflicting markedness constraint demanding deletion of [–​gov] is ranked between these two constraints in Nez Perce, which produces absence of [–​gov] in the morphological component (hence, a zero exponent) in intransitive contexts and retention of [–​gov] (hence, a non-​zero exponent /​ne/​, which is specified for this feature) with absolutive DPs in transitive contexts. (30) Absolutive allomorphy in Nez Perce: *DP[−gov]/vt & Max(case)

I: /ne/

*[–gov]

*DP[−gov]/vi & Max(case)

II: /Ø/

12.3.3.3 Ergative Marking The scales that play a role in accounting for ergative allomorphy are the person and case scales. Local conjunction of the constraints derived by harmonic alignment with Max(case) gives rise to a strict hierarchy: (31) *DP[+gov]/​3 & Max(case) ≫ *DP[+gov]/​loc & Max(case) Next, interleaving of *[+gov] leads to zero exponence in local contexts (given that insertion of the ergative marker /​nim/​depends on the presence of [+gov]): (32) Ergative allomorphy in Nez Perce: *DP[+gov]/3 & Max(case)

I: /nim/

*[+gov]

*DP[+gov]/loc & Max(case)

II: /Ø/

 

Three-way systems do not exist    297

12.3.4  Upriver Halkomelem 12.3.4.1 Data The evidence from Upriver Halkomelem (Salish) differs from what we have seen so far in two respects. First, argument encoding proceeds by agreement rather than by case-​marking; this is unproblematic against the background of the approach in section 12.1:  If assignment of case by a functional head (v, T) to a DP is viewed as an instance of Agree (Chomsky (2001)), i.e. as a probe-​goal relation, then the relevant case information will show up on both the head (v, T) and the DP, and can thus be morphologically realized either on the former or on the latter. Second, Upriver Halkomelem exhibits an accusative rather than an ergative basic pattern; in the approach adopted here, this implies that v structurally encodes DPint in transitive contexts, and T encodes all other core arguments in transitive and intransitive contexts. The agreement paradigm is given in (33) (see Galloway (1977, 141)). (33)

Distribution of cross-​reference markers DPext-​Vt

1SG

1PL

2SG

2PL

3SG/​PL

-​cəl

-​cət

-​cəxw

-​cεp

-​əs

-​cεp

-​∅

-​álə

-​∅/​-ə​ xw

DPext/​int-​Vi

-​cəl

-​cət

-​cəxw

DPint-​Vt

-​áyy

-​álxw

-​ámə

As illustrated in (33), the initial evidence for postulating a three-​way system in Upriver Halkomelem is confined to zero exponence in 3rd-​person intransitive configurations. Under present assumptions, this split is treated as an instance of non-​zero/​zero allomorphy with nominative exponents.13

12.3.4.2 Nominative Marking The distribution of nominative exponents (on T, as part of a complex verbal category) reveals harmonic alignment of both the transitivity scale and the person scale with the case scale, yielding the two-​dimensional system in (34). (34) a. b. c. d.

*DP[−gov]/​loc/​vt ≫ *DP[−gov]/​3/​vt *DP[−gov]/​loc/​vi ≫ *DP[−gov]/​3/​vi *DP[−gov]/​loc/​vt ≫ *DP[−gov]/​loc/​vi *DP[−gov]/​3/​vt ≫ *DP[−gov]/​3/​vi

After order-​preserving local conjunction with Max(case) (which applies to case features on both DP and T/​v), a system of partially free, and partially fixed, rankings is 13  There is also an allomorphy with 3rd-​person DP arguments in transitive environments. This is arguably int not of direct interest in the present context because it depends on the type of transitivity marker on the verb.

 

298    Gereon Müller and Daniela Thomas derived. As for the interleaving of *[–​gov], it is clear that this constraint must outrank *DP[−gov]/​3/​vi & Max(case) since zero exponence occurs in this context, and must in turn be dominated by *DP[−gov]/​loc/​vi & Max(case) since the non-​zero nominative exponent shows up here; in contrast, in transitive contexts all argument types are overtly encoded, so *[–​gov] must be outranked by all pertinent faithfulness constraints. All of this is shown schematically in (35), with the areas marked I and II capturing the two domains.14 (35) Nominative allomorphy in Upriver Halkomelem *DP[–gov]/loc/vt & Max(case)

*DP[–gov]/3/vt

*DP[–gov]/loc/vi

& Max(case)

& Max(case)

*DP[–gov]/3/vi & Max(case)

II: /Ø/

12.3.4.3 Accusative Marking Accusative marking is consistent and not subject to scale effects; the constraint *[+gov] demanding case feature deletion on v is ranked below all faithfulness constraints resulting from harmonic alignment and local conjunction.

12.3.5  Dyirbal 12.3.5.1 Data Dyirbal (Pama-​Nyungan) is also typically claimed to instantiate a three-​way system, based on the distribution of argument-​encoding exponents on DPs in (36) (see Dixon (1972; 1994)). We would like to suggest that Dyirbal case marking is best analyzed in terms of scale-​dependent allomorphy on the basis of a simple ergative system.

14  Note incidentally that in this particular case, a standard impoverishment rule would also suffice since it is just one specific context in which case feature deletion needs to be brought about. Still, even here, the other two arguments against classical impoverishment given above remain valid.

 

Three-way systems do not exist    299 (36)

Distribution of Case markers 1st/​2nd pronouns 3rd pronouns proper names common nouns DPext-​Vt

-​∅

-​ŋgu

-​ŋgu

ŋgu

DPext/​int-​Vi -​∅

-​∅

-​∅

-​∅

DPint-​Vt

-​∅

-​∅

-​∅

-​na

12.3.5.2 Ergative Marking The relevant scales for ergative marking are the case scale and the person scale. Harmonic alignment plus local conjunction with Max(case) produces the constraints in (37). (37)

*DP[+gov]/​3 & Max(case) ≫ *DP[+gov]/​loc & Max(case)

By interleaving *[gov] between these two constraints and assuming /​ŋgu/​to be a vocabulary item specified as [+gov], allomorphic realization of the ergative is correctly predicted; see (38). (38)

Ergative allomorphy in Dyirbal: *DP[+gov]/3 & Max(case) *[+gov]

*DP[+gov]/loc & Max(case)

II: /Ø/

Note that this system is completely identical to the system of ergative allomorphy in Nez Perce.

12.3.5.3 Absolutive Marking Turning to variation in the realization of absolutive case next, the first thing to note is that the same scales are relevant as they are with nominative allomorphy in Upriver Halkomelem: The basic case scale is harmonically aligned with both the transitivity scale and the person scale, and subsequently, local conjunction with Max(case) applies to the two constraint hierarchies thus generated, yielding the multidimensional system in (39). Assuming that the exponent /​na/​is specified for [–​gov], the markedness constraint *[–​gov] must determine the optimal output (forming the input for morphological realization) in all intransitive contexts, and in 3rd person transitive contexts; i.e. *[+gov] is only dominated by *DP[−gov]/​loc & Max(case). The resulting system is shown in (39). The area signaled by I shows non-​zero absolutive exponence; the area marked by II has zero exponence.

 

300    Gereon Müller and Daniela Thomas (39) Absolutive allomorphy in Dyirbal *DP[−gov]/loc/vt I: /na/

& Max(case)

*DP[−gov]/3/vt

*DP[−gov]/loc/vi

& Max(case)

& Max(case)

II: /Ø/

*DP[−gov]/3/vi & Max(case) As in several of the cases addressed before, no single impoverishment rule could capture this distribution of zero and non-​zero absolutive markers since the contexts in which deletion takes place do not form a natural class. Note also that even though Upriver Halkomelem and Dyirbal differ radically with respect to the basic argument encoding system employed in the language (accusative vs. ergative, agreement vs. case), deletion of the unmarked case (nominative/​absolutive) involves (a) the same case feature ([–​gov]), (b) identical derived faithfulness constraints with identical ranking restrictions ((39) and (35) are basically the same), and an identical markedness constraint demanding case feature deletion (*[–​gov]). The only difference concerns the cut-​off point between preservation and deletion; to highlight this difference, we have repeated the partition employed by Upriver Halkomelem in (39), in the form of a grey dotted line.

12.3.6  Interim Conclusion This concludes the first, major part of the analysis: We have shown that it is straightforwardly possible to reanalyze three-​way case systems as standard two-​way (ergative or accusative) case systems, with all the marker variation derived by scale-​driven optimization operations at the syntax/​morphology interface. Given that case marker allomorphy based on person, animacy, and definiteness is independently known to occur, and given that these effects also show up in all the languages that we have addressed here, the only additional assumption that is needed to capture all effects in a maximally simple way—​i.e. the postulation of a transitivity scale on a par with other Hale/​Silverstein scales—​strikes us as well motivated.15 Furthermore, the fact that putative three-​way systems are typically accompanied by Hale/​Silverstein scale effects, and that the fact that 15 

Also, the crosslinguistic rarity of three-​way systems (see n. 4) can now be explained under the assumption that non-​homogeneous post-​syntactic case-​feature deletion is inherently marked.

 

Three-way systems do not exist    301 these effects, though subject to implicational generalizations, are not uniform across languages, together pose what we take to be an enormous challenge for a syntactic approach recognizing three different cases; ambitious recent attempts notwithstanding (see in particular Deal (2014)) we would like to contend that it is hardly possible to come up with a comprehensive syntactic approach to the phenomenon that qualifies as both simple and elegant, and that covers both an individual language’s pattern in detail, and captures cross-​linguistic variation as well. However, that said, the question of whether a morphological or a syntactic approach to apparent three-​way systems is correct is also an empirical one; we address this issue in the final section of this chapter.

12.4  Syntactic Evidence The present morphology-​based approach to three-​way systems differs from syntactic approaches in that it reanalyzes what at first sight looks like an accusative DP (in Kham, Djapu, Nez Perce, and Dyirbal) as a non-​zero-​encoded absolutive DP, and what looks like an ergative DP (in Upriver Halkomelem) as a non-​zero-​encoded nominative DP. The prediction thus is that there might be independent evidence for the status of the pertinent DPs as absolutive/​nominative (i.e. [–​gov,–​obl], assigned by T). More generally, we expect to find evidence for a morphological approach in terms of case allomorphy based on identical syntactic behavior of the non-​zero-​marked and zero-​marked DPs; in the same way, different syntactic behavior might provide counter-​evidence against the proposal. Let us address some relevant phenomena.

12.4.1  Topic Chaining in Dyirbal A first piece of evidence comes from the topic chaining construction in Dyirbal, an instance of what has sometimes been called syntactic ergativity (Dixon (1972; 1994)). As shown in (40), if the coordinative process of topic chaining combines a transitive clause (with DPext bearing ergative case and DPint bearing absolutive case) and an intransitive clause where the sole DPext/​int argument is not overtly realized (here encoded by an empty pronoun pro, for expository purposes), then this latter DP, which must bear absolutive case, must be coreferent with the absolutive argument of the transitive clause, not with the ergative argument; see (40). (40) [CP1 ŋuma yabu-​ŋgu bura-​n]   father-​ abs mother-​erg see-​nonfut ‘Mother saw father and he/​*she returned.’

[CP2 pro pro-​abs

banaga-​nyu] return-​nonfut

Thus, there is a case-​matching requirement active in Dyirbal topic-​chaining constructions, and we can test the predictions made by the two approaches. On the one hand, if

 

302    Gereon Müller and Daniela Thomas local person DPint arguments marked by /​na/​bear accusative case (as assumed under the standard three-​way approach), they should not be able to corefer with a non-​overt sole DP argument of an intransitive clause (which bears absolutive case) in a topic chaining construction; on the other hand, if local person DPint arguments marked by /​na/​bear absolutive case (as assumed under the present analysis), they should be able to corefer with a non-​overt DP argument of an intransitive clause. The data in (41) (from Dixon (1972; 1994)) show that the latter prediction is in fact the correct one; as noted by Morgenroth & Salzmann (2013), this provides strong evidence against a separate accusative case, and hence against a three-​way system, in Dyirbal. (41) a. [CP1 ŋana-​Ø banaga-​nyu] [CP2 nyurra-​Ø pro    we-​abs return-​nonfut    you all-​erg pro-​abs ‘We returned and you all saw us.’

bura-​n] see-​nonfut

b. [CP1 nyurra-​Ø     ŋana-​na  bura-​n]       [CP2 pro    banaga-​nyu]    you all-​erg  we-​abs    see-​nonfut     pro-​abs  return-​nonfut ‘You all saw us and we returned.’ Note that (41-​b) shows that the three-​way approach not only fails to derive the possibility of coreference of the absolutive argument of the intransitive clause with the DPint argument of the transitive clause (because the latter shows non-​zero exponence); assuming that non-​zero exponence on DPext in the transitive clause indicates absolutive case (but see footnote 16 below), it also wrongly predicts coreference of the absolutive argument of the intransitive clause with the DPext argument of the transitive clause to be possible (since the latter exhibits zero-​exponence).

12.4.2  Modifiers and Relative Clauses in Dyirbal This latter consequence with respect to zero-​marked DPext arguments in transitive contexts also arises with modifiers in Dyirbal. The example in (42) (see Dixon (1972, 133), Mel’čuk (1979, 54)) illustrates that modifiers and relative clauses of a local person DPext argument of a transitive verb bear ergative case despite the lack of ergative marking of the pronoun itself. (42) ŋaɖa wuygi-​ŋgu balan I.nom old-​erg ncII.there.abs ‘I, old, hit the woman.’

ɖugumbil woman.abs

balga-​n hit-​nfut

This clearly shows that ergative is indeed assigned to DPext in (42) (even though it is not overtly realized by an ergative exponent on the head), and passed on to other DP-​internal items via concord (with subsequent morphological non-​zero realization on D). In the present analysis, this state of affairs can be addressed straightforwardly, by invoking standard assumptions about the order of operations involved here (see Müller (2009), Keine (2010),

 

Three-way systems do not exist    303 Georgi (2014)). First, there is assignment of ergative ([+gov,–​obl]) case to DPext (hence, D) by v; we assume this case feature to be transferred automatically to a potential NP selected by D (this is irrelevant in the case of pronouns, as in (42), but it would be relevant with non-​pronominal 3rd person DPs). Second, there is another operation (‘concord’) that copies the feature from the nominal spine (D, N) to AP and CP modifiers. And third, scale-​ driven optimization (leading to deletion of [+gov]) targets DP, deleting the feature on D but leaving it intact on AP and CP modifiers. This counter-​bleeding effect (deletion would bleed concord but comes too late to do so) is shown in (43). (43)

a. Case assignment

b. Concord vP

vP

D:[+gov]

v'

DPext

v'

DPext AP v

VP

... DPint ...

D:[+gov]

AP:[+gov]

v

VP

... DPint ...

c. Scale-driven deletion vP

v'

DPext D: –

AP:[+gov] v

Ø

VP

... DPint ...

In contrast, if the case feature assigned to local person DPext of transitive verbs in Dyirbal is absolutive rather than ergative, the source of the ergative exponents on DP-​internal items in (42) must remain a mystery.16 16   To be sure, adopting a three-​way approach involving ergative, accusative and nominative/​ absolutive for the distribution of markers in (36) does not per se imply that the source of zero exponence with local person DPext-​Vt is the same as the source of zero exponence with DPext/​DPint-​Vi (or with 3rd person DPint-​Vt, for that matter); in other words, one could in principle assume that Dyirbal has three structural cases, but some instances of zero marking involve a morphologically unrealized ergative rather than syntactically assigned nominative/​absolutive. If so, an approach along the lines just sketched would also be available in a three-​way approach, but it would miss an obvious generalization: The non-​ zero/​zero alternation with DPext-​Vt is clearly governed by the same factor that also restricts the non-​ zero/​zero alternation with DPint-​Vt (i.e. the person scale).

 

304    Gereon Müller and Daniela Thomas

12.4.3  Complex DPs in Djapu Further independent evidence for treating an apparent coexistence of absolutive and accusative as allomorphic realization of a single case (the absolutive) comes from Djapu. The following example from Morphy (1983, 136)  illustrates that a non-​zero exponent can show up on a DPint-​Vt (glossed here as acc, following the literature) whereas a modifier phrase agreeing with this DP exhibits zero exponence (glossed here as abs). (44) dhuwal ŋarra yothu-​n this.abs 1sg.nom child-​acc ‘I can hear a child crying.’

yukurra lie.unm

ŋ-​ä-​ma hear-​unm

[ŋäthi-​nyar] [cry-​nmlz.abs]

Similarly, Legate (2008, 77) observes that a demonstrative will show zero exponence where a non-​zero exponent occurs with a (human) DPint-​Vt. Following Legate, we take this to indicate that the same case feature is involved on the noun and the modifier or demonstrative, and that the variable exponence is a purely morphological phenomenon, reducible to the availability or lack of compatible vocabulary items. However, our analysis differs from Legate’s (2008) account in one crucial respect: Legate postulates a three-​way system for Djapu and considers the source of zero exponence with DPext/​int in intransitive contexts to be different from the source of zero exponence with DPint in transitive contexts (absolutive vs. unmarked accusative), which gives rise to the conceptual problem mentioned in footnote 16.

12.4.4  Relative Clauses in Kham Like Dyirbal, Kham has non-​zero ergative exponence on relative clauses associated with local person DPext-​Vt arguments that do not exhibit any morphological case marker themselves; cf. (45) (cf. Watters (2002, 201, 205)). (45) geː ma-​ba-​o-​ra-​e ge-​ma-​dəi-​ye we neg-​go-​nml-​pl-​erg 1pl-​neg-​receive-​impfv ‘Those of us who didn’t go didn’t get any.’ Again, this asymmetry follows directly under the present approach, along the lines of the derivation in (43); but it raises severe problems under a three-​way approach.

12.4.5  Modifiers in Nez Perce Recall from section 12.4.2. that cases where zero and non-​zero exponence of case cooccur in a single DP follow under the morphological approach but pose a challenge

 

Three-way systems do not exist    305 for a syntactic three-​way analysis. Conversely, Deal (2014) presents an argument based on the reverse configuration in support of a syntactic rather than morphological analysis. In Nez Perce, modifiers can (optionally) take ergative marking. From a morphological perspective, one might therefore expect this marking to occur on the modifier if it is combined with a local-​person transitive subject which does not bear ergative marking. As shown in (46), this is not the case (glossing follows Deal (2014, 17)). (46) a. Yú’s-​nim ’iceyéeye-​nm, wéet’u minma’í ’itúu-​ne pée-​p-​se-​∅ poor-​erg coyote-​erg neg prt what-​acc 3/​3-​eat-​imperf-​pres ‘Poor coyote isn’t eating anything.’ b. Yu’c       /​*Yú’s-​nim (pro),      wéet’u  q’o    minma’í  ’itúu-​ne  poor.nom  /​*poor-​erg pro.1sg  neg     prt  prt     what-​acc ’ee-​pí-​se-​∅ 3obj-​eat-​imperf-​pres ‘Poor me isn’t eating anything.’ However, closer inspection reveals that (46-​b) is in fact unproblematic under the approach adopted here. Recall from (43) that we expect marker preservation on modifiers if the order of operations is (i)  case assignment (of [+gov] to DPext by v), (ii) concord (of D and AP/​CP), and (iii) scale-​driven deletion, an instance of optimization. Suppose now that the last two operations can also apply in reverse order, such that feature deletion can precede concord; and that this is the case in Nez Perce (but not in, say, Dyirbal).17 If so, the absence of a case exponent on the modifier in (46-​b) is derived: [+gov] is deleted on D before concord with AP is effected (an instance of bleeding).

12.4.6  Coordination in Nez Perce Deal (2014, 20f.) gives a second argument in support of a three-​way analysis of Nez Perce. Coordination in Nez Perce is not subject to any restrictions when it occurs with DPext/​int-​Vi arguments or with DPint-​Vt arguments. In these contexts, all kinds of person combinations are allowed. (Case marking may appear on both coordinates or just the final one.) However, there are restrictions in the case of coordination of DPext-​Vt arguments. Coordinations of two local person pronouns and of two 3rd person pronouns/​nouns are unproblematic (cf., e.g., (47-​a)), whereas the combination of local person and 3rd person turns out to be ungrammatical (see (47-​b)):

17 

See Keine (2010), Doliana (2013) on early, inner-​syntactic impoverishment-​by-​optimization.

 

306    Gereon Müller and Daniela Thomas (47) a. Kátie(-​nim) kaa Hárold-​nim pée-​’pewi-​six-​∅ Múna-​ne. Katie(-​erg) and Harold-​erg 3/​3-​look.for-​imperf.pl-​pres Muna.acc ‘Katie and Harold are looking for Muna.’ b. *Ángel-​ nim kaa  ’iin     ’e-​ nées-​ tecukwe-​ cix-​ ∅           (pro). Angel-​erg     and 1sg.nom  3obj-​o.pl-​teach-​imperf.pl-​pres  (pro.3.pl) ‘I and Angel are teaching them.’ This can be derived if some DPext-​Vt arguments bear absolutive case in an otherwise ergative system (based on conflicting demands of case-​assignment); but closer scrutiny reveals that the present analysis (where all coordinated DPs in (47) bear ergative case) also does not face any particular difficulties. Assuming that & is the head of a coordinate structure &P, ergative case assignment by v will first instantiate [+gov] on &. As argued in section 13.4.5, scale-​driven deletion applies next in Nez Perce, before &-​internal concord (which transfers [+gov] to the two DPs in coordinate structures). However, at this point, there is incompatible contextual information: For the second DP in (47-​b), optimization applying to & would rely on a local person feature; for the first, it would rely on a 3rd person feature; and the outcome of optimization would be different (deletion vs. preservation of [+gov]). It is conceivable that languages in principle may have the option to give preference to one of the two outcomes of optimization in this kind of situation (giving rise, e.g., to first-​vs. last conjunct agreement); however, in Nez Perce, it leads to a breakdown of the derivation. In contrast, in other coordination environments (as in (47-​a), or with DPext/​ int-​Vi, or with DPint-​Vt), the two DP contexts for & uniformly demand deletion or preservation of the case feature; so no problem of conflicting instructions will arise. More generally, we may venture the hypothesis that the available syntactic evidence either directly supports the hypothesis that there are no three-​way systems, or can at least be addressed on the basis of a syntactic approach relying on only two structural cases in an insightful way. From an even more general perspective, if what precedes is on the right track, we can conclude that by addressing instances of variation in case exponence in the morphological component of the grammar, the syntactic component can be kept simple and elegant, with three-​way systems emerging as an artefact.

Acknowledgments For helpful comments, we would like to thank Sebastian Bank, Aaron Doliana, Doreen Georgi, Peter Staroverov, Michael Yoshitaka Erlewine, Lisa Travis, and one anonymous reviewer.

 

Three-way systems do not exist    307

Abbreviations abs, absolutive; acc, accusative; Anim, animate; Def, definite; DPext, external Determiner Phrase; DPint, internal Determiner Phrase; erg, ergative; gov, governed; HU, human; Hum, human; imperf, imperfective; impfv, imperfective; Inanim, inanimate; intrans, intransitive; loc, local person; Max, Maximality (constraint); neg, negation; nml, nominalizer; nmlz, nominalizer; nom, nominative; Nspec, nonspecific; nonfut, non-future; obl, oblique; perf, perfect; pl, plural; pldo, plural direct object; PN, proper name; pres, present; Pro, pronoun; prt, particle; pst, past; Spec, indefinite specific; trans, transitive; unm, unmarked; Vi, intransitive verb; Vt, transitive verb.

 

Chapter 13

Antipas si v e Maria Polinsky

13.1 Introduction Valency alternations follow from the mapping of logical arguments of a predicate into different clausal constituents. In morphologically accusative languages, active-​voice clauses typically have a nominative subject and accusative object (1). Morphologically ergative languages, by contrast, have ergative subjects and absolutive objects; this alignment can be expressed by case-​marking and agreement (2) or by agreement alone (3): (1) Did sxopyv grandfather.nom.masc grab.pst.masc ‘Grandfather grabbed the turnip.’

ripk-​u. turnip-​acc

(2) Anguti-​up annak taku-​janga. man-​erg woman.abs see-​3sg.subj:3sg.obj.prs ‘The man sees the woman.’ (3) Max-​Ø s-​tx’aj ix compl-​abs3 erg3-​wash clf ‘The woman washed the clothes.’

ix woman

an clf

pichilej. clothes

Ukrainian

Labrador Inuit (Smith 1982a: 164) Q’anjob’al

In passive-​voice clauses in both accusative and ergative languages, the logical subject remains unexpressed or appears as a by-​phrase; the subject position is filled by either the logical object ((4a), (5), and (6)) or by an expletive subject (4b): (4)

a. Repk-​ a       bu-​ l-​ a      sxople-​ n-​ a       (did-​ om). Ukrainian turnip.nom.fem  be-​pst-​fem  grab-​pst.ptcp-​fem  grandfather-​ins ‘The turnip was grabbed (by the grandfather).’

 

Antipassive   309 b. expl repk-​u      bu-​l-​o       sxople-​n-​o           (did-​om). Ukrainian     turnip-​acc be-​pst-​neuter grab-​pst.ptcp-​neuter grandfather-​ins ‘The turnip was grabbed (by the grandfather).’ (5)

  Labrador Inuit

Annak (anguti-​mut) taku-​jau-​juk. woman.abs man-​dat see-​pass-​3subj.prs ‘The woman is seen (by the man).’

(Smith 1982a: 165)

(6) Max-​Ø tx’aj-​lay an pichilej (y-​uj ix ix). Q’anjob’al compl-​abs3 wash-​pass clf clothes gen3-​rel.noun clf woman ‘The clothes were washed (by the woman).’ Broadly speaking, the defining property of the passive construction is the “suppression” or “demotion” of the highest-​ranking argument to a by-​phrase or a null existential argument without altering the truth conditions of the sentence (Goodall 1993, 1999; Bruening 2013). The logical object may appear as the syntactic subject of the resulting construction, but, as (4b) shows, it need not. In the foundational work on voice structures in Relational Grammar, “demotion” was a theoretical primitive (see Perlmutter 1980; Blake 1990b for overviews). In this chapter, I use this term pre-​theoretically to indicate that an expression has been lowered vis-​à-​vis the hierarchy of grammatical roles: (7)

subject > object > non-​core argument > non-​argument

While subject demotion is the unifying factor among passive constructions cross-​ linguistically, motivation for this demotion varies across languages and structures; differences among passives are often attributed to distinct features of the passive morpheme or voice head. In the antipassive construction, the logical object of a transitive verb is demoted: either suppressed or represented by an expression lower on the grammatical hierarchy (7). In (8), the verb appears in the intransitive (reflexive) form, and the logical object receives accusative case from the preposition za. This prepositional object fails various direct-​ object diagnostics; for example, it cannot bind depictives or float quantifiers. (8)

Did sxopyv-​sja grandfather.nom.masc grab.pst.masc-​refl ‘Grandfather grabbed at the turnip.’

za at

ripk-​u. turnip-​acc

Ukrainian

In (9), the intransitive counterpart of (2), the logical object appears in an oblique form. The case of the subject changes from ergative to absolutive, and the verb shows only subject agreement. (9)

Angutik anna-​mik taku-​juk. man.abs woman-​ins see-​3sg.subj.prs ‘The man sees a woman.’

Labrador Inuit (Smith 1982a: 164)

 

310   Maria Polinsky Likewise, in (10), the counterpart of (3), the verb loses its object agreement.1 (10)

Max-Ø tx’aj-​w-​i (an) compl-​abs3 wash-​ap-​itr clf ‘The woman washed (the) clothes.’

pichilej clothes

ix clf

ix. woman

Q’anjob’al

The term antipassive was coined by Michael Silverstein in his description of Chinook: I  have termed this -​ki-​form the ANTIPASSIVE construction, playing upon its inverse equivalence to a passive of accusative languages, because the sense is clearly equivalent to a transitive, though the form is intransitive, with the grammatical function of the remaining NP reversed (ergator becomes nonergator). (Silverstein 1972: 395)

At the time of Silverstein’s writing, it was widely believed that the absolutive DP of a transitive clause was the grammatical subject (“nonergator” in the Silverstein quote) in ergative languages. This analysis has since been rejected (cf. Anderson 1976, and much subsequent work), but the term “antipassive” has remained. Many papers have lamented the term at length; in what follows, I set aside terminological debate and focus on analysis. Note that the term “antipassive” has two related uses in the linguistics literature; throughout this chapter, “antipassive” will be understood as in (11a). (11)

a. Antipassive: a clause with a transitive predicate whose logical object is demoted to a non-​core argument or non-​argument b. Antipassive: the form of a two-​place predicate that appears in the antipassive clause

The antipassive predicate is semantically transitive, but does not project a direct object; hence, it is morphosyntactically intransitive.2 In what follows, I will be using the term “antipassive object” to signify the logical object of a clause. This logical object appears as a non-​core argument of the antipassive construction (8), (9) or is suppressed (12b). (12) a. Aid opa-​n matses that.one dog-​erg people.abs ‘That dog bites people.’

pe-​e-​k. bite-​npst-​ind

b. Aid      opa       pe-​an-​e-​k. that.one dog.abs  bite-​ap-​npst-​ind ‘That dog bites (me/​us).’

1 

Matses

(Fleck 2006: 559)

Also note that, unlike (3), the classifier on the object in (10) is optional, further signaling a change in its grammatical status. 2  The valency of an antipassive verb can be increased by adding an applicative head; this head then licenses an object, but crucially not the same direct object as the one selected by the non-​applicative form (for applicative antipassives in Chukchi, see Nedjalkov 1980; Kozinsky et al. 1988).

 

Antipassive   311 Antipassives have been documented in a number of languages, but not all antipassives meet the same operational criteria. Two criteria are typically used to establish the existence of an antipassive: (i) oblique marking on the object and absolutive (not ergative) marking on the subject; (ii) structural/​discourse “foregrounding” of the agent (see section 13.3; cf. Foley and Van Valin 1984 on foregrounding and backgrounding antipassives; Cooreman 1988 on discourse-​based and structural antipassives). These criteria, based on morphological exponence and discourse effects, are not sufficient to fully characterize the antipassive construction. Concerning criterion (i), antipassive morphology need not be consistent within a given language: a particular morphological exponent may be used to mark antipassive in one use, but not in others; conversely, more than one type of morphological exponence may be associated with the antipassive. Concerning criterion (ii), it is problematic to define a construction on the basis of discourse effects, although such effects may arise as a consequence or entailment of a particular construction; furthermore, the same interpretative or discourse effect may be associated with more than one construction. Thus, neither morphological nor discourse effects seem independently capable of defining the antipassive construction. The demotion or removal of the object argument, however, is definitional of the antipassive—​just as the demotion or removal of the external argument is definitional of the passive. While detailed descriptions of antipassives have been offered for individual languages (I reference some in the chapter; without such resources, this chapter could not have been written), the data are often insufficient to draw far-​reaching conclusions. In such instances, I simply offer observations and make suggestions for further study. Consequentially, this chapter may be read as both an overview and promissory note; it is my hope that future analytical work on antipassives will follow. Section 13.2 describes typical morphological hallmarks of the antipassive; section 13.3 addresses its main interpretative characteristics. Section 13.4 summarizes the main existing analyses of the construction. Section 13.5 discusses syntactic side effects observed under antipassivization. Section 13.6 contrasts the true antipassive with certain “lookalike” constructions. Finally, section 13.7 addresses the question of whether the antipassive is unique to ergative languages.

13.2  Morphological Indicators of the Antipassive Morphological signs of antipassivization vary widely both within and across languages (see below on Diyari). No single morphological diagnostic exists, and morphology alone may not be sufficient to identify the antipassive. However, a number of noteworthy patterns exist. Let us begin with case-​marking. An oblique case often signals the non-​core status of the antipassive object, as in (8) and (9). In ergative languages, a change in object encoding (absolutive > oblique) typically corresponds to a change in subject encoding (ergative >

 

312   Maria Polinsky absolutive). However, at least Warlpiri (Hale 1973a: 366), Djaru (Tsunoda 1981a: 149), and Goonyandi (Tsunoda 1988: 627) preserve ergative marking in the antipassive: (13) a. Njuntulu-​lu npa-​tju 2sg-​erg 2sg-​1sg ‘You speared me.’

pantu-​nu spear-​pst

ŋatju. 1sg.abs

Warlpiri

b. Njuntulu-​lu  npa-​tju-​la    pantu-​nu   ŋatju-​ku. 2sg-​erg     2sg-​1sg-​ap  spear-​pst  1sg-​dat ‘You speared at me.’ Some researchers argue that the structure in (13b) is in fact an instance of D(ifferential) O(bject) M(arking), which does not entail detransitivization (Campana 1992; Malchukov 2006: 347; Malchukov, Chapter 11, this volume). I will return to DOM in section 13.6.1. Oblique object marking is neither necessary nor sufficient to identify an antipassive. The logical object may appear under pseudo noun incorporation (PNI), a syntactic process that generates a caseless object NP (not a DP) adjacent to the verb (Massam 2001). Unlike regular incorporation, PNI incorporates expressions larger than N°. PNI is more easily detectable in languages with overt absolutive marking. In the following examples, the PNI object includes an adnominal PP; in the Tongan example, a modifier (lelei) scopes over the entire NP. Overt case-​marking of this expression is impossible.3 In both cases, the logical object (internal argument) is case-​marked differently than in the corresponding transitive clause: (14) a. ‘Oku puke ‘e he pepe [DP‘a e me’a va’inga prs hold erg det baby    abs det thing playing [PP mo e pulu] lelei].   com det ball good ‘The baby is holding a/​the nice toy and ball.’ (transitive)

Tongan

b. ‘Oku puke (*‘a) [NP e     me’a    va’inga  [PP mo e    pulu] lelei] prs   hold  abs       det  thing playing    com  det  ball    good ‘a e pepe. abs det baby ‘The baby is holding a nice toy and ball.’ (PNI antipassive) (15)

3 

a. Ka hoko te matua [DP ī ngā tīkiti aor buy det parent    acc det.pl ticket [PP mo ngā    tamariki]]. for    det.pl  child ‘The parent buys (the) tickets (intended) for the children.’

Māori

Additionally, the PNI expression cannot be separated from the verb. Due to space constraints, I do not show the ungrammatical data here.

 

Antipassive   313 b. Ka   hoko [NP tīkiti    [PP mo  ngā     tamariki]]  te      matua. aor  buy       ticket    for  det.pl  child   det  parent ‘The parent buys tickets (intended) for the children.’ PNI is not limited to objects, and the analysis of PNI is a complex issue (Massam 2001; Ball 2008; Clemens 2014). For our purposes, the relevant point is that object PNI is a particular instance of antipassivization: the object does not receive structural case, and the resulting construction is intransitive. In languages without overt object marking (e.g. Diyari, an ergative language with unmarked absolutives), additional diagnostics are required to distinguish direct objects from NPs under PNI. Diyari has a general detransitivizing suffix -​ṯadi that forms reflexives and antipassives (Austin 1981b: 71–​72). In a subclass of verbs (Austin’s class 2C), both the logical subject and the logical object of the antipassive seem to be in the absolutive form, with SOV word order:4 (16)

a. ṉnulu kaṇa-​li ṉiṉa ŋaṉṯi dem.erg person-​erg dem.abs meat.abs ‘The man is eating this meat.’(transitive)

ṯayi-​yi. eat-​prs

Diyari

b. ṉawu kaṇa ṉiṉa ŋaṉṯi ṯayi-​ṯadi-​yi. dem.abs person.abs dem.abs meat.abs eat-​ap-​prs ‘The man is having a feed of this meat.’ (antipassive) (Austin 1981b: 154) However, significant differences exist between the (apparent) absolutive objects in (16a,b). While the object in (16a) can be separated from the verb and scrambled, the object position in (16b) is fixed. (16a) answers the question, “Who ate the meat?”, whereas (16b) answers, “What is the man doing?” (Austin 1981b:  154). The nominal demonstrative may be adjectival; the NP+verb unit seems syntactically inseparable, yet does not form a lexical item. This suggests that in (16b), the object ṉiṉa ŋaṉṯi is not absolutive, but a caseless PNI object—​which explains its immobility.5 Finally, under regular noun incorporation (NI), the logical object is inaccessible to case-​licensors because it has been absorbed into the predicate; several researchers have suggested that NI is a morphological realization of antipassivization (Baker 1988; Kozinsky et al. 1988). In morphologically ergative languages, NI is accompanied by a shift in subject marking from ergative to absolutive. Consider the following triplet from Chukchi: the 4 

Another predicate type (Austin’s verb class 2B) undergoes alternation: ergative–​absolutive marking with SOV word order in the transitive; absolutive–​oblique marking with SVO order in the antipassive (Austin 1981b: 153–​154). 5  Example (16b) may be a bi-​absolutive (bi-​nominative) construction. Forker (2010; 2012) analyzes such constructions across Nakh-​Dagestanian languages uniformly as PNI, while Gagliardi et al. (2014) and Coon (2013a) show that the syntax of bi-​absolutives varies cross-​linguistically. Some bi-​absolutives may be antipassives.

 

314   Maria Polinsky transitive construction in (17a) corresponds to a regular antipassive in (17b) and to an antipassive with NI in (17c). Assuming that the Diyari example in (16b) is a genuine case of PNI, we also note an important difference between PNI and NI: PNI (as in Diyari) is compatible with an antipassive suffix on the verb, but NI (as in Chukchi) is in complementary distribution with the antipassive prefix. (17) a. ʔətt-​e melota-​lɣən piri-​nin. dog-​erg hare-​abs catch-​aor.3sg:3sg ‘The dog caught a/​the hare.’

Chukchi (Kurebito 2012: 184)

b. ʔətt-​ən    ine-​piri-​ɣʔi     (melot-​ etə). dog-​abs  ap-​catch-​aor.3sg  hare-​dat ‘The dog caught (a/​the hare/something).’ c. ʔətt-​ən    milute-​piri-​ɣʔi. dog-​abs  hare-​catch-​aor.3sg ‘The dog caught a/​the hare.’ d. *ʔətt-​ən    ine-​milute-​piri-​ɣʔi. dog-​abs  ap-​hare-​catch-​aor.3sg Since the logical object of the antipassive is a non-​core argument, it can always be omitted without significant change in meaning, as indicated in (17b). This optionality is yet another morphosyntactic sign of antipassivization. In terms of the grammatical role hierarchy, the antipassive object is always lower than the syntactic object in (7); object marking and (P)NI make this asymmetry visible. The antipassive can also be signaled by verbal affixation. In (17b), the Chukchi verb has the antipassive prefix ine-​(note also the difference in word order between (17a) and (17b)). In the majority of languages that mark the antipassive verbally, the affix indexes other categories as well. Two typical patterns of syncretism are attested: (18) a. antipassive is syncretic with detransitivizing affixes such as anticausative, reflexive/​reciprocal, middle, or passive markers, e.g., in Chukchi (Kozinsky et al. 1988), Diyari (Austin 1981b), Halkomelemem -​Vm (‘middle’, per Gerdts and Hukari 2005, 2006), some Pama-​Nyungan languages (Dixon 1972; 1977; Terrill 1997), Kiowa (Watkins 1984) b. antipassive is syncretic with aspectual markers, most commonly inchoative, inceptive, or iterative e.g. in Bezhta (Comrie et al. 2015) or Eskimo/​Inuit6 (Spreng 2012; Basilico 2012) I have not observed languages with a non-​syncretic antipassive marker. In principle, such a language could exist; however, the patterns in (18) are common enough to justify 6 

I am using “Inuit” as a cover term for several languages/​dialects in the North American Arctic and Labrador for which similar data have been reported by different researchers.

 

Antipassive   315 the prediction that a given antipassive marker will also serve as a general detransitivizing/​aspectual affix. The syncretic patterns in (18) correlate with two main syntactic analyses of the antipassive, discussed in section 13.4.2. More than one antipassive verbal marker can exist in a single language. These markers can be stacked (doubled), although the effects of such doubling are poorly understood. In Chukchi, the productive antipassive prefix ine-​and the semi-​productive antipassive suffix -​tko-​can co-​occur (Kozinsky et al. 1988: 661); likewise, in Halkomelem, the antipassive suffixes -​m-​(“middle,” per Gerdts and Hukari 2006) and -​els-​(“activity”) stack: (19)

Niʔ ḱwł​eʔ-​əm-​els ʔə aux pour-​ap-​ap obl ‘John served some tea.’

ḱw det

ti tea

tθə John. Halkomelem det John (Gerdts and Hukari 2005, ex. (26))

The stacking of antipassive markers has no apparent syntactic consequences. By itself, the suffix -​m serves to detransitivize the verb; the contrast in meaning between the stacked and single antipassives is unclear and warrants further investigation. Next, the antipassive is morphologically visible through agreement. Antipassives typically show intransitive agreement, since the object, as a syntactically non-​core argument, cannot agree—​compare (3) and (10) for Q’anjob’al; (17a) and (17b) for Chukchi. Changes in agreement are typically accompanied by changes in case-​marking and/​or the addition of a verbal marker; all three morphological cues appear in (17b). However, in some languages, the change in agreement is the only sign of antipassivization. Finally, antipassives sometimes differ from corresponding transitive constructions in word order (compare (17a,b)); however, this is a very weak diagnostic, especially because the oblique object in the antipassive is generally dispensable. Word order differences may be a side effect of the adjacency requirement imposed by PNI (see examples (14) and (14).

13.3  Interpretative Properties of Antipassives Several interpretative properties are commonly associated with the antipassive; none of these properties is essential, and none applies to all languages with purported antipassives. Thus, these properties are best viewed as concomitant, rather than as defining, characteristics of the antipassive, much as thematic discourse prominence is a concomitant but not defining property of the passive. To reiterate the discussion in section 13.1, antipassivization is a syntactic operation whose defining properties are independent of its interpretative characteristics and discourse uses. Antipassives often have special aspectual meaning:  inchoative, inceptive, durative, progressive, imperfective, or even iterative (Tsunoda 1981b; Tchekoff 1987; Dowty 1991; Cooreman 1994; Dixon 1994; Spreng 2010). In some languages, the antipassive is

 

316   Maria Polinsky reanalyzed as durative and loses its syntactic detransitivizing properties (Comrie et al. 2015: 554). In each case, the antipassive may be associated with atelicity (the antipassive ~ imperfective correlation).7 This association, however, is not present in all languages; for example, it is not found in Chukchi, Chamorro, or Polynesian languages. The typicality of the antipassive ~ imperfective correlation predicts the following: (20) If an antipassive construction can have a perfective (telic) interpretation, it must also have an imperfective (non-​telic) interpretation Another common interpretative effect of the antipassive is semantic or pragmatic SUBJECT prominence, a phenomenon referred to as “agent foregrounding,” “agent focus(ing),”8 or “agent maintenance.” Agent foregrounding likely follows from the activity interpretation of the antipassive and the absence/​backgrounding of the logical object. If the object is omitted, then the subject (agent) naturally becomes the only salient participant. When the object is expressed, its oblique encoding has subtle interpretative consequences. In the antipassive’s transitive counterpart, new discourse referents are typically introduced as internal (accusative, absolutive) arguments, especially in presentational constructions (Du Bois 1987b; Prince 1992; Birner and Ward 1996, 1998), indicating that the new referent will likely be featured in the upcoming discourse (Givón 1983b). If a new referent appears as an oblique (or incorporated nominal), it signals that the referent will not be maintained in subsequent discourse (Polinskaja and Nedjalkov 1987). Thus, the antipassive construction facilitates a low-​individuation interpretation (indefinite, non-​specific) of the object participant (Foley and Van Valin 1984; Cooreman 1994): in other words, the object is presented as a prototypical ‘anti-​topic’—​see Kalmár (1979), Berge (2011), and Johns and Kučerova (Chapter 17, this volume) for Inuit. The foregrounding of the agent and concomitant backgrounding of the object is nevertheless far from a universal characteristic of the antipassive. While in some languages, the antipassive is impossible with highly individuated (first or second person) objects (see Gerdts 1988a: 157 for Halkomelem), some languages require the antipassive construction specifically with these roles (see section 13.5.3). Semantic and discourse functions of antipassives differ both within and across languages (Heath 1976; Comrie 1978; Cooreman 1988; 1994).

13.4  Analysis of the Antipassive Several analyses of the antipassive have been advanced within a number of theories. What follows is a brief survey; the reader is advised to consult the respective references for more 7  The correlation antipassive ~ imperfective on the one hand and passive ~ perfective on the other is yet another sign of the mirror-​image correspondence between antipassive and passive. 8  Agent focusing should not be confused with the special agent focus construction, documented in Mayan languages (see section 13.6).

 

Antipassive   317 details. Although terminology and theoretical tools vary across approaches, analyses of the antipassive construction can be broadly divided into lexicalist and syntactic camps.

13.4.1 Lexical(ist) Approaches Even in languages with a robust antipassive, use of the construction is lexically specified. Typically, the antipassive occurs with “manner verbs”, i.e. verbs that denote actions performed in a particular manner with no entailed result-​state (Rappaport Hovav and Levin 2010 and references therein). Objects of manner verbs can be readily “suppressed” or left implicit, producing an agent emphasis that parallels the focus effect reported for the antipassive. Arkadiev and Letuchiy (2008) systematically test and confirm the restriction of the antipassive to manner verbs in Adyghe, but supporting work in other languages remains to be done. In some languages, only a handful of verbs are incompatible with the antipassive construction. In Chamorro, Chung (1998: 39) reports three verbs that do not form antipassive predicates (without noting how many verbs there are in total). Such scenarios raise the question: what is the minimally sufficient number of exceptions necessary to motivate a lexical generalization?9 Given the availability of lexical rules, productivity of a construction does not necessitate a syntactic analysis; conversely, confinement of a construction to a particular semantic class does not rule out a syntactic analysis—​such restrictions can be built into the features or meanings of the heads involved. Lexical(ist) approaches to the antipassive postulate a lexical rule that affects clausal argument structure, demoting the object/​theme to an oblique nominal. The antipassive morpheme is then added to the verbal base as a concomitant to the lexical rule. See Grimshaw and Mester (1985) for the application of this approach to Inuit languages and Woodbury and Sadock (1986) for refinements and modifications. A clear discussion can be found in Farrell’s (1992) lexical account of the Halkomelem antipassive. Working in the framework of Relational Grammar, Farrell argues that antipassivization cannot be derived syntactically; significantly, he observes that non-​eventive nominalizations can be formed from antipassive verbs but not from passive verbs:10 (21)

a. niʔ cən t’il-​əm ʔə aux 1subj sing-​ap obl ‘I sang the hymn.’  b. s-​t’il-​əm nmlz-​sing-​ap ‘song’

9 

tə det

st’iʔwiʔəł. hymn

Halkomelem (Farrell 1992, ex. (16a))

(Farrell 1992, ex. (17a))

For some languages, the data vary dialectically. For instance, the variety of Chukchi described by Skorik (1948; 1961; 1977) and the one I have observed have a fully productive antipassive; meanwhile, according to Dunn (1999: 200–​201, 70), the Telqep dialect has a low-​productivity antipassive. 10  The same detransitivizing suffix is used to derive the passive and the antipassive (see (18a)), but in the passive, that suffix must follow the transitive suffix -​Vt.

 

318   Maria Polinsky c. xws-​t’il-​əm agt.nmlz-​sing-​ap ‘singer’

(Farrell 1992, ex. (17b))

d. *s-​t’il-​ət-​əm nmlz-​sing-​tr-​pass If both the passive and antipassive constructions were formed syntactically, this contrast would be mysterious. If, however, antipassivization is a rule that modifies the argument structure of an input word, these facts follow straightforwardly. Gerdts and Hukari (2005; 2006) advance a lexicalist analysis that associates the antipassive with “agent maintenance” or prominence (see section 13.3); they propose a mapping rule that specifies antipassive verbs as two-​place predicates selecting only an agent argument. (22) Do not link object to an argument and cancel the inflectional position for object if there is one Head-​driven Phrase Structure Grammar (HPSG) characterizes the antipassive as a complex argument structure of the type shown in (23). The outer a(rgument)-​subject, i.e. the leftmost argument on the (local) argument list, maps to the syntactic subject; optionally, a distinct inner argument maps to an oblique object: (23) Antipassive ARG-​STR > These are just some available lexical analyses of the antipassive; other devices are also possible. The crucial consideration in each case concerns the language-​internal evidence for treating the antipassive as lexically specified. In those languages where the construction is (at least partially) productive, syntactic alternatives warrant attention.

13.4.2 Syntactic Approaches to the Antipassive The absence of a single, unifying syntactic analysis of the antipassive should not be surprising, given the significant cross-​linguistic variation we saw earlier. (Disagreements among researchers over the definition of “antipassive,” which I alluded to earlier, do not improve matters.) Analyses converge on the understanding that the difference between the transitive and the antipassive is in the status of the logical object. According to one family of analyses, accusative/​absolutive case is absorbed in the vP; other analyses contend that the antipassive allows object licensing, but that the structural locus of that licensing differs from the transitive.

13.4.2.1 Case Absorption “Demotion” of the logical object can be achieved by base-​generating a nominal element inside the VP. This abstract nominal constituent (AP) undergoes head movement and

 

Antipassive   319 absorbs the accusative/​absolutive case. Since the VP-​internal argument position is saturated, the verb cannot assign case to its object. Thus, if an antipassive object appears, it is analyzed as an adjunct. Schematically, the derivation proceeds as follows; the functional head v can assign only one case, and that case is assigned to the external argument. (24)

vP v’ v

VP ti

APi-V

Baker (1988) first observed the parallels between genuine NI and antipassivization, describing the antipassive morpheme as a type of incorporated object. Under Baker's analysis, the internal argument of the verb is filled by a non-​maximal nominal expression (most likely N, although some researchers identify it as D; cf. Bittner and Hale 1996a, b). Several morphological realizations are possible: an abstract “antipassive” nominal (expressed through the antipassive morpheme), an incorporated internal argument, or an implicit argument. This nominal antipassive marker undergoes c-​selection-​triggered head movement into the lexical verb (some researchers suggest that phrasal movement may be implicated as well). The incorporated nominal expression does not receive case, and the V+AP complex is syntactically intransitive. External case assignment (by the inflectional head or the functional head v that dominates the V+AP complex) proceeds according to language-​specific requirements. When the logical object is realized as an oblique expression, it is a genuine adjunct, never an argument; compare to the adjunct analysis of the by-​phrase in passives (but see Goodall 1993, 1999 and Bruening 2013, for a different view of passives). The nominal analysis is particularly appealing for antipassives whose marker is syncretic with reflexive or middle morphology (see (18a)), as reflexive and middle morphemes saturate the internal argument position of a two-​place predicate (Basilico 2004). Additionally, this approach predicts that NI and antipassivization are two sides of the same coin. Recall that the two operations are mutually exclusive in Chukchi: in fact, the Telqep dialect of Chukchi (Dunn 1999) uses NI in most instances where northern dialects employ a genuine antipassive (see n. 11). The nominal antipassive analysis is open to various interpretations with respect to the obligatory implicit argument—​a covert element whose syntactic representation is far from clear. It is generally assumed, following Rizzi (1986), that there is a difference between syntactically projected objects (which can be realized as pro, for example, in Italian, or as bound variables, as posited for Mandarin or Portuguese), and another, less-​ well-​defined category of lexically saturated implicit objects (IMPs), which need not be syntactically projected (as in English). Lexical saturation may proceed from different sources, including verbal semantics. The use of IMPs is often more lexically constrained than the use of syntactically represented objects.

 

320   Maria Polinsky The nominal antipassive analysis would be problematic if the silent object were expressed by a null pronominal; such a configuration would entail that a single theta-​ role be filled twice in the same predication. Since antipassives are often associated with lexical exceptions, this latter property suggests that objectless antipassives may in fact include IMPs. This possibility leads to a number of structural predictions extending from the differences between syntactically projected objects and IMPs. Specifically, only the former can participate in control, bind reflexives and be modified by adjunct small clauses. Although this aspect of antipassive constructions has not been systematically investigated, the available predictions could be put to the test in languages with implicit-​ object antipassives.

13.4.2.2 Different Object-​Licensing Positions Although the case-​absorption analysis has been popular, some researchers, in particular those working on Inuit, reject the connection between the antipassive and nominal head movement (Schmidt 2003; Spreng 2006; 2012). The main argument for this approach comes from the observation that in Inuit, noun incorporation and the antipassive marker are not mutually exclusive. Their co-​occurrence is illustrated in the following example from West Greenlandic:11 (25) a. Meeqqat tujuulu-​ler-​pai. children.abs sweater-​provide.with-​ind.3sg.3pl ‘S/​he dressed the children in sweaters.’ b. Meeqqat-​nik tujuulu-​li-​i-​voq. children-​ins sweater-​provide.with-​ap-​ind.3sg ‘S/​he dressed the children in sweaters.’

West Greenlandic

(Schmidt 2003: 388)

In the absence of a nominal AP head, the transitive/​antipassive distinction can be derived based on two related criteria: the way in which an object is licensed and the semantic contribution of the verbal root in terms of result or manner. Basilico’s (2012) analysis of Inuit ergative–​antipassive alternation is a particularly clear instance of such an approach. In Basilico’s analysis, the absolutive object is introduced and licensed by the functional head v (the ergative is considered an inherent case on that analysis). The verb root in this structure provides the result component to the interpretation of a given event, and the absolutive object serves as the measuring-​out argument (cf. (26a)). In the antipassive, the verb root provides the manner component, not the result component to the interpretation of a given event; this root combines with the light verb encoded by the antipassive morpheme, forming an intransitive. The resulting intransitive verb can optionally combine with a PP, whose interpretative 11  A number of researchers have suggested that what is referred to as noun incorporation in Inuit may involve different mechanisms than those found in familiar cases where the internal argument of a verb adjoins to the verbal head (Baker 1988; Johns 2009). If this proposal is on the right track, the argument against antipassivization as head movement loses its power.

 

Antipassive   321 contribution is in presenting the scale of a given event—​consider the schematics in (26b): (26) a.

vP v’

DPobject

vP

v [θ: theme] [abs]

v

√ [RESULT]

vP

b. v

PP v antipassive affix

√ [MANNER]

Some other analyses postulate the presence of an extra functional projection in the antipassive verbal complex (Alexiadou 1999; Spreng 2012; Compton, Chapter 34, this volume). The extra projection licenses the internal argument and is aspectual in nature. Details of interpretation differ, but most scholars hold that this abstract antipassive functional head is aspectually specified as [-​telic] or [+imperfective] (cf. (18b)). Its aspectual value is responsible for the persistent aspectual characteristics of the antipassive (section 13.3). This functional head v/​Asp selects the verbal root and licenses either inherent case or accusative case on the internal object. The accusative analysis of the antipassive object has been particularly prominent in work on Inuit (Bok-​Bennema 1991; Bittner 1994; Spreng 2012). Finally, the external argument of the antipassive receives case from the v/​Asp head or from a higher head: vP

(27)

v’

DPexternal v

AspP

DPinternal

Asp’ Asp

√/V

While details of implementation may differ, this approach relies on the contrast between absolutive licensing (high in the structure) and accusative licensing (low). The antipassive object is always “deeper” in the structure than the absolutive object—​but

 

322   Maria Polinsky if a case-​licensing system allows for both accusative and absolutive licensing, both objects receive structural case (cf. Woolford 1997; Aldridge 2008a; 2011; Legate 2008). However, this analysis does not explain why antipassive objects often appear in an oblique case. It is also unclear what motivates object drop in the antipassive; even if the two objects are licensed in different places in the verb phrase, the regular omission of one object but not the other remains unmotivated. The object-​omission problem does not arise on the case absorption analysis, which treats the antipassive object as an adjunct. The parallel between the object case assigned to the antipassive object and the more traditional accusative case is important in light of historical processes. It is often assumed that ergative languages can become nominative–​accusative if the ergative construction is lost and the antipassive is reanalyzed as a transitive (Anderson 1977; 1988; Comrie 1978; Harris and Campbell 1995: ch. 9; Spreng 2010; Aldridge 2011). The structural similarity between the antipassive object case and the true accusative could facilitate such a reanalysis.

13.4.2.3 Subject Case and Antipassives In addition to the demotion of the logical object, in morphologically ergative languages, the case of the subject also changes in the antipassive construction, from ergative to absolutive. Most researchers agree that the subject of a transitive verb and its antipassive/​intransitive counterpart are in the same position.12 Two main approaches have been explored. Some authors contend that the choice of subject case is a side effect of transitivity (Woolford 2006) and propose a special transitivity requirement which states that the ergative is incompatible with basic or derived intransitives. Bruening (2007) offers a detailed critique of this approach, protesting that “an additional transitivity requirement simply restates the definition of ergative case, making the explanation for its distribution include a statement of its distribution; it is inconsistent in its treatment of accusative case.” An alternative analysis (Marantz 1991; Bittner and Hale 1996b; Levin and Preminger 2015; Baker, Chapter  31, this volume; Baker and Bobaljik, Chapter  5, this volume) contends that case is determined by the presence of a case competitor. In this type of theory, nominative/​absolutive case is unmarked, and ergative and accusative are only assigned in opposition to a nominative. If a predicate has only one argument, that argument will receive the unmarked nominative/​absolutive case. If it has two arguments, one will receive the marked case—​ergative or accusative—​and the other will receive nominative/​absolutive. Ergative is assigned to the higher of two co-​arguments; accusative is assigned to the lower. An advantage of this approach is that it treats passive and antipassive consistently: in both structures, one argument is demoted (subject in the passive, object in the antipassive), leaving the other to receive default nominative/​ absolutive. 12 

See Bobaljik (1993a) for a different view, according to which transitive subjects move to a higher A-​position than transitive or antipassive subjects.

 

Antipassive   323

13.5  Syntactic Side Effects of Antipassivization This section discusses syntactic phenomena associated with the use of the antipassive. Although many of these effects are still poorly understood, some may motivate antipassive use in particular languages.

13.5.1 Way-​Station Effects In a number of languages, only (intransitive) subject or absolutive DPs can undergo A-​ bar movement, participate in control/​ raising chains, and bind anaphors. Antipassivization can serve as a “way-​station,” allowing the sole argument of a detransitivized verb to participate in such grammatical processes. In Chukchi, for example, relativization in participial clauses is possible only for absolutive arguments (Polinsky 1994): intransitive subjects and transitive direct objects can relativize directly with a gap, as shown in (28b) and (29b). For a transitive subject to relativize, it must first become an absolutive via antipassivization (30a); the subject of the antipassive is then relativized as in (30b).13 (28) a. ŋinqey pəkir-​gʔi. boy.abs arrive-​aor.3sg ‘The boy arrived.’

Chukchi

b. [ti  pəkərə-​lʔ-​ən]   ŋinqeyi   arrive-​ ptcp-​abs  boy.abs ‘the boy that arrived’ (29) a. Tumg-​e ŋinqey rəyegtetew-​nin. friend-​erg boy.abs save-​aor.3sg.3sg ‘The friend saved the boy.’ (transitive) b. [tumg-​ e  ti  rəyagtala-​lʔ-​ən] ŋinqeyi friend-​erg  save-​ptcp-​abs     boy.abs ‘the boy that the friend saved’ c. *[ ti  ŋinqey    rəyagtala-​lʔ-​ən] tumgətum   boy.abs  save-​ptcp-​abs    friend.abs (‘the friend that saved the boy’) 13  Chukchi is often used in illustration, but it is by no means unique in its relativization behavior. For example, Lower Grand Valley Dani (Bromley 1981) and Inuit (Bittner 1994; Manning 1996) also require antipassivization as a way-​station for subject extraction with a gap.

 

324   Maria Polinsky (30) a. Tumgətum ŋinqey-​ək ine-​nyegtele-​gʔi. friend.abs boy-​loc ap-​save-​aor.3sg ‘The friend saved the boy.’ (antipassive) b. [ti ŋinqey-​ ək  ine-​nyegtelewə-​lʔ-​ən] tumgətumi  boy-​ loc  ap-​save-​ptcp-​abs      friend.abs ‘the friend that saved the boy’ Relatedly, in a number of languages (not all of them ergative), non-​finite control can only target the intransitive subject position. Again, the antipassive “way-​station” can license control. In the following two examples from Jakaltec, control complement clauses must be intransitive (Craig 1977: 311–​327). The embedded clause in (31a) is antipassive; transitive non-​finite clauses with overt absolutives, as in (31b), are ungrammatical. (31) a. Ch-​ach to [PRO col-​wa-​l asp-​abs go help-​ap-​nom ‘You are going (there) to help him.’

y-​iñ gen-​to

Jakaltek

naj]. him

b. *Ch-​ ach to   [PRO ha-​ col-​ al       y-​ iñ    naj]. asp-​abs  go        erg-​help-​nom  gen-​to  him

(Craig 1977: 318)

A similar restriction appears in control complements in Chukchi (Nedjalkov 1976; Skorik 1977) and in some languages of Australia (Dixon 1979; 1994). Aldridge (2008a; 2008c; 2011) extends this restriction to a number of Austronesian languages (e.g. Seediq, Tagalog) and suggests that it may be tied to ergativity; however, the ergative status of these languages is debatable.

13.5.2 Scope Effects Antipassives and transitives reportedly differ in scope readings in Inuit. Inuit antipassive objects are restricted to narrow scope, while transitive absolutive objects can take wide or narrow scope (Bittner 1987, 1994; Benua 1995). In (32a), the absolutive object takes wide scope, and in (32b), the antipassive object must take narrow scope: (32) a. Atuartu-​t ila-​at ikiur-​tariaqar-​p-​a-​ra. student-​pl.erg part-​3pl.sg help-​must-​ind-​[+tr]-​1sg.3sg ‘There is one of the students that I must help.’

West Greenlandic

b. Atuartu-​t       ila-​an-​nik      ikiu-​i-​sariaqar-​p-​u-​nga. student-​pl.erg  part-​3pl.sg-​ins  help-​ap-​must-​ind-​[-​tr]-​1sg ‘I must help one of the students (any one will do).’ (Bittner 1994: 138)

 

Antipassive   325 Such scope differences, however, are not consistently associated with the antipassive cross-​linguistically. For example, in Adyghe, both transitive and antipassive clauses with an oblique object are ambiguous between surface and inverse scope:14 (33)

a. pŝaŝe-​m zeč’e-​r-​jә pjәsme-​(xe-​)r ∅-​ә-​txә-​ʁ. Adyghe girl-​erg all-​abs-​add letter-​pl-​abs 3abs-​3sg.erg-​write-​pst ‘A/​The girl wrote all the letters.’ (one > all; all > one) b. pŝaŝe-​r     zeč’e-​m-​jә    pjәsme-​(xe-​)m    ∅-​txa-​ʁe. girl-​abs  all-​obl-​add  letter-​pl-​obl     3abs-​write.ap-​pst ‘A/​The girl wrote all the letters.’ (one > all; all > one)

Likewise, in the English conative alternation (see section 13.7), both forms display scope ambiguities: (34) a. An athlete grabbed every curl-​bar. (a > every, every > a) b. An athlete grabbed at every curl-​bar. (a > every, every > a) Why is Inuit different from other languages? Let me  offer some considerations. The crucial test case for inverse scope is found in doubly quantified sentences where ‘one/​ a’ scopes over ‘every’ at the surface string. This configuration is present in (33) and (34) but not in (32); arriving at a robust comparison is difficult without additional West Greenlandic data. It is possible that fixed scope results from some interaction between quantification and modality (which is known to be quite complex; cf. Hacquard 2006: 118ff. and references therein), between quantification and negation, or between quantification and the defined aspectual properties noted for the Inuit antipassive (recall (27)). The bottom line is that differences in West Greenlandic scope readings may well be orthogonal to the transitive/​antipassive contrast. Assuming a structure such as (27), there is no reason to expect antipassive scope relations to differ from transitive scope relations. Our empirical knowledge of these phenomena must be considerably enriched before analysis can move beyond speculation.

13.5.3 Antipassive and Agreement Antipassives may appear when the subject and object are in an “inverse” relationship—​ i.e. when the subject is lower than the object on one or more hierarchies: person (1>2>3); number (sg > pl); animacy (human > animate > inanimate). Languages vary in the combinations they consider “inverse,” and careful examination of a given agreement system is always needed; in general, however, detransitivizing morphology—​passive 14  The ergative and the oblique in Adyghe have the suffix -​(V)m. Arkadiev and Letuchiy (2008) treat all instances of -​(V)m as oblique; here and in Caponigro and Polinsky (2011), I gloss -​(V)m differently depending on its function.

 

326   Maria Polinsky or antipassive—​can be used to mark inverse configurations. Inverse marking through antipassive is common in Australian languages. In Yukulta, a third person can act transitively upon another third person, but the antipassive is used when a third person acts upon a first person: (35) a. Kungul-​i=ka-​nt-​a mosquito-​erg=tr-​nprs-​realis ‘A mosquito bit him.’

paa-​tya. bite-​ind

Yukulta (Keen 1983, ex. (146))

b. Kungul-​ ta=thu=yingk-​ a         paa-​ tya. mosquito-​abs=1sg.dat-​nprs.intr-​realis  bite-​ind ‘A mosquito bit me.’ (Keen 1983, ex. (147)) Some instances of inverse ~ antipassive correlation may reveal a diachronic connection but be synchronically distinct. I will discuss one such case in section 13.6.2.

13.6  Related Constructions 13.6.1 Antipassive Lookalikes Above, I  mentioned certain linguistic phenomena that border on or subsume antipassives: some instances of PNI, object NI constructions, and some bi-​absolutive (bi-​ nominative) constructions. Here, I discuss two other constructions that resemble the antipassive but should be differentiated from it. The first is the agent focus (AF) construction, reported for a number of Mayan languages (see Stiebels 2006 for an overview). Although AF has some hallmarks of the antipassive, it is considered a transitive, not an intransitive, construction (Aissen 2011); AF is variously analyzed as a separate voice (Tonhauser 2003), an inverse form (Aissen 1999a), or a specialized agreement form for A-​bar subject extraction (Stiebels 2006). Some Mayan languages, such as Q’anjob’al, have both genuine antipassives and AF—​see (10) for the antipassive and Coon et al. (2014) for a discussion of AF. However, some early work, particularly Larsen and Norman (1979), identifies AF as an antipassive construction; it would be prudent to assess the AF ~ antipassive relationship in each language separately. Differential object marking (DOM) resembles the antipassive in that objects of transitive verbs receive different encoding depending on animacy, specificity, etc. However, in DOM, the logical object remains a syntactic direct object, regardless of overt marking (Aissen 2003b). Furthermore, semantic factors associated with DOM differ from those that condition antipassives. A  similar superficial parallel arises between antipassives and differential subject marking (DSM) constructions; although the case of the subject shifts in both constructions, only in the antipassive does this change in subject case-​ marking crucially depend on the status of the object.

 

Antipassive   327

13.6.2 Spurious Antipassive As noted, antipassives may be used to indicate an atypical/​inverse hierarchical relationship between agent and theme. Noting several unusual properties of this phenomenon, researchers have labeled the antipassive that serves this inverse function the “crazy antipassive,” “eccentric agreement” (Hale 2002), or “spurious antipassive” (Halle and Hale 1997). Spurious antipassives in Chukotko-​ Kamchatkan languages represent a well-​ documented case of such “inverse” agreement (Comrie 1979, 1980; Nedjalkov 1979; Bobaljik and Branigan 2006). Simplifying somewhat, these antipassives occur in contexts where a second-​or third-​person participant acts upon a first-​person participant. Compare (36a) and (36b); see Dunn (1999: 181–​184) and Bobaljik (2007) for complete paradigms. In (36b, c), the verb carries an antipassive prefix; agreement is intransitive, but pronominal case-​marking remains transitive. Both antipassive affixes, ine-​and -​tku-​, participate in inverse alignment: (36) a. ɣəm-​nan ɣət 1sg-​erg 2sg.abs ‘I saw you.’

tə-​ɬʔu-​ɣət. 1sg.subj-​see-​aor.2sg.obj

Chukchi

b. ə-​nan    ɣəm     ine-​ɬʔu-​ɣʔi. 3sg-​erg  1sg.abs  ap-​see-​aor.3sg ‘S/​he saw me.’ c. ɣət-​nan    muri    ɬʔu-​tku-​∅. 2sg-​erg  1pl.abs  see-​ap-​aor.2sg ‘You (sg.) saw us.’ Bobaljik and Branigan (2006) and Bobaljik (2007) argue that (36b, c) are regular transitive clauses, with both subject and object licensed in a typical manner by the inflectional head T. The “spuriousness” pertains to the morphological interpretation of these clauses. Since these authors understand agreement to be post-​syntactic, they locate antipassive insertion within the mapping of narrow syntax to the morphological component; the features of the lower argument (object) are deleted at T, rendering T apparently intransitive. This analysis eliminates “spuriousness” by maintaining the uniformity of the antipassive affix; however, it crucially depends on Bobaljik’s (2008) conception of agreement as a post-​syntactic operation. If object features play no role in derivation, then it is surprising that spurious antipassives distinguish between singular and plural objects. ine-​ is used only for singular first-​person objects, and -​tku-​ for plural (cf. (36b, c); see Dunn 1999: 184). In addition, an inverse form marked with the prefix ne-​ is used to encode a plural third-​person participant acting upon a (singular or plural) third-​person participant. Most importantly, regular antipassive markers are preserved in action nominalizations, but the “spurious” ones are lost, just like all other agreement markers in

 

328   Maria Polinsky nominalizations.15 This suggests that spurious antipassives are categorically different from true antipassives. (37) a. Gəm ʔituʔit-​ək 1sg.abs goose-​loc ‘I saw a/​the goose.’

ine-​ɬʔu-​ɣʔek. ap-​see-​aor.1sg

Chukchi

b. Gəm-​in     ʔituʔit-​ək/​*ʔituʔit-​in    ine-​ɬʔu-​wərg-​ən 1sg-​poss  goose-​loc/​goose-​poss  ap-​see-​nmlz-​abs ‘my seeing of a/​the goose’ (38) a. ətɬʔəg-​e ɣəm father-​erg 1sg.abs ‘Father saw me.’

ine-​ɬʔu-​ɣʔi. ap-​see-​aor.3sg

b. ətɬʔən-​ in   (ɣəm-​in/​*ɣəm)     ɬʔu-​wərg-​ən/​*ine-​ɬʔu-​wərg-​ən father-​poss  1sg-​poss/​1sg.abs  see-​nmlz-​abs/​ap-​see-​nmlz-​abs ‘Father’s seeing me’ Bobaljik and Branigan’s generalization—​that clauses such as (36b, c) are transitive—​ holds. However, ine-​ and -​tku-​ here are not antipassive morphemes and have nothing to do with valency-​changing morphology; instead, they are person/​number agreement markers (see Abramovitz 2015 for a similar conclusion). The details of such agreement are yet to be worked out. It is possible that the two functions of ine-​/​–​tku-​ are related historically; on the synchronic level, however, the Chukchi “spurious” antipassive is not antipassive at all.

13.7  Antipassive Is Not Unique to Ergative Languages Some researchers have suggested that antipassives are unique to ergative languages (Silverstein 1976; Dixon 1979; Spencer 1991: 24) or, even more narrowly, to syntactically ergative languages, that is, languages where the ergative cannot undergo A-​bar movement (e.g. Otsuka 2000). However, there is nothing in the basic definition of the antipassive that predicts this restriction. Choctaw (Davies 1984), Chamorro (Cooreman 1988, 1994), Kiowa (Watkins 1984), Māori (Bauer 1983), German (Müller 2011), Romance (Postal 1977; Masullo 1992; Mejias-​Bikandi 1999; Medová 2010), and Slavic (Say 2005;

15  Chukchi event (action) nominalizations typically allow just one possessive form, which can correspond to either the external or internal argument, making them ambiguous. The inclusion of both possessives within a nominalization is permitted but not preferable.

 

Antipassive   329 Medová 2010) are good examples of accusative languages with antipassives (for more examples from a survey of grammars, see Polinsky 2013). Within the framework of Relational Grammar, Postal (1977) specifically argues against the unique association of the antipassive with ergativity; see also Heath (1976), Givón (1984); Polinsky (2013). What is truly at issue here is visibility: in ergative languages, presence of the antipassive correlates with an obvious change of subject encoding from ergative to absolutive; thus, this construction is more noticeable in ergative languages than it is in accusative languages, but it is not limited to ergative languages. Conversely, passives are not impossible in ergative languages, contra Laka (1993b: 168), Dixon (1994: 152), van de Visser (2006). Quite a few languages exhibit both passive and antipassive constructions. Just within morphologically ergative languages, we can find both passive and antipassive in Halkomelem, Inuit, Georgian, several Mayan languages (Vapnarsky et al. 2012), and possibly Basque. It is true, however, that passives are generally less common in ergative languages (Kazenin 2001b: 926), leading Nichols (1992: 158) to propose the following universal: if a language has an antipassive but no passive, that language is ergative. In English, the closest parallels to the transitive ~ antipassive alternation are the conative alternation and the unexpressed object alternation, for which multiple lexical restrictions apply; Blight (2004) analyzes these alternations, as well as the preposition-​drop alternation, as English antipassives. The conative alternation contrasts two instances of the same verb: one that takes a direct object, and another that takes a prepositional complement. Only the direct object is construed as affected by the verbal event (Levin 1993: 5–​11, 41–​42; Beavers 2011; Vincent 2013). For example, (39a) entails that the hunter hit the target, rendering the continuation infelicitous; in (39b), this entailment is absent: (39)

a. The hunter shot the bear #but he missed. b. The hunter shot at the bear but he missed.

Levin associates the conative alternation with verbs that describe change of state achieved through motion and contact (as opposed to verbs that denote pure change of state, such as break, verbs that denote contact only, and verbs that denote motion only). It may be possible to extend this generalization to antipassives with an explicit oblique complement, but more cross-​linguistic work on change-​of-​state verbs and implicit-​ argument verbs is needed to determine the relevant lexical restrictions. It is true that, whereas use of a prototypical transitive verb entails a change of state in the object participant, the corresponding antipassive cancels such an entailment, and this cancellation correlates with the oblique marking of the object (Hopper and Thompson 1980; Tsunoda 1981b; Dowty 1991; Van Valin 1991; among others). Omission of the affected object participant can have two consequences. First, when a clause lacks overt mention of a participant affected by the event (incremental theme), the event is likely to be interpreted as incomplete (Dowty 1991; Spreng 2010; Basilico 2012). This consequence echoes the established correlation between the antipassive construction and the imperfective

 

330   Maria Polinsky interpretation. Second, in the absence of an affected patient, it becomes possible to interpret the agent as an affected participant of the event. The reflexive construction similarly permits a verbal event to cause change in the agent; as discussed earlier, antipassive and reflexive mechanisms do overlap in some languages. The unexpressed object alternation in English brings together the following types of sentences (Levin 1993: 33–​38): (40) a. I have ironed the clothes. b. I have ironed. Levin (1993: 33) observes that this alternation occurs with a wide range of activity verbs where the missing object is understood as the “typical object” of a given verb; of course, the notion of “typical object” is vague, but it seems that a similar observation can be made for the unexpressed object of the antipassive construction. All things considered, it is probably better to account for the English conative alternation and unexpressed-​object verbs via lexical rules (section 13.4.1); however, a syntactic approach like the one illustrated in section 13.4.2.1 is also possible.

13.8 Conclusions Antipassives have long been considered “exotic”—​found in exotic languages and associated with exotic syntax. One of my goals in this chapter has been to illustrate that the antipassive is in fact well behaved, observable wherever the logical object of a transitive predicate appears as a non-​core argument or an adjunct. Too often, we seek antipassives in places where they do not exist while ignoring obvious instances in our own “Standard Average European” backyard. Once we agree on a more manageable landscape for navigation, the task of identifying antipassives becomes much simpler. On the one hand, a number of constructions that resemble the antipassive are truly distinct; even if two constructions have similar discourse functions or similar morphological hallmarks, they need not be syntactically equivalent. On the other hand, certain neglected constructions, such as (P)NI objects, meet the structural definition of the antipassive and should be duly considered under that rubric. Based on the criteria discussed here, we must reject the notion that antipassives occur only in ergative languages. Likewise, it is unreasonable to maintain that antipassives and passives are mutually exclusive; nothing in the definition of the antipassive makes such a prediction, and a number of languages provide empirical evidence in support of anti/​passive compatibility. The antipassive bears analysis either as a case frame of individual lexically specified verbs that alternate with regular transitives, or as the output of a syntactic operation. Only in the latter case can one assume, following Silverstein (1972), that “the sense [of the antipassive—​MP] is clearly equivalent to a transitive.” Lexical analysis is plausible when only a subset of a language’s verbal lexicon participates

 

Antipassive   331 in the alternation and when independent evidence supports such an analysis, as in Halkomelem nominalizations. Within syntax, two main approaches are taken, which differ along two related dimensions: the character of the antipassive affix (nominal/​verbal), and object licensing. Under the first approach, a nominal element saturates the internal argument of a two-​place verb, and as a result, the verb cannot assign case to its object. Under the second approach, the difference between the transitive and the antipassive is reduced to a licensing distinction between absolutives (high objects) and accusatives (low objects). Often, such differential licensing is associated with an extra functional head with aspectual connotations. These two approaches attempt to account for the semantic and pragmatic effects of the antipassive without defining the construction by these properties.

Acknowledgments A portion of this work was supported by funding from NSF (BCS-​114223, BCS-​137274, BCS-​ 1414318), Harvard University, and the Max-​ Planck Institute for Evolutionary Anthropology. Thanks to Rafael Abramovitz, David Basilico, Grant Goodall, Ivona Kučerová, Beth Levin, Gesoel Mendes, Eric Potsdam, Jerry Sadock, Adam Singerman, and Bettina Spreng for helpful comments; I am solely responsible for any remaining errors. Unless stated otherwise, examples come from my own fieldnotes. I am grateful to Raxmet Yesheva (Adyghe), Peter Inenlikey and Vladimir Raxtilin (Chukchi), Winifred Bauer (Māori), Pedro Mateo Pedro (Q’anjob’al), Sisilia Lutui (Tongan), and Roksolana Mykhaylyk (Ukrainian) for linguistic examples.

Abbreviations abs, absolutive; acc, accusative; add, additive; agt, agent; ap, antipassive; asp, aspect; aux, auxiliary; clf, classifier; com, comitative; compl, completive; dat, dative; det, determiner; erg, ergative; fem, feminine; gen, genitive; ind, indicative; ins, instrumental; itr, intransitive; loc, locative; masc, masculine; n, non-​; ni, noun incorporation; nmlz, nominalizer; nom, nominative; obj, object; obl, oblique; pass, passive; pl, plural; pni, pseudo noun incorporation; poss, possessive; prs, present; pst, past; ptcp, participle; refl, reflexive; rel.noun, relational noun; subj, subject; tr, transitive.

 

Chapter 14

Rem arks on th e re l at i on bet w een case -​a l i g nme nt an d c onstitue nt  orde r Tarald Taraldsen

14.1 Introduction This chapter focuses on a descriptive generalization which claims that no language both has basic SVO order and Erg(ative)/​Abs(olutive) case-​alignment (sometimes called “Mahajan’s Generalization”). I’ll call this the *SErgVOAbs generalization. I do not intend to offer a theoretical account of this descriptive generalization, which according to Richard Kayne (p.c.) goes back to Schwartz (1972).1 My aim is rather to isolate some empirical issues that bear on the proper formulation of the generalization, and to examine to what extent various theories of case-​assignment provide a way of understanding it. Along the way, we will see how different theoretical interpretations enable us to isolate classes of languages that are expected not to fall under the generalization.

14.2  What is the *SErgVOAbs Generalization a Generalization about? There are different ways for an SVO order to arise, e.g. by V-​movement into the CP-​ layer, by V-​movement to T or some other functional head inside TP or by not raising 1 

Schwartz (1972) does not indicate how he arrived at the generalization. Paumarí (see Aikhenvald 2009) and Zoque (see Faarlund 2012) seem to be the only known exceptions to it. There is a brief comment on how Paumarí and Zoque might be dealt with in section 14.4.5.

 

The relation between case-alignment and constituent order    333 complements of V in situ (assuming universal H-​complement order as in Kayne (1994, 2010)). This is one factor in determining what the empirical coverage of the *SErgVOAbs generalization might be expected to be: Does it apply regardless how SVO order arises in a given language or sentence type, or is it restricted to languages/​sentence types where SVO order arises in one specific manner? For each of these possibilities, some theoretical accounts of the generalization will be more plausible than others. Likewise, there may be different ways for Erg/​Abs alignment to arise. For example, there may be an empirically motivated distinction between syntactic ergativity and morphological ergativity. If so, does the *SErgVOAbs generalization hold regardless how Erg/​Abs alignment arises, or does it hold only when Erg/​Abs alignment arises in one of the possible ways? Settling this issue will also impact on the choice of a theoretical account of the generalization. In this section, we look at the different derivational paths that may lead to SVO order as well as the ways Erg/​Abs alignment might arise.

14.2.1 What Kind of SVO Might *SErgVOAbs Be a Generalization about? The first issue we will discuss, is what should be meant by SVO in relation to *SErgVOAbs. This linearization pattern may have different structural sources. Does *SErgVOAbs apply to all sentences with SVO order regardless of how the SVO order is created? If that were the case, Kashmiri is a counterexample, since Kashmiri can combine SVO order with Erg/​Abs case-​alignment in main clauses and a range of embedded finite clauses. The following examples adapted from Koul and Wali (2006) illustrate this, but they also illustrate that Kashmiri is a V2 language: (1)

Aslam-​an dits kita:b Mohn-​as Aslam-​Erg gave.fsg book.fsg. Mohan-​Dat ‘Aslam gave Mohan a book yesterday.’

(2)

Ra:th dits Aslam-​an kita:b yesterday gave-​fsg Aslam-​Erg book.fsg ‘Yesterday, Aslam gave Mohan a book.’

ra:th yesterday Mohn-​as Mohan-​Dat

Hence, we discard the possibility that *SErgVOAbs is about just any SVO order. Since Kashmiri has SOV order non-​finite clauses as well as well as in conditionals, alternative questions and relative clauses and, as we have seen, also has XVSO order in main clauses (and all embedded clauses that admit SVO order), it is natural to analyze it as a language with SOV order within TP and “verb second” V-​movement, much like German or Dutch. Then, main clauses with SVO order have the V in C or higher, and we can eliminate the counterexample by restricting *SErgVOAbs to sentences with SVO order within TP (at some stage of the derivation).

 

334   Tarald Taraldsen However, SVO order within TP might arise in a number of different ways. A priori, both V and O might remain in VP both in SVO and SOV languages, in which case VO vs. OV would be determined by the setting of a directionality parameter: (3) a [VP V O ] b [VP O V ] Alternatively, the V remains in VP in both SVO and SOV languages, but the O is raised out of VP only in SOV languages, as in Kayne (1994): (4) a [VP V O ] b O… [VP V ] It is equally plausible, however, that there are SVO languages where an O raises out of the VP, but this is masked by V-​movement across the raised O; cf. Johnson (1991), Taraldsen (2000): (5) O … [VP V ] → V … O … [VP ] It may be that the verb moves to T in languages where it ends up preceding sentence adverbs, as in finite clauses in French (Pollock 1989). But in SVO-​languages like English and most Mainland Scandinavian varieties, the V must follow all adverbs inside TP, and even in those Mainland Scandinavian varieties that may have the V preceding adverbs inside TP, the V arguably doesn’t raise to T; cf. Bentzen (2006). In principle, there might be as many types of TP-​internal V-​movement as there are functional heads between C and VP. If all these possible derivations of TP-​internal SVO order are actually realized in one language or another, it seems rather unlikely that the *SErgVOAbs generalization is valid for all SVO languages. Therefore, if the generalization does hold for all SVO languages, and we think this is not accidental, we will be led to restricting the set of possible derivations leading to TP-​internal SVO order. Conversely, if the generalization turns out not to hold for all SVO languages, we will maintain the full range of possible ways of being SVO TP-​internally and restrict the empirical scope of the *SErgVOAbs generalization so that it only should hold for SVO orders derived in certain specific ways.

14.2.2 The Empirical Scope of the Generalization and Theoretical Accounts Suppose the empirical scope of the *SErgVOAbs generalization includes SVO orders with V inside VP. If SOV languages also have both V and O inside VP and OV is determined merely by the setting of a directionality parameter, the account offered by Mahajan (1994) (see section 14.4.2) is the only existing proposal capable of differentiating between

 

The relation between case-alignment and constituent order    335 SVO and SOV sentences with respect to Erg/​Abs alignment, since it is the only account in which only linear order plays a role. If SOV orders only result from moving O out of VP, and the derivation of SVO orders cannot involve moving the O out of VP, theoretical accounts may capitalize on a structural difference between SVO sentences and SOV sentences. There may be different ways that raising the object could interact with case-​assignment. For example, it might enable T to assign nominative/​absolutive case to the object instead of the subject; cf. the discussion in section 14.5.1. If the scope of the generalization includes TP-​internal SVO orders derived by V-​ movement across O previously extracted from the VP, the raising of the object cannot be the sole decisive factor, but another analytical option becomes available. Raising V might make Erg/​Abs alignment impossible. Since VSO-​languages may have Erg/​Abs alignment, one must then say that only V-​movement placing the verb between the S and the O is incompatible with Erg/​Abs alignment. This is the basic idea pursued in Taraldsen (2010), where it is developed in such a way that it is immaterial exactly which of the heads within TP the verb raises to, as long as it ends up in a position higher than a previously raised object, but lower than the subject position. In fact, Taraldsen’s account predicts that even an SVO sentence with both the V and O within VP disallows Erg/​Abs alignment, since the subject is an intervener.

14.2.3 Ways of Being Ergative Just as there is more than one way to create an SVO order, there may well be more than one way to create an Erg/​Abs case-​alignment. This, in part, is due to the distinction between syntactic case and its morphological realization. A syntactic Erg/​Nom/​Acc pattern may surface as an Erg/​Abs pattern as a result of Nom/​Acc syncretism, a possibility investigated in Legate (2008, 2012a), for example. In addition to this, syntactically determined Erg/​Abs alignments may fall into different categories. For example, ergative case may be assigned as a “semantic case” to any DP in Spec-​vP in certain languages, while in other languages a DP in Spec-​vP is assigned ergative case only if there is also an object. In the first type of ergative language (“active languages”), the subjects of certain intransitive verbs have ergative case, but in the second type of language, the subject of an intransitive verb never has ergative case. For languages of the second kind, both Bittner and Hale (1996a) and Baker (2015) appeal to the notion of “case competition” introduced by Marantz (1991); cf. section 14.4.4. But Bittner and Hale also assume that ergative case is assigned by I, and that an object can become a case-​competitor visible to I in two different ways—​either by raising the object to Spec-​IP or by raising the V to I. In the former case, the language displays properties associated with syntactic ergativity, in Bittner and Hale’s sense, and in the latter case it does not.2 2 

But Dyirbal, one of Bittner and Hale’s examples of a syntactically ergative language, is analyzed as syntactically tripartite Erg–​Nom–​Acc by Legate (2012a).

 

336   Tarald Taraldsen If surface Erg/​Abs alignment can arise in all the different ways as mentioned, it becomes quite difficult to make theoretical sense of a descriptive *SErgVOAbs generalization understood to rule out all surface combinations of SVO order and Erg/​Abs alignment. Thus, either there is essentially only one single way of being ergative or *SErgVOAbs should be expected to have counterexamples within some independently identifiable type of ergative languages.

14.2.4 Summary I have looked at different ways SVO orders may be created as well as different ways Erg/​ Abs alignment might arise. The conclusion is that either there should be empirical evidence that the *SErgVOAbs generalization fails to hold in a range of well-​defined cases, or else there should be only one way of obtaining Erg/​Abs case-​alignment and perhaps only one way being SVO TP-​internally.

14.3  More on the Empirical Coverage of the Generalization The *SErgVOAbs generalization may be thought of as making a claim only about SVO sentences of a certain kind combining with an overtly marked Erg/​Abs case-​alignment. But its empirical scope might be broadened beyond this, and I will now consider the possibility that the descriptive generalization might extend to tripartite languages and to neutral languages with Erg/​Abs agreement patterns.

14.3.1 Tripartite Languages One may wonder whether the *SErgVOAbs generalization holds (if it does) because ergative case cannot be assigned in (certain) SVO-​structures, or because the object cannot receive absolutive case in such structures. Information about the so-​called tripartite languages would bear on this. In tripartite languages, a transitive subject is assigned ergative case and an intransitive subject is assigned nominative/​absolutive, as in languages with Erg/​Abs alignment, but an object is assigned accusative. Hence, if *SErgVOAbs really is about the availability of absolutive case for the object, the generalization is consistent with the existence of tripartite SVO-​languages, but if the generalization is about ergative case, there should be no such languages. In the WALS sample, there are no tripartite languages with dominant SVO order vs. two tripartite SOV languages. But there are also no tripartite VSO languages (vs. one

 

The relation between case-alignment and constituent order    337 ergative/​absolutive VSO language). Two languages with no dominant order are classified as tripartite. One of these is Nez Perce, although the Nez Perce examples reproduced in Baker (2015) all seem to show SVO order. We also note that if there are tripartite SVO languages, it seems unlikely that *SErgVOAbs can be surface true, since there is no way of excluding a surface Erg/​Abs alignment from arising from a Nom/​Acc syncretism in an underlyingly tripartite language, and there is little reason to suppose that Nom/​Acc syncretism would be sensitive to SVO vs. SOV and VSO. Assuming, for example, the case-​theory of Caha (2009), where cases are represented as sets of features standing in a subset relation to each other, the syntax would produce (6) in a language L with underlying tripartite alignment:3 (6) a A-​{ X, Y, Z } [VP V O-​{ Y, Z } b S-​{ Y } [VP V ] But if L has the exponents B ←→ { X, Y, Z } and C ←→ { Y, Z }, but no D ←→ { Z }, (6) will surface as (8), by the Superset Principle: (7)

The Superset Principle A morpheme A with the lexical entry A ←→ F, where F is a feature set, can spell out any feature set F’ which is a subset of F.

(8)

a A-​B [VP V O-​C ] b S-​C [VP V ]

14.3.2 Ergative/​Absolutive Agreement Patterns One may also wonder whether the *SErgVOAbs generalization should be expected to hold only for languages with overt case-​marking. In particular, it seems conceivable that it might also hold for neutral languages with Erg/​Abs agreement patterns. An agreement system is said to follow an Erg/​Abs agreement pattern, if it treats S on a par with O to the exclusion of A. I will focus on the situation where some head only agrees with absolutive S and O. A Nom/​Acc agreement pattern is one in which A and S are treated the same way to exclusion of O. For example, some head only agrees with A and S. Languages with Ergative/​ Absolutive case-​ alignment can have either Ergative/​ Absolutive agreement (three languages in WALS, e.g. Central Yup’ik) or Nominative/​ Accusative agreement (12 languages in WALS, e.g. Berber). Importantly, neutral languages are also compatible with both agreement patterns (five languages in WALS

3  To avoid introducing orthogonal issues at this point, I don’t adopt here Caha’s claim that the sets of case features should be replaced with syntactic trees where each case-​feature is a head. However, this ingredient of Caha’s proposal will assumed in section 14.4.3.

 

338   Tarald Taraldsen with Ergative/​Absolutive agreement, e.g. Jacaltec (discussed in Woolford 1999), vs. 52 languages with Nominative/​Accusative agreement, e.g. Nahuatl), but languages with Nominative/​Accusative case-​alignment cannot have Ergative/​Absolutive agreement (no such languages in WALS). For our purposes, it is important to consider the possibility that agreement patterns are determined by prior case-​assignment, i.e. agreement is case-​sensitive in the sense of Bobaljik (2008) and Baker (2015). For example, a probe H may be able to agree only with DPs bearing absolutive case, leading to an Erg/​Abs agreement pattern in any language with Erg/​Abs case-​alignment. To allow for the combination of Erg/​Abs case-​alignment with a Nom/​Acc agreement pattern, one may follow Baker in adapting Bobaljik’s (2008) proposal: There is a case-​hierarchy … Erg > … Acc > Nom/​Abs and that if P can agree with a DP with case K in … Erg > … Acc > Nom, it can agree with a DP with any K’ below K. A language with Erg/​Abs case-​alignment and a Nom/​Acc agreement pattern would then have H able to agree with a DP with ergative case, hence also with DPs with Absolutive case. That is, the existence of (9)a (where the subscript x marks agreement) in a language entails the existence of (9)b in the same language: (9) a Hx … b Hx …

[vP Ax-​Erg [VP V O-​Abs ]] [vP Sx-​Erg [VP V ]]

By Relativized Minimality, H will then agree with both A and S, but never with O (unless O moves before H is merged).4 This type of account excludes the combination of Nom/​Acc case-​alignment with an Erg/​Abs agreement pattern. For H to agree with O, H must be able to agree with accusative DPs, but then it can also agree with nominative DPs, and therefore Relativized Minimality guarantees that a Nom/​Acc agreement pattern ensues (again provided the O doesn’t move before H is merged). Unlike accounts where agreement is subject to Relativized Minimality, but is not case-​ sensitive, this approach also allows one to account for failure of agreement with ergative S in languages where H agrees with O and absolutive S. It is sufficient to stipulate that H only agree with absolutive DPs. However, the descriptive generalization that no language has both Nom/​Acc case-​ alignment and Erg/​Abs agreement is based on surface case. Therefore, case-​sensitive agreement will not capture the generalization if an underlying Erg/​Abs case-​alignment can surface as a Nom/​Acc alignment. But there seems to be no way this could happen: The exponent associated with both ergative and absolutive case for subjects, will presumably also be associated with absolutive objects, so that the result is a neutral alignment. Things are more delicate if we start out from an underlying tripartite case-​ alignment. On the approach to case-​sensitive agreement taken here, it should be 4 

Notice that these assumptions make it impossible to derive a tripartite agreement pattern. I don’t know whether this is a good result.

 

The relation between case-alignment and constituent order    339 possible to have H agreeing with accusative O and nominative S, but not with ergative A. So, if the same exponent could be chosen both for ergative and nominative case, but not for accusative case, the result would be a surface Nom/​Acc case-​alignment combined with an Erg/​Abs agreement pattern. But this possibility too is ruled out, if we adopt the case-​hierarchy as suggested and Caha’s (2009) account of case-​syncretism. On this account, each case K in the hierarchy corresponds to set of features properly including the set of features corresponding to any K’ below K in the hierarchy. For example, if the nominative corresponds to { Z }, the accusative corresponds to { Y, Z } and the ergative minimally corresponds to { Z,Y,Z }. A syncretism over the ergative and the nominative arises from the Superset Principle (repeated below) just in case there is an exponent A ←→ { Z, Y, Z }, but no B ←→ { Y, Z } and no C ←→ { Z }(because of the Elsewhere Principle in (10)): (7)

The Superset Principle A morpheme A with the lexical entry A ←→ F, where F is a feature set, can spell out any feature set F’ which is a subset of F.

(10)

The Elsewhere Principle If two morphemes A and B with entries A ←→ F and B ←→ G can both spell out a feature set H by the Superset Principle, but G is a proper subset of F, B must be chosen.

If so, an ergative/​nominative syncretism must necessarily also include the accusative. Again, the result is a neutral language. If indeed an Erg/​Abs agreement pattern can only arise in languages with an underlying Erg/​Abs or Erg/​Abs/​Acc case alignment, but an underlying Erg/​Abs(/​Acc) case-​ alignment may surface as neutral alignment as a result of case-​syncretism, the next question is whether the *SErgVOAbs generalization extends to neutral languages with Erg/​Abs agreement patterns.5

14.3.3 Summary We have suggested that one should investigate whether some appropriately sharpened version of the *SErgVOAbs generalization which is valid for languages with overtly marked Erg/​Abs case-​alignment, also holds for tripartite languages and neutral 5 

Past participle agreement in Standard Italian might be said to follow an Ergative/​Abs (or Acc) agreement pattern, since the participle only agrees with O and unaccusative S. The order of the participle and the object is VO, but the object doesn’t actually trigger agreement unless it moves when the participle also has an external argument. That is, only object clitics and wh-​moved objects may trigger agreement in transitive sentences; cf. Kayne (2000a). However, this might be because the ergative subject is a “defective intervener” for agreement, unless the object is raised to a position between the subject and the agreeing probe; cf. section 14.5.3.

 

340   Tarald Taraldsen languages with Erg/​Abs agreement patterns. If the generalization does hold for tripartite languages, we conclude that the generalization really is to be understood as *SErgVO, but if it doesn’t, the generalization must be *SVOAbs. If the generalization is valid for neutral languages with Erg/​Abs agreement patterns, we conclude that the generalization is really about syntactic case, as one would be inclined to suspect to begin with. However, although there are no neutral SVO languages with Ergative/​ Absolutive agreement in the WALS sample, there are 3 neutral SVO languages with active agreement (Taba, Mountain Arapesh, Apurinâ).6 If the active agreement pattern reflects underlying Active/​Inactive (or “Split-​S”) case-​alignment in these three languages, one may conjecture that *SErgVOAbs doesn’t hold for active languages, plausibly because of the way the ergative is assigned in such languages. But then, the fact that there are no SVO languages with overt Active/​Inactive case-​alignment in WALS, comes as a surprise.

14.4 *SErgVOAbs = *SErgVO? If a suitably precise version of the *SErgVOAbs generalization holds, one will ask why. We will assume that any interesting explanation will be based on assumptions regarding syntactic case rather than its morphological realization.7 If so, the generalization may hold either because the subject cannot have syntactic ergative case in certain structures linearizing as SVO (*SErgVO), or because the object cannot bear syntactic absolutive case in such structures (*SVOAbs) (or both). As mentioned earlier, information about tripartite languages should be relevant to the choice between these options. If SVO order is merely incompatible with assignment of syntactic absolutive case to the object, e.g. by T, there should be tripartite SVO languages (as well as SVO languages with surface Erg/​Abs alignment arising from underlying Erg/​ Nom/​Acc by syncretism). But if SVO order is incompatible with the subject surfacing with ergative case, there should be no tripartite SVO languages (and no SVO languages with surface Erg/​Abs alignment arising from underlying Erg/​Nom/​Acc by syncretism). At present, however, no conclusion can be drawn from this, since the data available to me seems insufficient. In this section, we will look at the major proposals regarding the assignment of syntactic ergative case and evaluate them with respect to the prospects of providing an account of *SErgVOAbs as *SErgVO.

6  Notice that participle agreement in Standard Italian also follows an Active/​Inactive agreement pattern, since there is no agreement with the S unergative verbs. 7  Woolford’s (1997) rebuttal, endorsed by Legate (2012a), is based on the observation that datives too may exhibit a transitivity restriction, but her argument gratuitously presupposes that the dative is necessarily an inherent case rather than a dependent case requiring a case-​competitor; cf. Baker (2015).

 

The relation between case-alignment and constituent order    341

14.4.1 Ergative Case Assigned by v Ergative case may plausibly be taken to be assigned by (certain subspecies of) “little v” to its Specifier; cf. Woolford (1997, 2006); Anand and Nevins (2006); Legate (2012a); among others. That is, the ergative is an “inherent case” in the sense that it is assigned together with a specific Θ-​role. The main argument in favor of this view comes from the existence of languages in which ergative case is associated with every subject arguably born in Spec-​vP, but never to derived subjects (in passives and with unaccusative verbs). On the other hand, this proposal is less well suited to account for languages in which only the subjects of transitive verbs can have ergative case.8 I also note that there may be counterexamples to the generalization that derived subjects cannot have ergative case, at least if Erg/​Abs agreement requires a syntactic Erg/​Abs case-​alignment. Holmberg and Odden (2004) show that the Kurdish language Hawrami has Erg/​Abs agreement in passivized double object constructions (but not in passivized monotransitive sentences), although the derived subject does not preserve dative marking. Likewise, Rezac et al. (2014) argue for raising-​to-​ergative in Basque and conclude that the ergative is not inherent in Basque. Conceivably, ergative case is assigned in different ways in different languages. Here, I will simply discuss how *SErgVO might be accounted for if ergative case is assigned by v. It seems highly unlikely that the properties of structures linearizing as SVO could interfere with assignment of ergative case by v to its Specifier. However, it is possible that such structures might force previously assigned ergative case to be canceled in ways I will now sketch.

14.4.2 Mahajan (1994) The basic idea developed in Mahajan (1994) is inspired by Kayne’s (1993) account of the have/​be alternation with past participles in Romance. Kayne takes have to be the spell-​ out of BE with a preposition originating in C incorporated into it: (11)

[AuxP BE [CP P [IP DP [IP I … → [AuxP P+BE [CP P [IP DP [IP I …

Kayne’s system is set up so as to require a complementizer P only with transitive and unergative participles. Since this is reminiscent of a certain type of Erg/​Abs alignment, Mahajan suggests extending Kayne’s analysis to Erg/​Abs case-​alignment in languages like Hindi by saying that P doesn’t incorporate into the auxiliary in such languages, and that an unincorporated P assigns ergative case to the subject. Adding that P universally incorporates into a left-​adjacent auxiliary, he derives *SErgAuxVO. 8 

Mahajan’s account is tailor-​made for languages where Erg/​Abs alignment only occurs with perfective verbs, and perfective verbs combine with auxiliaries, as in Hindi and Romance.

 

342   Tarald Taraldsen Since *SErgAuxVO does not entail *SErgVO, a further step must be taken if *SErgVO is indeed the right descriptive generalization.9 For example, one might say that the head X hosting the auxiliary in the compound tenses is also present in simple tenses, and that the verb adjoins to it, when there is no auxiliary, assuming that P+X is then spelled out by the verb. To capture the dependence of the P on the participle being transitive or unergative, one may replace (11) with (12) taking the P to be selected by v: (12) [AuxP BE [vP [PP P DP] [vP v … → [AuxP P+BE [vP [PP P DP] [vP v … But this seems equivalent to viewing P as a case-​head K (= “ergative”) licensed by v. In this sense, Mahajan’s analysis provides a mechanism for depriving a subject of an ergative case-​feature previously assigned to it. A potential weakness in Mahajan’s account is its reliance on linear adjacency, which is now generally supposed not to play a role in syntax. But even if linear adjacency does play a role in syntax (see Kayne 2010), there is a problem, if the merger of a head with a phrasal complement universally results in H-​compl order (Kayne 1994, 2010). Then, an auxiliary will always be left-​adjacent to the ergative P/​K at the point where the auxiliary merges with a participial phrase. Since P-​incorporation into a left-​adjacent auxiliary must be obligatory in Mahajan’s system, there should therefore be no ergative languages, if P-​incorporation applies at that point. Rather, P-​incorporation must apply at a later stage of the derivation to give the participial phrase time to slip across the participle. But the timing is rather delicate. Even though the participle regularly precedes the auxiliary in German, the German have/​be alternation seems rather similar to the one seen in Romance, which, from Mahajan’s perspective, must mean that P-​incorporation must apply before the participial phrase is moved across the auxiliary in German. In other words, the movement of the participial complement of an auxiliary in German must be to a higher position than in ergative SOV languages like Hindi. The question is whether there is any independent evidence that this is in fact true. Another issue is why P-​incorporation into a left-​adjacent auxiliary should be universally obligatory.10 For that matter, one might also wonder whether the disappearance of ergative case can be blamed on P/​K-​incorporation into the auxiliary at all, since Basque seems to have a have/​be alternation, but the subject carries an ergative case-​affix even in the presence of have. 9  Notice that if interpreted this way, Mahajan’s account also provides a way of making an ergative surface as a nominative without appealing to syncretism. Hence, languages combining surface Nom/​Acc alignment (rather than neutral) with an Erg/​Abs agreement pattern depending on underlying Erg/​Abs/​ Acc case-​alignment will not be ruled out by the assumptions about syncretism invoked in section 14.3.2. But past participle agreement in Romance may in fact be an instance of an Erg/​Abs agreement pattern in languages with overt Nom/​Acc case-​alignment at least for pronouns. 10  Kayne (1993) suggests that P-incorporation must apply in languages which lack a spell-out for the oblique case that would otherwise be assigned by the P, i.e. the ergative on Mahajan’s interpretation. But if this were to be the only thing driving P-incorporation, no language with the morphological means to

 

The relation between case-alignment and constituent order    343

14.4.3 Taraldsen (2010) The approach taken by Taraldsen (2010) also assumes that a nominative DP may emerge in the syntax from a DP previously assigned ergative case in Spec-​vP. But instead of becoming a nominative by losing its ergative K to incorporation, it strands its ergative case-​layer by subextraction, a process dubbed “peeling” in Caha (2009): (13)

[NomP Nom … [vP [ErgP X [AccP Y [NomP Z DP ]]] [vP v … → [NomP [NomP Z DP] [NomP Nom … [vP [ErgP X [AccP Y [NomP Z DP ]]] [vP v …

This is based on an assumption already mentioned:  The different cases are decomposed into sets of privative features linearly ordered by the subset relation. In addition, Taraldsen (2010) also follows Caha by assuming that each case-​feature is a syntactic head so that sets of case-​features are represented as trees. In (13), Nom is a head on the verbal spine attracting a NomP to its Specifier from the nearest DP. It is also assumed that an object always raises to an Acc-​position between Nom and vP except in anti-​passives, where the object surfaces inside VP with the oblique case initially assigned to it; see Medová (2009): (14)

[NomP Nom … [AccP Y [NomP Z DPO]] … [vP [ErgP X [AccP Y [NomP Z DPA ]]] [vP v …

If nothing further happens, the object will therefore be the DP closest to Nom, and an Erg/​Abs alignment emerges with the object’s nominative layer in Spec-​NomP: (15)

[NomP [NomP Z DPO ] [NomP Nom … [AccP Y [NomP Z DPO]] … [vP [ErgP X [AccP Y [NomP Z DPA ]]] [vP v …

But SVO within TP is taken to arise only from moving the remnant vP in between Nom and the previously raised object for reasons discussed in Taraldsen (2010): (16)

[NomP Nom …[vP [ErgP X [AccP Y [NomP Z DPA ]]] [v’ v … [AccP Y [NomP Z DPO]] … [vP [ErgP X [AccP Y [NomP Z DPA ]]] [vP v …

In (16), the nominative layer of the external argument is assumed to be closest to Nom, and by locality Nom/​Acc alignment must emerge.11,12 Hence, no language with spell out ergative case would have P-incorporation and might therefore combine Erg/Abs alignment with SVO order. 11  It is not actually clear that NomP contained in the subject is closer to Nom than the object’s NomP, since the former is embedded under more case-heads than the latter. 12  Notice that this raises the question how Icelandic, an SVO-​language, can have quirky subjects and nominative objects.

 

344   Tarald Taraldsen TP-​internal SVO order can have Erg/​Abs case-​alignment. That is, *SVO+Erg follows even if v universally assigns ergative case to its Specifier.13 The most obvious problem for this account is perhaps that it predicts that the object should be higher than the external argument in all ergative languages, as in Bittner and Hale’s (1996a) account of Dyirbal and Inuit, but this does not seem correct. If the classical diagnostics for syntactic ergativity respond to structural properties of the sort assumed for ergative sentences by Taraldsen, not all ergative sentences can have these structural properties, since not all ergative languages display the hallmarks of syntactic ergativity.14 Notice also that while Taraldsen’s account makes the right prediction for case-​ alignment in those ergative languages in which only the subjects of transitive verbs have ergative case, it also incorrectly predicts that there should be no language in which both subjects of transitive verbs and subjects of unergative intransitive verbs have ergative case, unless unergative verbs come with a hidden object as assumed by Bittner and Hale (1996a) for Georgian. In this respect, it is similar to the case-​competition accounts to which we now turn.15

14.4.4 Case-​Competition The case-​competition approach assumes that certain cases can only be assigned to a case-​less DPi in a case-​domain XP if XP also contains another case-​less DPj, a “case-​ competitor.” A case of this sort is usually referred to as a “dependent case.” If DPi asymmetrically c-​commands DPj, some languages will assign the dependent case to DPi, while others assign it to DPj.16 To illustrate, let us assume that vP is a case-​domain. If a vP contains only a single case-​ less DP, as in (17), dependent case is not assigned and the DP must be case-​licensed in one of the alternative ways sketched in section 14.5: (17) [vP DP [vP v [VP V ]]] But if vP contains a case-​less object DP in addition to the external argument, as in (18), dependent case will be assigned either to the external argument or to the object depending on which way the language sets the relevant parameter: 13 

Notice that on this account, case-​sensitive agreement will be able to produce Erg/​Abs agreement in a Nom/​Acc language if the probe is sufficiently low, as it might be in Romance past participle agreement. 14  Aikhenvald’s (2009) observation that the “pivot” for coordination is always S or O, never A, in Paumarí regardless of whether the case-​alignment is Erg/​Abs or Nom/​Acc, suggests that case-​alignment is in fact quite independent of the properties underlying syntactic ergativity. 15  A related problem is that Taraldsen’s analysis cannot easily accommodate the tripartite languages. 16  Notice that this seems to go against the spirit of Kayne’s (1994) antisymmetry proposal, since saying that the dependent case go to either A or B when A asymmetrically c-​commands B, subject to a language-​specific choice, is similar to saying that A asymmetrically c-​commands B can map onto A precedes B or A follows B, as dictated by a language-​specific parameter setting.

 

The relation between case-alignment and constituent order    345 (18)

[vP DPi [vP v [VP DPj V]]]

Assume that the DP that remains case-​less at the end of the vP-​cycle, will eventually be assigned the same case as the case-​less external argument in (17). Then, if dependent case is assigned to the external argument in (18), the language will have Erg/​Abs case-​alignment. If dependent case is assigned to the object DP, we will have Nom/​Acc alignment.17 If the dependent case can be assigned both to the higher and the lower DP at the same time, as suggested by Baker (2015), a tripartite language emerges. This analysis comes with the transitivity condition built in. It predicts not only that an object will not have accusative case in the absence of an external argument, but also that an external argument has ergative case only in the presence of a case-​less object.18 Therefore, this line of analysis has no problem with ergative languages where only the subject of a transitive verb can be ergative, but must posit covert objects for languages in which all external arguments bear ergative case. On the conceptual side, one may wonder where exactly the procedure for assignment of dependent case and the parameter associated with it resides in the grammar. The proposal in Taraldsen (2010) can be viewed as an attempt to answer these questions, i.e. as a specific implementation of the case-​competition view. A different answer is provided by Bittner and Hale (1996a) who assume that ergative case is assigned by I while accusative is assigned by V. Given that a case-​competitor must be case-​less, it follows from this that ergative case will only be assigned in sentences where the verb fails to assign accusative case to the object. Hence, whether Erg/​Abs or Nom/​Acc alignment emerges, ultimately depends on a property of verbs.19 If the domain for assignment of dependent case is vP, the dependent case approach offers limited promise of understanding a *SVO+Erg generalization. Whether the dependent case is assigned to the higher or to lower one of two case-​less DPs inside vP, should be totally independent of how the ingredients of the vP end up being linearized with respect to each other. Only analyses in which ergative case previously assigned can be lost, as illustrated, might do the trick. If the relevant case-​domain is I/​TP, however, additional options become available. For example, Bittner and Hale (1996a) assume that ergative case is assigned by I, and that a case-​less object becomes a case-​competitor enabling I to assign ergative case to the subject when either the object raises to Spec-​IP or the verb raises to I. In the first case, the subject cannot also raise to Spec-​IP so that the direct outcome within IP is an OSV

17 

In Taraldsen’s (2010) analysis, the alignment type depends on which of the two DPs succeeds in reaching the Nom position. The effect of requiring a case-​competitor comes from the fact that when there is a single DP, that DP will always reach the Nom position. 18  The promoted first object in a passivized double object construction is also expected to have ergative case, as in Hawrami, if both objects are case-​less at the stage of the derivation where dependent case is assigned. 19  On this approach, too, tripartite languages call for additional assumptions, as in section 11 of Bittner and Hale (1996a).

 

346   Tarald Taraldsen order. In the second case, the subject doesn’t raise to Spec-​IP, because it is case-​licensed in situ and movement to Spec-​IP applies only when needed for case-​licensing. In this case, the outcome is a VSO order. Either way, underlying SVO order within VP will not be preserved at the IP-​level. In a language with Nom/​Acc alignment, on the other hand, the subject is case-​less and may move to Spec-​IP to be governed by C (depending on whether V has raised to I or not). Therefore, a sentence with Nom/​Acc alignment may surface with SVO order, assuming the order VO within VP. Notice, however, that in order to accommodate ergative languages with rigid SOV, Bittner and Hale’s account must acknowledge the possibility that a subject may raise to a position above Spec-​IP for reasons not related to case-​licensing. This movement will derive SOV orders from OSV (within IP), but must still be unable to derive SVO orders from VSO (within IP). Therefore, even Bittner and Hale’s version of the case-​ competition approach needs extra assumptions to capture the *SErgVOAbs generalization, unless, of course, it turns out that SVO sentences in fact allow Erg/​Abs alignment precisely when it can be shown that S is in the same position above Spec-​IP where it must be placed in ergative SOV languages.

14.4.5 Kayne (2000b) Kayne (2000b) has a different way of connecting V-​movement to ergative case-​assignment as part of an attempt to understand *SErgVO in terms of general properties of adpositions. The basic axiom is that adpositions are functional heads on the verbal spine and therefore combine with what we take to be their complements via movement. The first step in a derivation putting together the sequence to John, for instance, would be as in (19): (19) [PP to [vP I [VP spoke John ]]] → [PP John [PP to [vP I [VP spoke John ]]]] Then, a head W is merged, and the P incorporates into W yielding: (20) [WP to+W [PP John [PP to [vP I [VP spoke John ]]]]] Finally, P+W attracts vP: (21) [WP [vP I [VP spoke John ]] [WP to+W [PP John [PP to [vP I [VP spoke John ]]]]] If W is not merged, the adposition surfaces as a postposition, and to capture the preference for postpositions over prepositions in verb-​final languages, Kayne proposes that P+W universally attracts vP.20 20 

An SXV language can then have prepositions just in case the V is extracted from vP, before vP is raised to Spec-​WP.

 

The relation between case-alignment and constituent order    347 Suppose now that ergative case is assigned by a special P to a DP in its Spec. The derivation of a sentence with an ergative subject must then include a step similar to (19): (22) [PP P … [vP DPA [VP V DPO ]]] → [PP DPA [PP P … [vP DPA [VP V DPO ]]]] In a language where all Ps are prepositions, the derivation continues as in (20)–​(21): (23)

→ [WP P+W [PP DPA [PP P … [vP DPA [VP V DPO ]]]] → [WP[vP DPA [VP V DPO ]]    [WP P+W [PP DPA [PP P … [vP DPa [VP V DPO ]]

Since the last step produces a V-​initial order (provided, of course, the subject does not subsequently move across the raised vP), *SVO+Erg follows for any language with prepositions rather than postpositions. In particular, it follows for all SVO languages, if adpositions are always prepositions in SVO languages.21 Notice, however, that nothing excludes the existence of V-​initial sentences with ergative subjects in languages that otherwise have SVX order. If such mixed languages don’t exist, additional assumptions are called for. Notice also that there are a few SVO languages with postpositions. Hence, there might also be a few ergative SVO languages, i.e. *SErgVO should find exceptions among those SVO languages that have postpositions. In the exceptional languages, Kayne’s ergative P should be a postposition. This expectation is in fact met in the two exceptional languages I know of, viz. Paumarí and Zoque. Both languages have SVO-​sentences with Erg/​Abs alignment, but the ergative case-​marker is enclitic to the noun. Judging from Faarlund (2012), all case-​markers are in fact postpositional in Zoque. The following example from Paumarí comes from Aikhenvald (2009:113) (Tr = transitivity marker, Th = object agreement): (24) Kodi-​jomahi-​a bi-​a-​vi-​kha-​‘a-​ha ada jao’oro 1sg.-​dog(m)-​Erg 3sgTr-​get-​Tr-​get-​Asp-​Th.m Dem.m-​agouti (m) ‘Our dog caught an agouti.’ Faarlund (2012:120) provides the following Zoque example: (25)

21 

Te’ n-​galyo= ‘is y-​ni-​tüp-​‘üy-​u       te’ wedu DET M-​cock= Erg 3A-​front-​jump-​V-​CP  DET fox ‘The cock attacked the fox.’ (3A = 3rd person series A prefix, V = verbalizer, CP = completive aspect)

The proposal in Taraldsen (2010) might be recast in similar terms by saying that accusative case is associated with a preposition above vP in SVO-​languages.

 

348   Tarald Taraldsen

14.4.6 Summary We have examined different proposals about the licensing of ergative case in order to determine whether they provide any hope of seeing the *SErgVOAbs generalization as reflecting the impossibility of assigning ergative case in structures that give rise to SVO order. We have seen that if ergative case is assigned vP-​internally, either as a “semantic case” or as a dependent case, only analyses where a subject is deprived of previously assigned ergative case in the course of derivations leading to SVO order, have any reasonable chance of explaining the generalization as the effect of *SErgVO. But the two existing accounts that pursue this line of analysis have both been shown to be defective or incomplete. If ergative case is assigned by I, as in Bittner and Hale’s account, there is a better chance of reducing *SVO+Erg/​Abs to *SVO+Erg, and in fact Bittner and Hale claim to have achieved exactly that, but we have also seen that their account too is incomplete in ways that threaten to undercut their conclusion. Finally, we have looked at Kayne’s (2000b) proposal, which makes a very specific prediction about the scope of *SErgVO—​a prediction which in fact is consistent with the two known exceptions to the generalization.

14.5 *SErgVOAbs = *SVOAbs? We will now look at the major proposals for absolutive case with a view to determining whether they supply a plausible starting point for attributing *SErgVOAbs to failure of absolutive case-​licensing for the object in structures that give rise to SVO order, i.e. for seeing *SErgVOAbs as *SVOAbs.

14.5.1 Abs Assigned by T In the analysis proposed by Marantz (1991), the absolutive is automatically available as a default for any DP to which inherent or dependent case has not been assigned. In this scenario, it is hard to see how the positioning of the verb might interfere with the licensing of absolutive case. If, on the other hand, absolutive case is assigned by T (like nominative case on standard accounts), locality may come into play. As mentioned in section 14.2.2, moving the object across the subject, might be a precondition for the object to be probed by T, and if the verb cannot be moved to a position between T and the raised object and the subject winds up in Spec-​TP, *SVOAbs would follow. Even if the verb can move across a raised object, but only by remnant movement of vP carrying along the subject, as in Taraldsen (2010), *SVOAbs would follow provided this turns the subject into an intervener between T and the object.

 

The relation between case-alignment and constituent order    349 In addition to Relativized Minimality, phasehood might be relevant. This possibility is exploited by Lahne (2008). Taking the standard view that vP is a phase, she points out that if the PIC takes effect as soon as the vP-​phase has been completed, T will not be able to value the object’s case-​feature unless the object has moved to the vP-​edge. If so, the object must always become an “inner Specifier” of v in ergative languages. This in turn means that there is no head position between the subject and the object which the verb could move to, provided the subject doesn’t move out of vP. Hence, *SVOAbs. But there is a problem with this. Icelandic is SVO, and therefore the S and the O cannot both be specifiers of vP at the end of the day. But in sentences with quirky subjects, an object has nominative case. If the nominative is valued by T, this should be impossible on Lahne’s assumptions. More generally, if the absolutive case and the nominative case are in fact the same case and this case is always licensed in the same fashion, Icelandic sentences with SVO order and nominative case on the O show that there is no hope of deriving *SErgVOAbs from *SVOAbs.

14.5.2 Nominative Is Independent of Agreement in Icelandic Actually, agreement facts suggest that the Icelandic nominative is best analyzed as a default case. In Icelandic sentences with quirky subjects, the verb may show overt number agreement with the nominative object, as in (26): (26) Mér líku∂-​u me.Dat liked-​3pl ‘I liked the horses.’

hestarnir horses-​Nom-​the-​Nom.m.pl

But in some cases there is no agreement even though the object has nominative case.22 (27)

a Honum líka∂-​i him.Dat liked-​3sg ‘He liked us.’

vi∂ we.Nom

b Mér    fannst   hestarnir       vera venir me.Dat seemed.3sg horses-​ the.Nom be     beautiful-​ Nom.pl ‘I found the horses beautiful.’ And as shown in Holmberg and Hróarsdóttir (2000), a postverbal quirky subject blocks agreement with the nominative object: (28) †a∂ fannst/​ *funnust einhverjum  student tölvurnar ljótar It    seemed.3sg/​*3pl  some-​Dat  student computers-​the.Nom ugly-​ Nom.f.pl ‘Some student found the computers ugly.’ 22 

Sentences like (27)a where the nominative object is not 3rd person are unacceptable to some speakers. But they become unacceptable to all speakers with plural agreement on the verb.

 

350   Tarald Taraldsen This strongly suggests that nominative case is not licensed under Agree with T in Icelandic, while it might perfectly well be analyzed as the default case. But if nominative case and absolutive case are the same thing, as is often assumed, then one would also expect that the absolutive is a default case in some language with Erg/​Abs case-​ alignment. If we can identify such languages and find that the *SErgVOAbs generalization doesn’t hold for them, while it does hold for languages in which there is independent evidence that absolutive case is licensed by T, we should conclude that the generalization really is *SVOAbs after all. Conversely, if the *SErgVOAbs generalization is shown to hold also for languages with the absolutive as a default case, the conclusion must be that the correct generalization is *SErgVO.

14.5.3 Defective Intervention In (28), the dative DP acts as a “defective intervener” with respect to agreement with the nominative object. The probe is case-​sensitive and cannot agree with datives, but nevertheless the intervening dative prevents the probe from reaching the nominative. The phenomenon of defective intervention is also relevant to the analysis of sentences with Erg/​Abs alignment. To account for an Erg/​Abs agreement pattern with agreement only with absolutive DPs, it is not sufficient to say that the probe is case-​sensitive, once defective intervention is taken into account. Unless an absolutive object is raised to a position P above the subject before the probe X is merged, the subject will be a defective intervener, as in (28). The object may then also be assigned absolutive case by X or a head higher than X without the subject acting as an intervener. (If ergative case is assigned by a probe, this probe must be v or a head located between v and the raised object.) A Nom/​Acc agreement pattern, however, will only arise if the object does not raise to P. But then absolutive case cannot be assigned to it by any head outside vP, since the ergative subject would act as an intervener. This suggests that absolutive case may be licensed by a vP-​external head in ergative languages with agreement only with O and absolutive S, but is a default case in ergative languages with agreement only with A and S, just like the Icelandic nominative. Consequently, if *SErgVOAbs is to be understood as *SVOAbs, we may expect that the generalization doesn’t hold for ergative languages with Nom/​Acc agreement patterns, since it seems impossible to make sense of *SVOAbs when the absolutive is a default case.

14.5.4 Summary The *SErgVOAbs generalization can be understood as *SVOAbs only if absolutive case must be licensed by a head outside vP, e.g. T. But even so, *SVOAbs seems implausible, if nominative case is also always licensed by T, since Icelandic has SVO sentences with nominative subjects. However, intervention facts suggest that the Icelandic nominative

 

The relation between case-alignment and constituent order    351 is really a default case, and therefore the possibility remains that *SVOAbs holds to the extent that the absolutive is licensed by T. But it seems likely that there are ergative languages in which the absolutive is a default case, just like the Icelandic nominative, and *SVOAbs is not expected to hold for such languages. If *SErgVOAbs nevertheless turns out to hold for all ergative languages, we conclude that *SErgVOAbs is best understood as *SErgVO, but if it holds only for those ergative languages where the absolutive is arguably not a default, *SErgVOAbs is really *SVOAbs. We have also suggested that the absolutive may be licensed by T in ergative languages with Erg/​Abs agreement, but must be a default in ergative languages with Nom/​Acc agreement.

14.6 Conclusion One conclusion is that if there are different ways for Erg/​Abs alignment to arise as well as different ways for a sentence to come out with SVO order, one should expect the *SErgVOAbs generalization to have exceptions, hopefully of a well-​defined kind. But if the generalization is valid for all sentences with Erg/​Abs alignment, we are led to conclude that Erg/​Abs alignment must arise in essentially the same way in all languages. We have also suggested that tripartite languages and neutral languages with Erg/​Abs agreement patterns should be investigated with a view to determining whether such languages should fall under the scope of the *SErgVOAbs generalization. In particular, determining whether tripartite languages fall under the generalization, is crucial to deciding whether the generalization should be seen as *SErgVO or *SVOAbs. In the final sections of the chapter, I have tried to evaluate the major theoretical accounts of ergative and absolutive case with respect to how well they allow us to ground a suitably sharpened version of the *SErgVOAbs generalization in general principles of syntax.

Abbreviations 3A, third person series A prefix; Abs, absolutive; asp, aspect; CP, completive aspect; Dat, dative; Dem, demonstrative; DET, determiner; Erg, ergative; f, feminine; m, masculine; Nom, nominative; pl, plural; sg, singular; Th, object agreement; Tr, transitivity marker; V, Verbalizer.

 

 

Extensions

 

 

Chapter 15

Ergativi t y i n nom inaliz at i on Artemis Alexiadou

15.1 Introduction The correlation between ergativity and deverbal nominalization crosslinguistically has been discussed extensively in e.g. Alexiadou (2001), Salanova (2007), Imanishi (2014), among others; and see Williams (1987), Bok-​Bennema (1991), and Johns (1992) for earlier similar ideas. According to these authors, nominalizations across languages show an ergative case-​marking pattern, which is attributed to the presence of a defective v (or Voice) in their syntactic representation. In this chapter, after briefly summarizing the main arguments in favor of this view, I will address the following question: what is it about noun-​ness that obligatorily triggers the presence of an ergative case pattern?

15.2  Deverbal Nominalizations Have Ergative Case Patterns Alexiadou (2001) explores the hypothesis that certain types of nominalizations exhibit ergative case patterns. This claim has been made on the basis of data from a variety of languages, which realize the theme argument of the nominalization in the genitive. Consider the Greek examples in (1). In (1a), the Agent is introduced by a prepositional phrase. This PP has the same form as the one we find with verbal passives (1b). (1) a. i katastrofi tis polis apo tus varvarus mesa se tris meres the destruction the city.GEN by the barbarians within three days ‘The destruction of the city by the barbarians within three days’

 

356   Artemis Alexiadou b. i poli       katastrafike    apo tus  varvarus. the city.NOM destroyed.PASS by  the barbarians ‘The city was destroyed by the barbarians.’ Irrespectively of the presence of the PP, the genitive argument bears the theme/​patient theta-​role. What is excluded is an agent bearing genitive case and co-​occurring with a genitive theme, no matter in which order the two genitives appear (see Horrocks & Stavrou 1987): (2) a. *i silipsi the capture

tu Jani the John.GEN

tis astinomias the police.GEN

b. *tis astinomias i silipsi tu Jani In case of nominalizations derived from intransitive verbs, both unaccusative and unergative ones, their single argument surfaces with genitive:1 (3) a. i afiksi   ton pedion the arrive the children.GEN b. to the

treksimo running

tu the

athliti athlete.GEN

It appears then that the patterns in (1) and (3) are reminiscent of an ergative case pattern, in which the single argument of an intransitive predicate surfaces with the same case as the internal argument of a transitive predicate, and the agent argument bears a different marking. Similar patterns are found in other languages as well, as can be seen by the following set of examples. For instance, in English internal arguments of nominalizations derived from transitive predicates as well as the single argument of intransitive nominalizations 1  Note that Grimshaw (1990) argued that nominalizations of type (3b) should not be considered as complex event nominals in her terms or argument supporting nominals in e.g. Alexiadou, Iordachioaia, & Soare’s (2010) or Borer’s (2013) terminology. In fact, in Alexiadou (2001), I provided ample evidence that unergative nominalizations should not be treated as argument supporting nominals, suggesting that the genitive case we see in these nominals is similar to that of the possessor in non-​derived nominals (e.g. John’s book). If this is correct, nothing special needs to be said with respect to (3b) in the sense that the genitive is introduced in the nominal part of the structure, i.e. presumably by the n head that embeds a verbal substructure, and receives case in D, see also Imanishi (2014). The alternative would be to assume that data such as the ones in (3) suggest that nominalizations have case patterns similar to those ergative languages, such as Niuean, in which subjects of both unergative and unaccusative predicates are marked with absolutive. Later on in the chapter, I will discuss the requirement that nominalizations lack an external argument, see (39). In order for (3b) to conform with this requirement, one would need to assume that, in agreement with Massam (2009), subjects of unergatives are introduced within vP, at least in nominalizations. I will come back to this point in the last section of the chapter.

 

Ergativity in nominalization    357 are introduced by of and the agent in the case of the deverbal nominalization of a transitive predicate is introduced via a by-​phrase: (4)

a. the destruction of the city by the barbarians b. the arrival of the policemen c. the jumping of the cow

What is not allowed is that both arguments are introduced via of:2 (5)

*the destruction of the city of the barbarians

Picallo (1991) reports that a similar pattern is found in nominalizations in Catalan, where it is not possible to realize both arguments of the nominalization in the genitive, as is illustrated by the unacceptability of (6). (6) *l’afusellament de l’escamot d’en Ferrer Guardia the execution of the squad of Ferrer Guardia ‘the squad’s execution of Ferrer Guardia’ Genitive arguments are always interpreted as themes in such cases. Thus, the following examples are grammatical, but only in the interpretation where the squad and the police are executed or captured respectively. The agent, when realized, must be realized as a PP (see 7c): (7)

a. l’afusellament de l’escamot the execution of the squad b. la captura de la policia the capture of the police c. l’afusellament d’en Ferrer Guardia per part de l’escamot the execution of Ferrer Guardia on part of the squad

Theme/​*Agent Theme/​*Agent

2  There is, however, an important difference between Greek and English nominalizations: the agent argument in English nominalizations derived from transitive verbs can also appear in the Saxon genitive, a pattern that is often referred to as the transitive variant of the nominalization (see Chomsky 1970, Kayne 1984, among others; cf. Koptjevskaja-​Tamm (1993) who labels this pattern double possessive):

(i)  a.  John’s destruction of the city b. the judge’s reversal of the decision In this, Modern Greek differs from Classical Greek, which allowed double possessive nominalizations, see (ii), from Manolessou (2000):   (ii)  hê Phaiakôn proenoikêsis       the.NOM Phaecians.GEN occupation.NOM  the Phaeacians’ occupation of Corcyra T. 1.25.

tês Kerkuras the.GEN Corcyra.GEN

I will briefly discuss this pattern in section 15.5. Note that even in English there are certain restrictions on the formation of transitive nominalizations related to the type of external arguments these permit, see Alexiadou et al. (2013) for discussion and references.

 

358   Artemis Alexiadou In French, Spanish, and Italian (see Cinque 1980; Milner 1982; Zubizarreta 1987; Bottari 1992; among others) two ‘de-​phrases’, cannot appear within nominalizations, and the agent argument is introduced by par, por and de parte de (8). (8) a. *la destruction de la ville des soldats the destruction of the city of the soldiers

French

b. la destruction de la ville par les soldats the destruction of the city by the solders c. *la captura de los fugitivos de los soldados the capture of the fugitives of the soldiers

Spanish

d. La captura de los fugitivos por los soldados the capture of the fugitives by the soldiers e. *La cattura del soldato del enemico the capture of the soldier of the enemy

Italian

f. La cattura del soldato da parte del enemico the capture of the soldier by the enemy As was the case in Greek and English, the agent is introduced by a PP which has a similar form to the PP found in passive sentences introducing the agent. In fact, the ergative pattern in nominalization is really pervasive. Salanova (2007) shows that even if a language permits ergative splits in the verbal domain, the nominal domain is obligatorily ergative. As shown below, Mẽbengokre event nominalizations, which are employed in a variety of constructions in the language, display an ergative–​ absolutive pattern, examples from Salanova (2007): (9) a. b^ kam i-​mõr kuni forest in 1-​go.PL.N all “all my goings into the woods” b. ijɛ  ^ktirɛ    krõr        jã 1ERG  hawk.people make.peace.N this “this (occasion in which) I was making peace with the Àktire” In the case of Mẽbengokre, contrary to what happens in English and in the other languages discussed here, there is morphological identity between the cases employed in nominalizations and those employed in ergative main clauses. Salanova stresses that ‘an important fact to note is that, contrary to other ergativity splits, which languages may or may not have, action nominalizations are normally ergative.’ The cross-​linguistic observations with respect to ergativity in nominalization are summarized in Table 15.1, from Alexiadou (2001: 166).

 

Ergativity in nominalization    359 Table 15.1 Cross-​linguistic observations and ergativity in nominalization N/​A system

E/​A system

Nominalization

A-​argument

NOM

ERG/PP

S-​argument

NOM ABS

GEN

P-​argument

ACC ABS

GEN

This ties in nicely with observations made in other literature, where it is suggested that the ergative syntax observed in most e.g. Philippine and Formosan languages is the result of diachronic reanalysis of clausal nominalizations as finite verbal clauses, see e.g. Aldridge (2013a, Chapter 21, this volume), Kaufmann (Chapter 24, this volume) for recent discussion and references. Hence there must be something deeper explaining the connection between nominalization and ergativity. Before I  raise the questions to be addressed in this chapter, three remarks are in order. First, not all types of nominalizations exhibit ergative case patterns, as we will see, and see Koptjevskaja-​Tamm (1993) for a typological survey. Thus, ergativity characterizes only a subset of nominalizations, namely those that show a pattern of the type in Table 15.1. Second, it should be pointed out that in e.g. Mayan languages ergative is homophonous with genitive. In a subgroup of Mayan languages all subjects align with ergative/​genitive. This is the case for example in Chol/​Q’anjob’al, as discussed in Coon (2010b), but not in Kaqchikel, as discussed in Imanishi (2014), which has an alignment more similar to the one we find in Greek. The Chol pattern, by contrast, is closer to nominalized forms of the type encountered in English verbal gerunds, see section 15.4.3. for some discussion. Third, as we see in Table 15.1 the similarity between nominalization and ergativity emerges for those ergative languages that treat single arguments of all types of intransitive predicates alike. In other words, what is comparable are case patterns in Greek type nominalizations with ergative case patterns in languages such as Niuean, in which single arguments of both unergative and unaccusative predicates bear absolutive, and it is only the agent of a transitive predicate that surfaces with a distinct case marking, see n. 1 and the conclusions section. The question that arises then is: what is special about nominal forms that obligatorily gives rise to an ergative case pattern? In the next section, we will see that ergative patterns arise only in a subset of nominalizations in those nominative–​accusative languages, which have more than one nominalization pattern. Crucially, such patterns obligatorily contain an n head. We will then turn to the question what is special about the morphosyntax of these nominalizations. The answer will be, as anticipated in Alexiadou (2001) and Salanova (2007), and see also Imanishi (2014), that it relates to the case assigning heads in nominal structure. Specifically, n-​based nominalizations make room for one structural Case within the nP phase, and require a deficient VoiceP/​vP complement, i.e. a Voice/​vP projection that does not project an external/​agent argument.

 

360   Artemis Alexiadou

15.3  Two Types of Deverbal Nominalizations 15.3.1 Mixed Extended Projections In the recent literature, several authors have discussed the criteria that can be used to determine that nominalizations come either with a verbal internal structure of the type in (12a) or with a mixed internal structure of the type in (12b); for instance, Alexiadou et  al. (2011) have argued that given their eventive nature, all deverbal Argument Structure nominalizations (ASNs) involve vP, but differ with respect to the type of functional structure above the vP layer. Certain types of nominalizations contain only verbal functional projections and no nominal layers, while others involve nominal layers on top of optional extended verbal projections (see Borsley and Kornfilt 2000; Alexiadou 2001; Ehrich 2002): (10) a. [ DP [ extended Verbal FP [vP ... ]]] b. [ DP [ Nominal FP [ (extended Verbal FP) [vP ... ]]]]

verbal internal structure mixed internal structure

According to (10a–​b), all ASNs have the syntactic distribution of a noun (i.e. which is provided by the DP layer), but they retain the syntactic and semantic characteristics of the input verb to a varying degree (vP and extended Verbal FPs), to which they may or may not add further Nominal FPs. The type of evidence that was used to diagnose the presence of these layers relates to the verbal and nominal properties of the different types of nominalizations. Under nominal properties, I understand the availability of adjectival modification, gender and plural marking, the presence of several types of determiners as well as the presence of genitive subjects. Under verbal properties, we can include the presence of arguments, the presence of auxiliaries, the licensing of adverbial modification, the availability of accusative Case on the internal argument, and nominative Case on the external argument. On this view, the distinction between Vs and Ns is not absolute, but gradual in nature: the V/​N cut-​off point of a nominalization can be located at any point in these scales. In the next section, I will apply these criteria to English gerunds.

15.3.2 English Gerunds As Greek only has one type of ASN, namely the type (10b), leaving nominalized clauses aside, and thus it is not so illuminating as to the division of labor between nominal and verbal properties within nominalizations, let us look at the English patterns in some detail. As is well known, English distinguishes between verbal and nominal gerunds. Nominal gerunds but not verbal gerunds accept all kinds of determiners.

 

Ergativity in nominalization    361 (11)

a. *That/​*the/​*a criticizing the book annoyed us. b. The/​that/​?a reading of the manuscript pleased us.

Moreover, the verbal gerund assigns accusative case (11a), while the nominal gerund, similar to the -​ation nominal examples discussed in the previous section, realizes its theme as an of-​PP (see Lees 1960; Chomsky 1970). As has been discussed by many authors, and is recently summarized in Alexiadou et al. (2011), further properties correlate with this distinction: the subject of the verbal gerund cannot be replaced by a determiner, but this is possible with the nominal gerund (12); the verbal gerund only allows adverbial modifiers, the nominal gerund only allows adjectives (13): (12)

a. *That/​*The criticizing the book annoyed us. b. The jumping of the cows annoyed us.

(13)

a. Pat disapproved of my quietly/​*quiet leaving the room. b. The careful/​*carefully restoring of the painting took six months.

Under the standard assumption that adverbs modify verbal structures and adjectives nominal structures, this contrast suggests that the verbal gerund contains a verbal internal structure (Abney 1987; Borer 1993; Kratzer 1994; and others), while the nominal gerund has a nominal internal structure. This difference is maintained with respect to aspectual modifiers, as illustrated in (14). While the verbal gerund can license aspectual adverbs this is not the case for the nominal gerund: (14)

a. I am sick of her constantly/​*constant criticizing me. b. He could not tolerate her constant/​*constantly criticizing of him.

Following Alexiadou (1997) and Cinque (1999), this suggests that the verbal gerund hosts an AspP, which is absent in the nominal gerund. Thus, nominal gerunds cannot license this type of adverbials. For the purposes of this chapter, I  assume, following Pustejovsky (1995), Alexiadou (2001) and Borer (2005a) that the verbal gerund -​ing suffix carries imperfective aspect. To conclude, verbal gerunds lack nominal internal properties. As a consequence, they also lack plural morphology:3 (15)

*He could not stand her criticizings me.

3  For now, I will consider derived nominals to be identical to nominal gerunds, although I am aware of the fact that important differences exist between them. Since, however, they behave alike as far as the case patterns are concerned, I will assume that they are structurally rather similar. In section 15.4, however, I will offer some more details as to their internal structure.

 

362   Artemis Alexiadou In view of what was mentioned in section 15.3.1, we would expect the nominal gerund to allow plural morphology. However, as has been discussed in detail in Alexiadou et al. (2010), though plural morphology is indeed possible with some nominal gerunds, it is impossible with others. As shown in (16), some nominal gerunds allow plural, others disallow it (16b), and in some other cases, the nominal gerund is excluded in favor of some other nominalizer: (16) a. I heard of repeated killings of unarmed civilians. b. *The repeated fallings of the stock prices induced their further decline. c. The frequent late arrivals/​*arrivings of the train made me take the bus. Alexiadou et al. argued that this is because the nominalizer -​ing (different from the verbal gerund suffix -​ing) is sensitive to the inner Aspect of the original verb and its competition with other deverbal nominalizers. This was noted in Borer (2005a, vol. 2: 239–​245), who claimed that the nominalizer -​ing is ungrammatical with telic events like the achievements in (17), but it is fine with atelic events (activities and semelfactives), as in (18): (17) a. *the arriving of the train b. *the erupting of Vesuvius c. *the exploding of the balloon (18) a. the sinking of the ships b. the falling of the stock prices c. the jumping of the cows Other affixes (19a), the verbal gerund (19b) and the progressive (19c) accept telic verbs: (19) a. the arrival of the train b. The train arriving at 5 pm is unlikely. c. The train is arriving. Alexiadou et al. (2010), following Jackendoff (1991), assume that only [+bounded] entities may pluralize. If the nominalizer -​ing tends to select atelic events, which are [-​bounded], we expect it to reject plural marking. This explains the contrast above. The ungrammaticality of (16c) is a consequence of the incompatibility of the nominal gerund with telic verbs, so the nominalizer -​al is preferred (see also (19a)). Thus, crucially the inner aspect sensitivity of the nominalizer -​ing also has to do with the availability of other nominal suffixes for the same structure.

15.3.3 Greek Nominalizations If we now compare the Greek nominalization data to the two types of English nominalizations, the following picture emerges:  in Greek, all nominalizations show

 

Ergativity in nominalization    363 gender marking, and this seems to be sensitive to the type of nominalizer involved, -​ m-​nouns are neuter, while -​s-​nouns are feminine. In agreement with e.g. Alexiadou (2004), Kramer (2014b), Alexiadou et al. (2015) and references therein, gender features are on n: (20) a. kathariz-​o   kathariz-​m-​a clean-​verb  cleaning b. anagnorizo recognize

anagnori-s-i recognition

The behavior of Greek -​m-​nouns is similar to English nominal gerunds, see Alexiadou (2011) for extensive discussion. As Kolliakou (2003) observed, prototypical state and accomplishment predicates do not give grammatical nominalizations when they combine with the affix -​m-​. (21)

a. * agapima (loving) *skepsimo (thinking) b. *dolofonima (assassinating) *katastrema (destroying)

Kolliakou (2003: 179)

She notes that a subset of -​m-​nouns denotes activities, as in (22a), while others denote concrete nouns, as in (22b): (22) a. Activities perpatao walk sprohno ena karotsi push a cart b. Concrete nouns paraskevazo produce

to perpati-​m-​a the walk to sprok-​sim-​o tu karotsiu the pushing of the cart paraskevas-m-a product/​concoction

However, a closer look at the possible and impossible formations reveals finer details. Certain accomplishment predicates can build -​m-​nouns, as illustrated below. On the other hand, achievement predicates cannot build -​m-​nouns at all: (23)

a. Accomplishments htizo ena spiti build a house zografizo ena kiklo draw a circle

to htisimo enos spitiu the building of a house to zografisma enos kiklu the drawing of a circle

 

364   Artemis Alexiadou b. Achievements anagnorizo recognize ftano arrive ekrignio explode

i anagnorisi/​*anagnorisma tu klefti the recognition of the thief i afiksi/​*to afigma the arrival/​the arriving i ekriksi/​*to ekrigma the explosion

Achievement nominalizations with -​m-​do not receive a different interpretation. They are simply unacceptable. This suggests that -​m-​affixation does not introduce aspectual shift, it rather introduces non-​culmination, extension of activity. Hence it gives grammatical results only with those accomplishments which can receive an extended interpretation. As Kolliakou notes, even if the source predicate denotes a bounded event, the -​m-​ nominalization is interpreted as a process, i.e. as a non-​bounded event. Note that the -​ s-​affix seems to happily co-​occur with achievements, being what one might call a telicity marker. We can conclude then that -​m-​nouns are similar to the English nominalizer -​ing.

15.3.4 External Arguments in Nominalization Finally, let me turn to issues pertaining to the presence of external arguments in nominalizations. English nominal gerunds arguably contain Voice, which, however, does not project an overt external argument, i.e. it has passive like properties. As can be seen in (24), the ing-​of gerund (24b) patterns with the verbal passive in (24a) in excluding a self-​action interpretation, the standard diagnostic for verbal passives in Kratzer (1996, 2003). By contrast, derived nominals in (24c) allow a self-​action interpretation indicating the lack of VoiceP, see Alexiadou et al. (2013) for discussion. (24) a. The children were being registered.   i. *Th = Ag: The children registered themselves ii.  Th ≠ Ag: The children were registered by someone b. The report mentioned the painfully slow registering of the children. Th ≠ Ag /​*Th = Ag c. The report mentioned the painfully slow registration of the children. Th ≠ Ag /​Th = Ag

Greek nominalizations behave somewhat different. In some cases, the genitive DP is interpreted only as an internal argument (Alexiadou 2001). Thus (25) is not compatible with reflexive interpretation: (25)

i the

anagelia announcing

ton the

kalesmenon guests.GEN

(agent≠theme)

 

Ergativity in nominalization    365 In some other cases, Greek nominals derived from alternating verbs are structurally ambiguous between the passive and the unaccusative form, which I take to signal the absence of Voice. Evidence for this view comes from the following domains, discussed in Alexiadou, Anagnostopoulou, & Schäfer (2009). First, nominals can but do not need to contain Voice. Nominalizations of internally caused verbs never license an agent PP/​VoiceP: (26)

to sapisma the rotting

(27)

a. Ta   fila   sapisan apo  tin  igrasia    / ​ *apo ton kipuro the leaves rotted  from the humidity / ​ by  the gardener b. *I  the

ton the

filon apo tin leaves.GEN from the

igrasia /​ humidity /​

*o the

kipuros gardener

igrasia /​ *apo humidity /​ by

sapise rotted

ta the

ton kipuro the gardener

fila leaves

Second, modification by adjectives/​adverbs like spontaneous/​suddenly gives us distinct interpretations: spontaneous gives a passive interpretation, while sudden is ambiguous between a passive and an unaccusative one. In the case of internally caused roots, however, it only has the unaccusative interpretation: (28)

(29)

(30)

a. to   afthormito     anigma    tis    portas the spontaneous opening the door.GEN

→

passive interpretation

b. to   ksafniko  anigma    tis     portas the sudden  opening the door.GEN

→

ambiguous interpretation

a. to   anigma    tis     portas        afthormita → the opening the door.GEN   spontaneously  b.  to   anigma    tis     portas      ksafnika →   the opening the door.GEN suddenly

passive interpretation ambiguous interpretation

a. *to afthormito  sapisma ton filon the spontaneous  rotting     the   leaves.GEN

→

b. to  ksafniko sapisma ton filon the sudden  rotting  the  leaves.GEN

→ unaccusative interpretation

*passive interpretation

On the basis of the above diagnostic, we can conclude that while English nominal gerunds contain passive Voice, Greek nominals are ambiguous between an interpretation that contains passive Voice and one that lacks Voice. To conclude, both Greek nominalizations and English nominal gerunds (and certain English derived nominals, see section 15.4) can then be accounted by the structure in (30), and in addition (31) is available for certain Greek nominals and English derived nominals. Arguably, the structure (33) is involved in the formation of English verbal gerunds: (31)

[ DP [ (NumP) [ ClassP[±count] [nP [VoiceP [ vP …]]]]]]

 

366   Artemis Alexiadou (32)

[ DP [ (NumP) [ ClassP[±count] [nP [ vP …]]]]]

(33)

[ DP [ AspectP [ VoiceP [ vP [ Root ]]]]]

We can now establish the generalization that ergative case patterns in nominalization are related to the presence of n. As far as I can tell, there is no type of nominalization that contains a nominalizer and has a nominative–​accusative Case pattern.4 The question is why should this be the case?

15.4  Explaining the Source of Ergativity in the Nominal Domain 15.4.1 General Remarks Coon & Salanova (2009) have argued that the source of ergativity is the separation of the predicate head from T°. Specifically, they propose that because of the presence of an intervening head between the predicate and TP, case assignment takes place within the nP domain, which results in the ergative case pattern. From their perspective, when v and T are in a local relationship, v activates T for obligatory nominative case assignment. When v and T are not in a local relationship, v itself triggers obligatory absolutive case. They argue that in the case of Mebengokre the disruption of the local relationship emerges through nominalization. I think that the basic intuition of their analysis is correct, i.e. the emergence of ergative case patterns has certainly something to do with the number of case licensing/​checking heads within nP, and naturally the presence of nP itself; thus, their idea will be very helpful to understand why nominalizations which include n necessarily give rise to ergative case patterns. However, some refinements are necessary. Importantly, we cannot attribute the ergativity pattern to a disruption of a local relationship between vP and TP, as all types of nominalization discussed in section 15.3 lack T. For instance, the verbal gerund lacks T, but still it does not show an ergative case pattern, and disallows the internal argument to surface with genitive. Crucially, then, it is the presence of nP that is responsible for the ergative pattern. Before I turn to my assumptions concerning Case in general and in nominalization in particular, let me briefly remind the reader that Grimshaw (1990) offered a way to capture this. She argued that nominalization is a process akin to passivization. According to this 4 

There is one exception to that known to me, namely the Romanian supine. This nominalization behaves similar to the English verbal gerund in all properties but one, namely the one related to its Case pattern, which is similar to that English derived nominals and nominal gerunds. Alexiadou, Iordachioaia, & Soare (2010) discuss this and relate it to the properties of the Romanian enclitic article, which creates a nominal defective environment, unlike English D. Note also that this type of nominalization is based on a verbal participle.

 

Ergativity in nominalization    367 view, nominalization, like passivization, includes a process of suppressing or demoting the external argument. Support for the proposal that these arguments are suppressed comes from the observation that both the possessive and the by-​phrase are optional, see (34–​35). (34) a. The enemy’s destruction of the city was unexpected. b. The destruction of the city was unexpected. (35)

a. The city was destroyed by the enemy b. The city was destroyed

If nominalization is like passivization, then we would expect an implicit argument to be present in nominals. Roeper (1987) argues that control properties indicate the presence of an agent in the subject position of nominals, see (36): (36) a. the use of drugs to go to sleep b. the PROi use of drugs [ to PROi go to sleep ] Borer (2013: 184) presents a series of arguments that nominals should be compared with verbal passives. For instance, as in clausal passives, “a generic interpretation for the implicit subject in ASNs is available in appropriate contexts, where it is, presumably, licensed by a generic operator”; this is shown in (37–38), Borer’s (10–11): (37)

a. In the middle ages, old people were particularly appreciated [by all]. b. In some parts of the world, girls are excluded from school [by all].

(38)

a. the appreciation of old people in the middle ages [by all] b. the exclusion of girls from school in some parts of the world [by all]

As we have seen in the previous two sections, nominalizations in other languages realize the external argument as a PP, similar to the one found in verbal passives, and allow for disjoint reference interpretations, exactly like verbal passives. We can thus conclude that indeed some nominalizations are passive-​like. Others are derived from intransitive predicates and arguably lack an external argument introduced in VoiceP. From this perspective, then, nominalizations realize ergative structures basically because they are passive/​unaccusative, as argued for in Alexiadou (2001), see (39a), from Alexiadou (2001: 111). A related formulation is given in (39b), from Imanishi (2014: 123), but see the remarks in n. 1: (39)

a. Nominalizations lack external arguments. b. The Unaccusative requirement on nominalization Nominalized verbs must lack an external argument.

Recently, Bruening (2013) proposed that NOM heads (or n in our terminology) in addition to changing category have properties very similar to those of passives. Thus,

 

368   Artemis Alexiadou in the presence of an n head, its task, when embedding a transitive verb, is to turn it into a passive. In particular, according to Bruening (2013: 31), the head Nom, which embeds a VoiceP, is like his Pass head “in requiring that all arguments be saturated. If there is an open argument, Nom, unlike Pass, will project it in its own specifier.” This would account for the double-​possessive pattern of English nominalization, but not necessarily for the pattern containing a by-​phrase. “If there is no open argument, Nom, like Pass, is semantically vacuous. The way the open argument of Voice can be saturated prior to merger of Nom is by adjoining a by-​phrase”. In this chapter, I will assume a version of Bruening’s characterization. Specifically, I will assume that deverbal nominalizations are subject to principle (39), which could be implemented as having a particular kind of n in the structure. This type of selects for a structure that does not project an external argument in VoiceP.

15.4.2 Case In this section, I will briefly summarize my assumptions concerning Case, and in particular ergative Case. Building on Marantz (1991/​2000) and Baker (2015), I assume that case is assigned on the basis of the principles of “dependent case” (cf. Yip, Maling, & Jackendoff 1987), see also Baker & Bobaljik (Chapter  5, this volume). Specifically, Marantz (1991/​2000) argues that the distribution of morphological case is determined at PF, subject to the case realization hierarchy in (40): (40) case realization disjunctive hierarchy: i) lexically governed case, ii) “dependent” case (accusative and ergative), iii) unmarked case (environment-​sensitive), iv) default case The more specific a case is it is assigned first taking precedence over the cases lower in the hierarchy. In this system, structural accusative Case is “dependent case” subject to the definition in (41):5 (41) Dependent case is assigned by V+I to a position governed by V+I when a distinct position governed by V+I is: a. not “marked” (not part of a chain governed by a lexical case determiner) b. distinct from the chain being assigned dependent case Baker (2015; 48–49), building on Baker and Vinokurova (2010), proposes a reformulation of this as in (42), cf. Alexiadou, Anagnostopoulou & Sevdali (2014: 24) for a reformulation of (41) in terms of Agree: 5  Originally dependent case assignment was regulated by the formation of a chain between V + I, see Marantz (1991/​2000). More recent analyses embrace a phase-​based model and attempt to restate the conditions on dependent case assignment, see e.g. (42) and Schäfer (2012a) among others and references therein.

 

Ergativity in nominalization    369 (42) a. If there are two distinct NPs in the same spell out domain such that NP1 c-​commands NP2, then value the case feature of NP2 as accusative unless NP1 has already been marked for case. b. If there are two distinct NPs in the same spell out domain such that NP1c-​commands NP2, then value the case feature of NP1 as ergative unless NP2 has already been marked for case. While Marantz (1991/​2000) as well as Baker (2015) considered ergative case to be a dependent case assigned on the subject, see (40), other authors have argued that ergative case is not structural. For instance, Nash (1995) proposed that in accusative languages the subject is projected external to the VP as the specifier of a functional projection which selects VP, Voice in our structures. In ergative languages, however, the subject is projected VP-​internally, as the highest adjunct of the lexical VP projection. Since no external argument is introduced in Spec,VoiceP, following Bok-​Bennema (1991), ergativity then means that transitive verbs cannot assign dependent structural accusative case. The single argument receives unmarked absolutive Case, see also Alexiadou (2001), Woolford (2006), and more recently Legate (2012a). See also Legate (Chapter 6, this volume), Nash (Chapter 8, this volume), Sheehan (Chapter 3, this volume), and Baker & Bobaljik (Chapter 5, this volume) for further discussion on the nature of ergative case. What is important for us is that the morphological realization of case is sensitive to a particular local domain in which case is assigned. In the verbal domain, the relevant domain is Voice+v+I or rather the CP, see Baker (2015). In the nominal domain, in the absence of I, n and D are candidates for creating the relevant domains, i.e. the nP and DP layer.

15.4.3 Zooming in on n Let us then take a closer look at the properties of this n head. Following Embick (2010), I take n to be a phase head, in the sense that it fixes the interpretation of the structure as a nominal one. As soon as n is inserted, higher functional heads with verbal properties are not allowed, as n and the functional layers above n have to Agree in morphosyntactic features, cf. Imanishi (2014). n is a nominalizer, but also has properties similar to a passive head. Due to its category changing properties, n now creates a nominal Case domain. Since n either embeds a passive VoiceP or simply a vP, see (39), it now counts as the relevant head that determines case. According to (39), the unmarked case for arguments within nP is GEN; this is briefly mentioned in Marantz’s (1991/​2000) article, see also Alexiadou (2001), and Baker (2015). The intuition behind the dependency view of ACC is based on the presence of another argument being assigned unmarked case. Let us assume that this intuition is correct, and moreover assume that dependent Case patterns arise when we have more than one head that can enter Agree with an argument, see Schäfer (2008), Bobaljik (2008). Crucially, what is excluded is a head entering multiple Agree with two arguments, see also Haegeman & Lohndal (2010) for discussion.

 

370   Artemis Alexiadou In the verbal domain, the internal argument of a predicate enters Agree with Voice, and surfaces with accusative, while the external argument enters Agree with T and surfaces with nominative (or alternatively receives unmarked nominative, see Baker 2015). A nominal reflex of this pattern is found in the case of the verbal gerund, the external argument enters Agree with D and surfaces in the Saxon genitive, while the internal argument enters Agree with Voice and surfaces with accusative. Consider now what happens in the case of nominal gerunds or derived nominals. In these cases, our domain of Agree for Case is now the nP. In all cases, we have examined, there is only one argument within the nP, which surfaces with genitive. (43)

DP nP n

vP/VoiceP-ext.arg./ DPgen

The only option allowed for the other argument, in the case of a nominalization derived from a transitive predicate, is either to surface as a by-​phrase below nP, since strictly speaking the complement of nP has undergone passive formation, or to be merged higher than the nP domain, thus escaping nP. The latter is what we see in (44), where the external argument is in Spec,DP, and the single internal argument is in the genitive: (44) John’s destruction of the manuscript Data such as the ones in (44) crucially suggest that English D is different from that in other languages, e.g. the Romanian supine data briefly discussed in note 4. Unlike Romanian D, for instance, English D can enter Agree, and thus be responsible for licensing an extra Case within the nominal clause. In contrast, Romanian D is simply a nominalizer, and functions similar to n in (43). When the option in (44) cannot be chosen, due, for example, to the properties of the DP projection, as is the case in Greek, where Spec,DP is an A’-​position, or because movement has not taken place, there is no other option available for the external argument but to resort to a PP realization. The nominalization structure in this case, unlike the verbal structure, simply does not make two heads within the nP available for the two arguments to Agree with and surface with distinct cases, thus the external argument will surface as a PP, i.e. take an ergative realization. This is precisely why we cannot have two genitives within the same nP domain, i.e. in one case domain. This seems to be a more general phenomenon that can be captured under the subject-​in-​situ generalization proposed in Alexiadou & Anagnostopoulou (2001), see (45), and also the Multiple Case Condition in (46), from Collins (2003: 16), cf. the notion of Distinctness in Richards (2010):

 

Ergativity in nominalization    371 (45)

The subject-​in-​situ generalization (SSG) By Spell-​Out, vP can contain only one argument with a structural Case feature. Alexiadou and Anagnostopoulou (2001)

(46) Multiple Case Condition By Spell-​Out VP can contain no more than one argument with a (valued) undeleted Case feature. In both (45), and (46), replacing vP and VP with nP will give us the required result. Since verbal gerunds lack n, they are not subject to (45) or (46), and the external argument obligatorily moves to Spec, DP in this case; dependent Case can be assigned as the external argument enters Agree with D, and no process of passivization has taken place: (47) John’s destroying the manuscript The important question that arises is why nominalizations that contain n or other elements that function as nominalizers cannot tolerate Tense. I believe the answer to this question is relatively simple and rather straightforward: derived nominals simply lack tense, as they cannot be located in time; only events can. Even for the languages that have been argued to contain nominal tense, it has been shown that this is either adverbial in nature (Wiltschko 2003), or that it takes over functions associated with determiners in languages that lack nominal tense (Lecarme 1996, and subsequent work; Alexiadou 2008). Summarizing the analysis: in nominalization, the presence of n locks the interpretation and the properties of the derived element as being nominal. Importantly, the presence of n has as a consequence that higher functional heads are nominal in nature, and not verbal. n creates a case domain requiring the DP argument of ASNs to surface with genitive. In the case of ASNs derived from transitive predicates, unless the external argument moves outside nP, it will be realized via a PP, and as a result an ergative pattern is enforced.

15.5 Conclusion In this chapter, I showed that the presence of a nominalizer head in a syntactic structure creates an environment in which only one argument can receive structural Case. As a result, an ergative case pattern emerges. Let me now briefly return to unergative nominalizations. As stated in footnote one, traditionally such nominalizations are not treated as argument supporting. Thus, the fact that their single argument bears genitive could be explained by assuming that this is introduced in the nominal part of the structure, i.e. it is not an argument of the nominalized verb. From this perspective then the fact that their single argument bears genitive is simply a result the default case pattern within nominals, and the nominalization itself is not really subject to the requirement in (39).

 

372   Artemis Alexiadou If the single argument of an unergative predicate is introduced in VoiceP, similar to the external argument of a transitive predicate, (39) would enforce a syntax similar to that of impersonal passives, which is clearly not what we have in the case of examples such as (3b). Thus we would be led to assume that the single argument of unergative nominals is introduced in nominal structure. If we, however, assume that ergativity always treats the subject of an intransitive and the object of a transitive predicate alike, while it treats the agent of a transitive predicate differently, we would be led to a different conclusion. The case patterns would be interpreted as suggesting that intransitive nominalizations do not distinguish between unergative and unaccusative predicates. This means that they show a case system of the type described for Niuean by e.g. Massam (2009) and reference therein. In this type of language, the single argument of both unergative and unaccusative predicates bears absolutive case, see Table 15.1. This is in contrast with what has been reported for e.g. certain dialects of Basque, where the subject of unergative predicates surfaces with ergative and the subject of unaccusatives bears absolutive case, see Tollan (2013) for a recent discussion and references. Massam (2009) explicitly argued that the Niuean case pattern can be derived by assuming that only transitive agents are introduced in Voice, while arguments of unergative verbs are introduced within the vP. If unergative nominalizations behave alike, then (39) basically enforces the blurring of the unergative vs. unaccusative distinction in the nominal domain. One could interpret this as suggesting that (39) should explicitly state that nominalizations must lack external arguments of the type introduced in VoiceP. The question that arises then is why not all ergative languages conform to the case patterns in Table 15.1 as far as their verbal domain is concerned (e.g. Niuean vs. dialects of Basque). While treating this issue is well beyond the scope of this chapter, two observations seem relevant here. First, as already discussed in Massam (2009), such patterns raise questions as to the treatment of (all) unergatives as concealed transitives across languages (Hale & Keyer 2002). Second, it points to a difference in the syntactic representation of the event/​theta role of the subject of a transitive predicate as opposed to that of an unergative one, see Massam (2009) for some discussion.

Acknowledgments I am indebted to the editors of this volume, Terje Lohndal, and one anonymous reviewer for their comments. My research has been supported by a DFG grant for project B1 The Form and Interpretation of Derived Nominals of the collaborative research center 732 Incremental Specification in Context at the University of Stuttgart.

Abbreviations 1, first person; ABS, absolutive; ACC, accusative; ERG, ergative; GEN, genitive; NOM, nominative; PASS, passive; PL, plural; PP, prepositional phrase.

 

Chapter 16

Ergativit y a nd Au stronesia n-​t ype voice syst e ms Michael Yoshitaka Erlewine, Theodore Levin, and Coppe van Urk

16.1 Introduction Many languages of the Austronesian family exhibit what has been called a “voice system”:1 a particular pattern of alternations in word order, case marking, and verbal morphology, which also interacts with Ā-​extraction. The voice system has been a central concern in the study of Austronesian syntax. One influential proposal for such languages treats them as morphologically and syntactically ergative. The ergative hypothesis is attractive because it offers a way of mapping some of the morphosyntactic properties that look uniquely Austronesian, such as its voice morphology, to familiar features of non-​Austronesian languages. In this chapter, we critique the ergative analysis of Austronesian-​type voice system languages, using data from well-​ studied voice system languages, including Tagalog, Malagasy, and Atayalic languages, along with new data from Balinese and Dinka (Nilotic), a non-​Austronesian language with all of the hallmark properties of an Austronesian voice system.2 On the basis of dissociations between case, voice, and

1 

The “voice system” has been made famous by Philippine languages, although the basic description (section 16.1.1) also applies to a range of Formosan and Western Malayo-​Polynesian languages. The voice system has been reconstructed for Proto-​Austronesian (Wolff 1973; Starosta et al. 1982/​2009). See also n. 4 on the term “voice system.” 2  We note that there are also Austronesian languages which are morphologically and syntactically ergative but do not exhibit the voice system described here. See for example Otsuka (Chapter 40, this volume) for discussion of Tongan.

 

374    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk extraction, we argue that there must be mechanisms other than ergativity that will yield the behavior associated with Austronesian voice. The chapter is organized as follows. We first introduce Austronesian voice systems and their treatment as morphologically and syntactically ergative. In section 16.2, we present new data from the Nilotic language Dinka, a non-​Austronesian language with a voice system, in which dissociations between voice and case reveal a consistently nominative-​ accusative alignment. Section 16.3 documents evidence from multiple extraction in the Malay/​Indonesian languages Balinese and Bahasa Indonesia that a voice system can exist in the absence of strict syntactic ergativity. In addition, we show that there are surprising restrictions on the licensing of in situ subjects in these languages, unexpected under an ergative analysis.

16.1.1 Properties of a Voice System Voice systems are characterized by the fact that a single argument of the clause—​possibly a non-​core argument, as we will see—​is privileged in certain ways. This argument may be in a certain linear position or receive a particular morphological marking, and dedicated morphology on the verb indicates which argument of the verb was chosen for this special status. Furthermore, Ā-​extraction is often limited to this privileged argument. By way of example, consider the sentences in (1). The sentences all describe Tali eating fish, but vary in word order, case marking, and verbal morphology. (1)

Voice alternation in Squliq Atayal (Liu 2004):3 a. M-​ aniq qulih qu Tali’. sv-​eat     fish     qu  Tali ‘Tali eats fish.’

Subject Voice (SV)

b. Niq-​un  na’   Tali’ qu qulih  qasa. eat-​ov    gen  Tali    qu  fish     that ‘Tali ate the fish.’

 Object Voice (OV)

c. Niq-​an  na’  Tali’ qulih qu ngasal  qasa. eat-​lv   gen  Tali   fish     qu  house    that ‘Tali eats fish in that house.’

Locative Voice (LV)

d. S-​ qaniq na’  Tali’ qulih qu qway. iv-​eat    gen  Tali   fish     qu  chopsticks ‘Tali eats fish with chopsticks.’          Benefactive/​Instrumental Voice (B/​IV)

In each example, one argument of the verb (in italics) is in sentence-​final position preceded by the marker qu. Voice morphology on the verb (in bold) reflects this choice of 3 

Glosses and translations are modified. It is most common in the Philippine and Formosan literature to refer to Subject Voice and Object Voice as “Actor Voice” and “Patient Voice,” respectively.

 

Ergativity and Austronesian-type voice systems    375 argument. It is common for Philippine and Formosan languages to have four or five distinct voices. Note that the subject in Non-​Subject Voices is preceded by the genitive case maker na, which is also used for nominal possessors. This genitive marking of Non-​ Subject Voice subjects will become important later. We will refer to the special argument as the “pivot.” A  terminological warning is immediately in order: we mean to use the terms “pivot” and “voice” as pre-​theoretical labels for the privileged argument cross-​referenced by verbal morphology in these languages and the morphology cross-​referencing it. The use of “voice” should not be conflated with familiar active/​passive alternations in non-​Austronesian languages.4 An important property of voice systems is that Ā-​extraction is typically restricted to the pivot argument, as illustrated by Squliq Atayal wh-​questions (2–​3).5 Subject Ā-​extraction requires Subject Voice morphology (2a–​b), while object Ā-​extraction requires Object Voice morphology (3a–​b). This pattern extends to other voices as well. (2)

Subject wh-​question ⇒ Subject Voice: a. Ima   (qu)  wal    m-​ aniq sehuy qasa? who  qu   past  sv-​eat   taro    that ‘Who ate that taro?’ b. *Ima who

(3)

(qu) qu

wal past

niq-​un eat-​ov

sehuy taro

qasa? that

Object wh-​question ⇒ Object Voice: a. *Nanu (qu) wal m-​aniq (qu) Yuraw?  what qu past sv-​eat qu Yuraw b. Nanu (qu) wal  niq-​ un  na    Yuraw? what   qu  past  eat-​ov  gen  Yuraw ‘What did Yuraw eat?’

Atayal exhibits all of the hallmark properties of an Austronesian-​type voice system. These properties are summarized in (4). It is worth noting, however, that not all Austronesian languages which could be or have been described as having a voice system clearly exhibit all four of these characteristics.

4  A range of different terms have been used in previous Austronesian literature for these same notions. For example, the terms “subject,” “focus,” “topic,” and “trigger” have all been used by some authors for what we call the “pivot” here. Similarly, the “voice system” is often called a “focus system,” among other terms. See Blust (2002); Ross and Teng (2005) for an overview of terminological use in the literature, also discussed in Blust (2013: sec. 7.1). 5  Wh-​questions in Atayal and other Austronesian voice system languages have been variously analyzed as Ā-​movement of the wh-​word itself or a pseudocleft construction with the wh-​word predicating a headless relative to its right; either way, we assume these examples involve Ā-​extraction over the pivot argument.

 

376    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk (4) Characteristics of Austronesian-​type voice systems: a. A privileged argument: One argument is designated the “pivot,” and is realized in a particular morphological form and/​or structural position, regardless of its original grammatical function. b. Articulated voice morphology: Morphology on the verb varies with the choice of pivot, including options for taking certain oblique arguments as pivots. c. Extraction restriction: Ā-​extraction (wh-​movement, relativization, topicalization, etc.) is limited to the pivot argument. d. Marking of non-​pivot subjects: Non-​pivot subjects are morphologically marked, often coinciding with the form of possessors (i.e. genitive case). One of the main challenges of Austronesian syntax is to explain this unique constellation of properties. One prominent attempt to do so, which we will now review, is to analyze voice systems as morphologically and syntactically ergative. (See also Kaufman (Chapter 24, this volume) for further discussion of Austronesian voice systems and their analysis.)

16.1.2 The Ergative Hypothesis In the late 1970s and early 1980s, a line of work emerged suggesting that voice system languages should be analyzed as morphologically and syntactically ergative (De Guzman 1976, 1988; Payne 1982; Cooreman et al. 1984; Gerdts 1988b), and the hypothesis has been modernized and championed in the past decade by Aldridge (2004 and subsequent work). Payne (1982), for example, draws explicit parallels between the clause structure of Tagalog and that of the ergative language Yupik Eskimo. The central tenet of the ergative hypothesis is that the privileged argument of the clause (our “pivot”) carries absolutive case. Marked subjects in Non-​Subject Voices are ergative arguments. Subject Voice clauses with transitive roots are analyzed as antipassive clauses, so that Voice morphology is, in the simplest cases, a marker of the verb’s syntactic transitivity. We illustrate the ergative hypothesis using the Squliq Atayal voice system described in section 16.1.1.6 Example (5a) repeats the Squliq Atayal examples (1b), reglossed according to an ergative analysis, together with an intransitive clause from Liu (2004) in (5b). Alterations are bolded. (5)

6 

Squliq Atayal as an ergative language: a. Niq-​un      na’   Tali’  qu     qulih  qasa. eat-​trans  erg  Tali   abs  fish    that ‘Tali ate the fish.’

Object Voice (1b) = transitive

The presentation here follows ergative analyses of Atayal as in Huang (1994) and Starosta (1999); and the ergative analysis of the sister language Seediq (Atayalic) in Aldridge (2004).

 

Ergativity and Austronesian-type voice systems    377  b.  Cyux m-​’abi’  prog intrans-​sleep  ‘Tali is sleeping.’

qu abs

Tali’. Tali

Subject Voice = intransitive

In an ergative analysis, Object Voice clauses are analyzed as simple transitive clauses, in which the object is marked with absolutive case and the subject with ergative case (5a). Recall that in Atayal Object Voice—​and more generally in Non-​Subject Voices—​the subject is morphologically marked in the same way that nominal possessors are, with the marker na. This is treated as a syncretism between ergative and genitive case, a common pattern cross-​linguistically (Trask 1979). In this analysis, Subject Voice marks a syntactically intransitive clause, so that the prototypical case of SV is an example like (5b), in which the intransitive subject is morphologically marked in the same way as the transitive object in (5a): with the absolutive marker qu. The voice morphology glossed as OV and SV in (5) are then markers of the clause’s syntactic transitivity, transitive and intransitive, respectively. This analysis can be extended to Subject Voice clauses with transitive roots by treating them as antipassive constructions. (See Polinsky (Chapter 13, this volume) for an overview of the antipassive cross-​linguistically.) The antipassive alternation takes the transitive verb in (5a) and demotes the object qulih ‘fish’ into an oblique, resulting in a syntactically intransitive verb with a single argument, Tali. The result is (6): the verb is now intransitive and therefore bears intransitive morphology (m-​). Tali is now the subject of an intransitive verb and thus carries absolutive (qu). No morphology is associated with the antipassivization proper.7 The argument ‘fish’ which was demoted is, under this view, now an oblique. No oblique marking is observed in (6), but note that this argument would be preceded by a distinct marker in other Atayalic languages such as Mayrinax Atayal (Huang et al. 1998; Huang 2000). (6) Subject Voice with a transitive root is analyzed as an antipassive (ap): M-​aniq qulih qu Tali’. intrans-​eat(ap) fish(obl) abs Tali ‘Tali eats fish.’ Subject Voice (1a) = antipassive Additional voices beyond Subject and Object Voice can be analyzed as applicative constructions (Aldridge 2004), which introduce an adjunct or indirect object as the highest internal argument. It is this argument that is then picked out as the absolutive. Note that we will mainly concentrate on the contrast between Subject and Non-​Subject Voices, taking Object Voice as our representative case. The final ingredient in an ergative analysis is syntactic ergativity. Recall that in Austronesian-​type voice systems, only the pivot argument cross-​referenced by voice morphology (in the descriptive terminology used in the previous section) can be 7  The lack of overt morphological evidence for the ergative hypothesis led some researchers to develop alternative, usage-​based diagnostics for ergativity, for example based on the corpus frequency and acquisition of different voices. Such arguments will not be discussed here. See Cumming and Wouk (1987) for review and discussion.

 

378    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk Ā-​extracted. Under the ergative hypothesis, this is described as a restriction that only absolutive arguments can be Ā-​extracted. For example, a transitive object wh-​question requires Object Voice, which is the regular transitive clause form, as the object is then the absolutive argument. A  transitive subject wh-​question requires Subject Voice, which uses antipassivization to turn the transitive subject into an absolutive argument.8 This type of extraction restriction is independently observed in many (though not all) morphologically ergative languages (e.g. Manning 1994). A classic example of this extraction restriction in an unambiguously morphologically ergative language comes from Dyirbal (Pama-​Nyungan; Australia), in (7a–​c). (7) Dyirbal relativization targets the absolutive (Dixon 1979: p. 128): a. ŋuma-​ ŋgu  [duŋgara-​ ŋu]-​ ru yabu     bura-​ n. father-​erg  cry-​rel-​erg    mother(abs) see-​ past ‘Father, who was crying, saw mother.’       Intransitive subject relative b. ŋuma     [yabu-​ ŋgu    bura-​ ŋu] duŋgara-​ nyu. father(abs) mother-​ erg  see-​rel   cry-​past ‘Father, who mother saw, was crying.’        Transitive object relative c. ŋuma     [bural-​ ŋa-​ ŋu yabu-​ gu]    duygara-​ nyu. father(abs) see-​ ap-​rel    mother-​dat  cry-​past ‘Father, who saw mother, was crying.’   Transitive subject relative ⇒ antipassive

Intransitive subjects and transitive objects (absolutive arguments) can be relativized without restrictions, but relativization of a transitive subject requires an antipassivization step first (7c), in order to make the subject an absolutive. This property that only absolutive arguments can be Ā-​extracted is often referred to simply as “syntactic ergativity,” although the label originally referred to the presence of any syntactic process sensitive to the ergative/​absolutive distinction. We will adopt this terminological choice here and refer to this syntactically ergative extraction asymmetry as “syntactic ergativity.” Consider how the ergative hypothesis captures each of the core properties of voice systems (4): (8) The ergative hypothesis for Austronesian-​type voice systems, following (4): a. A privileged argument: Every clause has one absolutive argument. b. Articulated voice morphology: Morphology on the verb reflects the transitivity of the clause and any argument structure alternations, which correlate with the choice of absolutive argument. Applicatives introduce an additional argument as the highest internal argument, which will thus be absolutive.

8  See for example Aldridge (2004) for a detailed derivation of this extraction restriction. Note that, for Aldridge, the structural position of absolutive arguments makes it the unique target for movement; the extraction asymmetry is not a ban on movement of ergative or oblique arguments per se.

 

Ergativity and Austronesian-type voice systems    379   c.  Extraction restriction: The language is syntactically ergative: Ā-​extraction (wh-​movement, relativization, topicalization, etc.) is limited to the absolutive argument.   d.  Marking of non-​pivot subjects: Transitive subjects are ergative. Ergative may be syncretic with genitive. (In Subject Voice, antipassivization makes the subject absolutive.) The ergative hypothesis was illustrated here using the Atayalic language of Squliq Atayal, but it has also been considered in contemporary literature for other well-​studied Austronesian voice system languages, including Tagalog (Aldridge 2004), Malagasy (Paul and Travis 2006), and Indonesian (Aldridge 2008b). The strength of the ergative hypothesis lies in the fact that it offers an explanation of voice system behavior that does not require postulating mechanisms that are unique to Austronesian. Under the assumption that Austronesian languages are syntactically ergative, this view allows us to recast voice systems as a particular grouping of argument structure alternations which are otherwise cross-​linguistically well attested. In the next two sections, however, we show that there are voice systems which cannot plausibly be analyzed as ergative. In Dinka, a Nilotic language with all the properties of an Austronesian voice system, dissociations between voice and case uncover a consistently nominative-​accusative alignment. In Balinese and Bahasa Indonesia, two Malay/​Indonesian languages, multiple extraction in questions, topicalization, and relative clauses reveals the absence of syntactic ergativity in OV clauses. These facts suggest that neither morphological nor syntactic ergativity are necessary ingredients for an Austronesian-​type voice system. We conclude then that, even if ergativity might underlie some Austronesian voice systems, there must be mechanisms other than ergativity that will bring about a voice system.

16.2  Dissociating Case and Voice in Dinka In this section, we introduce the voice system of the Nilotic language Dinka. As in Austronesian languages, voice morphology in Dinka correlates with restrictions on extraction and changes in case relations. Most importantly for our purposes, Dinka subjects display the same case pattern that has provided the impetus for the ergativity view of Austronesian voice, alternating between unmarked case in the Subject Voice and a marked case also used for possessors in all other voices, variously referred to as genitive, oblique, or marked nominative (Andersen 1991, 2002; König 2008b, Chapter 37, this volume). However, in Dinka, case marking on subjects is dissociable from voice morphology. We will show that there are several syntactic environments in which the voice system is suppressed, triggering the appearance of Subject Voice as a morphological default. In

 

380    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk these contexts, however, subjects still appear in the genitive case, the case that would be treated as ergative in an ergative analysis. In fact, these environments show a consistent nominative-​accusative alignment. As a result, there must be a mechanism that yields these case marking patterns that is independent of voice morphology.9

16.2.1 The Dinka Voice System Dinka is a Nilotic language spoken in South Sudan. Data in this chapter is from Dinka Bor, the major dialect in the southeastern dialect group. Dinka is a V2 language, which, following van Urk (2015), we take to reflect a requirement of C that it must have a specifier, with concomitant movement of the highest verb/​auxiliary up to C.  Dinka, like Malagasy (e.g. Pearson 2001, 2005), has three voices, which reflect the grammatical function of the noun phrase in Spec-​CP, or the pivot. Subject Voice is used when the subject is the clause-​initial pivot (9a), Object Voice when it is the object (9b), and the Oblique Voice is employed for all other choices of pivot (9c). (9) Voice on second position verb: a. Bòl à-​ càm     cuîin  cé̤  pǎal.  Bol 3s-​eat.sv  food   p   knife ‘Bol is eating the food with a knife.’

  Subject Voice (SV)

b. Cuîin   à-​cέεm  Bôl      nè   pǎal.  food    3s-​eat.ov   Bol.gen  p  knife ‘The food, Bol is eating with a knife.’ c. Pǎal     à-​cέεmè​   Bôl      cuîin   knife 3s-​eat.oblv  Bol.gen  food ‘With a knife, Bol is eating the food.’

  Object Voice (OV)

     Oblique Voice (OblV)

Voice morphology appears on the verb or auxiliary in C, which is the main verb in (9a–​c). However, if an auxiliary is present, the highest auxiliary moves to second position, just as in Germanic V2 languages. In such constructions, voice distinctions are made on the auxiliary and not the verb. The examples in (10a–​c) illustrate this for the perfect auxiliary cé.10  (10) Voice on second position auxiliary: a. BÒl à-cé̤          cuîin  câam     nè   pǎal.  Bol 3s-​perf.sv  food    eat.nf  p  knife ‘Bol has eaten food with a knife.’

      Subject Voice

9  See also König (2008b, Chapter 37, this volume) for descriptions of Nilotic languages with the same case pattern without voice morphology. 10  Dinka differs in this respect from many Austronesian languages, in which voice morphology appears on the verb. However, it is known that even related languages can vary in whether certain

 

Ergativity and Austronesian-type voice systems    381  b.  Cuîin à - c í i       Bôl      câam    nè  pǎal.    knife  food    3s-​perf.ov  Bol.gen  eat.nf  p    ‘Food, Bol has eaten with a knife.’  c.  Pǎal     à-cénè    Bôl      cuîin  câam.     knife  3s-​perf.oblv  Bol.gen  food   eat.nf  ‘With a knife, Bol has eaten the food.’

   Object Voice

   Oblique Voice

As in Austronesian languages, voice has repercussions for case marking and extraction. The pivot always occurs in the unmarked case, regardless of its grammatical function, as evident in the examples in (9a–​c) and (10a–​c). In addition, voice restricts extraction, so that only the pivot can undergo wh-​movement, for example (11a–​c). (11)

Voice restricts extraction: a. Yeŋà  càm/​*cέεm/​*cέεmè       cuîin  nè  pǎal?    knife who   eat.sv/​eat.ov/​eat.oblv  food   p   ‘Who is eating the food with a knife?’

   Subject Voice

b. Yeŋó  cέεm/​*càm/​*cέεmè   Bôl      nè  pǎal?  eat.ov/​eat.sv/​eat.oblv Bol.gen    knife what  p   ‘What is Bol eating with a knife?’

    Object Voice

c. Yeŋó   cέεmè/​*càm/​*cέεm     Bôl      cuîin?   eat.sv/​eat.ov  Bol.gen  food  eat.oblv/​ what   ‘What is Bol eating the food with?’

    Oblique Voice

Just as in other voice systems, the case marking on subjects alternates with voice. In Subject Voice, the clause-​initial subject occurs in the unmarked case (12a), but, in Object Voice or Oblique Voice, subjects appear in the genitive case (12b).11 (12)

Voice determines case marking on subjects: a. Àyén  à-​càm   cuîin  nè  pǎal.   Ayen  3s-​eat.sv  food   p   knife ‘Ayen is eating the food with a knife.’ b. Cuîin   à-​cέεm   Áyèn      nè  pǎal.   knife food   3s-​eat.ov  Ayen.gen  p     ‘The food, Ayen is eating with a knife.’

morphology is expressed on the verb or auxiliary. We therefore believe this difference does not preclude us from considering Dinka in the context of a broader discussion of Austronesian-​type voice systems, as Dinka does exhibit the core properties summarized in (4). 11 

Case marking in Dinka involves tonal alternations. See Andersen (2002) for an overview.

 

382    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk We refer to this case as genitive, because it also appears on possessors (see Andersen 2002 and König 2008b for discussion). In any case, the similarity with Austronesian voice systems is striking. It should be clear then that Dinka has all the properties that make an ergativity approach to voice appealing: a case alternation with subjects and restrictions on extraction that correlate with voice morphology.

16.2.2 Against an Ergative Analysis of Dinka We will now show that we can rule out an ergative analysis of Dinka. This leads us to the conclusion that, despite the advantages of the ergative analysis, there must be a different mechanism for arriving at an Austronesian-​style voice system. The first problem with an ergative analysis of Dinka is that morphology encoding argument structure alternations has a different distribution than voice morphology. For example, Dinka has an antipassive construction, which is independent of the voice system described in section 16.2.1. As documented in detail by Andersen (1992), antipassive morphology appears on the verb and the object is demoted to an optional PP (13a–​b).12 (13) Dinka has an independent antipassive: a. Àyén  à-cé       cuîin  câam.   Ayen 3s-​perf.sv  food   eat.nf ‘Ayen has eaten the food.’ b. Àyén à-cé càm  Ayen 3s-​perf.sv eat.ap.nf ‘Ayen has eaten food.’

(è cuîin).   p food

Antipassive morphology always appears on the lexical verb, even when an auxiliary is present, as (13b) illustrates. Voice morphology, in contrast, shifts to the highest auxiliary if one is present, as previously discussed. This difference is problematic for a view in which voice morphology is argument structure morphology, particularly if we treat Subject Voice as an antipassive. Another problem faced by an ergative view of Dinka is that the mechanisms behind voice morphology can be shown to be independent of the mechanisms behind genitive case marking. In particular, there are several syntactic environments in Dinka in which

12 

There is also an independent applicative construction, which introduces benefactive arguments. Like the antipassive, this morphology is restricted to the verb. It is not unreasonable to think, however, that Dinka might have two distinct applicatives (along the lines of Pylkkänen 2002). Mark Baker (p.c.) asks whether we could think of Dinka as having two different antipassives, as has been claimed for some Mayan languages. However, it is not clear that all of the constructions that have been analyzed as antipassives in Mayan in fact are true antipassives—​see e.g. Smith-​Stark (1978), Grinevald Craig (1979), Aissen (1999a), Stiebels (2006) for arguments against viewing Agent Focus (formerly described as the “agentive/​focus antipassive”) as an antipassive.

 

Ergativity and Austronesian-type voice systems    383 V1 order is possible, and where no phrase overtly moves to Spec-​CP. In these contexts, subjects surface in the genitive case, but the clause is marked with Subject Voice. The V1 pattern is found obligatorily in yes-​no questions and optionally in wh-​in situ questions or after the finite complementizer è. Both environments involve full finite  in allowing V1 order (14a–​c). clauses, but differ from matrix declarative clauses (14)

V1 order in yes-​no and in situ questions and embedded clauses: a. Cé   Áyèn   cuîin  câam?   perf.sv  Ayen.gen  food   eat.nf ‘Has Ayen eaten the food?’ b. Cám   Bôl       ŋó?  eat.sv  Bol.gen  what ‘What is Bol eating?’ cuîin câam]. c. A-yùukù luêeel [è cé Áyèn      3s-​hab.1pl say.nf C  perf.sv Ayen.gen food eat.nf ‘We say that Ayen has eaten the food.’

We propose that the V1 order arises because these constructions involve C heads that do not require V2.13 When these V1 orders are possible, every nominal in the clause is case-​marked just as when it is not the pivot. Subjects are genitive, as in Object Voice and Oblique Voice clauses. Importantly, however, a V1 clause only allows Subject Voice morphology. This mismatch is surprising under an ergative analysis. If we take Non-​Subject Voice morphology to reflect ergative alignment, genitive should not be able to surface in the absence of this morphology. It should not matter whether V2 is possible, since these clauses are big enough to host the requisite argument structure alternations. It is worth reflecting briefly on what kind of approach to voice morphology might fare better with regard to the facts in (14a–​c). We think that, at least for Dinka, this pattern argues strongly for an analysis in which voice morphology is treated as extraction marking, as in wh-​agreement or case agreement approaches (e.g. Chung 1994; Richards 2000; Pearson 2001, 2005; Rackowski 2002). If voice morphology is a by-​product of extraction, then there should not be voice distinctions in clauses without extraction. We can then interpret the Subject Voice just as the default form in the voice morphology paradigm.14 The final problem for an ergative analysis we would like to discuss is that the genitive case on non-​pivot subjects shows no sensitivity to properties of the verb, such as transitivity and 13 

Another option is that some of the constructions involve silent operators that satisfy V2 but do not participate in the voice system, either because they are not nominal in nature or because they are base-​ generated in the left periphery and so have not undergone movement. 14  Certain nonfinite clauses in a range of Formosan languages can only occur in Subject Voice (Chang 2010), which suggests that Subject Voice is a morphological default in these languages as well. This view is strengthened by the analysis of such embeddings as restructuring, and the availability of so-​called long passive constructions which show that the embedded Subject Voice morphology is not syntactically real. See Chen (2010, 2014) for such arguments from Mayrinax and Squliq Atayal and Wurmbrand (2015) for discussion.

 

384    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk unaccusativity.15 As long as the subject is not clause-​initial, such as in a V1 clause, genitive case may surface in unaccusatives (15a), unergatives (15b), and antipassives (15c).16 (15)

Genitive case occurs with all intransitives: a. Bé       lèc      dhuôoŋ?   fut.sv  stick.gen  break.inch.nf ‘Will the stick break?’ b. Thε´t      Bôl?  cook.sv  Bol.gen ‘Is Bol cooking?’ c. Bé       Bôl   càm      (è̤  cuî̤in)?  fut.sv  Bol.gen  eat.ap.nf  p   food ‘Will Bol eat food?’

This pattern too is surprising under an ergative analysis, because it reveals a consistent case marking for subjects according to a nominative-​accusative alignment. Taken together, these facts suggest that there are mechanisms other than ergativity that will yield an Austronesian-​type voice system. Specifically, it seems clear that there are syntactic processes independent of ergativity that may lead to voice morphology as well as a case alternation involving genitive for subjects.

16.2.3 Genitive as a Repair In this section, we consider the question of what mechanism might lie behind the assignment of genitive case in Dinka. As discussed in the previous section, we assume that voice morphology in Dinka should be treated as a form of extraction marking, as in wh-​agreement or case agreement proposals (Chung 1994; Richards 2000; Pearson 2001, 2005), given its independence from the processes behind case marking on subjects. The view of genitive case we want to pursue here is that it represents a strategy for licensing nominals not in a case position, and so functions as a type of repair (cf.

15 

See Rackowski (2002) for similar argumentation in Tagalog. One of the ways in which we can tell that these are unaccusatives is that verbs like dhuòoŋ (‘break.inch’) participate in a inchoative/​causative alternation (ia–​b). 16 

(i)

Inchoative/​causative alternation: a. Léc   à-bé      dhuôoŋ.  stick 3s​fut.sv  break.inch.nf ‘The stick will break.’ b. Bòl à-bé léc  Bol 3s-​fut.sv stick ‘Bol will break the stick.’

dhôoŋ. break.nf

 

Ergativity and Austronesian-type voice systems    385 Donohue and Donohue 2010; Imanishi 2014). To be precise, we follow Halpert (2012) in assuming that, in some languages, case morphology may be merged directly to a nominal to license it, if no other licensing strategy is available.17 See also Stowell (1981) on English of-​insertion. We apply this to Dinka as follows. We propose, following van Urk (2015), that Dinka Spec-​CP fulfills both of the functions traditionally associated with Spec-​CP and Spec-​TP, so that it is the landing site of Ā-​movement, but also a case position. In Subject Voice, subjects receive case in Spec-​CP and so appear in the unmarked case. In Non-​Subject Voices, however, the subject needs to be licensed in a different way, because Spec-​CP is occupied and T is not a case assigner. This is the role of genitive case morphology. In Dinka, this strategy is not necessary for other nominals. As van Urk and Richards (2015) show, there is a position for objects inside of the verb phrase where they may receive unmarked case. This analysis extends well to Austronesian languages. A number of Austronesian systems can be described in the same terms as Dinka. An example is the (Squliq) Atayal system described in section 16.1 (16a–​c). (16)  Voice in Squliq Atayal, repeated from (1):      a. M-​aniq qulih qu Tali’. sv-​eat fish qu Tali ‘Tali eats fish.’

Subject Voice (SV)

     b. Niq-​ un na’  Tali’ qu qulih qasa. eat-​ov   gen  Tali   qu  fish     that ‘Tali ate the fish.’

Object Voice (OV)

     c. Niq-​ an na’  Tali’ qulih qu ngasal qasa. eat-​lv    gen  Tali    fish   qu  house    that ‘Tali eats fish in that house.’

Locative Voice (LV)

In Atayal, voice morphology references the XP that moves to the position marked by qu.18 Aside from this, however, we see the same case alternations as in Dinka. Subjects are unmarked in Subject Voice and genitive otherwise, while objects are always unmarked. We might also expect to find voice languages in which the object may also be in need of such a licensing strategy when not in pivot position. This appears to be the case in Tagalog. In Tagalog, any subject or object not cross-​referenced by voice morphology is marked genitive (17a–​c).19 17 

See Rezac (2011) for a technical implementation of the notion of repair, based on similar repairs in the context of violations of the Person-​Case Constraint. 18  The details of the analysis of Atayal is further complicated by the fact that there are cases where qu marks an argument other than the pivot. Such cases constitute an argument against viewing qu as a case marker. See Erlewine (to appear) for details. 19  A notable exception is a process of differential object marking that targets proper names and pronouns in the context of subject extraction (e.g. McFarland 1978).

 

386    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk (17) Tagalog voice marking (adapted from Guilfoyle et al. 1992): a. Sino  ang   bili    ng    damit para sa    bata? who  ang  sv.asp-​buy  gen  dress    for    obl  child ‘Who bought the dress for the child?’ Subject Voice (SV) b. Ano   ang    bili     ng    tao  para sa    bata? what  ang  ov.asp-​buy  gen  man  for    obl  child ‘What did the man buy for the child?’              Object Voice (OV) c. Sino ang i-​bili ng tao ng damit? what ang bv.asp-​buy gen man gen dress ‘Who was bought the dress (for) by the man?’       Benefactive Voice (BV) This fits well with the notion that genitive is available as an alternative case-​licensing strategy. Under this view, the only difference between Tagalog and Dinka or Atayal is that Tagalog also lacks a licensor for objects in non-​Object Voices, triggering genitive morphology there as well.20, 21 In fact, we can find evidence in Tagalog as well that the distribution of genitive is independent of voice morphology. As pointed out by McGinn (1988) and Schachter (1996), Tagalog, just like Dinka, has constructions without voice distinctions. In the recent perfective, no XP is marked with ang-​morphology and no voice morphology shows up on the verb (18).22 (18) Kabi-​bigay lang ng maestra ng libro sa bata. rec.perf-​give just gen teacher gen book dat child ‘The teacher just gave a book to the child.’         (Schachter 1996: p. 7) Importantly, both the subject and the object still receive genitive marking. This construction is then essentially analogous to the V1 constructions of Dinka, and shows that the Tagalog genitive, regardless of whether it appears on the subject or object, is also not dependent on voice. This section has shown that the mechanisms behind voice morphology can be dissociated from those behind case marking on subjects. This is evidence that there must be routes independent of ergativity that lead to a voice system. We suggested that the marked case on Non-Subject Voice subjects reflects the presence of case morphology directly merged to the nominal, following Halpert (2012) (see also Imanishi 2014), in order to provide a way of case-​licensing subjects outside of the voice system. 20 

As Aldridge (2012b) points out, genitive marking on Tagalog objects has interpretive consequences. We adopt Aldridge’s proposal that this is result of the interaction between inherent licensing and the application of existential closure. See Aldridge (2012b) for details. 21  This kind of system could also be a source of Austronesian languages in which objects in non-​Object Voices surface with accusative, if we allow for the “repair” case to have a different spell-​out inside the VP. 22  As discussed in McGinn (1988) and Schachter (1996), any XP in a recent perfective clause may undergo extraction. This seems to fit well with the view, implicit in our discussion, that voice morphology is in essence cosmetic and does not impose extraction restrictions.

 

Ergativity and Austronesian-type voice systems    387

16.3  Dissociating Voice and Extraction in Malay/​Indonesian As discussed in the previous section, one of the defining characteristics of Austronesian voice systems is that only the nominal cross-​referenced by voice morphology, the pivot, is eligible for extraction. Within an ergative analysis of Austronesian voice, this correlation is attributed to syntactic ergativity. Extraction of the subject is only possible in Subject Voice where the subject receives absolutive case. In this section, we examine wh-​movement in the Malay/​Indonesian languages Balinese and Bahasa Indonesia. We show that contrary to the expectations of an ergative analysis, non-​pivot subjects are not immobile. This indicates that syntactic ergativity is not a necessary condition of Austronesian-​type voice systems. In addition, there are surprising conditions on the realization of non-​pivot subjects in Balinese and Bahasa Indonesia, which suggests that they are subject to a strict head-​head adjacency requirement with the verb. We argue that this reflects an alternative licensing strategy, much like genitive case in Atayal, Dinka, and Tagalog.

16.3.1 The Malay/​Indonesian Voice System Balinese and Bahasa Indonesia, as well as other Malay/​Indonesian languages like Javanese and Madurese, have three voices: Subject Voice, Object Voice, and an Indo-​European-​ style passive voice. We will only be concerned with Subject and Object Voices here. In Balinese, Subject Voice is marked by a nasal prefix ng-​, whose form is phonologically conditioned by the initial segment of the verbal stem (19a). Object Voice is marked by the absence of this prefix (19b). The preverbal position is the canonical pivot position, to the left of all auxiliaries. Non-​pivot arguments are realized to the right of the verb. (19)

Balinese Subject Voice and Object Voice a. Tiang lakar numbas bawi-​ne punika. I will sv.buy pig-​def that ‘I will buy that pig.’

Subject Voice

b. Bawi-​ ne punika lakar tumbas tiang. pig-​def  that     will   ov.buy I ‘I will buy that pig.’

Object Voice

In Bahasa Indonesia, the prefix meng-​, whose form is also phonologically conditioned, marks Subject Voice (20a). As in Balinese, Object Voice is marked by the absence of this morphology (20b). Unlike in Balinese, non-​pivot subjects appear immediately to the left of the verb.

 

388    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk (20) Bahasa Indonesia Subject Voice and Object Voice (adapted from Cole et al. 2008): a. Tono tidak mem-​beli buku di toko buku. Tono neg sv-​buy book loc store book ‘Tono didn’t buy the book at the book store.’ Subject Voice b. Top ini    sudah saya beli. hat    this  perf     1sg     ov.buy ‘I bought this hat.’

Object Voice

Compared to the voice systems of Philippine and Formosan languages, the voice inventory of Malay/​Indonesian languages is diminished. These languages do not employ separate voice morphology for cross-​referencing benefactive, locative or recipient arguments. Rather, they utilize applicative marking that promotes such arguments to direct objects. As a result, any argument other than the external argument that serves as pivot is cross-​referenced by Object Voice.

16.3.2  Wh-​extraction in Indonesian Languages Balinese and Bahasa Indonesia, like other Austronesian-​type languages discussed in the previous section, show wh-​extraction asymmetries.23 When the verb bears Subject Voice, only the subject can be extracted. This restriction is easy to see in Bahasa Indonesia in which the marker yang co-​occurs with wh-​movement. Consider the dichotomy in (21a–​b): (21) Subject Voice restricts extraction to subjects (Cole and Hermon 2005): a. Siapa yang _​_​_​ mem-​beli buku di toko buku? who yang sv-​buy book loc store book ‘Who bought a book at the book store?’ b. *Apa     yang    Tono  mem-​beli  _​_​_​  di     toko   buku?    what  yang  Tono  sv-​ buy    loc  store book Intended: ‘What did Tono buy at the book store?’

23 

In addition to wh-​extraction, Malay/​Indonesian languages permit wh-​in situ constructions. In such constructions, either argument can be questioned regardless of voice, as illustrated for Balinese by the data in (ia–​d). (i) In situ wh-​questions show no voice asymmetries: a. Nyen  meli  montor anyar? who     sv.buy car     new b. Montor anyar beli car new ov.buy ‘Who bought a new car?’

nyen? who

 

Ergativity and Austronesian-type voice systems    389 Similar facts obtain in Balinese, although there is no analog of yang in the language. In Subject Voice, only the subject may be extracted (22a–​b).24 (22) Subject Voice blocks object extraction: a. Nyen ngalih bawi-​ne punika ditu who sv.seek pig-​def that there ‘Who looked for that pig there yesterday?’

ibi? yesterday

b. *Apa   ci      ngalih    ditu     ibi?   what  you  sv.seek there yesterday   Intended: ‘What did you look for there yesterday?’ In Object Voice clauses, only the object can be wh-​extracted. The dichotomy is again easily seen in Bahasa Indonesia (23a–​b), and also holds in Balinese (23a–​b).25 (23)

Object Voice restricts extraction to objects (Cole and Hermon 2005): a. Apa    yang  _​_​_ ​akan  kamu  lihat? what  yang      will    2sg    ov.see ‘What will you see?’ b. *Siapa yang buku ini akan _​_​_​ lihat?    who yang book this will ov.see    Intended: ‘Who will see this book?’

(24) Object Voice blocks subject extraction: a. Apa    beli    Nyoman? What  ov.buy Nyoman ‘What did Nyoman buy?’

c. Cicing  ngugut nyen? dog    sv.bite who d. Nyen gugut   cicing? who    ov.bite dog ‘Who did the dog bite?’ Similar observations are made for Bahasa Indonesia by Cole et al. (2008). These facts indicate that voice only restricts extraction, and not whether the non-​pivot argument can be questioned. 24  We could imagine analyzing (22a) as an instance of wh-​in situ (as in n. 23). However, more complex examples involving object scrambling show that wh-​subjects can undergo movement in the Subject Voice, as discussed by Arka (2004). 25  Here, we cannot actually be sure that (24a) involves wh-​extraction rather than wh-​in situ, since scrambling the subject before the verb in OV is independently ungrammatical (Artawa 1994; Clynes 1995; Wechsler and Arka 1998). Furthermore, given the relatively free word order of adverbial elements, we cannot be sure that the relative position of the wh-​phrase with respect to an adverbial indicates overt movement. Nevertheless, given the availability of overt wh-​movement in SV clauses, we take such movement to be possible here as well.

 

390    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk b. *Nyen montor anyar beli _​_​_​?    who car new ov.buy    Intended: ‘Who bought a new car?’ Based upon the extraction asymmetries illustrated in (23)-​(24), it is often reported that wh-​ extraction only targets the nominal cross-​ referenced by the verb (e.g. Wechsler and Arka 1998, Arka 2004 for Balinese; Cartier 1979, Hopper 1983, Verhaar 1988, Arka and Manning 1998, for Bahasa Indonesia). Like the extraction asymmetries in Philippine and Formosan Austronesian languages which display more articulated voice systems, these facts are amenable to an ergative analysis of Austronesian voice, in which these extraction restrictions are attributed to syntactic ergativity although Balinese and Bahasa Indonesia both lack overt case morphology. However, the extraction restriction in Balinese and Bahasa Indonesia is not as rigid as it is in Atayal or Dinka. It is possible to extract both subject and object arguments, so long as the appropriate Voice morphology is realized.26 Unlike the ungrammatical (22b), the object and subject may simultaneously be extracted in Balinese when the verb bears OV morphology (25a–​b). (25) OV permits non-​subject extraction: a. Apa     ci   alih   _​_​_​   ditu     ibi? what  you    ov.seek  there yesterday ‘What did you look for there yesterday?’ b. Buku nyen Nyoman lakar book which Nyoman will ‘Which book will Nyoman read?’

baca   ov.read

_​_​_​?

In (25), neither argument is in situ, because both are realized to the left of the verb.27 Such examples are problematic for an analysis of Austronesian voice which involves syntactic ergativity, because they show that non-​pivot arguments may undergo movement. Extraction of an object over an already extracted subject is also attested in topicalization. In an SV clause, both the canonical SVO word order and the marked OSV word order are well formed (26) (Arka 2004).

26 

See e.g. Cole et al. (2008) and Yanti (2010) for similar observations in related languages. Edith Aldridge (p.c.) asks whether the preverbal subjects in (25) could be clitics on the verb. They cannot be, for two reasons. First, Balinese does have a series of pronominal clitics but they follow the verb, as in (i) below from Wechsler and Arka (1998: p. 21). Second, these pronominal clitics are always hosted by the lexical verb, but the preverbal subjects can precede auxiliaries, as in (25b). 27 

(i) Buku-​ne jemak=a. book-​def ov.take=3 ‘(S)he took the book.’

 

Ergativity and Austronesian-type voice systems    391 (26) Non-​pivot object topicalization: a. Tiang  nunas     kopi-​ne    niki. 1   sv.take coffee-​ def  this ‘I took this coffee.’ b. Kopi-​ne niki tiang coffee-​def this 1 ‘This coffee, I took it.’

nunas. sv.take

Again, in (26), neither argument is in situ, because both are realized to the left of the verb, illustrating that non-​pivot arguments may undergo movement. A similar argument has been made from Bahasa Indonesia relative clauses (Chung 1976, 1978; Cole and Hermon 2005). As in matrix wh-​questions, relative clauses display extraction asymmetries. An object cannot be relativized if the predicate of the embedded clause bears SV morphology and the subject is in pivot position (27). (27)

Subject Voice restricts extraction to subjects: *[Buku [yang    Budi tidak akan mem-​baca]] sangat menarik.    [buku  [yang  Budi  neg     will   sv-​read  ]]  very      interesting Intended: ‘The book that Budi will not read is very interesting.’

However, if the verb appears in OV, object relativization can accompany subject fronting (28). (28) Object Voice permits multiple extraction: [Buku [yang Budi tidak akan baca   ]] [buku [yang Budi neg will ov.read]] ‘The book that Budi will not read is very interesting.’

sangat very

menarik. interesting

We can see that the subject has undergone movement, because it is realized to the left of auxiliaries and negation (cf. (20)a–​b). Like Balinese matrix wh-​questions, the behavior of Bahasa Indonesia relative clauses reveals that OV is dissociable from extraction. The observation that multiple arguments can be extracted in Malay/​Indonesian languages indicates that syntactic ergativity is not a necessary condition on the formation of voice systems. Voice does not determine which arguments are available for extraction, as would be expected under a strict implementation of syntactic ergativity. Rather, Voice seems to indicate which arguments have been extracted to which positions.28 This characterization is, like the Dinka data in section 16.2, amenable to a view that Voice morphology is extraction marking, as in wh-​agreement or case agreement approaches (e.g. 28  This second point is critical. Not all extraction is marked equally. Wh-​extraction of the object over the subject requires OV morphology, as in (25) and (28). Topicalization requires SV morphology (26). We suggest that the positions targeted by these movements are distinct. Movement to the former results in a change of Voice; movement to the latter does not. Chamorro wh-​agreement also displays a change in verbal morphology triggered by wh-​movement (Chung 1994).

 

392    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk Chung 1994; Richards 2000; Pearson 2001, 2005). If voice morphology is a by-​product of extraction, then multiple arguments should be able to extract, as is attested. Voice simply marks the results of the extraction process.29 What remains to be explained is why certain combinations of extracted arguments, like (24b), are unattested, which will be addressed in the following section.

16.3.3 The Behavior of Non-​Pivot Subjects The ergativity hypothesis faces further complications when considering restrictions on non-​pivot subjects. In this section, we show that there are constraints on what nominals are well formed as non-​pivot subjects in Balinese.30 Specifically, such nominals must display head-​head adjacency between the nominal head and verb (e.g. Baker 2014b; Levin 2015). We suggest that this represents an alternative method of subject licensing, in lieu of the genitive case in Atayal, Dinka, and Tagalog.31 In Balinese, in situ subjects do not appear in a dedicated case, as in Dinka or as in many other Austronesian languages. In fact, there is no overt case morphology in the language. Instead, in situ subjects are constrained in entirely different ways. These subjects can only be realized as pronouns (29a),32 proper names (29b), and indefinite NPs (29c). Definite descriptions are blocked from appearing in post-​verbal position (29d) (Wechsler and Arka 1998). (29) The form of Balinese in situ subjects: a. Be-​e daar ida. fish-​def ov.eat 3sg ‘(S)he ate the fish.’ b. Be-​e daar Nyoman. fish-​def ov.eat Nyoman ‘Nyoman ate the fish.’ c. Be-​e daar fish-​def ov.eat ‘A dog ate the fish.’

cicing. dog

d. *Be-​e daar cicing-​e. fish-​def ov.eat dog-​def ‘The dog ate the fish.’ 29 

This position is taken in Saddy (1991) and was later adopted by Cole and Hermon (1994, 1998) and Soh (1996). However see Aldridge (2008b) for an alternative proposal. 30  Similar facts hold of Bahasa Indonesia (Guilfoyle et al. 1992; Sneddon 1996). However, non-​pivot subjects are limited to pronouns and proper names. 31  See Baker (1988) for a specific implementation of how adjacency of a nominal to a verb, or more accurately the adjunction process which yields adjacency, i.e. Head Movement, can license that nominal in the absence of case assignment. 32  The pronominal element is a clitic in low register and a free pronoun in high register speech.

 

Ergativity and Austronesian-type voice systems    393 This is not an instance of differential subject marking, because it groups together indefinite subjects, pronouns, and proper names, to the exclusion of definite subjects, unlike any process of differential argument marking (e.g. Aissen 2003b). Levin (2015) argues instead that what the acceptable subjects in (29a–​d) have in common is that the highest nominal head (D0 in the case of (29a–​b) and N0 in the case of (29c)) is surface adjacent to the verb. In contrast, definite subjects are headed by the suffix -​e; the NP then intervenes between the verb and the highest nominal head. This reflects a more general pattern. Whenever linear adjacency of the verb and the highest nominal head is disrupted, ungrammaticality arises. Such intervention can be caused both by material outside of the nominal or by material inside of the nominal. Adverbs, which generally show freedom of placement in the clause (e.g. Wechsler and Arka 1998), cannot appear between the verb and OV subject (30). (30) NP-​external intervention: *Be-​e daar keras-​keras ida/​Nyoman/​cicing.  fish-​def ov.eat quickly 3sg/​Nyoman/​dog ‘(S)he/​Nyoman/​A dog ate the fish quickly.’ Similarly, while modifiers are canonically realized to the right of the nominal they modify, some can be realized to the left (31a). Modifier-​noun order is impossible with OV subjects, however, because the modifier intervenes between the verb and the subject (31b).33 (31) NP-​internal intervention: a. (Liu) cicing (liu) ngugut Nyoman many dog many sv.bite Nyoman ‘Many dogs bit Nyoman. b. Nyoman  gugut     (*liu)     cicing  (liu) Nyoman  ov.bite   many  dog      many ‘Many dogs bit Nyoman. We propose that this reflects a strategy of subject licensing under adjacency, following Levin (2015). In particular, we suggest that, as in Dinka, there is no case position for non-​pivot subjects, and so these subjects require an alternative method of licensing. This approach lets us capture the adjacency facts, but can also explain the limited set of nominals which can occur as non-​pivot subjects. Only those nominals in which the nominal head is immediately adjacent to the verb will be well formed. This restricts definite DPs to pronouns and proper names, because any other DP will have (overt) 33  See Baker (2014b) and references cited therein for similar observations regarding pseudo-​noun incorporation. In such constructions head-​head adjacency is also required between the nominal head of a caseless NP and the verb. Intervention effects arise when NP-​external and NP-​internal material disrupts the required adjacency.

 

394    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk NP-​material intervening between the D0 head, which appears to the right of the NP, and the verb. Furthermore, while NPs can be larger than N0, as in (31b), their head must be the leftmost element in the phrase. As in Dinka, there are then two possible means of licensing the subject in Balinese (and Bahasa Indonesia). Recall from our previous discussion of the genitive repair in Dinka that we take the pivot position to be the landing site of Ā-​movement, but also a case position. In Subject Voice, subjects receive case in this position. In Non-​Subject Voices, however, the subject needs to be licensed in a different way, because the pivot position is occupied. This is the role of licensing under adjacency. We believe that the general logic of Baker’s (1988, and subsequent work) account of licensing via Head Movement can be extended to these data. Specifically, we suggest, following Levin (2015), that adjunction of a nominal head to a verbal head renders it invisible to the Case Filter.34 Crucially, ungrammatical instances of multiple extraction can be captured under this view of licensing via adjunction. As noted, multiple extraction is possible when the object is extracted over an already extracted subject in the case of wh-​movement (25) and topicalization (26). However, wh-​movement (24b), repeated in (32a) or topicalization (32b) of a subject over an already extracted object is ungrammatical (32). (32) Subjects cannot extract across a fronted object: a. *Nyen montor anyar beli _​_​_​?  who car new ov.buy   Intended: ‘Who bought a new car?’ b. *Cicing  ia  uber   _​_​_​.  dog  3    ov.chase   Intended: ‘A dog, it chased him/​her.’ The ungrammaticality of these sentences can be captured as a failure to case-​license the subject. We propose that only the pivot position is a case-​position. All other positions in the left periphery are strict Ā-​positions, unable to case-​license arguments. If the subject is not extracted to pivot position, it must be licensed under adjunction in immediately postverbal position. Subsequent movement operations either cannot target the subject at all due to this requirement, or move the subject to a position in which licensing under adjunction is impossible, yielding ungrammaticality. Similar facts obtain in Austronesian languages with more voices, such as Malagasy. As discussed in Paul (1996) and Keenan (2000), non-​pivot subjects undergo a form of compounding with the initial verb referred to as N-​bonding,35 as the examples in (33a–​c) demonstrate. 34 

Levin (2015) assumes that adjunction can occur at various points in the derivation (e.g. Halle and Marantz 1993; Bobaljik 1995). See Embick and Noyer (2001) for an articulated account of the interaction of adjunction and derivational timing. In the case of Balinese, adjunction occurs very late in the derivation after linear order has been established, capturing the strict, linear head-​head adjacency requirement. 35  N-​bonding is also attested on possessors, again highlight the similarity of form shared by (non-​pivot) subjects and possessors attested in many of the languages discussed in this chapter. This may suggest that possessors and non-​pivot subjects in Malagasy both lack a licensor.

 

Ergativity and Austronesian-type voice systems    395 (33)

N-​bonding in Malagasy: a. Hitan-​ny lehilahy ny trano. ov.see-​det man det house ‘The house was seen by the man.’ b. Hitan-​ao     ny    trano. ov.see-​2sg  det  house ‘The house was seen by you.’ c. Hitan-​dRabe  ny    trano. ov.see-​Rabe     det  house ‘The house was seen by Rabe.’

Just as in Balinese, these subjects cannot be separated from the verb (34a–​b). (34) Malagasy non-​pivot subjects must be right-​adjacent to the verb (adapted from Pearson 2005): a. Nohanin-​ny     gidro    haingana ny  voankazo omaly. past.ov.eat-​det  lemur  quickly      det  fruit   yesterday ‘The lemur ate the fruit quickly yesterday.’ b. *Nohanin’  haingana  ny  gidro  ny voankazo omaly. We propose that these subjects are licensed in the same way as Balinese non-​pivot subjects, under adjunction, which yields head-​head adjacency. Unlike Balinese, non-​pivot subjects can be definite in Malagasy, as (33a) and (34a) show. Importantly, Malagasy differs from Balinese in that the D head is leftmost in the noun phrase and so is immediately adjacent to the verb.36

16.4  Concluding Remarks In this chapter, we have shown that there are languages with Austronesian-​type voice systems that do not display any ergativity. We introduced novel data from the Nilotic language Dinka, a non-​Austronesian language with a voice system, which has a consistent underlying nominative-​accusative alignment. In addition, we documented a dissociation between voice and extraction in Malay/​Indonesian languages, which argues against the idea that all voice systems display syntactic ergativity. On the basis of these facts, we conclude that ergativity cannot be the only route to a voice system. 36  Lisa Travis (p.c.) observes that in Malagasy Oblique Voice constructions, the non-​pivot subject displays head-​head adjacency with both unergative and unaccusative verbs. We assume that in both cases there is only one argument licensor. Burzio’s generalization holds. When a non-​core argument is extracted to pivot, the subject, regardless of base position, must be licensed under adjunction with the verb.

 

396    Michael Yoshitaka Erlewine, Theodore Levin, Coppe van Urk At the same time, there is an important generalization in the behavior of these different voice system languages:  non-​pivot subjects are treated differently from other arguments. In Atayal, Dinka, and Tagalog, non-​pivot subjects appear in genitive case. In Balinese and Malagasy, non-​pivot subjects require adjacency with the verb. We can give a unified characterization to these two types of behaviors through a requirement that non-​pivot subjects require a special form of licensing (Case). The two strategies observed are simply two different ways of licensing the non-​pivot subject. This licensing requirement is shared between voice system languages which are more amenable to an analysis as morphologically ergative and those which are not. A remaining open question is why and how languages differ in the availability of these two repairs: a last-​resort genitive case and licensing by adjacency. One final issue we would like to discuss relates to the analysis of voice morphology. The dissociations between voice and extraction we observed in Dinka and Malay/​ Indonesian support a treatment of voice as extraction marking (e.g. Chung 1994; Richards 2000; Pearson 2001; Rackowski 2002). In ongoing work, we develop a theory for Austronesian-​type voice systems as extraction marking, which also explains the need for exceptional licensing of non-​pivot subjects.

Acknowledgments The work here is part of our ongoing work on the syntax of Austronesian-​type voice systems within Austronesian and beyond; see also Erlewine, Levin, and van Urk (2015). We thank Julie Legate, David Pesetsky, Masha Polinsky, Norvin Richards, and audiences at the 21st meeting of the Austronesian Formal Linguistics Association, the 2015 LSA, and at McGill University for discussion and comments. We especially thank Edith Aldridge, Mark Baker, and Lisa Travis for extensive comments on this manuscript. Part of this work is supported by an NSF Dissertation Improvement Grant for the third author, BCS-​1440427. Errors are ours.

Abbreviations abs, absolutive; ap, antipassive; asp, aspect; bv, Benefactive Voice; def, definite; erg, ergative; fut, future; gen, genitive; hab, habitual; inch, inchoative; intrans, intransitive; iv, Instrumental Voice; loc, locative; lv, Locative Voice; neg, negation; nf, nonfinite; obl, oblique; oblv, Oblique Voice; ov, Object Voice; p, preposition; past/​pst, past; prf/​pfct, perfect; rel, relative clause marker; sv, Subject Voice; trans, transitive.

 

Chapter 17

On the morpho syntac t i c re f l exes of in format i on structure i n t h e e rgative pat te rni ng of Inu it l ang uag e Alana Johns and Ivona Kučerová

17.1 Introduction The Inuit language1 is often characterized as an ergative language (Bok-​Bennema 1991; Johns 1992, 2000, 2006; Manning 1996; Compton, Chapter  34, this volume; among others). Interestingly, the Inuit language exhibits a case assignment variability which, unlike traditional split ergativity, does not affect argument alignment, but instead concerns which—​and how many—​arguments trigger ϕ-​feature agreement on finite verb. This chapter asks what is the nature of the relevant grammatical property, and how does it relate to the agreement properties that result. This case variability has not gone unnoticed in the current literature. Some proposals attribute the split to aspect, for instance, Bittner (1987), Spreng (2006, 2010, 2012), Clarke (2009), while others to information structure, or scope (Kalmár 1979; Bok-Bennema

1  The Inuit language is a branch of the Eskimo languages of the Eskimo-​Aleut language family. There are four major dialect groupings: Inupiaq, Western, Eastern, and Greenlandic. The majority of linguistic work has been on the Kalallisut (West Greenlandic) dialect. Throughout this chapter we will refer to the Inuit language rather than use the term Inuit, as some do. Inuit is literally ‘people’ inu(k)-​it person-​plural. We follow the Nunavut government in our usage.

 

398    Alana Johns and Ivona Kučerová 1991; Bittner 1994; Manga 1996; Bittner and Hale 1996b; Hallman 2008; Berge 2011; among others). All these proposals point out some important property of the observed case variability. Yet, it is not always clear how the proposals derive the morphosyntactic properties of the Inuit language, neither do they provide insight into how these seemingly disparate properties relate to each other. We argue—​closely following the proposal of Berge (2011)—​that the ergative clause structure of the Inuit language is conditioned by information structure properties, more precisely by its topic-​comment properties. We propose a formal model in which this morphosyntactic make-​up directly follows from this information structure trigger. Furthermore, we will show that not only does the model account for the morphosyntactic properties of the Inuit language, namely, split case and agreement properties, but it also subsumes other relevant properties discussed in the literature, i.e. scope properties of objects and aspect. These will be seen to be byproducts of the information structure underlying the ergative split. Before presenting our argument, we will first provide evidence that the difference between so-called single (also called intransitive) and double (also called transitive) agreement is a difference of ϕ-​agree vs cliticization (Johns to appear pace Compton 2014). We will link this conclusion to the second core observation (see Berge 2011), namely, that absolutive objects must be topics.2 More precisely, we will show that they must be (aboutness) topics in the sense of Reinhart (1981), which is a narrower notion of topics than that used in functionalist literature and does not necessarily entail discourse notions such as recency of mention or persistence in subsequent discourse.3 Crucially, this type of topic—​which we will call sentential topic—​is strictly realized at a sentential level and has a dedicated syntactic representation. This move will have several immediate consequences: Since topics have been associated with cliticization (Dočekal and Kallulli 2012), the analysis that double agreement is formed by cliticization, along with the independent analysis of the corresponding argument as sentential topic, provides an explanation for the clitic nature of double agreement (and for double agreement itself). Furthermore, since sentential topics must be at the edge of a phase,4 we will argue for a VP shell-​like (applicative) structure of Inuit ergative clauses as a necessary precondition for objects being marked as topics (Basilico 2003). With this structural distinction in place, we will show that case marking straightforwardly follows from locality and the morphological case hierarchy (Marantz 1991), in

2  Berge argues that absolutive subjects can also be topics but we focus here on absolutive objects which are uniquely associated with double or transitive agreement. Related ideas, that some sort of discourse familiarity is relevant, are found in Kalmár (1979) and Manga (1996). 3  Berge (2011) refers to this type of topic as local topics and distinguishes them from global topics, which are persistent across a wider domain of discourse. 4  Our analysis is framed within the phase theory of Minimalist program (Chomsky 2001, 2008, 2013, among others). A phase in this theory refers to a syntactic derivational unit which is in a technical sense syntactically and semantically complete. The notion thus roughly corresponds to the notion of cycle in the earlier stages of generative grammar.

 

Morphosyntactic reflexes of information structure in Inuit    399 a way familiar from case alternations in English double-​object constructions. Finally, since topics create the illusion of wide scope (Endriss 2009), and have ‘maximization’ properties, this allows us to explain the scope observations and the appearance of an aspect-​based split (Borik 2002, Filip and Rothstein 2006, Filip 2008, Ramchand 2008). For reasons of space, portions of our analysis will only be briefly outlined. We hope, however, that the overall structure of the argument will emerge with sufficient clarity. Aside from this language-​specific exploration, our proposal contributes to a more general debate on the nature of cross-​linguistic differences in argument alignment. We argue that even though the underlying factors (e.g., agentivity, topic/​comment, given/​ new) may vary among languages, once we isolate the critical alignment factor, the rest of the system (locality, movement properties, case assignment properties) may be stated in universal terms (see also in this volume Baker and Bobajlik (Chapter 5), Du Bois (Chapter 2), Cheliah (Chapter 38), and Coon and Preminger (Chapter 10), for a similar position).

17.2  The Core Facts 17.2.1 Double Agreement as Marker of Ergativity In the Inuit language, a transitive verb may appear in two distinct patterns: the so-​called ergative and the so-​called antipassive.5 They differ not only in their case assignment but also in their agreement properties. As for the case properties, the subject of antipassive is morphologically marked as absolutive, while the object is marked with an oblique case (mik). In the ergative pattern, the subject is marked with the so called relative case6—​a term that corresponds to ergative in the description of other ergative languages—​and the object as absolutive. Note that while the relative case has an overt morphological realization, absolutive is zero marked. In this chapter we will use the term relative when referring to the case, and ergative when referring to the ergative clause pattern (including case and agreement of the transitive verb). As for finite agreement, the ergative pattern is associated with an agreement pattern which we will call here transitive double agreement. Unlike its intransitive counterpart, given in (1a), the transitive verb in (1b) inflects for person and number of two arguments, the subject and the object.

5 

By antipassive we refer both to transitive verbs which require an antipassive morpheme to show intransitive agreement and to transitive verbs that do not require such a morpheme to show intransitive agreement. 6  Formally, the morphological form of relative case is identical to possessive case found in the nominal domain.

 

400    Alana Johns and Ivona Kučerová (1)

Baker Lake (Johns 1992, 58-​59) a. angut    ani-​ juq man-​abs.s  walk-​INTR.PArt.3s ‘The man is walking.’ b. arna-​up angut kuni-​ga-​a woman-​erg.s man.abs.s kiss-​tr.part-​3s/​3s ‘The woman kissed the man.’

(single)

(transitive double)

The antipassive pattern (Spreng 2012), sometimes also called semitransitive (Fortescue 1984), triggers an agreement pattern, which we could call transitive single agreement.7 Abstracting away from other properties of antipassives, the fact that interests us here is that the verb agrees only with the subject. This is seen in example (2) from Labrador Inuttitut. (2) Margarita Kuinatsa-​i-​juk Ritsati-​mik Margarita.abs.s tickle-​ap-​intr.part.3s Richard-​mod.s ‘Margarita is tickling Richard.’ (transitive singular) Labrador (Johns 2001b, 211) The description of the system is slightly complicated by the fact that the Inuit language is a pro-​drop language. As a result, in many instances there is no overt DP, i.e. there is no overt case marking. In addition, the plural of the relative case is homophonous with the plural of the absolutive. Consequently case distinctions are less overt than agreement distinctions, which we contend are central to transitive constructions. Furthermore, there appears to be a dialect difference between Eastern and Western dialects of the Inuit language in Canada: while in Western dialects, objects may be overt DPs in either of the two transitive constructions (double or single agreement), in Eastern dialects, the object DP of a transitive double agreement construction is not usually overt.8 Finally, the ergative pattern is significantly less frequent in Eastern dialects than Western dialects. While we will have nothing to say about the frequency effect, a closer look at the morphosyntactic properties of the double pattern in section 17.2.2. will shed light on why absolutive object DPs might be dropped in one dialect group but not the other.

17.2.2 Double Agreement as Cliticization The intuition behind this and previous accounts is that while an object in the ergative pattern is a core argument and therefore able to relate to the verb directly, an object in 7  We will continue to use the term antipassive to refer to constructions with subject/​object but only single agreement. 8  This distinction is not absolute and requires more research. See (Johns to appear) for more details.

 

Morphosyntactic reflexes of information structure in Inuit    401 the antipassive pattern is an oblique and must be outside of the domain of agreement. The question that interests us here is what structural representation underlies these morphosyntactic configurations. In order to answer this question we start by investigating the morphosyntactic properties of the agreement properties. Johns (to appear) suggests that objects in the ergative pattern must be salient from the previous discourse. The observation is based on data such as (3)  from Labrador Inuttitut. Here we see that the first occurrence of Kajotta ‘cup’ as an object appears in the antipassive pattern, (3a) with single agreement. It is only in (3c) that the same object, this time not overtly realized, triggers double agreement on the verb. (3) The discourse requirement on double agreement (Labrador): a. John kata-​i-​juk Kajotta-​mik John.abs.s drop-ap-​intr.part.3s cup-​mod.s b. amma-​lu Kajottak siKumi-​mmat, also-​and cup.abs.s break-​caus.3s c. âkKi-​sima-​janga nipi-​ti-​guti-​mmut fix-​perf.-​tr.part. 3s/​3s adhere-​cause-​instrument-​instr.s ‘John dropped the cup and then when the cup broke, he fixed it with the glue.’ Johns (2013, to appear) uses this observation to argue that the transitive double agreement is not a result of ϕ-​feature agree.9 Instead she proposes that object inflection is a clitic, following recent reanalyses of object agreement (Preminger 2009; Nevins 2011; Kramer 2014a). Since clitics require some form of saliency—​her argument goes—​the information structure restriction demonstrated in (3) immediately follows. Furthermore, an analysis of Inuit object inflection as clitic immediately affords a perspective on the dialect difference introduced in 17.2.1 as a familiar difference in clitic doubling (Anagnostopoulou 2006). As Anagnostopoulou shows, languages differ in whether or not the clitic’s full DP associate is overt. In other words, while all clitic languages morphologically realize the clitic, only some realize simultaneously both the clitic and the doubled DP. Thus if we analyze the object agreement as a clitic, we can explain the restriction on overtness of object DPs as a dialect variation in clitic doubling: while Western dialects allow clitic doubling and hence double agreement with overt object DPs, clitic doubling in Eastern dialects is severely limited, so we expect to see double agreement only when there is no overt DP.10 We follow Johns’s proposal and provide additional evidence that agreement with the object of the ergative construction contains a clitic (contra Compton 2014).11 Our 9  A combination of pro and ϕ-​agree analysis was proposed by Merchant 2011 for Aleut, a distant relative of the Inuit language. 10  Interestingly, this entails that Labrador Inuttitut is no longer ergative by definition, since transitive clauses and ergative patterning are not equivalent. 11  Though in section 17.2.3 we will disagree with Johns (to appear) on her information-​structure characterization of objects in the ergative pattern.

 

402    Alana Johns and Ivona Kučerová argument is based on the discussion of ϕ-​agree versus cliticization in Nevins (2011). Nevins argues that clitics, unlike ϕ-​agree, are tense invariant, display gaps in morphological paradigms, are reminiscent of PCC effects, and have omnivorous number. As we will see, even though there are non-​trivial issues in determining whether double agreement is tense-​invariant, the double agreement pattern in the Inuit language displays the other two properties characteristic of cliticization. An additional argument will come from the domain of semantic interpretation (Dočekal and Kallulli 2012). Let us start with the question of tense invariance. As Compton (2014) points out, even though there is no tense-​sensitivity in the Inuit language, there is a morphophonological sensitivity to mood. For example, transitive participial (declarative) mood inflection for 2s/​1s is jar-​ma, while transitive interrogative equivalent is -​vi-​nga. Both ma/​-​nga indicate first person singular object. While for Compton this sensitivity indicates ϕ-​agree, we find problems with this argument. First, tense invariance is not a direct by-​product of category (pronoun vs agreement). Nevins (2011) indeed argues that clitichood needs to be based on syntactic, not on purely morphophonological evidence. For the conclusion about tense invariance to hold, it must be a consequence of which type of element is in closer proximity to T. In languages discussed by Nevins, object clitics which undergo object shift tuck in under the subject in spec,vP. They are higher, but still not as close to T as subjects are. As Compton (2014) points out, however, Mood (not T) is the major clausal category in the Inuit language. Furthermore, as Compton and Pittman (2010) argue, a word in the Inuit language is a phase, bounded by mood at the phase edge. Arguments evacuate before the final phase is completed. Inflectional arguments will therefore adjoin to mood. Assuming Distributed Morphology (Halle and Marantz 1993), at the point at which the form of the object pronoun undergoes morphological insertion, it is adjacent to the mood morpheme. It is therefore not surprising if there is some morphophonological sensitivity based on variance between mood morphemes. Because of these complexities, tense variance cannot be used as a determining property for the status of double agreement.12,13

12 

There are additional issues; for example while double agreement of independent moods resembles possessive agreement found on possessums, double agreement of dependent moods resembles possessive agreement found when the possessor is itself possessed, as in (26) from the Labrador dialect. (26) anêna-​ma mother-​poss.1s/​3s ‘my mother’s hats’

nasa-​ngit hat-​3s/​p

Finally subject agreement varies between single and double agreement, so it is not inconceivable that this could affect object form. 13  The reader may have noticed that double agreement is not always clearly segmentable into subject and object at the morphophonological level, unlike some other languages where object agreement has claimed to be clitic (Kramer 2014a). We do not view this as an issue, given that we rely on syntactic evidence, following Nevins (2011). Merchant (2011), in his analysis of Aleut, describes the portmanteau effects of clitics, citing similar effects in Basque (Arregi and Nevins 2008).

 

Morphosyntactic reflexes of information structure in Inuit    403 Let us now turn to the question of Person Case Constraint (PCC) effects, i.e. a restriction on 1/​2 person objects in certain environments (Béjar and Rezac 2003; Rezac 2008; among others). The Inuit language displays systematic person constraints within object marking. This is entirely unexpected if the object marker were a genuine instantiation of ϕ-​agree. As observed in Johns (1996), the Labrador dialect indeed displays a restriction on the person marking of the object in the ergative pattern. While 1>3 may be found in both participial and indicative moods, (4), 3>1 is possible only in the indicative mood, as in (5).14 (4) a. nigi-​jaga eat-​TR.PArt.1s/​3s b. nigi-​vaga eat-​TR.ind.1s/​3s ‘I ate it.’ (5) a. *taku-​jânga see-​tr.part.3s/​1s b. taku-​vânga see-​tr.ind.3s/​1s ‘He saw me.’ The example in (6), from the South Baffin dialect shows a similar restriction but with a different result. The first example shows a 1>3 transitive participial mood verb form with double agreement. In the second example, however, the *1>3 prohibition forces a single agreement verb form with the object as an independent pronoun (in fact an antipassive construction). (6) a. mali-​langa-​si-​jara follow-​going.to-​incept.-​tr.part.1s/​3s ‘I am going to follow him.’ b. Jaani uvannit ikaju-​ruma-​nngit-​tuq John.abs 1s.pro-​mod help-​want-​neg.-​intr.part.3s ‘John does not want to help me.’ Compton (2014) argues against a clitic analysis for verbs based on his claim that possessive agreement, which is closely related to transitive double agreement, shows default agreement in the South Baffin dialect in oblique case environments. Default agreement

14  The indicative mood and the participial mood are very close semantically, with the indicative mood usually adding a sense of vividness. Where the participial mood is grammatically prohibited, the indicative mood is used in its place, but without the vividness.

 

404    Alana Johns and Ivona Kučerová is a hallmark of ϕ-​agree (Preminger 2009), so if we accept that possessive agreement and transitive verb agreement are closely related, this would be an argument against transitive verb agreement having a clitic. The crucial data involving possessives is in (7). (7) a. iksivauta-​ra chair-​1s.poss/​s ‘my chair’ b. uvanga  iksivauta-​nga-​nit 1s.pro   chair-​3poss-​ablative ‘from my chair’ In (7a) we see possessive agreement on the head nominal that indicates both features of the possessor and number of the nominal. In (7b) where the nominal is in oblique case, we see a periphrastic construction, but also a third person possessor marking on the head nominal. Compton claims that this is a default possessor marking, leading to his conclusion that possessor/​transitive double agreement cannot involve clitics. While the data are interesting, we do not believe it refutes an analysis of transitive verb agreement as involving clitics. Note that the morphology shown in (7b) does not appear on verbs. Furthermore, Yuan (2014) provides critical data that shows that (7b) is an instantiation of a PCC repair, a property consistent with clitics, not agreement (Rezac 2008). The default agreement analysis makes a clear prediction. If the third person possessive marking in (7b) indeed is default agreement, the number marking on the nominal head should always be third person singular. Yuan (2014) shows that the third person marking on the head nominal can be plural as well. The examples in (8), from Yuan (2014), show an absolutive possessed noun on the left, with an oblique case marked possessive on the right. Only the latter has the periphrastic form with possessor inflection ngit-​ on the possessum, indicating not only that it is a possessed form, but that it is a plural possessed form. (8) qimmi-​kka → uvanga qimmi-​ngin-​nut dog-​1s/​3p 1s.pro dog-​3p-​allative ‘my dogs’ ‘to my dogs’ Consequently, Yuan argues that this is not default agreement but simply the reflex of the possessum features. As a result, Yuan characterizes this construction as a PCC effect where the oblique case blocks co-​occurrence with first/​second person objects. From this perspective the presence of a first person independent pronoun on the right in (8) can be seen as a repair, a property consistent with clitics, not agreement (Rezac 2008). If possessive forms are relevant evidence to clitics in transitive double agreement, then this evidence supports a clitic analysis of transitive agreement. In summary, there is ample evidence of PCC effects involving object marking (and also possessor marking), leading to the conclusion they are clitics.

 

Morphosyntactic reflexes of information structure in Inuit    405 Let us turn to the third syntactic property associated with object clitic marking in Nevins (2011), namely omnivorous number, where the same plural marking may denote either the subject or the object as plural. While omnivorous number is clear in Aleut (Merchant 2011), it is not as clear in the Inuit language. However consider the partial transitive indicative paradigm from Harper (1974) for North Baffin dialects in (9). (9)

Subject Object 3s 3d 3p

3s

3d

3p

vaa or vanga vaangik vait or vangik vangak

vaangik

vangik

vaat or vangat vaangik vait or vangit

In (9) we see that the regular third plural marker for nominals -​(i)t marks not only the 3plural of the object in 3/​3p -​vait (singular subject) but also the plural of the 3plural agent in 3p/​3s -​vaat (singular object). Thus the Inuit language shows some evidence of omnivorous number, supporting the clitic analysis. Our final argument supporting the clitic nature of object marking in the Inuit language is that its presence is associated with a special meaning. This property has been acknowledged as a diagnostic of clitichood by Dočekal and Kallulli (2012), Anagnostopoulou (2006), Kramer (2014a), among others. The presence of ϕ-​agree never relates to special meaning. For example in Albanian, as Dočekal and Kallulli (2012, p. 117) show, the object of a verb requires clitic doubling in contexts where a topical interpretation exists, as in (10)–​(11). At the same time, clitic doubling is prohibited in contexts where the object of the verb is not topic, as in (12). Note that this example involves existential ‘have.’ (10) A: Who read the book? B:

Ana *(e) lexoi librin. Anna cl.acc.3s read  book.the

(11) A: What did Ana do with/​to the book? B: Ana *(e) lexoi librin. Anna cl.acc.3s read  book.the (12) (*I) kishte minj n. gjith. apartamentin. cl.acc.3p had mice in all apartment.the ‘There were mice all over the apartment.’ The data from the Labrador dialect that parallel Dočekal and Kallulli’s examples (10)–​(12) demonstrate the topic status of the absolutive objects. As we can see in (13), if the object is a topic, the agreement pattern is ergative (double agreement) and there is no overt DP.15 15 

Note that names often do not reflect relative case, but this is not important, as our claim focusses on the single/​double agreement distinction.

 

406    Alana Johns and Ivona Kučerová (13)

A: kina iKalu-​mmik iga-​sima-​jong? who.abs fish-​mod.s cook-​perf-​intr.part.interr.3s ‘Who cooked the fish?’ B: Peta iga-​sima-​janga Peter.abs cook-​perf-​tr.part.interr.3s/​s ‘Peter cooked it.’

At the same time we see that definite DPs which are not topics are found in the antipassive (single agreement), as in (14). (14) A: Su-​sima-​jong what.do-​perf-​intr.part.interr.3s ‘What did Sally do?’

Sally? Sally

B: Sally atua-​tsi-​sima-​juk alla-​mik. Sally.abs read-​ap-​perf-​intr.part.3s book-​mod.s ‘Sally read the book.’ Since the Labrador dialect does not favor clitic doubling, a topic with an overt DP will also be in the antipassive (single agreement), as can be seen in (15). (15)

A: kina Kuki-​nni-​jong nanu-​mming? who.abs shoot-​ap-​intr.part.interr.3s polar.bear-​mod.s.interr ‘Who shot the polar bear?’ B: Davide Kuki-​nni-​juk nanu-​mmik David.abs shoot-​ap-​intr.part.3s polar.bear-​mod.s ‘David shot the polar bear.’

Kramer (2014a) shows that clitic doubling in Amharic is related to special meaning as well. It can be found with wh-​words but only if d-​linked, and is also found as a sort of emphasis. Kramer (2014a, p. 624) suggests that topichood may be an underlying factor, but acknowledges that more fieldwork is needed. The Inuit ergative construction is also known to have a special meaning. The exact nature of the meaning difference and the structure underlying varies. It has been attributed to specific (Manga 1996), wide scope (Bittner 1987, 1994), given (Kalmár 1979, Johns to appear), and topic (Berge 2011). The exact semantic nature of the double agreement pattern will be discussed in section 17.2.3 but for now it suffices to say that, that unlike objects of transitive verbs which do not bear special nominal interpretation as a result of being transitive objects, the nominal in absolutive case in the ergative pattern in the Inuit language has a distinct interpretation from that it would get otherwise. We conclude that the semantic properties of the double object agreement yet again point in the direction of a clitic, not ϕ-​agree.

 

Morphosyntactic reflexes of information structure in Inuit    407 To summarize, we have shown that the object markers in the Inuit language clearly display all the characteristics of object clitics. This conclusion supports Nevins’ (2011, p. 967) tentative contention, based on a suggestion by Woolford, that in all languages displaying agreement with both subject and object, object agreement should be reanalyzed as a pronominal clitic.

17.2.3 The Case for Topics The fact that objects in the ergative and antipassive pattern are not semantically equal has not gone unnoticed in the literature on the Inuit language and other ergative languages. In this section we follow Berge (2011) and her analysis of Western Greenlandic and argue that objects in the ergative pattern are best characterized as topics. More precisely, we will argue that absolutive objects which Berge refers to as locals topics correspond to (aboutness) topics in the sense of Reinhart (1981). We will call these topics sentential topics in order to indicate that unlike their broader counterpart, they have a designated clause-​bound syntactic representation. Thus our notion of sentential topic will be strictly used as a label for the part of a structure the sentence is about, instead of the broader notion of topic often used in the functionalist literature (global topics or non-​topics in Berge’s terminology).16 In contrast, objects of antipassive are best characterized as anti-​topics (non-​topics in Dočekal and Kallulli 2012), i.e. backgrounded elements.17 In section 17.3 we will show that not only is this characterization empirically more accurate than other previously suggested characterizations but it also straightforwardly derives the core morphosyntactic properties of the Inuit case and agreement system we investigate here. Bittner (1987) and following work observed that absolutive objects in the Inuit language, unlike their oblique counterparts, take wide scope. The example in (16), originally from Bittner (1987, (40)–​(41)), demonstrates this through the relation between an indefinite noun phrase ‘kayak’ and sentential negation. In the ergative pattern, as in (16a), the absolutive object scopes above negation, which results in a specific reading. In contrast, whether definite or indefinite, its oblique counterpart in (16b) must be interpreted within the scope of negation (‘He/​she doesn’t use a kayak anymore’). While this fact is undisputed, it is not clear what underlying grammatical property is responsible for the scopal interaction. Note the contrast could be a result of a genuine wide scope, but equally it might have arisen from another semantic factor, 16 

Berge’s notion of local topic is related to the notion of theme in the Prague school terminology. The notion of global topic, which Berge uses to characterize objects in the ergative pattern, is closer to Schwarzschild (1999)’s notion of givenness. As for aboutness/​sentential topics, we use Reinhart (1981)’s formalization—​more precisely we will adopt the formal implementation of Endriss (2009)—​because it is easier to implement within the generative framework we assume here. We refer the interested reader to Hajičová et al. (1998) for an attempt to reconcile these two distinct generative traditions in semantic terms. As for the information status of subjects, we refer the reader to Berge (2011)’s description. 17  Similar claims have been made by other authors for other ergative languages, mostly in the functionalist tradition (Dixon 1972; Blake 1976; Mallinson and Blake 1981; Du Bois 1987b; Authier and Haude 2012; among others).

 

408    Alana Johns and Ivona Kučerová e.g. specificity, definiteness, or perhaps givenness modeled as anaphoricity (Kalmár 1979; Manga 1996; Johns to appear).18 (16) a. qajaq atur-​unnaar-​paa kayak.abs use-​no_​longer-​ind.3s/​3s ∃ [x is a kayak & ¬(he uses x)] b. qaanna-​mik atur-​unnaar-​puq kayak.mod.s use-​no_​longer-​ind.3s ¬∃x[x is a kayak & he uses x]

abs: ∃ > ¬

ins: ¬ > ∃ (Hallman 2008, 10)

Let us have a closer look at the hypothesis that the difference between absolutive and oblique objects is indeed based on a genuine scope interaction. If the semantic contribution of the oblique case is narrow scope, then we expect that non-​scoping elements such as proper names, personal pronouns, or rigid designators, such as ‘my father,’ should always appear only in absolutive case as they obligatorily scope over the type of scope operators Bittner bases her argument on.19 However, as these Canadian Inuit examples in (17) from Hallman (2008) demonstrate, this prediction is not borne out. Thus the fact that non-​scoping items may appear with both case markings is entirely unexpected under a genuine scope hypothesis. The explanation of the appearance of wide scope must lie elsewhere. (17)

a. qimmiq taiviti-​mit kii-​si-​qqau-​juq dog.abs David.mod.s bite-​ap-​past-​INTR.part.3s ‘A dog bit David.’ b. qimmiq uvannit kii-​si-​qqau-​juq dog.abs me.mod.s bite.-​ap-​past-​INTR.part.3s ‘A dog bit me.’ c. qimmiq ataata-​nit kii-​si-​qqau-​juq dog.abs father-​my.mod.s bite-​ap-​INTR.part-​ind.3s ‘A dog bit my father.’

(Hallman 2008, 51)

18  Hallman (2008) proposes that the transitive object is either anaphoric or introduces new arguments because an NP in absolutive case (a) has existential assertion, rather than supposition, and (b) a uniqueness presupposition. 19  The logic of our argument crucially relies on Fox (2000)’s notion of scope economy. (See also Fox 1995; Reinhart 1995, 2006.) According to Fox, nominal expressions can undergo a scope-​taking operation, only if the scope-​taking structural change (be it quantifier raising or something else) yields an interpretation that would not be available otherwise. Since here, the interpretation associated with the wide scope is already available in situ, there is no reason for a structural change to take place. One could argue that there is another feature present in the structure that is in and of itself independent of wide scope; however, we prefer the simpler analysis presented in the main text.

 

Morphosyntactic reflexes of information structure in Inuit    409 Could definiteness or givenness, modeled as discourse anaphoricity (Schwarzschild 1999), be a possible explanation, as suggested in Kalmár (1979), Manga (1996), Johns (to appear)? Interestingly, as Hallman (2008) points out, although absolutive case often reflects definite or anaphoric meaning, absolutive object DPs can also appear in out of the blue contexts, as in (18), an example from Kalaallisut (Bittner 1987, (12)), taken from the Gospel according to Matthew in the New Testament. (18) figiqussuar-​lu aqqusirnup sania-​niit-​tuq taku-​gamiuk fig.tree.abs-​and road.erg its.side-​loc-​part.3s see-​when.3r/​3s ‘and as he saw a fig tree standing at the side of the road’ The ‘figtree’ in absolutive case is a new entity in the discourse, assuming the English translation is a guide. This is unexpected if absolutive objects have to be discourse salient. In other words, notions such as given/​new, discourse salient do not seem to provide an accurate description of the facts. Neither does the distinction between definiteness and indefiniteness, at least not in the sense assumed for English (see the discussion in Hallman 2008). Furthermore, as Compton (Chapter 34, this volume) argues, oblique case (MOD) cannot be equated with indefinite properties since it is found on names of objects in some dialects and it is found on names under some circumstance in all dialects. The examples given in (17) make the same point. We conclude that neither a genuine scope analysis, nor ones based on definiteness or discourse saliency are empirically accurate. Instead we propose that absolutive objects are sentential topics, following the core proposal of Berge (2011). However, before we demonstrate the validity of this hypothesis, we find it useful to first outline what we mean by sentential topics, as there has been a significant amount of confusion about the semantic contribution of topics and how they relate to notions such as givenness or discourse saliency. As pointed out by Molnár (1993), there are in principle three distinct levels of information packaging (the following formulation is based on Endriss 2009). (i) a level that distinguishes what is said (the comment) from what this comment is about (the topic of the utterance); (ii) a level that differentiates between things that are new to the hearer (the rheme) and things that are already known (the theme); and (iii) a level where the utterance is divided into what is important or relevant from the speaker’s viewpoint, i.e. the focus, and what is not as important and thus constitutes the background. We concur with Reinhart (1981), Molnár (1993), and Endriss (2009), among others, that a sentential topic is what a sentence is about.20 Thus this is a notion that strictly operates at the sentential level and may be directly encoded in the morphosyntactic representation. From the semantic point of view, this means that for something to be a sentential topic, it needs to be associated with a referential address which in and of itself is

20 

The distinction between topic and comment thus syntactically corresponds to the distinction between thetic and categorical statements (Kuroda 1972; Ladusaw 2000; Basilico 2003; among others). We will return to this distinction in section 17.3, where we discuss the syntax of topics.

 

410    Alana Johns and Ivona Kučerová associated with the common ground. The common ground association may come about in two distinct ways: either the item is already in the common ground, or its existence needs first to be asserted, and then added to the common ground. The final interpretation results once predication takes place over the anchored item.21 A side point, which is going to be important in section 17.3 is that when the address is being established, a ‘maximized’ interpretation is necessary (Endriss 2009). The crucial point for us is that topics must be associated with a referential address, i.e. they are modeled as an address for the context update. The consequence of this is that topical material cannot be interpreted in the predicative part of the sentence. As Endriss (2009) carefully explores, this results in an appearance of wide scope and/​or specificity, often associated with topics. Note that the notion of referential address often coincides with familiarity/​discourse saliency but it is not identical. Thus this narrow notion of topicality is compatible with indefinites and other non-​salient elements such as quantifiers. In contrast, the broader notion of topics, often used in the functionalist literature, requires topics to be recently mentioned, persistent in subsequent discourse, etc., and as a result is not a good fit for the data we investigate here. We are now in a position to come back to our Inuit language data. Recall that we have seen that absolutive objects take wide scope (although oblique objects may take wide scope as well), and we have also seen that although absolutive objects tend to be definite and discourse salient, they can be novel as well, as in the example in (18). While this example was problematic for the hypothesis that absolutive objects are discourse salient or given, the example is consistent with the object being a topic. Under such analysis, ‘figtree’ may be understood as ‘cataphoric,’ i.e. being a sentential topic, with the rest of the utterance being its comment (see also Berge 2011, who describes an example of double agreement as cataphoric). A topic analysis will allow absolutive objects to be new, but only if the rest of the utterance predicates over them (e.g. predicates of appearance on the stage or noteworthy in the sense of Ionin 2006). The English example in (19) demonstrates this type of interaction: strictly speaking, the definite DP this guy is discourse new, however, the demonstrative form indicates that the definite DP is going to be commented upon (see Ionin 2006 for a more detailed discussion). (19) I entered the subway and there was this guy. a. #I was lucky to find a seat and I read a book until we reached my station. b. He immediately started talking to me. In the same way, the fact that the ‘figtree’ will and must play a role in the narrative is indicated through the presence of absolutive case. To summarize, we argue that all the facts we have seen so far are compatible with absolutive objects being sentential topics. What about oblique objects? According to Berge (2011) they are anti-​topics. What that means is that they are part of the comment structure of an utterance. If this is correct, 21 

For a formal implementation, see Endriss (2009, 245).

 

Morphosyntactic reflexes of information structure in Inuit    411 we expect that they will take narrow scope, unless their lexical content will make them scope outside of the predicative structure. This is precisely what we have seen in (16) and (17), respectively. Furthermore, even though comment is often new information, it is compatible with a given/​discourse salient element as well, as long as the given element is not the sentential topic. This prediction is borne out as well, as seen in (20). Here, ‘Peter’ is the subject of the first sentence, i.e. a global topic if we were to use the terminology of Berge (2011). However, in the following utterance, ‘Peter’ occupies a non-​topic position and even though it is clearly discourse-​salient/​given, it appears as an oblique adjunct. (20) a. pita-​up qukiq-​qau-​ngit-​tanga peter.erg shoot-​past-​neg-​TR.part.3s/​3s ‘Peter didn’t shoot a seal.’

natsiq seal.abs

b. qimat-​si-​qqau-​juq pita-​mit qukiuti-​nga-​nit flee-​ap-past-​INTR.part.3s peter.abl gun-​his.mod.s saku-​li-​gasua-​liq-​tillugu cartridge-​make-​try-​prog-​conj.4/​3s ‘It fled from Peter while he was trying to put a cartridge in his gun.’ (Hallman 2008, 46) To summarize, all these properties are compatible with the topic/​comment distinction. Further supporting evidence for this conclusion comes from Murasugi (2014). Murasugi conducted a behavioral study showing that 1/​2 person objects are more likely to be in absolutive case than in oblique case.22 This finding is compatible with the idea that oblique objects are comment, while absolutive objects are topics, as the speaker/​ hearer is more often the topic of a sentence than the comment.

17.3  Putting the Pieces Together Let us summarize what we have learned so far. First, the Inuit language exhibits an information-​structure driven case and agreement split. More precisely, the object of a transitive structure may be in absolutive or oblique case. If it is in absolutive case, then the object must be a sentential topic. If it is in oblique case, then the object must be in the comment part of the structure. Other semantic effects associated with absolutive objects, such as wide scope or the tendency to be definite or discourse salient, are a direct consequence of sentential topics being associated with a referential address. Furthermore, we have shown that if the object is sentential topic, then it triggers double agreement on the verb. Crucially, the object marker is not a 22 

This finding needs to be compared with seemingly contradictory restrictions, such as those in (4)–​(6).

 

412    Alana Johns and Ivona Kučerová morphological exponent of ϕ-​agree (pace Compton 2014). Instead, it is a nominal clitic, adjoined to the verbal complex. In this section, we will use the sentential-​topic analysis of absolutive objects to explain the inflectional properties and case properties of the Inuit split. As for double agreement, our findings confirm other proposals that argue that clitic doubling is always conditioned by information structure, including object markers on verb (Anagnostopoulou 2006; Kramer 2014a; among others). Specifically, we follow the proposal made in Dočekal and Kallulli (2012) who argue that only sentential topics trigger clitic doubling.23 Thus under the clitic analysis of the verbal object marker, the fact that only absolutive objects trigger double agreement is unsurprising, since all absolutive objects are sentential topics. This conclusion also straightforwardly derives another fact, namely, the dialectal difference between Eastern and Western dialects. As we discussed in section 17.2.1, while in Western dialects objects may be overt in either transitive pattern (ergative or antipassive), in Eastern dialects, the object DP of a transitive double agreement construction is not usually overt. As Anagnostopoulou (2006) and work cited there shows, there is indeed a large body of cross-​linguistic and dialectal variation in the domain of clitic doubling which affects whether or not the full DP is overt. We argue that the dialectal variation attested in the Inuit language dialects can be subsumed under this common variation in the domain of clitic doubling. We thus conclude that the topic analysis of absolutive objects captures not only their semantic properties, but it also provides an insight into the agreement split and the nature of dialectal variation associated with it. We will now turn to the more fundamental question underlying the current discussion which asks what it is about absolutive objects that requires them to be interpreted as sentential topics. We argue that this property is another direct consequence of their semantic import, that is of their requirement to be associated with a referential address. Specifically, we argue that the referential requirement forces sentential topics to be at the edge of a phase since otherwise they would not be accessible to the interpretive module (Chomsky 2000, 2001, 2013; among others). The associated XP movement changes locality properties for the purposes of case assignment, and in turn yields a change in the morphological realization of direct objects. Rizzi (1997, 2004), Grohmann (2003), among others, established that topic-​like interpretations, including contrastive topics,24 move to CP, i.e. the edge of the CP phase. Interestingly, there is evidence that even within vP, topics must move to the phase edge 23  Dočekal and Kallulli (2012) closely follow the formalization proposed in Endriss (2009). The core of their argument comes from the observation that only a very specific type of quantifier (for the semantically savvy reader, only those that can be mapped on a minimal witness set) can function as sentential topics—​and consequently can be clitic doubled. Crucially, no other information-​structure dimension, e.g., familiarity, correctly identifies the right group of quantifiers. 24  Note that even though contrastive topics by name resemble sentential topics, semantically they are quite different, as contrastive topics bring about focus interpretation. See, for instance, Kučerová and Neeleman (2012) for syntactic consequences of the additional semantic import.

 

Morphosyntactic reflexes of information structure in Inuit    413 as well. This has been most convincingly shown for German by Frey (2000) and related work. Furthermore, we know that even if topics stay lower in the structure, they must move to the edge of the phase at LF (see, for instance, Polinsky and Potsdam (2001) for intriguing evidence from long-​distance agreement). The question is why this should be so. We argue that the obligatory movement to the edge of the phase is a direct consequence of topics requiring to be associated with a referential address. We argue that since the referential address is created through anchoring to the common ground in the interpretive component (CI), topic movement makes items accessible to minimal search by the CI component at the point of Transfer (Chomsky 2008, 2013, Narita 2011; cf. von Fintel 2004).25 While in languages like English, if there is more than one argument, sentential topics coincide with grammatical subjects, and consequently, the relevant locality domain is CP, in the Inuit language sentential topics coincide with objects, and consequently the relevant locality domain for establishing referential anchoring is vP.26 This conclusion raises a non-​trivial issue. If we take seriously the semantic analysis of sentential topics such as that of Endriss (2009), then sentential topics are first anchored, and then they are predicated upon by the rest of the structure. If sentential topics were anchored at CI before CP is built (which follows from vP being sent to Transfer), and only then are they predicated upon, the resulting derivation would be counter-​cyclic. We argue that the solution lies in an intuition that underlies much of previous and existing research on ergative languages, namely, the idea that ergative systems are at some level of abstraction passive or unaccusative structures (Fillmore 1968; Hale 1970; Marantz 1984; Bok-Bennema 1991; Johns 1992; among others). Technically, what this amounts to is that an ergative subject is not merged in the same position as that of the external argument in nominative/​accusative systems. The reason is that either υ is entirely missing (L. Nash 1995, 1996), or it is defective (Bok-​Bennema 1991; Johns 1992; Alexiadou 2001). We can rephrase this conclusion in terms of phases and their Transfer (Chomsky 2001, 2008).27

25  Minimal search is the current Minimal Program formalization of the idea that syntactic objects are accessible to further operations only if they appear at the edge of their local domain, cf. the Phase Impenetrability Condition of Chomsky (2000) or the concept of specified subject in the GB framework. The main difference is that the minimal search allows for feature interactions across modules, i.e. not only between syntactic domains. See Narita (2011) for an extensive exploration of consequences of this very property. 26  Note that if an internal argument is to be interpreted as a sentential topic in nominative/​accusative languages, the structure often undergoes a significant change to yield the necessary alignment. For instance, in English, topical internal arguments are typically realized as subjects of passives, while in languages like German or Czech they are A-​scrambled to the edge of their local domain. See, for example, Kučerová (2007) and references cited therein. 27  Note that a similar generalization underlies Compton and Pittman (2010)’s proposal that for the purposes of the morphophonological mapping, CP is the smallest accessible domain in the Inuit language.

 

414    Alana Johns and Ivona Kučerová (21) Absolutive topic generalization: For an internal argument to be minimally searchable and anchored to the common ground, it must be sent to Transfer at the propositional level (CP), i.e. vP phase is not sent to Transfer prior to the completion of CP.28 Note that this conclusion is complicated by the fact that the lower structure of the transitive clause in the Inuit language is not a familiar vP of nominative/​accusative languages, but either an applicative29 or a nominalized structure (Johns 1992 and much following work; cf. also Alexiadou 2001 and Chapter 15, this volume and Haig, Chapter 20, this volume). Recall that to achieve the topic/​comment sentential partition, the topic must be asserted and the comment subsequently must be predicated over it. According to some authors, it is this very partition that underlies the distinction between so-​called categorical and thetic statement distinction (Kuroda 1972; Ladusaw 2000; Basilico 2003; among others). Here we adopt the view advocated in Basilico (2003), namely, that the topic/​comment structure requires a syntactic partition (see Diesing 1992 and following work for the idea that semantic partitions map directly on syntactic structure). As Basilico points out, while the topic interpretation of a subject can be achieved by raising the subject from vP to TP, it is not immediately clear how to create the same syntax-​semantics partitioning effect with an object. We argue that for an internal argument to be interpreted as a sentential topic, it must raise from its base-​generated position to a higher functional projection within the same phase—​analogous to the raising of subjects. We further argue that for such a movement to be possible, the internal argument cannot be merged within a simple VP projection. Instead, the VP part of the structure must be more complex in order to facilitate the required raising. We follow a suggestion made in Basilico (2012) and argue that the internal argument is merged in an applicative-​like structure. More precisely, the internal argument is merged as a sister of a low applicative head (Pylkkänen 2002). If the internal argument is not a sentential topic, then it remains within the applicative projection. If, however, it is to obtain a sentential topic interpretation, it must raise to the specifier of VP—​analogous to the raising of topical subjects from the specifier of vP to the specifier of TP. The trees in (22) schematize the basic structural distinction between the ergative and the antipassive pattern. Note that the truncated structure does not provide a direct representation of the nominalized character of the vP/​VP part, neither does it contain higher functional projections, such as MoodP and TP.30

28 

Meaning there is no v at all, or it is defective. For a recent discussion of applicatives, see Carrier (2014). 30  Notice that our applicative structure introduces an internal argument of the direct object sort, unlike high applicatives proposed for indirect objects in the Inuit language (Carrier 2014). Though we do not fully understand the relation between nominalizations and applicative structures, it may be that the reason that the Inuit language has only high applicative structure for indirect objects is because there is a competing low applicative structure which introduces direct object. 29 

 

Morphosyntactic reflexes of information structure in Inuit    415 (22) a.

Ergative pattern: vP v

VP INTERNAL ARG

V’ V

ApplP Appl

t

b. Antipassive pattern: vP v

VP ApplP

V   

Appl

INTERNAL ARG

What does this structural difference entail for case assignment? We argue that the structural difference is similar to that of Dative shift in English. If the object remains within the applicative projection, its case is determined by the applicative head. The result is an oblique case (MOD). In contrast, if the internal argument raises to the spec,VP, it can be assigned case by whatever the appropriate structural case assigner is.31 As a result, the internal argument surfaces as absolutive. (23) The Dative shift analogy a. Give money to him. b. Give him money.

∼ antipassive pattern ∼ ergative pattern

Note that we assume a Distributed-​morphology style of case assignment (Marantz 1991); that is, the morphological realization of case reflects the morphological-​case hierarchy, and hence only indirectly the underlying syntactic structure. Furthermore, after 31 

It is possible that the corresponding case assigner is v. However, since v is defective, it depends on the implementation of defectiveness whether or not it might act as a case assigner.

 

416    Alana Johns and Ivona Kučerová the internal argument raises out of the applicative projection, it must trigger cliticization within the higher functional complex. The proposed analysis thus straightforwardly ties the information-​structure properties of the Inuit structures and their case realization. Before we conclude, we will shortly return to a couple of outstanding properties, namely, aspectual properties of the ergative and antipassive structures, and dialectal differences in terms of what type of DPs may appear in the antipassive structure. It has been suggested that the ergative vs antipassive split is aspect-​based (Spreng 2006, 2010, 2012; Clarke 2009). Even though a careful investigation of this suggestion extends far beyond this chapter, we would like to suggest that the role of aspect is secondary, and in fact it relates to the topic/​comment distinction we argue for here. See also Berge (2011). There is a large body of syntax-​semantics literature that makes a connection between aspect, more precisely telicity,32 and some form of definiteness (Krifka 1998, Borik 2002; Filip and Rothstein 2006; Basilico 2008; Filip 2008; Ramchand 2008). The basic intuition can be approximated by a comparison of English and Czech strictly incremental verbs (Filip 2008), as can be seen in examples (24) and (25). The readings in (24a) and​ (25a) are atelic, they are neutral as to whether or not the event in question was completed, e.g. we don’t know if all the apples were eaten. On the other hand, the readings in (24b) and (25b) are associated with a telic interpretation, i.e. all the apples were eaten. (24) a. Peter ate apples b. Peter ate the apples (25) a. Petr jedl    jablka Petr ate.imperf  apples ‘Peter ate apples.’ b. Petr snědl jablka Petr ate.perf apples ‘Peter at the apples.’ When we closely look at these examples, we see that in Czech the grammatical source of the telic interpretation is the perfective marking on the verbal morphology. Even though the English translation indicates that the object is to be interpreted as definite, the noun phrase itself does not have any definiteness marking. In other words, the aspectual marking in Czech and the definiteness marking must have the same—​or a very similar—​semantic denominator. We follow Filip and Rothstein (2006) and Filip (2008) and argue that the common denominator is best modeled in terms of the maximization 32  Though the terms perfectivity and telicity may coincide, they are distinct concepts (Giorgi and Pianesi 2001). While perfectivity denotes the inclusion of the event time in the topic time—​using the terminology of Klein (1994) and the denotation of Paslawska and von Stechow (2003) and others—​ telicity determines whether the actual event has been completed.

 

Morphosyntactic reflexes of information structure in Inuit    417 of the event/​direct object. Note, furthermore, that for an event to be maximized, it must be first mapped to a scale. In fact, Basilico (2012) proposes that it is the scalar properties of events which underlie the formation of antipassives in the Inuit language. How does this relate to topics? According to Endriss (2009), a sentential topic requires a maximized interpretation, irrespective of whether it involves a definite, indefinite, or quantifiers.33 We suggest that it is this very property that drives the telic interpretation of the ergative pattern and yields an imperfective-​like interpretation of its antipassive counterpart. This is similar to the English telic effect involving the maximization interpretation of the definite argument which we observed in (24b). In other words, there is a direct connection between topical interpretations and aspect, a relation which likely underlies the aspectual properties attested in the Inuit patterns. We thus conclude that the topic analysis subsumes the aspect analysis, and consequently, a direct reference to aspect is unnecessary. The last remaining issue has to do with dialectal variation in the domain of oblique DPs that may appear in the antipassive construction. While in the Western Inuit dialects and Western Greenlandic, the antipassive construction requires the DP in oblique case (MOD) to be indefinite—​or more precisely it excludes referential DPs, there appears to be no such restriction in the Eastern Canadian dialects. We find this reminiscent of restrictions on Scandinavian Object Shift (Thráinsson 2001 and literature cited there). Object Shift is A-​movement of certain object DPs to the specifier of vP. Even though Object Shift somewhat resembles A-​scrambling in Germanic and Slavic languages, it seems to be structurally more restricted (Holmberg 1986, 1999). Furthermore, while some Scandinavian languages allow Object Shift only of pronouns (for example, Danish); others (e.g. Icelandic) allow optional movement of full definite DPs as well. Finally yet other varieties (for instance, some Norwegian dialects, Nilsen 1997) additionally allow Object Shift of indefinites. Furthermore, languages differ as to whether they allow just Object Shift or whether they extend it to allow semantically motivated movement (A-​scrambling), as in Icelandic. In contrast, there are languages, such as English that allow topic movement to the left periphery, but their semantic movement within vP is restricted only to certain ditransitive verbs. The full exploration of this possible connection however awaits future research. To summarize, we have argued that the most adequate characterization of absolutive objects in the ergative pattern in the Inuit language is in terms of sentential topics. Even though our analysis has not fully explored all consequences of this hypothesis, we have shown how it derives the major morphosyntactic properties of the ergative and the antipassive pattern, and we have sketched how the topic analysis ties together various seemingly independent observations about the nature of the split, i.e. its information structure properties, scope properties, and aspectual properties. Even though more work needs to be done, especially in the domain of dialectal variation, the proposed

33 

The main contribution of Endriss (2009) is that she models topicalized quantifiers to a minimal witness set representation, i.e. the maximal set of which a certain property must hold.

 

418    Alana Johns and Ivona Kučerová analysis raises various issues about the nature of ergative languages in general. First of all, we suggested that there is a connection between the passive-​like properties of the Inuit language and the necessity for the vP phase not to be sent to Transfer prematurely, if the internal argument is to be interpreted as a sentential topic. One question that immediately arises is what is the trigger and what is the consequence. The other question is whether a similar connection might hold in other ergative languages. Another conclusion with possible consequences for other ergative languages is that the topic interpretation might be the source of certain aspectual interpretations associated with the split. Aspect is a common property associated with ergative splits in general, and it is possible that the proper characterization of these splits indeed lies elsewhere. Before one jumps too quickly to this conclusion, though, it is important to keep in mind that aspectual splits tend to involve re-​alignment of arguments, which is not what we see in the Inuit language. In other words, aspectual distinctions in this language do not condition whether we get ergative or antipassive patterning. If the connection to other types of ergative splits is real, it is not trivial. Yet in our mind it is worth exploring in future research, as is the relation to nominalized structures, only touched upon here.

Acknowledgments We want to express our appreciation to Elizabeth Cowper and especially to Judith Aissen for comments and suggestions on this chapter. We also wish to thank Saila Michael, as well as Katie E. Winters and other Nunatsiavut Inuit for data. Thanks also to SSHRC for funding this research project in a grant (435-​2015-​0979) to Johns, Kučerová and Lampe and a grant (435-​ 2012-​1567) to Kučerová.

Abbreviations 1, 1st person; 3, 3rd person; ABS, absolutive case; ACC, accusative case; AP, antipassive; CAUS, causative; CL, clitic; CONJ, conjunctive mood; D, dual; ERG, ergative case; IMPERF, imperfective; INCEPT, inceptive; IND, indicative mood; INSTR, instrumental case; INSTRUMENT, instrument affix; INTERR, interrogative intonation; INTR, intransitive; LOC, locative; MOD, modalis case; NEG, negation; P, plural; PART, participial mood; PERF, perfective; POSS, possessive; PRO, pronoun; PROG, progressive; R, reflexive; S, singular; TR, transitive.

 

Chapter 18

Ergative c onst e l l at i ons in the stru c t u re of speech ac ts Martina Wiltschko

18.1 Introduction Ergativity is a heterogeneous phenomenon (Bittner & Hale 1994). The surface constellations associated with it indicate that we need to distinguish two structural layers: a layer where arguments are introduced (henceforth argument–​structure) and a layer where grammatical relations (case) are introduced (henceforth case–​structure) (cf. Williams 2003). The goal of this chapter is to demonstrate that we find similar structural constellations in the layer of structure where speech act relations are introduced (Speas & Tenny 2003; Haegeman 2013). In particular, I argue that speech act structure (henceforth SA-​ structure) consists of a grounding layer, where the speaker’s or the addressee’s commitment toward the proposition are encoded. The second layer of SA-​structure is dedicated to the response system of language:  e.g. what the speaker (henceforth S) wants the addressee (henceforth A) to do with the utterance. Each of these layers can come in different guises, in much the same way as argument–​structure can be transitive, ergative, or unergative. This chapter is organized as follows. In section 18.2, I introduce ergative constellations in the domain of argument–​structure. In section 18.3, I introduce background assumptions regarding the syntax of speech acts. And in sections 18.4 and 18.5 I show that the logic of ergativity can be equally applied to SA-​structure. In section 18.6, I conclude.

 

420   Martina Wiltschko

18.2  The Heterogeneity of Ergativity and the Logic behind It An ergative/​absolutive system is typically described as follows: intransitive subjects are marked like transitive objects but differently from transitive subjects. In particular, transitive subjects are marked as ergative while intransitive subjects and objects are marked as absolutive. This contrasts with nominative/​accusative systems in which transitive and intransitive subjects are identically marked as nominative while objects are marked differently, namely as accusative (Dixon 1979). While all ergative systems have in common that transitive subjects differ from intransitive subjects in some ways, there are also many ways in which such systems differ from each other. Hence, it has become clear that ergativity is a surface phenomenon that can come about in different ways and it does not come as a surprise that many different analyses have been proposed (see Coon & Adar 2013; Deal 2015, for an overview). From a generative perspective, it is not surprising that surface ergativity comes about in different ways: the key notions that identify ergative constellations are not primitives. That is, transitive subjects, transitive objects, and intransitive subjects are all derived concepts. To understand ergativity from a generative point of view, we need to understand these notions. Within the generative tradition, subjects are not a unified concept. Rather, they are defined across various levels of structure. That is, one of the key insights within the Principles and Parameters framework (e.g. Chomsky 1981) is the distinction between thematic roles and grammatical (case) roles. Thematic roles are assigned to arguments by the verbs that introduce them, while grammatical roles are assigned by functional categories (such as INFL). This assumption accounts for the fact that arguments that are realized as grammatical subjects can bear different thematic relations to their verbs. For example, in the active voice, the agent is realized as the grammatical subject (1)a, while in the passive voice it is the theme or patient which is realized as the grammatical subject (1)b. (1)

a. The dog was catching the ball. b. The ball was caught by the dog.

This type of mismatch between thematic and grammatical roles is precisely the motivator for separating the thematic domain from the domain of grammatical relations and structural case-​assignment. The separation between thematic and grammatical roles was not always complete. Rather, initially, it was assumed that in passive sentences the thematic object role and the grammatical object role were both assigned in the complement of V, while the

 

Ergative constellations in the structure of speech acts    421 thematic subject role and the grammatical subject role were both assigned in SpecIP. The mismatch between these roles observed in passives was analyzed as involving the absorption of accusative case via passive morphology (Baker et al. 1989), which in turn results in the unavailability of the agent role (Burzio 1986). Given the assumption that all sentences must have subjects and that all arguments must receive case, the object argument was assumed to move to SpecIP where it receives nominative case. Hence, this is a situation where the thematic role is assigned in a different position than the grammatical role. (2)   

The GB-style analysis of passive IP VP

Subj AG

V' V

TH

Since this first deconstruction of thematic and grammatical roles, however, there have been two seminal assumptions, which resulted in a complete structural distinction between thematic and grammatical roles. These two assumptions are the VP-​internal subject hypothesis (Koopman & Sportiche 1991)  and the assumption that accusative case, like nominative case, is assigned by a functional category above the VP (Borer 1994, 2005b; Megerdoomian 2000).1 As for the VP-​internal subject hypothesis, in its current incarnation it is usually assumed that agents are introduced by a semi-​functional head v (Chomsky 1992; Kratzer 1996). Thus, there is a one-​to-​one correspondence between heads and their arguments. This yields a complete separation of argument–​structure from case–​structure as illustrated in (3) (Williams 2003; Wiltschko 2014).

1 

Not everyone subscribes to this conceptualization of the case domain. In particular, it is often assumed that accusative is assigned by v—​the same (semi-​) functional head which introduces the external argument (e.g. Chomsky 1995). However, evidence from systems where accusative case-​ assignment is sensitive to aspectual properties suggests otherwise (Kiparsky 1998a). Even if accusative case is assigned by v, the points about ergativity to follow still hold. What is important is that thematic roles are structurally separate from grammatical roles.

 

422   Martina Wiltschko (3)  

A two-layered system IP

Subj

Case-structure AspP vP

Object

argument-structure

AG

VP TH

Thus, we can identify two notions of subjects: thematic subjects realized in SpecvP and grammatical subjects realized in SpecIP. And similarly, we can identify two notions of objects: thematic objects, realized within VP and grammatical objects realized in SpecAsp. The assumption that both subjects and objects come in (at least) two different types (thematic and grammatical) has important implications for our understanding of ergativity. Note that the classic description of ergativity makes no distinction between thematic and grammatical roles. Once this division is in place, however, we can distinguish between two notions of ergativity. On the one hand, we can define ergativity in terms of thematic relations only. And, on the other hand, we can define ergativity in terms of grammatical relations. In his seminal paper, Perlmutter 1978 identifies two types of intransitive verbs. Those whose sole argument starts out like the thematic object of a transitive verb. These are known as unaccusative verbs. In a nominative/​accusative system, such arguments still behave like transitive subjects, at least on the surface. This is because verbs that lack an external argument fail to assign accusative case (hence the label unaccusative). Consequently, the underlying object moves to the position of grammatical subjects, just like the object of a passive verb as in (4)b. In contrast, there is also a class of verbs whose sole argument starts out like the subject of a transitive verb and hence this argument shares some properties with transitive subjects (4)a. These are known as unergative verbs, a label which reflects the fact that the behavior of the contrasting class of verbs (unaccusative verbs) is akin to ergativity. And indeed some scholars refer to unaccusative verbs as ergative verbs (e.g. Burzio 1981; den Besten 1981), a convention I will follow. (4)    a. unergative verbs

b. unaccusative (=ergative) verbs

IP Subj

IP VP

AG

Subj

VP

V' V

V' V

TH

 

Ergative constellations in the structure of speech acts    423 Within the minimalist program, theory-​internal considerations lead Chomsky 1992 to adopt an alternative conceptualization of unergatives, namely that of Hale & Keyser 1993 (henceforth H&K). According to H&K, unergative verbs are best analyzed as concealed transitives: they come with an internal argument, which incorporates into the verb as in (5). (5) 

Unergatives as concealed transitives vP v'

AG v

VP V

TH

Thus, even languages that are otherwise nominative/​accusative show effects of ergativity: in terms of argument–​structure properties, intransitive subjects may behave either like transitive subjects or like transitive objects. Since this type of ergativity is exclusively a matter of argument structure, I refer to it as argument–​structure ergativity. It contrasts with case–​structure ergativity which arises as a matter of grammatical marking. That is, case–​structure ergativity is not a matter of how many arguments are introduced at the argument–​structure layer, but instead it concerns the marking of those arguments in terms of their grammatical relations (often in the form of case or agreement). In a nominative/​accusative system the subject of an intransitive is marked in the same way as the subject of a transitive (i.e. nominative) and the transitive object receives special case (i.e. accusative). In contrast, in an ergative/​absolutive system the subject of an intransitive is marked in the same way as the object of a transitive (i.e. ergative) and the transitive subject receives special case (i.e. ergative). Thus, case–​structure ergativity has in common with argument–​structure ergativity the fact that transitive subjects are special. As discussed at length in this volume, there are three core analyses to account for the difference between ergative and accusative case systems. According to early approaches (Campana 1992; Murasugi 1992; Bobaljik 1993a) case-​assignment is correlated with agreement triggered by functional heads. In a nominative system, the higher functional head is always active whereas the lower functional head is only active if there is an argument associated with the higher head. Hence transitive objects receive a different case (ACC) than transitive and intransitive subjects, as in (6). In an ergative system, the lower functional head is always active whereas the higher head is only active if there is an argument associated with the lower head. Hence transitive subjects will receive a different case than intransitive subjects (which remain associated with the lower position), as in (7).

 

424   Martina Wiltschko (6) nominative system a. [F2 argnom [F1      [arg v [VP V    ]]]] b. [F2 argnom [F1 argacc [arg v [VP V arg]]]]

→ intransitive → transitive

(7) ergative system a. [F2      [F1 argabs [arg v [VP V    ]]]] b. [F2 argerg [F1 argabs [arg v [VP V arg]]]]

→ intransitive → transitive

Several problems have been identified with this line of approach, including the fact that case doesn’t always correlate with agreement (see for example Baker 2014a) and some of the movements necessary for the analysis are not motivated under minimalist assumptions. The two prevalent approaches in current minimalist theorizing are the dependent case approach and the inherent case approach. According to the former (which has its roots in Marantz 1991), the ergative/​accusative contrast follows from a parameter akin to (8). There is a structural case which can only be assigned if another argument appears in the same domain (e.g. see Bobaljik 2008; Baker & Vinokurova 2010, Baker 2014a; Coon & Preminger, Chapter 10, this volume) (8)

The dependent case analysis If there are two distinct NPs in the same clause (“governed by V+I”) then: (i) Mark the lower one with dependent case (accusative) and/​or (ii) Mark the higher one with dependent case (ergative). (iii) Otherwise, mark NPs with unmarked/​default case (called nominative or absolutive). Adapted from Baker 2014a (3)

According to the inherent case approach, ergative case is treated as an inherent case rather than a structural case and thus correlates with thematic role assignment (see L. Nash 1996; Woolford 1997, 2006; Aldridge 2004, 2008b, 2012b; Anand and Nevins 2006; Legate 2006, 2008, 2012; Laka 2006a; Massam 2006; Mahajan 2012). Hence on this view, ergative case is assumed to be assigned in the domain of argument–​structure, much lower than any of the structural cases. In the remainder of this chapter I explore the question as to whether ergative properties are found in SA-​structure as well.

18.3  Introducing Speech Act Structure Speech acts are often considered a purely pragmatic phenomenon. That is, traditionally syntax is taken to be the module that regulates the composition of meaningful units of

 

Ergative constellations in the structure of speech acts    425 language into larger constituents and phrases. The unit of investigation for many syntacticians is typically a sentence expressing a proposition, which is in turn the maximal unit of analysis of many semanticists. There is however a body of research that seeks to incorporate elements of speech acts into the domain of syntax. The purpose of this section is to review this literature. It will serve as the backdrop against which we explore ergative constellations in SA-​structure.

18.3.1 Ross’s (1970) Performative Hypothesis Within the generative literature, the primacy of the sentence for syntactic investigations has first been called into question by Ross (1970). He proposes a layer of structure that dominates the root clause and encodes the illocutionary force of a given utterance. Ross takes this layer to be another type of clause, consisting of a predicate (e.g. a verb of saying) a subject (the speaker) and an object (the addressee). This analysis is known as the performative hypothesis because its core insight is that even declarative clauses can be viewed as being performative, i.e. by uttering a declarative S does something, namely performing an act of speaking (see also Sadock 1969, 1974). Ross’s analysis is schematized in Figure 18.1. Deep Structure (interpretation)

Surface Structure (form)

S SAstructure I tell you that

S

Performative Deletion

pstructure I have a dog

S pstructure I have a dog

Figure 18.1  Ross’s (1970) performative hypothesis

The surface declarative clause is analyzed as being embedded in a superordinate structure, which can roughly be paraphrased as I tell you that. Since none of the postulated components of the superordinate structure are overtly marked, Ross (1970) assumes a rule of performative deletion according to which the superordinate structure is deleted deriving the surface form I have a dog.2 If there is no overt marking of the superordinate structure, how do we know that it is still part of syntax, rather than simply being a matter of pragmatic inferencing? Ross (1970) discusses a series of arguments having to do with (i) the presence of a higher 2 

This analysis is couched within a framework that took Deep structure (DS) to be the input for interpretation.

 

426   Martina Wiltschko first person subject (the speaker); (ii) the presence of a verb of saying above the matrix clause; (iii) the presence of a higher 2nd person indirect object (the addressee); and (iv) the possibility to modify the performative clause. If there is indeed another layer above the root of the clause, and if this layer can be characterized in terms of another predicate–​argument–​structure, then we expect to find the hallmark of ergative constellations. However, before we can explore whether this is indeed the case, we need to be sure that we have an adequate framework within which to pursue this question. That is, it is well known that Ross’s (1970) analysis faces serious problems, which were sufficient for the field to reject the performative hypothesis (Anderson 1971; Grewendorf 1972; Fraser 1974; Gazdar 1979; Mittwoch 1977; Newmeyer 1986). However, the arguments Ross presented, as well as the general insight that speech acts ought to be syntactically represented didn’t die out completely. Its revival is in part made possible by the discovery of functional categories. That is, much like the insights of generative semantics into the decomposition of events has been reanalyzed in terms of postulating a series of (semi-​)functional categories within the VP-​domain (e.g. see Ramchand 2011) the decomposition of speech acts has been reanalyzed in terms of postulating a series of functional categories above the CP domain. Among the categories postulated by different scholars we find PragP (Hill 2006), SpeechActP (SAP; Hill 2007a, 2007b; Krifka 2013), AttitudeP (Paul 2014), and PartP (Haegeman 2014; Zu 2015). I turn to the most prominent incarnation of this type of analyses in the next subsection.

18.3.2 Functional Categories in the Speech Act Domain The current revival of the performative hypothesis—​let us call it the neo-​performative hypothesis—​presents some overt evidence for the existence of SA-​structure. In particular, recent syntactic analyses focus on units of language that directly encode SA-​ structure. These include evidential markers (Speas & Tenny 2003), sentence-​peripheral particles (Haegeman 2013), Vocatives (Hill 2013), and response particles (Krifka 2013), among others. These units of language are incorporated into syntactic structures in ways that indicate the workings of the syntactic component: they display linear ordering restrictions, pronominalization patterns, scope effects, and agreement patterns. For reasons of space, I limit the discussion to agreement effects. As mentioned in Ross (1970), some languages display S-​agreement in gender. Thai is such a language as shown in (9). The sentence peripheral marker khráp is used with male speakers while kâ is used for female speakers. (9) a. Khaw maa khráp. he come spkr=male ‘He is coming.’

 

Ergative constellations in the structure of speech acts    427 b. Khaw maa kâ. he come spkr=female ‘He is coming.’ (from Oyharçabal (1993) presented in Miyagawa (2012): ex 5) On the assumption that agreement is syntactic (Chomsky 1981, 1995, 2001) S must be represented in the clausal architecture (see Giorgi 2009 for explicit arguments that S coordinates are represented in the CP-​layer). And similarly, there are languages that display A-​agreement (i.e. allocutive agreement; Miyagawa 2014). For example, in the Basque examples in (10), the sentence-​final auxiliary differs in form depending on the gender and number of A as well as the nature of the relation between S and A such that formal status is explicitly encoded (10). (10) Basque A-​agreement a. Pettek lan egin dik Peter.erg work.abs do.prf aux.​2masc ‘Peter worked.’ b. Pettek lan egin Peter.erg work.abs do.prf ‘Peter worked.’

din aux.​2fem

c. Pettek lan egin Peter.erg work.abs do.prf ‘Peter worked.’

dizü aux.​2formal

d. Pettek lan egin Peter.erg work.abs do.prf ‘Peter worked.’

du aux.​2pl (Miyagawa 2012: (8), cited in Oyharçabal 1993)

Again, if agreement is syntactically conditioned, it must be the case that A is represented syntactically (see Zanuttini 2008). In sum, evidence from agreement patterns lends support to the syntactic encoding of speech act participants. Thus, despite the problems that the performative hypothesis faces, there are still empirical generalizations that need to be accounted for. This is precisely what neo-​performative hypotheses seek to achieve: they capture (among other things) the evident presence of speech act participants in the syntactic structure but they are not vulnerable to the same criticism as Ross’s original proposal (see Speas & Tenny 2003: 338 for explicit comparison). While Ross (1970) took the superordinate structure to be a run-​of-​the-​mill matrix clause of the type S[NP VP], neo-​performative analyses take the superordinate structure to be an extension of the functional projection of the clause. For example, Speas & Tenny 2003 propose a complex speech act phrase (saP) consisting of two layers, as illustrated in (11). The higher head introduces S in its specifier

 

428   Martina Wiltschko position and takes a lower saP as its complement, which in turn hosts the utterance content in its specifier and A (labeled Hearer in Speas and Tenny 2003) in its complement. (11)  

Speech act structure in the extended clausal projection saP sa

(Speaker) sa

(Utterance content)

sa* sa* sa*

(Hearer)

This model preserves the main insight of the performative hypothesis, in that it postulates a superordinate SA-​structure. While this structure is part of the functional architecture, Speas & Tenny 2003 suggest that it still follows the same logic as argument–​structure, adopting the framework of Hale and Keyser 1993. According to this proposal, SA-​structure is likened to the double object construction:  S serves as the agent of the speech act; the utterance content serves as the theme and A as its goal. Thus, they maintain the assumption that—​in the context of an assertion—​SA-​structure encodes something like ‘I give the utterance to you.’ Speas & Tenny 2003 further argue that the restriction on the number of speech acts we observe across languages follows from the available structural configurations. In particular, interrogatives are derived by a passive-​like movement such that the hearer moves above the speaker, and imperatives and subjunctives are defined by their non-​finite utterance content. Note that subjunctive is not typically included in the list of speech acts. It appears that Speas & Tenny 2003 take clause-​type or mood to be the defining property of speech acts. However, things are more complicated in that clause-​type alone is not a reliable predictor of illocutionary force. For example, rising declaratives, extensively discussed in Gunlogson 2003, 2008, are formally declaratives but are associated with rising intonation which triggers a questioning interpretation. Thus, while Speas & Tenny’s analysis of SA-​structure doesn’t face the problems of Ross’s (1970), it comes with its own problems (Gärtner & Steinbach 2006). Here I address one potential problem that arises in the context of the present hypothesis. Specifically, given the argument–​structure approach toward SA-​structure we would expect to find argument–​structure ergativity in the domain of SA-​structure (see Haegeman 2013). Speas and Tenny 2003 however do not discuss this possibility. And given their particular implementation of SA-​structure, this is not surprising. What we would expect based on their analysis is to find two types of intransitive speech acts: those that have an external (S) argument only would parallel unergative predicates while those that have an internal (A) argument only, would parallel ergative (aka unaccusative) predicates. Given that every utterance requires the presence of S, the latter

 

Ergative constellations in the structure of speech acts    429 option seems non-​sensical. And this may be the reason Speas & Tenny 2003 didn’t pursue this possibility. Thus, either ergativity plays no role in SA-​structure or else we have to rethink the composition of SA-​structure. In what follows, I pursue the latter option showing that there are independent reasons to postulate a different model of SA-​structure.

18.3.3 The Complexity of Speech Acts There are two problems with the neo-​performative hypothesis as conceived of in Speas & Tenny 2003 and subsequent work. First, there is empirical evidence that A is structurally higher than S (Lam 2014, Heim et al. 2014). For example, Cantonese, has a series of sentence-​peripheral discourse particles. Relevant for the present purpose are two such particles. Me13 combines with declaratives and derives a request for confirmation with a negative bias on behalf of S, as in (12)b. Hence, me1 is classified as an S-​oriented particle. The second particle, ho2 also derives a request for confirmation, but it introduces an additional meaning component, namely that S assumes that A believes p, (12)c. Hence ho2 is classified as an A-​oriented particle (see Lam 2014 for detailed discussion). (12) a. zi3ming4 jau5 fu6ceot1 gwo3 si4gaan3 Jimmy have devote asp time ‘Jimmy has spent time (on the project).’ b. zi3ming4 jau5 fu6ceot1 gwo3 si4gaan3 me1? Jimmy have devote Asp time prt ‘Jimmy has spent time (on the project), has he?’ c. zi3ming4 jau5 fu6ceot1 gwo3 si4gaan3 gaa3 ho2? Jimmy have devote Asp time prt prt ‘Jimmy has spent time (on the project), right?’ Crucially, the two particles can co-​occur but their ordering is restricted such that A-​ oriented ho2 has to follow S-​oriented me1, as in (13). (13) Jimmy is the first of a long taxi queue. A taxi is coming, but someone not from the queue opens the door of the taxi saying loudly that he is in a hurry. Everyone in the queue is angry. Jimmy says this to the second person in the queue: a. daai6 seng1 zau6 dak1 gaa3 laa3 me1 ho2 loud voice then okay prt prt prt prt ‘What can one get by just by being loud? I assume you’d agree it’s a valid question, right?’ 3 

Numbers indicate tone.

 

430   Martina Wiltschko

b. *daai6 loud

seng1 zau6 dak1 gaa3 laa3 ho2 voice then okay prt prt prt

me1 prt

Lam 2014: 64 (6)

Assuming that the further to the right a particle appears, the higher it must be in the hierarchical structure (Lam 2014), we have to conclude that the A-​argument is structurally higher than the S-​argument.4 This suggests that SA-​structure is organized differently than assumed under current neo-​performative analyses.5 This further implies that the common conceptualization of declarative clauses in terms of I give p to you cannot be on the right track. Interestingly, pragmatic analyses of speech acts have changed since Ross (1970). His view of declaratives reflects the assumption that declarative assertions have the following conditions of use: (14) Conditions of use for declarative assertions (i) S believes the proposition (p) conveyed by her utterance. (ii) S wants A to adopt p into her set of beliefs Bach and Harnish 1979 This view doesn’t capture all declaratives, however. In particular, there are (at least) two ways in which a declarative can be modified. First, S may modify the commitment to p thereby changing what is being said. Much of Speas’ (2003) work explores this type of speech act modification. For example, S can indicate, by means of an evidential marker, that they don’t have direct evidence for the truth of p. Second, S can also modify what s/​he wants A to do. According to (14), by uttering a typical declarative, S expects A to believe p and thus asks A to adopt p into their set of beliefs. However, this is not the only thing S can do with declarative clauses. As we have seen above, the addition of a sentence-​final particle in Cantonese renders a declarative into a request for confirmation. The particle modifies what S expects A to do with the utterance. Beyssade & Marandin (2006) refer to this aspect of the speech act as the Call on Addressee (henceforth CoA). We observe a similar pattern of speech act modification in English. Consider the difference between a regular declarative (15)a and one that is modified by the sentence-​final particle eh (15)b. Given its function, Heim et al. (2014) refer to this type of particle as conformational, and I follow this convention. (15) a. You are leaving. b. You are leaving, eh? 4  An anonymous reviewer points out that the linear order would also be compatible with an analysis whereby particles are head-​initial and the utterance content moves above the particles. In this case, S would have to be associated with the higher position, as in Speas & Tenny 2003. However, the interpretation of the particles suggests that the A-​oriented one scopes above the S-​oriented one: A is asked to respond to S’s belief. 5  See Thoma (2016) for additional arguments based on discourse particles in Bavarian German.

 

Ergative constellations in the structure of speech acts    431 According to the conditions of use for a declarative assertion (15)a would be used if the conditions in (16) hold. (16) Conditions of use for declaratives (i) Bel(S,p) (ii) CoA: Bel(A,p) Given that (15)a is a statement about A, its discourse conditions are slightly marked. Some special circumstances must hold for S to felicitously tell A something about themselves. For example, S may use (15)a as an indirect command. The conditions of use for the eh-​modified declarative in (15)b differ. By uttering (15) b, S does not inform A that p and hence it cannot be used as an indirect command. Rather (15)b is used to confirm that S’s belief is accurate. More precisely, S is not fully committed to the truth of p—​though S has a bias toward believing p. I indicate this with a disjunctive commitment to p with opposite polarity values. Boldface on the positive belief is used to indicate the bias. In addition to modifying the commitment toward p, eh also modifies CoA. In particular, with the use of eh, S requests A to confirm her biased belief (17(ii)). (17) Conditions of use for eh-​modified declaratives (i) Bel (S,p) ∨ Bel (S,¬p) (ii) CoA: Confirm (Bel (A,p)) In sum, eh appears to modify both S’s commitment as well as CoA. Based on the properties of speech act modifiers such as eh, I propose that there are two distinct layers that comprise SA-​structure (see Heim et al. 2014 for more detailed discussion): a layer which is responsible for encoding the commitment of the speech act participants to p and a second layer where CoA is encoded. As indicated in (18), I refer to the lower layer as GroundP and the higher layer as ResponseP. The label GroundP is meant to evoke Clark & Brennan’s 1991 notion of grounding as well as the notion of the common ground (cf. Heim et al. 2014, Thoma 2016, Wiltschko & Heim 2016). (18)    

A complex speech act structure RespP CoA Responding

GroundP propostional attitude

S "sentence"

 

432   Martina Wiltschko Moreover, speech act modifiers may also modify declarative clauses such that they express (S’s assessment of) A’s commitment to p. Consequently, Lam (2014) argues that GroundP comes in two guises: one is relativized to S’s set of beliefs (Ground-​S) while the other is relativized to A’s set of beliefs (Ground-​A). Accordingly, the two particles introduced in (12) and (13) occupy different syntactic positions: Ground-​S (me1) and Ground-​ A (ho2), respectively. (19) 

An articulated grounding layer RespP CoA

groundP

Ground-A

GroundP Ground-S

S p-structure

With this model of SA-​structure in place, we can now explore whether the hallmarks of ergativity can be detected in this domain. We shall see that argument–​structure ergativity plays a role in both the grounding layer as well as in the response layer. As we shall see, interpretable roles are assigned in each of these layers (grounding and responding roles); hence both structures are more akin to argument–​structure than to case–​ structure and hence we expect to find argument–​structure ergativity in both layers. However, before I show that this prediction is indeed borne out, I briefly discuss the relation between p-​structure, grounding structure, and response structure.

18.3.4 The Syntax of Assertions Recall that according to Ross (1970), assertions are performative speech acts which encode something akin to “I give p to you.” According to the present analysis, which is based on contemporary understanding of speech acts, assertions are complex moves. That is, in a typical assertion, S conveys their commitment to the truth of p. At the same time, they request from A to adopt the same belief. I assume, following Truckenbrodt 2003, that the latter is encoded by means of a falling intonation contour. Thus, by means of a declarative, S asserts that the proposition is in their belief set (Ground-​S) and the assertive falling intonation asks A to do the same thing. This contrasts with rising intonation, which asks A to respond to the utterance. The syntax of typical assertions is schematized in (20).

 

Ergative constellations in the structure of speech acts    433 (20) 

Syntax of assertions RespP CoA GroundP Ground-S

S p-structure

Note that not every utterance needs to include a CoA or an assessment about A’s belief set. That is, S may simply utter a declarative clause to indicate their commitment toward the truth of p, without requesting for A to share this commitment. Consider the example in (21). (21) John runs into Mary and notices that she is walking a young puppy. He knows that she had wanted to get a new dog for a while. He exclaims: You have a new dog! Here John does not wish to inform Mary that she has a new dog. Instead, with the use of this declarative, John lets Mary know that he now knows that she has a new dog. In this way, the declarative clause-​type serves as an exclamation. Note crucially that the intonational contour of a non-​informative assertion differs from that associated with informative assertions. This indicates that the response structure need not be projected. We can thus explore ergative constellations associated with the grounding structure independent of those associated with the response structure. In what follows I show that we do indeed find ergative constellations in both these structures. I start with a discussion of ergative speech acts based on the grounding structure (section 18.4) and then I explore ergative speech acts based on the responding structure (section 18.5).

18.4  Ergativity and the Grounding Structure Let us consider a basic declarative with Ground-​S only. These constructions will serve as the basis relative to which we explore ergativity because they present us with the grounding structure version of transitive predicates. The internal argument corresponds to p-​structure whereas the external argument corresponds to Ground-​S, as in (22).

 

434   Martina Wiltschko Transitive grounding structure

(22) 

GroundP Ground-S external argument

Ground Ground

S p-structure internal argument

Ground-​S thus parallels the initiator of the grounding event, just like in Speas & Tenny’s 2003 conceptualization of declaratives. However, the present proposal differs in the way the role of A is conceptualized. Within the grounding structure A is not viewed as the goal of the speech act but instead it may be understood as a role akin to a causative argument. It won’t play a role in the syntax of ergative speech acts to which we now turn. That is, we expect there to be two types of intransitive grounding structures: unergative and ergative (aka unaccusative) structures.

18.4.1 Unergative Intransitive Structure We expect unergative speech acts to consist of an external argument only, with the utterance being incorporated. I propose that this is the case in imperatives. In particular, I follow Portner (2004) in assuming that an imperative clause-​type denotes a property. Since properties cannot function as arguments, the complement is incorporated, as in (23). (23) 

Imperatives as unergatives. GroundP Ground-S

Ground Ground

S p-structure

Just as incorporated nouns denote properties rather than arguments, so does the propositional content of imperatives. In addition, I  assume that Ground-​S associated with

 

Ergative constellations in the structure of speech acts    435 imperatives contains a set of intentions, rather than a set of beliefs.6 Thus, the clause-​type of the complement has an effect on the interpretation of Ground-​S. For completeness note that a full-​fledged SA-​structure for imperatives would also contain the response structure, with a CoA which marks S’s request that A put the same p-​content into their set of intentions.

18.4.2 Ergative Intransitive Speech Acts Given the logic of our analysis, we expect that ergative speech acts lack the external argument of the grounding structure (Ground-​S) but instead consist of the internal argument only. I propose that this is the case in a speech act known as presentation. This is a type of speech act not often discussed. The term is due to Faller (2002: 16), who argues that sentences with evidential markers have presentative rather than assertive force (see also von Fintel 2003; Portner 2006; Déchaine 2007). According to Déchaine et al. (2014), this speech act has fewer commitments than assertions. In particular, its force is that of presenting a proposition for consideration without making a truth-​claim about it. In other words, S puts forth p without committing to the truth of p (Déchaine et al. 2014: 6). According to Faller 2002 this type of speech act is triggered by evidential marking. The example in (24) is one of Faller’s examples from Cuzco Quechua and Faller’s description is as follows: the speaker brings the embedded proposition into the conversation for consideration. That is the current speaker’s speech act is one of presentation of another speaker’s assertion. […]There is no condition that the speaker believes p, and the illocutionary act is that of present (2002: 198f.). (24) Para-​sha-​n-​si rain-​prog-​3-​si p= ‘It is raining” ill = present (p)

Faller 2002: 199 (165)

Within our analysis, this amounts to saying that S does not express a propositional attitude toward the proposition—​hence it doesn’t enter into Ground-​S. Thus, I hypothesize that a presentative is a clause-​type which lacks the external argument of the grounding layer (Ground-​S), as in (25). (25) 

Presentatives as ergatives GroundP Ground Ground

S p-structure

6 

This corresponds to Portner’s (2004) To-​Do-​list and Han’s (2001) Plan-​set.

 

436   Martina Wiltschko In this way, presentatives are the most basic speech acts: they contain the least amount of structure compared to all other speech act types. This echoes Portner’s 2004 semantic analysis of this speech act, which takes it to be the most basic kind of update.

18.4.3 Summary We have now seen that some of the commonly discussed speech acts (assertions and requests) as well as the less commonly discussed presentation can be analyzed as instantiating the three types of speech acts expected on the assumption that speech act structure parallels argument structure. Without taking into account the second layer above GroundP for now, we can summarize the parallelism as in Table 18.1. Table 18.1 The parallelism between argument–​structure and speech act structure Argument–​structure

Speech act structure

Transitive predicates

Declarative assertion

Unergatives predicates (qua concealed transitive)

Imperative

Unaccusative (ergative) predicates

Presentation

In what follows I show that similar considerations apply to the response structure.

18.5  Ergativity and the Response Structure In section 18.3.3, I have introduced the idea that the grounding structure is dominated by a dedicated layer that encodes CoA. CoA refers to the aspect of a speech act which can be viewed as requesting a response. Hence I refer to this structure as the response–​structure. I propose that the response–​structure itself is complex: it consists of a projection that hosts the subject of the response. In line with the convention established for argument–​structure, where the head which introduces the external argument is labeled (little) v, I use the label resp for this layer. The internal argument of the response layer hosts the object of the response. The full response structure is illustrated in (26).

 

Ergative constellations in the structure of speech acts    437 (26) 

A complex response-structure respP RespSUBJ

resp resp

RespP Resp

RespOBJ Resp

Comm p

If there is indeed another layer above the grounding structure, we expect to find two ergative constellations in the second layer as well. That is, just as there are two types of intransitive argument–​structures and grounding structures, we expect to find two types of intransitive response–​structures. In this section I show that this is indeed the case. I argue that transitive response structures are realized as requests for confirmation (5.1). Furthermore, I argue that polar questions are best analyzed as unergative structures (5.2). As for ergative response–​structures, I argue that this is instantiated by utterance-​ initial ja in Austrian German, which marks the utterance as a response (5.3).

18.5.1 Requests for Confirmation as Transitive Response Structures In a transitive construal, we expect that the utterance is marked for who is the responder as well as what it is that the responder is supposed to respond to. I argue that this is precisely the configuration we find in confirmation requests marked by sentence-​final confirmationals, such as in (27). (27) You are leaving now, huh? I assume that the particle huh marks what S wants A to respond to (namely her believe that p). Thus, huh marks the utterance as the object of the requested response. In this way, Resp has a similar function as v in that it marks the sister of its complement as an object (in this case the object of the response). I further assume that the rising intonation realized on the confirmational encodes CoA (see Heim et al. 2014 for more detailed

 

438   Martina Wiltschko discussion). I further assume that A is associated with the specifier of respP by external merge.7 This is illustrated in (28), where rising intonation is represented as /​. Transitive response structure

(28) 

respP resp

Adr resp-S /

Resp Resp-O huh

GroundP

Evidence that the sentence-​final particle huh marks the object of the response stems from the fact that there are different types of confirmationals. While all of them are used to request confirmation, they differ according to what S wants confirmation for (i.e. what A is asked to respond to). Consider the examples in (29). (29) Mary has been planning to go on a trip for a while but she had never set a date. One day, she decides to pack and hop on the next train. Her roommate John witnesses her packing. a. John: You are leaving now, {eh/​right?} b. Mary: I’m leaving now, {eh /​#right?} Consider first (29)a. In this context, John is pretty certain that Mary is leaving, though there is a chance that she is just packing her clothes to bring them to the dry-​cleaner. John can use (29)a to request confirmation of his belief that Mary is leaving. This differs from (29)b. Mary knows very well that she is leaving, and she is pretty certain that John also knows. However, there is a chance that John may think she is just preparing her clothes for the dry cleaner. Here Mary can use (29)b to confirm that John knows that she is leaving. In other words, S is requesting confirmation that her belief about A’s belief is correct. In both utterances, eh is used to request confirmation. The request itself is encoded by the rising intonation on the particle. However, the utterances differ in what they request to be confirmed. Eh has two possible uses. It can be used to request

7 

It is for this reason that response–​structure cannot be equated with case–​structure. In the domain of case–​structure, argument positions are filled by internal merge rather than external merge (with the exception of expletives).

 

Ergative constellations in the structure of speech acts    439 confirmation of S’s belief regarding p but it can also be used to request confirmation of S’s assumption about A’s belief. Thus, eh marks the propositional attitude of S or A as the object of the response. In contrast, right doesn’t mark the propositional attitude toward the proposition as the object of the response but instead the proposition itself. In other words, with the use of right, S requests confirmation that a given proposition is true. Therefore, the use of right cannot be relativized to what A believes. In sum, while requests for confirmation share several aspects of meaning with regular questions (i.e. both require a response and thus involve a CoA) they also differ from regular questions, as I discuss in the next subsection.

18.5.2 Polar Questions as Unergative Response–​Structures Given the complex SA-​structure in (26), we expect that unergative response–​structures mark the subject of response only. I argue that this is the case in standard questions marked with rising intonation. As in requests for confirmation, rising intonation marks the subject of the response (i.e. CoA), but the object of response is not explicitly marked. Hence, questions are compatible with contexts where S is asking for information because s/​he doesn’t know the answer as in (30)a. But questions are also compatible with contexts in which S knows the answer. This may be in contexts where S wants to test A and hence inquires whether A knows the answer, as in (31). Similarly, in (32) S knows perfectly well that A does not know the answer. This question is only asked to solicit a response from A. (30) Mary has been planning to go on a trip for a while but she had never set a date. Her roommate John wants to know whether she is still planning to go. He asks: Are you still going on your trip? (31)

History teacher to student in grade 5: Did Columbus discover America?

(32) “You know who lives there? Huh? No you wouldn't know who lives there I’m just saying, and you know who lives there?” (scene from Martin Scorsese’s Taxi Driver) In sum, I argue that yes/​no questions with rising intonation are used to mark the utterance as one that requires a response. That is, Ground-​S indicates that the p-​structure of the utterance is in the set of questions (recall from section 18.4.1 that the clause-​type determines the relevant discourse component in GroundP). Furthermore, I  assume that rising intonation associates with respP to mark that A should respond (Heim 2015). Since the object of the response is not explicitly marked (unlike in requests for confirmation marked by tags) this corresponds to an unergative intransitive constellation, as shown in (33).

 

440   Martina Wiltschko (33) 

Unergative intransitive response structure

respP resp

Adr resp-S /

GroundP Ground-S

S p-structure

18.5.3 Ergative Response–​Structure: Marking the Utterance as a Response Finally, we also expect to find utterances in which only the object of response is marked but not the subject. I argue that this is the case in utterances which are themselves marked as a response without requesting a response. Note that exclamatives are S-​oriented speech acts: no response is requested. Nevertheless, exclamatives can still be embedded under the response structure. In particular, in spoken German, any utterance (including exclamatives) can be introduced by a sentence-​initial response particle (jo). It marks the utterance it introduces as a response. To see this, consider the example in (34) and its representation in (35). In (34) it is the absence of a predicted event that triggers A’s exclamation. While the presence of jo is not obligatory, its use is strongly preferred. I suggest that this is because jo explicitly marks the utterance as a response. (34) Context. A and B are co-​workers. Their working hours are fixed and they always go home at 4.30. Typically, they get ready to leave at 4.25 so they can be out the door by 4.30. Today B is not showing any signs of getting ready even at 4.25. A comments: Jo  du   oaweit-​st   heit      long!? jo you work-​ 2sg today long ‘You’re working long today.’

 

Ergative constellations in the structure of speech acts    441 (35)  Ergative response-structure RespP Resp-O jo

GroundP Ground-S

S p-structure

Note that this use of the response particle ja differs from its use as an affirmative response marker in form and function. The sentence-​initial response marker discussed here is prosodically integrated with the utterance it introduces and it cannot stand alone. In its use as an affirmative response marker jo is associated with its own prosodic contour and can thus stand alone. Furthermore, as an affirmative response marker it can only respond to yes/​no questions whereas as a marker of response jo can respond to any speech act as well as non-​linguistic events as in (34). For further discussion of the grammar of this type of response particle see Wiltschko (in press).

18.6 Conclusion In this chapter I have explored the idea that ergative constellations are detectable in the syntax of speech acts. This idea is a logical consequence of Ross’s (1970) performative hypothesis and its more recent incarnations (Speas & Tenny 2003; Haegeman 2013). This is because the logic of ergativity operates on the composition of predicate–​argument structures and speech act structure can be conceived of as a special kind of such structure. In particular, the ergative constellations we observe mirror those found in the domain of argument–​structure. I have argued that SA-​structure is best analyzed as consisting of two distinct layers. The lower layer (GroundP) consists of the utterance and a level where the speech act participant’s commitment to that utterance is expressed, i.e. their propositional attitudes. We have also seen evidence for another layer (RespP) which is dedicated to encoding the Call on Addressee in the sense of Beyssade & Marandin (2006). In particular, I have suggested that this can be understood as a layer responsible for the responsive aspect of language. With this framework in mind, we were able to explore ergative constellations In the domain of argument–​structure ergativity it is widely acknowledged that we need to distinguish not only between transitive and intransitive predicates but also

 

442   Martina Wiltschko between two types of intransitive predicates: unergatives and ergatives (aka unaccusatives). I have argued that we find the same division in the domain of speech act structure. In particular, in the grounding domain the internal argument is the proposition whereas the external argument corresponds to the participant whose commitment toward p is expressed. On this view, declaratives are analyzed as transitive clause-​types, imperatives are analyzed as unergatives, and presentatives are analyzed as ergative speech acts. In addition, we have seen evidence for an articulated response structure, where both subject and object of response can be marked. Both are marked in requests for confirmation (by means of confirmationals) but only one the subject of the response is marked in regular questions. We have further seen that there is a dedicated marker in German which serves to mark the utterance as a response. In the absence of a CoA (as is the case with exclamatives) this results in an ergative constellation in that only the object of the response is marked. For obvious reasons, this chapter has to remain programmatic in nature. The main point to take away from it is that, everything else being equal, we expect ergative constellations not to be restricted to the domain of argument–​and case–​structure. If we are indeed dealing with a matter of structural configurations, as assumed by most work in the generative tradition, then these structural configurations should not be restricted to one particular domain. We have further seen that this avenue of research is fruitful in that it brings to light different types of speech acts including some that are not standardly discussed. However, due to the exploratory nature of this enterprise we only discussed a limited number of such speech ac types. However, if the proposed grammar of speech acts is indeed on the right track, we expect a more complex and fine-​grained typology of speech acts as is typically assumed. It is my hope that this chapter will inspire others to explore the structure of speech acts from this angle.

Abbreviations 2, second person; abs, absolutive; aux, auxiliary; erg, ergative; fem, feminine; Jo, response marker “jo”; masc, masculine; prf, perfective; prog, progressive; prt, particle; spkr, speaker.

 

pa rt i i i

A P P ROAC H E S TO E RG AT I V I T Y

 

 

Diachronic

 

 

Chapter 19

Gr amm atical i z at i on of ergative case ma rk i ng William B. McGregor

19.1 Introduction The grammaticalization of case markers has been discussed in a number of chapters in previous volumes in the Oxford Handbooks in Linguistics series (Heine 2003, 2009; Kulikov 2009; Johanson 2009; König 2011). The present chapter focuses on ergative case markers, which is also touched on in most of the previously mentioned chapters; here we provide a more elaborate and nuanced account. It is beyond the scope of this chapter to examine the grammaticalization of other aspects of ergative patterning, including other types of morphological ergativity (e.g. ergative agreement and cross-​referencing systems, as found in e.g. Sacapultec—​Du Bois 1987b) and syntactic ergativity (e.g. Dixon 1994: 143–​181). Grammaticalization has been conceived of in a wide variety of different ways by scholars (Narrog and Heine 2011: 2–​3). We are concerned here with the development of grammatical items. The standard textbook conceptualization is thus appropriate: grammaticalization is a diachronic processes whereby a lexical item becomes grammatical or a grammatical item becomes more grammatical (e.g. Kuryłowicz 1975: 52; Hopper and Traugott 1993: xv; Lehmann 2002: vii; Matthews 2007: 164). Two qualifications are in order. First, unlike some writers (e.g. Heine 2009), I will not assume that grammaticalization necessarily begins with lexical items: the available historical evidence simply does not support such a strong claim. Second, in discussions of the grammaticalization of case-​marking morphemes metaphor is frequently invoked as an explanatory mechanism. Conceptual motivation is often proposed whereby an abstract domain is construed in terms of a concrete domain (Heine et al. 1991). Not all instances of the grammaticalization of case markers are so motivated, or indeed cognitively motivated in any significant sense (see also Gildea 2004). As will be seen, the grammaticalization of ergative case markers reveals a far richer set of pathways than expected from the literature (e.g. Trask 1979; Estival and Myhill 1988; Dixon 1994; Palancar 2002). This diversity is consistent with the enormous variety within the phenomenon of ergativity itself (DeLancey 2006; Gildea 2004).

 

448   William B. McGregor Ergative case markers can also be the sources of other grammatical markers, as discussed in 19.3. Section 19.4 overviews what can happen to ergative case marking in language contact situations, including language shift and endangerment.

19.2  The Emergence of Ergative Case Markers Four sources of ergative case markers are attested in varying degrees of frequency and with varying degrees of plausibility: Lexical items Markers of other cases Indexical elements Directional elements We examine these sources in order in the following subsections. They are not the only sources that have been identified. One frequently mentioned source is in discourse tendencies: ergative case-​marking systems arise due to tendencies in discourse relating to the flow of information (Du Bois 1987b, 2003b; Hopper and Traugott 1993: 162). The problem is that case markers require some material source (see also Queixalós and Gildea 2010: 13, fn. 20). Nevertheless, 19.2.3 will reveal that a discourse tendency is sometimes relevant. Another possible source is in noun class markers. Sands (1996: 48–​ 49) proposes such an origin for ergative case markers in some non-​Pama–​Nyungan languages of Australia. However, there are more plausible sources in indexical items that simultaneously mark nominal class (Greenberg 1978/​1990; McGregor 2008: 307–​308; Kulikov 2009: 448; and section 19.2.3, this chapter).

19.2.1 Lexical Sources of Ergative Case Markers This is a poorly attested source of ergative case markers, and convincing examples are thin on the ground. Heine and Kuteva (2002: 165–​166) suggest that the lexeme ‘hand’ may develop into an agent marker in passives, though no subsequent development into an ergative marker is attested. Hopper and Traugott (1993: 163) suggest that the preposition oleh is an ergative marker in Malay, and has origins in a verb meaning ‘get, obtain, do, manage, return’ which originally occurred in a separate clause. However, it seems that this preposition is in fact a marker of the agent of a passive construction in Modern Malay. Two ergative markers in Nungali (Mirndi, northern Australia) may perhaps have a lexical source in the Wanderwort mayi ‘vegetable food’. One is the allomorph of the case and class marking prefix mi-​which occurs on adjectives modifying nominals of the vegetable class (Hoddinott and Kofod 1976b: 398; Green and Nordlinger 2004: 306). The other is the suffix -​mayi used on vegetable class nouns (Hoddinott and Kofod 1976b: 398–​399). Sands

 

Grammaticalization of ergative case marking    449 (1996: 48–​49) proposes a development for the latter allomorph from a class marker to an ergative marker,1 though as already remarked, a more plausible pathway of development is from a determiner or pronominal marking the vegetable class. In this scenario, the original lexical source is irrelevant to the development of the ergative marker. The crucial development lay in a single process whereby a paradigm of class-​marking indexical elements were reanalysed as case markers; the ultimate source of each had nothing to do with the process. Coupe (2011a: 32–​34) suggests a lexical source for the ergative marker in the Ao group in the Tibeto-​Burman family. He suggests that the most likely diachronic source of Mongsen Ao ergative -​nə and its cognates within the Ao group, is a relational lexeme ‘side’ in (pre-​)proto-​Ao. This generalized marker of location, Coupe suggests, subsequently developed into a general oblique marker *na in proto-​Ao, and ultimately into an ergative marker (via a process such as discussed under 19.2.2). Coupe (2011a: 33) cites Peterson (2002) on a similar lexical origin for the ergative morpheme in Chin languages.

19.2.2 Markers of Other Cases The most commonly proposed source for ergative case markers is other case markers, especially case markers with more local or concrete meanings. Four are reasonably well attested: instrumental, genitive, oblique, and ablative (e.g. Lehmann 2002: 98–​99; Heine and Kuteva 2002: 180; Palancar 2002; Heine 2009: 467; Malchukov and Narrog 2009: 525). Here we overview a sample; there is no attempt to be comprehensive. Other possible sources include locative and comitative case markers. The former is not infrequently mentioned, though convincing examples are hard to find. In the instances that I have seen, this historical development is presumed merely on the basis of a shared formal marker, and the assumption that local case markers will extend to core cases rather than the reverse. Moreover, the same form usually also marks instruments, so the change may well be via an intermediate instrumental stage. This also seems to be the case for all instances of comitative marker to ergative (Lehmann 2002: 99). Grammaticalization studies have, until recently, tended to focus on the development of individual morphemes in isolation, paying little heed of the constructions in which they occur. The grammaticalization of ergative markers represents an exception. In the early days of awareness of this case-​marking phenomenon linguists were inclined to see ergative constructions as passives (e.g. Pott 1873; Müller 1887; Schuchardt 1895; Uhlenbeck 1916; see further Seely 1977: 197–​199). Although such views are no longer widely held by linguists, even in recent times one finds some who maintain passive origins for ergative constructions. For instance, Estival and Myhill (1988: 445) propose ‘the hypothesis that in fact all ergative constructions have developed from passives’. Slightly less restricted proposals have been made by many others. Trask (1979) proposes two source constructions, passives (for what he calls Type A ergative systems) and perfectives (for Type B ergative 1 

Sands’ scenario is one of simple replacement. She assumes that all Australian languages are related, and that proto-​Australian was an ergative language. For some unknown reason the inherited ergative marker was lost in the development of Nungali, and the class marker was pressed into use as its replacement.

 

450   William B. McGregor systems). Palancar (2002: 221) suggests that ergative constructions developed from transitive interpretations of formerly intransitive (or reduced transitivity) constructions of various types, not necessarily passives or perfectives. Queixalós and Gildea (2010: 13–​17) propose that ergative patterns arise in main clauses via reanalysis of biclausal constructions such as nominalizations, and marked voice constructions such as passives. In these scenarios, ergative markers arise via reanalysis of markers of the (possibly optional) agent phrases in the marked constructions, which became increasingly common until they ultimately replaced the ordinary transitive construction. However, there are many languages for which there is no evidence that ergative markers arose within marked constructions such as passives or nominalizations. This is the case for ergative languages of Australia, where in no known language did an ergative case marker demonstrably arise in a passive or nominalization (e.g. Heath 1980: 907; McGregor 2009: 498). This also seems to be the case for ergative languages of the Eastern Sudanic groups (e.g. Nilotic, Surmic, and the Jebel language Gaahmg) and Tima (Niger-​ Congo) (Gerrit Dimmendaal, p.c.). On the other hand, languages of the Indic branch of Indo-​European are generally considered to provide a ‘textbook’ case for the development of ergative case marking through reinterpretation of a passive construction (Anderson 1977; Comrie 1978; Payne 1980; Dixon 1994: 190; see Stroński 2009: 78–​89 for a detailed overview). The received scenario has it that the mother language Sanskrit was an accusative language, and that the agent NP in an early passive construction was in instrumental case. The passive was reinterpreted as a perfect, and correspondingly the instrumental case of the agent was reinterpreted as ergative case of the Agent in the new transitive construction.2 This change did not occur in the imperfect, which retained nominative–​accusative case marking. Despite its plausibility there are serious difficulties with this proposal (Klaiman 1987; Bynon 2005; Butt 2006a, 2008; Stroński 2009). First, there are problems with the notion that the source construction was a passive. Second, historical–​comparative evidence reveals that the ergative of most modern Indic languages cannot be a reflex of the old instrumental marker (Butt 2006a: 77, 2008). Nevertheless, instrumental case markers are among the most commonly suggested sources of ergative markers, and represent the most plausible sources according to Heine and Kuteva (2002:  180). Garrett (1990) proposes that the Anatolian ergative marker *-​anti developed from an earlier instrument/​ablative marker. He proposes that this development occurred in a transitive construction with an instrumental NP but no subject NP, like ‘(s/​he) strikes it with a dagger’, which he suggests must have existed in Anatolian. This was interpreted, he suggests, as ‘a dagger strikes it’, with the instrument reanalysed as subject. As Garrett (1990: 285) observes, this development does not require that the instrument/​ablative marker was ever a causal marker (as is sometimes presumed—​e.g. Palancar 2002); nor does it invoke metaphoric transfer. 2 

In this chapter I follow the terminology I employed in McGregor (2009: 480) whereby Agent indicates the subject of a transitive clause, Actor the subject of an intransitive clause, and Undergoer the object of a transitive clause. This is a simplification of my proposals (McGregor 1997, 1998, 2002a, 2014), and these roles should not be equated with the familiar but problematic labels S, A, and O of linguistic typology—​see also Mithun and Chafe (1999). Lower-​case terms indicate etic relations.

 

Grammaticalization of ergative case marking    451 Foley (1986: 107) observes that in a number of Papuan languages (including Enga, Kewa, Kâte, Dani, and Selepet) the same case marker is used on Agent and instrument NPs, and suggests that the ergative usage represents an extension from the instrumental. The problem is that there is no supporting historical–​comparative evidence. Garrett (1990: 280–​285) presents some such evidence for the ergative marker in the Gorokan languages of the Eastern Highlands of Papua New Guinea. Internal evidence suggests that a single marker in the proto-​language had both ergative and instrumental uses. Garrett (1990: 285) suggests further that the Gorokan morpheme is the reflex of a Common Eastern Highlands morpheme that had only instrumental uses. Genitive markers are also common sources for ergative case markers. In the Iranian branch of Indo-​Iranian a possessive perfect construction was reanalysed as an active main clause; the genitive-​marked possessor in the original nominalized construction was correspondingly reanalysed as an ergative marker (e.g. Benveniste 1952/​1971; Allen 1964; Anderson 1977; Trask 1979; Dixon 1994: 191; Stroński 2009: 79–​80). Bynon (2005) suggests that the original ergative marker in Indic languages was also a genitive, and that it was subsequently replaced by an instrumental marker (see also Stroński 2009: 95–​96).3 Dimmendaal (2014: 12) suggests that ergative markers in the few morphologically ergative African languages derive from genitives (Nilotic languages such as Päri—​cf. however 19.2.2) or instrumentals (Surmic languages, e.g. Majang). Furthermore, the Shilluk (Nilotic) ergative preposition yɪ is identical in form with the widespread oblique preposition of Western Nilotic (Dimmendaal 2014: 13). Ablative sources of ergative markers have been proposed for some languages, often via a causal (sometimes genitive) intermediary. The metaphorical basis is apparent: agents are sources and causers of actions and events; the target domain of agentivitity is plausibly construed in terms of a spatial source domain. Palancar (2002: 253–​255) observes that the Basque ergative marker -​k is formally similar to the ablative -​tik, and suggests that the ergative derives from an earlier ablative, while the modern ablative involves additional phonological material. In support of this, he observes that remnantal ablative uses of -​k survive in the Guipuzkoan dialect and that a causal sense survives in some idiomatic expressions. In some languages, ablative markers assign focus to Agent NPs. Thus in the Mirndi languages Jaminjung and Ngaliwurru (Schultze-​Berndt 2000: 168), the ergative marker can be replaced by the ablative to assign contrastive focus to an Agent NP. In nearby Kija (Jarrakan), which is not morphologically ergative, the ablative case marker can be employed on an Agent NP to the same effect. These examples indicate that there is no need for a causative intermediate stage. The above instances concern developments from markers of non-​core cases to markers of core cases. In some ergative languages of Africa ergative markers may have sources in markers of core cases, in particular in markers of nominative case. In East Africa, a number of languages are found in which the nominative is the marked case (König 2006, 2008a, 2008b, 2009). The relatively few ergative languages of the African continent occur in the 3 

In view of the historical–​comparative facts mentioned this is improbable. However, it could well be that some other marker later replaced the genitive. This could be a dative marker, as suggested by Butt (2006a, 2008)—​though Butt cautions that the origin of the form is uncertain.

 

452   William B. McGregor same region. In some of them an origin of the ergative case marker in a nominative marker is plausible; the reverse direction may also have occurred in some languages—​see 19.3. Päri shows a number of ergative characteristics including ergative case marking of Agent NPs in transitive indicative clauses, as shown in (1) (Andersen 1988). The same marker, -​ɪ ̀ ~ -​ɩ ̀ ~ -​è ~ -​ɛ ̀ (here I gloss over details of allomorphy), is a nominative case marker in other clause types (including most subordinate clause types, imperative clauses, and indicative clauses with marked focus). (1)

dháagɔ̀ á-​yàaɲ ùbúrr-​ɩ ̀ woman C-​insult Ubur-​ERG ‘Ubur insulted the woman.’

Päri (Andersen 1988: 292)

Andersen (1988) argues that the marked nominative pattern was historically prior, previously found in all clause types, and that the ergative marker arose through reanalysis of the nominative marker. He suggests that the earlier system showed VSO and VS word orders, like some modern languages. A topicalization pattern emerged in which postverbal NPs were fronted, and lost their case marking. This topicalization pattern became normal for Actors and Undergoers, giving rise to the modern unmarked word orders SV (intransitive clauses) and OVS (transitive clauses). The result of these changes would have been an ergative system as found in modern Päri, with Agent marked by the old nominative case marker, and Actor and Undergoer not case marked. Consistent with this, the Päri ergative marker may be cognate with the nominative singular marker -​i ~ -​e of distantly related Murle. König (2006: 706–​708) agrees that the ergative marker of Päri derived from a nominative marker. She suggests a slightly different diachronic scenario:  from definite marker to nominative marker, and thence to ergative marker. She observes that the Päri ergative marker is cognate with a definiteness marker in Anywa (which is in a dialectal relation to Päri), that is optionally used on post-​verbal Actor NPs in both transitive and intransitive clauses.4 Significantly, Andersen (1988: 318) gives an example of an intransitive Actor NP in Päri marked by the ergative in what he suggests is a type of cleft construction. One wonders whether it might not be possible to account for the ergative and the marked nominative markers in Päri as independent developments from an earlier definiteness marker as per the scenario outlined in the next section.

19.2.3 Indexical Elements Oblique markers in passives, possessive perfects, and nominalizations are the most over-​rated sources of ergative case markers. Indexical items such as pronominal 4 

Dimmendaal (2014: 8–​9) questions the analysis of this marker as definiteness marker, and proposes that it is an ergative marker. The problem is that both Agent and Actor NPs can take the marker when in post-​ verbal position (Reh 1996: 137). This is not to suggest that I find Reh’s analysis entirely convincing. Her claim that the morpheme in question is not a case marker seems dubious: possibly Anywa has an optional marked nominative, or the marked nominative may be a portmanteau morpheme that also indicates definiteness.

 

Grammaticalization of ergative case marking    453 elements and determiners are by contrast the most underrated sources. Indeed, in many general survey treatments they either fail to be mentioned at all or are mentioned only in passing, and fail to be incorporated into the primary story (e.g. Dixon 1994: 182–​206; Heine and Kuteva 2002; Palancar 2002; Gildea 2004; Heine 2009: 467; König 2011). This is despite the considerable body of evidence that has accumulated over the past century or so, including in sources used in some of these works. Furthermore, it is remarkable that the very person who appears to have been the first to use the term ergative, albeit for a local case, the missionary George Taplin (1872: 85–​ 86; see Lindner 2014), remarks that in the Ngarrindjeri language of the Adelaide region, South Australia: The causative [i.e. ergative in modern terminology] is formed in the singular by the affix il—​evidently an abbreviation of the pronoun kili (by him); as, Kornil mempir napangk, or inangk nap—​The man struck his wife. (Taplin 1880: 8)

Support for Taplin’s proposal comes from the formal identity of the ergative marker and -​il, the third person ergative bound pronoun (Cerin 1994), cited in Simpson et al. (2008:  104). As this example indicates, the indexical item in question often marks something else as well as indexicality, often case, sometimes noun class. However, as suggested below, it is the indexical component that is principally relevant to this grammaticalization process. As mentioned in 19.2.1, in some members of the Mirndi family nominals take portmanteau affixes indicating both ergative case and noun class. For instance, Jingulu has ergative allomorphs -​(r)ni on masculine nominals, and -​nga on feminine nominals (Chadwick 1976: 394–​395; Pensalfini 1999: 227; Green and Nordlinger 2004: 306). These are plausible reflexes of earlier gender markers *(r)ni masculine and *ngayi feminine, that were simultaneously third person forms, possibly oblique case forms of demonstratives or pronouns (Chadwick 1976: 394; see also Blake 1988: 12–​13, 33). Similarly, Nungali has corresponding masculine and feminine ergative prefixes nyi-​and nganyi-​~ nyanyi-​ (Hoddinott and Kofod 1976b: 397–​398), which may also derive from earlier demonstratives or pronominals that distinguished masculine and feminine genders. This renders more plausible the suggestion already made that the ergative formatives occurring on vegetable class nouns and modifying adjectives might have originated in indexicals. For the remaining class in Nungali, things are not quite so clear, though there is a recurrent (w)u that might be a reflex of an indexical item marking this class (such a class is found in some other languages of northern Australia). In most other Mirndi languages ergative markers do not show different forms according to nominal class. Nonetheless, the elsewhere ergative allomorph is -​ni in Jaminjung and Ngaliwurru (Schultze-​Berndt 2000: 54–​56) and Wambaya (Nordlinger 1998a: 83). This is presumably cognate with the masculine form in Jingulu and Nungali. Nyulnyulan languages have a bound ergative postposition showing a variety of allomorphs, including -​ni ~ -​rni ~ -​in ~ -​nim ~ -​nimaa ~ -​nma ~ -​na. These forms could be reflexes of the *(r)ni masculine indexical referred to in the previous paragraph. Significantly,

 

454   William B. McGregor there is evidence from the modern languages suggestive of a third person singular pronominal form *(yi)ni in proto-​Nyulnyulan (Stokes and McGregor 2003; McGregor 2008: 309). Some Daly River languages also show an ergative form that might have derived from an earlier pronominal element ni, e.g. Ngangikurrunggur -​ningke and Ngan’gityemerri –​ninggi (cf. Blake 1988: 45, who interprets the initial ni element as a class marker). An interrogative indexical may be the source of the ergative marker in Garrwa, a non-​Pama–​Nyungan language of the Gulf of Carpentaria region (see further below). The elsewhere ergative allomorph has the form  –​wanyi, which is identical with the free nominative interrogative wanyi ‘what, who’. The other allomorph, -​nyi, could be a truncation. Proto-​Pama–​Nyungan was almost certainly an ergative language, and most languages of this family have ergative markers inherited from the proto-​language (Blake 1987a; Sands 1996: 1–​39). In a few languages, however, the ergative markers appear to be recent innovations, deriving from indexicals. Wangkumara distinguishes two noun classes:  masculine singular and non-​ masculine-​ singular, encompassing everything else (Breen 1976). Case-​ marking of arguments is by bound pronouns attached to the relevant nominal. Three case forms are distinguished for all nominals and free pronouns: nominative (for Actors), ergative (for Agents), and accusative (for Undergoers). The masculine singular ergative marker is -​(u)lu (a reduced form of the free masculine singular ergative pronoun nhulu)5 while the non-​masculine-​singular ergative is -​(a)ndru (a reduced form of the free form nhandru), as shown in (2) and (3). (2)

(3)

karna-​ulu kalka-​nga thithi-​nhanha person-​ERG hit-​PST dog-​ACC ‘The man hit the bitch.’ makurr-​andru nganha kabiliba-​nga tree-​ERG 1SG.ACC block-​PST ‘The trees blocked me.’

Wangkumara (Breen 1976: 337) Wangkumara (Breen 1976: 337)

In the Pantyikali and Southern Paakantyi dialects of Paakantyi bound forms of two ergative demonstratives, ‘this’ and ‘this here’, can be attached to nouns as ergative markers (Hercus 1982: 63), as in (4). The free form can be used instead, as in (5), which apparently express an emphatic meaning. (4) yartu-​thurru kaanti-​t-​urru-​ana wind-​DEM:ERG carry-​FUT-​3SG.NOM-​3SG.ACC ‘The wind will carry it along.’

Southern Paakantyi (Hercus 1982: 63)

5  This form of the ergative shows the need for caution in identification of forms on the basis of allomorphs in isolation. Seen by itself one might wrongly conclude that the -​(u)lu allomorph is a direct inheritance from the proto-​Pama-​Nyungan ergative marker *-​lu (e.g. Blake 1988: 27).

 

Grammaticalization of ergative case marking    455 (5) kajiluku marlibarlu irnuruu-​dinga wadu-​ji-​na Southern Paakantyi small boy DEM:ERG-​EMP take-​PST-​3SG.ACC ‘It was the small boy who took it.’  (Hercus 1982: 63) Similarly in Paarruntyi in addition to the inherited system another mode of expression is available for singular Agents: a free resumptive ergative pronoun alongside of (though not necessarily adjacent to) the unmarked Agent NP, as in (6). (6) nganha karli-​wa thatya wuthu-​rru nguma my dog-​EMP bite PST:3SG-​ERG 2:SG:ACC ‘My dog bit you.’

Paarruntyi (Hercus 1982: 62)

There are a number of less certain cases of this developmental pathway. The Umpithamu ergative marker -​mbal may be a reflex of a widespread demonstrative form pala (Jean-​ Christophe Verstraete, pers. comm.). The ergative markers in Wardaman are -​yi ~ -​ji ~ -​nyi; similar forms occur in Gunwinjuwan varieties:  -​yih (Dalabon, Rembarrnga, Ngalakgan), -​yi ~ -​yih (Kune, Manyallaluk Mayali), and -​yi (Warray). The first allomorph in Wardaman is identical in form with one of the three nominal class-​marking prefixes, yi-​for animates and humans. Moreover, yi occurs as a component of third person pronominal forms in many languages from the Kimberley to Arnhem Land, and as an agreement class-​marker in various languages, including the Gunwinjuwan language Nunggubuyu (yii-​). It seems more probable that these ergative forms developed from a class-​marked indexical element than that they are reflexes of a proto-​Australian ergative allomorph *-​DHu (McGregor 2008: 314; cf. Sands 1996: 66). McGregor (2008: 310–​313) suggests an explanation for this grammaticalization scenario. Things might have begun with a source construction comprising an NP in apposition with a determiner (definite or indefinite/​interrogative) or pronoun (see also Green and Nordlinger 2004: 304). Such a construction might have been used to single out unexpected Agents in discourse, while expected Agents could have either been ellipsed entirely, or have been represented by unmarked NPs. That such a construction might be dedicated to singling out Agent NPs to the exclusion of NPs in other grammatical roles is not unreasonable if we assume the preferred argument structure (PAS) constraint was in operation (Du Bois 1987b, 2003b). According to the PAS, the Agent role is strongly associated with given information. If an Agent NP conveys new information this marked pragmatic circumstance might well be formally marked. On the other hand, Actor and Undergoer are cross-​linguistically preferred sites for introduction of new items into discourse, and thus of unexpected NPs. There should thus be no strong need to signal that NPs in these roles are unexpected. Thus juxtaposition of a resumptive pronominal to NPs in these roles would be unnecessary. The PAS constraint concerns information status of the Agent. If the indexical item (and cleft construction) began to be used with new Agent NPs, it is not implausible that it might generalize to marking Agent NPs that are unexpected for other reasons, such as low animacy or relative powerlessness. We cannot exclude the possibility that the

 

456   William B. McGregor initial motivation for use of the cleft construction was not something else, for instance the animacy hierarchy (Silverstein 1976). It is possible that some traces of different initial points might remain. Thus, in the Cape York Pama–​Nyungan language Wik-​Mungkan the ergative marker for ordinary NPs is the clitic -​ang, which is formally identical to the accusative marker for pronouns (Kilham 1977: 56); similarly, in Kala Lagaw Ya -​n is an ergative marker on common nouns and an accusative on singular pronouns (Kennedy 1984: 156). Assuming that these are innovations that can be traced back to cleft constructions, an explanation is possible for these unusual systems. The cleft may have been used with phrases that were unexpected in Agent and Undergoer roles according to animacy—​inanimate NPs in the Agent role and pronominals in the Undergoer role. Right from the very beginning the indexical item in the cleft construction would have been closely associated with the grammatical role of Agent. Over time, this construction might have increased in frequency, and begun to lose its association with unexpectedness. The presence of the resumptive item might become the norm in Agent NPs. The resumptive item might not occur on Agents that are expected, but be used otherwise, as seems to be the case in Warrwa and other Nyulnyulan languages. Alternatively—​or subsequently—​it might become obligatory. The emergence of indexical-​based ergative markers in Paakantyi varieties appears to be recent. The inherited ergative marker -​rru is rare in most dialects (except Paarruntyi), possibly a consequence of other changes in the grammar such as rigidification of word order. The inherited marker may thus be in the last stages before complete disappearance. In this context, the PAS constraint might motivate the appearance of the new constructions involving resumptive items. Although many of the source forms proposed above involve markers of case and/​or class as well as indexicality, it is the latter component that evidently drives the grammaticalizations. The beginning is in a type of cleft construction that assigns focus to unexpected Agents. The other grammatical categories marked by the form are relevant to the development insofar as they are selected by the grammatical context. This scenario also provides a plausible account for the development of the ergative marker from an interrogative determiner in Garrwa. It is not suggested that this explanation accounts for all instances of the rise of ergative markers from indexicals. It does not explain the situation in Wangkumara, where nominative and accusative case-​markers have also grammaticalized from free pronouns. PAS does not predict marked treatment of unexpected (or expected) Actors or Undergoers. Perhaps in this language the inherited ergative marker wore down phonologically, and (in the absence of rigidification of word order), the pronominal items in the appropriate case forms came to be used to mark (indeed, over-​mark) grammatical relations. Alternatively, it may have been that unexpected or unpredictable NPs in any role came to be highlighted—​after all, according to the PAS Actor and Undergoer roles are associated with both given and new information, and there could well have been a need to mark the difference. Over time and with increased usage these markers could have lost their focal values and became the markers of the three core grammatical relations. Formal similarity of ergative case markers and indexical items is not restricted to the Australian continent.

 

Grammaticalization of ergative case marking    457 In Trans-​New Guinea Tauya the ergative suffix -​ni is formally identical with the third person singular pronoun ni (MacDonald 1990: 322). MacDonald (1990: 322–​323) suggests that transitive clauses showed SOV word order, and that those with third person singular objects in ni became increasingly frequent. This led to the reanalysis of ni as a suffix to the subject of transitive clauses with third person singular objects; this suffix was subsequently generalized to occur in all transitive clauses. The proto-​Polynesian ergative preposition *e is, according to Kikusawa (2002: 155), most likely the personal noun marker *i or *ii. The ergative preposition e in the Oceanic language Roviana may derive from the proto-​Western Oceanic personal article *e (Corston 1996: 61). Ergative case markers -​mə and -​n in the North-​West Caucasian languages Kabardian and Ubykh respectively may trace back to mə and jəna ‘this’ (Kumaxov 1971: 43, 158, 1989:31f, cited in Kulikov 2006, 2009). The Georgian ergative suffix -​ma ~ -​m evidently derives from Old Georgian -​man, which is identical in form with a third person ergative pronominal (Boeder 1979: 457–​458).

19.2.4 Directional Elements An unusual source of ergative marker is a directional element indicating ‘hither’. The Sahaptian language Nez Perce of north-​western USA has the ergative marker -​nim ~ -​nm—​as shown in (7)—​which has such a source in a ‘hither’ marker according to Rude (1991, 1997).6 Correspondingly, the accusative marker -​ne may have grammaticalized from a ‘thither’ marker. (7) wewúkiye-​ne pée-​’wi-​ye háama-​nm elk-​ACC 3ERG:3ACC-​shoot-​PST man-​ERG ‘The man shot an elk.’

Nez Perce (Rude 1991: 25)

Given the metaphoric source of ergatives in ablatives frequently alluded to (see 19.2.2) this is an unexpected pathway of grammaticalization. Rude (1991), however, provides a plausible explanation. He suggests (Rude 1991: 45) that the Sahaptian ergative marker can be traced back to a locative *m in pre-​Sahaptian-​Klamath. This developed in two directions: to a genitive case marker, and to the directional marker *ɨm ‘hither’ in proto-​ Sahaptian. The latter was subsequently reanalysed in two ways: as a verbal suffix ‘hither’, and separately as -​(n)im ergative when the object was a speech act participant, i.e. first or second person. In a transitive clause describing action directed to a speech act participant it is possible that the ‘hither’ marker was employed on both the verb and the Agent NP. In the former case, this would indicate direction towards the speech act participants. In the latter case, being restricted to transitive clauses, the marker could well be reinterpreted as 6 

Nez Perce does not show prototypical ergative–​absolutive patterning in its case-​marking system. First and second person NPs follow a nominative–​accusative case-​marking system. Third person NPs have a three-​way case-​marking system distinguishing ergative (marking Agents), nominative (marking Actors), and accusative (marking Undergoer) (Rude 1991: 25). This does not render the label ergative inappropriate for the case marker of the Agent NP: it is an ergative marker in a tripartite system.

 

458   William B. McGregor a marker of Agent when the event is directed to a speech act participant. Significantly, in modern Sahaptin the ergative marker -​(n)ɨm—​a reflex of -​(n)im—​is restricted to transitive clauses in which a third person Agent is acting on a first or second person Undergoer (Rude 1991: 27). In Nez Perce the nominal marker may have undergone further reanalysis and generalized to transitive clauses regardless of the person of the Undergoer. Consistent with this grammaticalization scenario first and second person pronouns do not take ergative marking. With a speech act participant as Agent the ‘hither’ directional element should not be used, the action being directed away from the speech act participant. This would remain an inexplicable restriction if the ergative had arisen via grammaticalization of the genitive.

19.3  Developments from Ergative Case Markers Ergative case markers can disappear virtually without trace, as seems to have happened to the inherited marker in some Pama–​Nyungan languages (as discussed in 19.2.3). They may also be reanalysed as markers of other cases and other grammatical categories. Below I discuss a few examples; considerations of space preclude comprehensive treatment. A few Pama–​Nyungan languages have lost their ergative morphology, and have become nominative–​accusative. This happened in many Ngayarda languages of the Pilbara region of Western Australia, including Panyjima, Ngarluma, Yinjiparnti, and Martuthunira; others, e.g. Ngarla, Nhuwala, Nyamal, have remained ergative. A plausible grammaticalization pathway is adumbrated in Dench (1982). He proposes that alongside the ordinary transitive construction there was a marked intransitive alternant, in which the erstwhile Agent was unmarked, while the Undergoer was in the dative case. Such a construction is attested in various languages of the region, and is often associated with lack of affectedness of the Undergoer. This may have become an increasingly popular means of expression and ultimately have been reanalysed as the transitive construction.7 The result would be a language with an accusative case-​marking system, where the accusative case was identical with the former dative case. This is shown in (8), where -​ku is also a marker of datives and benefactives, and a reflex of the proto-​Pama–​Nyungan purposive marker. (8) ngunha marlpa yukurru-​ku wiya-​rna that man dog-​ACC see-​PST ‘That man saw the dog.’

Panyjima (Dench 1982: 47)

Reflexes of the proto-​Pama–​Nyungan ergative marker remain in the accusative Ngayarda languages. One is the instrumental marker, which has a rather restricted usage—​e.g. in 7  Dench (1982: 58) suggests that the increasing popularity of this mode of expression may have been motivated by a pressure for the morphological case-​marking system to align with the syntactically accusative system.

 

Grammaticalization of ergative case marking    459 Panyjima it is used only on inalienably possessed instruments (Dench 1982: 47). Another is in the marker of the agent of the various types of passive constructions, which include two inflectional passives (as in (9) and (10)) and a derivational passive (as in (11)). As Dench (1982: 55–​56) observes, the inflectional passives highlight the affectedness of the semantic undergoer; they would have been incompatible with the semantics of the construction in which the undergoer was marked by the dative. Thus they retained the ergative case-​marking system of the proto-​language. After the dative construction had been reanalysed as the unmarked transitive construction, they were reanalysed as passives. The case marking of the derivational passive may have been modelled on the inflectional passives. If Dench is right, the marking of agent of the passive does not derive from the instrumental use of the original ergative marker, but from its Agent-​marking use.8 This would thus be a genuine instance of development from an ergative case marker. (9)

panha yalha ngayi-​rnaanu wirrpi-​ngku that shed throw-​PFV.PASS wind-​INS ‘That bough-​shed has been wrecked by the wind.’

Panyjima (Dench 1982: 49)

(10) ngatha pilanyja-​ku katama-​lpuru ngunha-​jirri-​lu 1SG.NOM frightened-​PRS hit-​POT.PASS that-​PL-​INS jilyantharri-​lu children-​INS ‘I’m frightened I might get hit by those children.’ (11)

yukurru ngunha-​ngku marlpa-​ngku wiya-​nnguli-​nha dog that-​INS man-​INS see-​PASS-​PST ‘The dog was seen by that man.’

Panyjima

Panyjima (Dench 1982: 47)

Dimmendaal (2010a: 34, 2014) suggest—​contrary to Andersen (1988) and König (2006, 2008b, 2009: 546–​547)—​that nominative markers in all Nilotic and Surmic languages have origins in ergative markers that extended in usage to cover Actors. He proposes that ergative marking of Agent NPs in post-​verbal position extended to marking of post-​verbal Actor NPs, thus giving rise to marked nominative systems. The latter order appears to be an innovation. Tennet subordinate clauses show pre-​verbal Actor NPs, and 8 

Most Nyulnyulan languages have a medio-​active construction which shows identical case marking as ordinary transitive clauses but the verb cross-​references the unmarked NP rather than the ergatively marked NP as in ordinary transitive clauses (McGregor 1999b). This construction is semantically almost identical with the perfective passive of Ngayarda languages. One wonders whether the perfective passive—​and perhaps also the potential passive—​might not be traceable back to earlier marked constructions that were not ordinary transitives. This does not affect the observation that it was the ergative use of the case marker that was reanalysed as marker of the agent in passives, not the instrumental. However, Dench’s scenario implies that an ergative marker of the Agent role has become a marker of the agent of a passive, which change must be a degrammaticalization, given the usual assumption that the change from marker of agent of a passive to an ergative marker is an instance of grammaticalization. My alternative does not invoke such a change, since the Agent of the medio-​active is not a core participant in the clause (see McGregor 2014 for further discussion).

 

460   William B. McGregor post-​verbal Agent NPs (Dimmendaal 2014: 9). Given that subordinate clauses are typically more conservative than main clauses, they may well reflect the typical word order of the proto-​language. It would show an ergative case-​marking system: the post-​verbal Agent NP would be marked, while the pre-​verbal Actor in intransitive clauses would not be marked morphologically. If main intransitive clauses innovated a word order in which Actors could occur post-​verbally (possibly motivated by information distribution), the ergative marker might occur on these NPs, resulting in a marked nominative system. While this seems plausible, there remain doubts about its universality in Nilotic and Surmic languages. Thus, as we have seen, in Päri evidence from some subordinate clause types—​exactly the same type of evidence employed by Dimmendaal (2014)—​suggests that the marked nominative system was earlier, and developed into an ergative system. This grammaticalization pathway is not restricted to the African continent. In the Kartvelian language Mingrelian the erstwhile ergative marker, restricted to the aorist, extended to cover intransitive Actor NPs, thus resulting in a marked nominative system (Dixon 1994: 202). Similarly, in Pitta-​Pitta the ergative marker is -​lu, a reflex of the proto-​ form; this is used only in non-​future contexts. In the future, -​ngu, evidently a reflex of the ergative allomorph *-​ngku of proto-​Pama–​Nyungan, is a nominative marker occurring on both Agent and Actor roles (Blake 1979c: 193). As remarked in 19.2.2 the usual interpretation of the situation in which Agent of a transitive clause is marked in the same way as an instrument is that the instrumental use is historically primary, and that this developed into an ergative marker. In some languages in which the ergative derives from an indexical item we also find a single form with both ergative and instrumental uses, e.g. Jaminjung (Schultze-​Berndt 2000: 54). Given the grammaticalization scenario of 19.2.3 it seems improbable that use as an instrumental marker preceded use as an ergative marker: it is unlikely that a cleft-​like construction would begin with instrumental NPs, and expand to Agent NPs. Moreover, PAS accounts for a beginning in Agent marking. In such circumstances the ergative marker appears to have extended to cover instrumental NPs. Wangkumara lends support to this scenario. Instrumental NPs in this language are marked by the non-​masculine-​singular ergative allomorph -​(a)ndru; the masculine-​singular is not attested in this function, apparently regardless of the gender of the instrument (Breen 1976: 338). It is improbable that the instrumental use, in which the bound pronominal is attached to an inanimate noun, would predate use on an animate noun. Furthermore, Breen (1976: 338) provides an example of this marker extending into the accompaniment domain usually marked by the comitative -​bartu: (12) karnia yanthangaga pakarranyi-​andru, marli-​andru kandigandi-​andru man go/​PST/​? boomerang-​ERG spear-​ERG axe-​ERG ‘The man came here with his boomerang, his spears, and his axe.’    Wangkumara Markers of ergative case sometimes give rise to markers of other grammatical categories, including subordinate clause markers and tense/​aspect markers; see 19.4 for possible development into a focus marker. In Newari (Tibeto-​Burman) the ergative/​instrumental case marker -​na is used as a temporal clause subordinator, as in (13). Genetti (1991) argues

 

Grammaticalization of ergative case marking    461 that the subordinate clause usage developed from the case-​marking use, though it is not clear whether or not this was via the ergative or instrumental use or something else such as comitative—​cf. ‘with his coming to the house …’. Heine (2009: 468) suggests that the case marker first extended from nouns to nominalized verbs, and then to subordinate clauses. (13) chẽ-​ku yer-​na wā ām-​e naku moŋ-​an Newari house-​LOC come-​SUB EMP 3SG.M-​GEN cheek swell-​PRT coŋ-​gu stay-​3SG.PST.HAB ‘When he came to the house his cheek was swollen.’ (Genetti 1991: 227) In a number of Australian Aboriginal languages, the ergative marker is also found on secondary predicates, typically on Agent NPs (McGregor 2005), external inalienable possessions (McGregor 1999a), and sometimes on manner adverbials (not necessarily in transitive clauses). This is the case in the Bunuban language Gooniyandi, where the same marker, -​ngga ~ -​ga, is also found on non-​finite clauses indicating attributes of Actor NPs (McGregor 1990: 178, 396–​397). A plausible line of development is that the ergative marker first came to be used on secondary predicates on Agents, perhaps to make it clear which role the secondary predicate applied to. From there an extension to marking body parts and the like in inalienably possessed constructions (a type of secondary predicate construction in Gooniyandi) and nominals indicating a quality of the Agent relevant to the event (e.g. speed, size, condition, mental state) is not unreasonable. The marker could then have been extended further to non-​finite verbs that indicated a concomitant state of the Agent, or a concomitant event the Agent was engaged in. Why this extended to non-​finite verbs indicating concomitant states of Actors in intransitive clauses remains a puzzle. Admittedly this scenario is entirely hypothetical; nevertheless, it is based on the type of changes that are not infrequently found in grammaticalization. In Kala Lagau Ya virtually all case markers, including the ergative, are used as markers of tense and aspect on verbs (Kennedy 1984). The ergative -​n marks completive aspect. However, the ergative also marks accusative on singular pronouns, and it is just as plausible that the extension into the verbal domain began with the accusative as the ergative (Blake 2001: 180).

19.4  Ergative Case Marking in Language Contact and Contact Languages There can be little doubt that language contact can be a factor in the emergence and development of ergative case markers and systems. The areal clusterings of languages with ergative case marking attests to this. It has been suggested that ergativity in

 

462   William B. McGregor Indo-​Aryan originated in the contact situation within the South-​Asia linguistic area (Stroński 2009: 103). However, there are difficulties with this proposal, and convincing morpho-​syntactic convergences with known possible substrate languages are lacking (Stroński 2009: 104). Contact with Dravidian languages may have contributed to the gradual loss of ergative patterning in Indic languages. Ergative case marking in non-​ Pama–​Nyungan languages of northern Australia may result from contact with Pama–​ Nyungan languages. The fact that ergative and marked nominative languages in Africa are areally restricted suggests that these phenomena may be contact-​based. Ergative case markers are perhaps sometimes borrowed. Anderson (1977) suggests that the Sinha ergative was borrowed from Tibetan. The Hindi/​Urdu ergative postposition ne has been attributed to borrowing from Tibeto-​Burman languages, a number of which show an ergative form like na (Zakharyin 1979). Coupe (2011a: 31) proposes that the ergative marker of Chungli (of the Angami-​Ao group in Tibeto-​Burman) was borrowed from Chang (Bodo-​Konyak-​Jinghpaw). In extreme circumstances of language contact morbidity of one or more of the languages of a speech community can result. In many Australian Aboriginal communities in post-​contact times there has been marked shift away from the indigenous languages towards English lexicalized varieties such as Kriol and Aboriginal English. The traditional languages sometimes undergo grammatical changes as speakers use them in increasingly reduced circumstances. Case markers, including ergative case markers, often show reduced allomorphy. Schmidt (1985: 383–​386) documents the reduction in allomorphy of the ergative case marker in the language of young Dyirbal speakers in the early 1980s, initially to a single allomorph, later to its complete disappearance in the youngest generation of speakers. Other Australian languages show similar losses of the ergative, with or without allomorphic reduction. In the case of Nyulnyul, now without mother-​tongue speakers, the last two fluent speakers appear to have used the traditional optional ergative system (McGregor 2002b). However, two semi-​speakers I interviewed in the 1980s showed new systems. One had apparently lost the ergative marker entirely, using English-​based word order to distinguish arguments. The other generalized ergative marking to all Agent NPs obligatorily. A different set of changes occurred in Jingulu. Traditional Jingulu was a morphologically ergative language in which the ergative marker was used obligatorily on all Agent NPs:  nouns showed an ergative–​absolutive system of marking, while pronouns were marked according to a tripartite system (Chadwick 1976; Pensalfini 1999: 226). In the late 1960s when Neil Chadwick began his work on Jingulu, emphatic or focal markers appeared that had not been present when Ken Hale did fieldwork on the language a decade earlier. These markers are formally identical with ergative and dative allomorphs. Pensalfini (1999: 238) suggests that they indeed derive from these case markers by reanalysis that was initially induced by contact with Kriol. In the Jingulu of the 1990s the erstwhile ergative marker occurs not just on Agent NPs, but also on Actors and Undergoers. However, when it occurs on an Agent NP, the focal marker follows an instance of the same marker. Pensalfini (1999: 238) suggests that when the present-​day speakers were children learning the language in an environment in which Kriol was becoming increasingly prominent they first analysed Jingulu as a nominative–​accusative language, and reanalysed the ergative and dative markers as markers of discourse prominence. Later, with more exposure to the traditional

 

Grammaticalization of ergative case marking    463 language they learnt that the ergative and dative markers were indeed case markers, but the discourse usages had already been established, and were not unlearnt. Thus ergative and dative markers coexisted alongside of the prominence markers that they gave rise to. The Amazonian language Tacana, unlike the other members of the Tacanan family, shows a system of optional ergative marking. Guillaume (2014) suggests that this is most likely an innovation of Tacana, possibly resulting from language obsolescence and/​or influence from Spanish, and that the proto-​language had a system of obligatory ergative marking. New languages sometimes arise in contact situations (Bakker and Matras 2013). There appear to be no instances of pidgins or creoles with ergative case markers. However, a few mixed languages with ergative case marking are attested. The Australian mixed languages Gurindji Kriol and Light Warlpiri are examples (McConvell and Meakins 2005; Meakins and O’Shannessy 2010; Meakins 2011; 2014a). These languages arose in contact between the traditional Aboriginal languages Gurindji and Warlpiri, both Pama–​Nyungan, and Kriol, an English-​lexicalized contact variety that arose in Northern Australia sometime in the twentieth century. Meakins and O’Shannessy (2010) argue that the competition between the two systems of argument marking, the ergative case-​marking systems of Gurindji and Warlpiri and word order in Kriol, gave rise to a new system in the mixed languages. This is an optional ergative system, in which use of the ergative marker accords prominence to the Agent NP. There are other differences between the ergative case-​marking system of the mixed and traditional languages. In Gurindji Kriol ergative marking has generalized to all Agent NPs—​any Agent NP may be ergatively marked regardless of its animacy—​whereas traditional Gurindji showed split ergativity, whereby free pronouns did not take ergative marking. In traditional Warlpiri, ergative marking is obligatory on all nominal types except first and second person singular pronouns, on which it is optional. Both Gurindji Kriol and Light Warlpiri show reduced allomorphy in the ergative case marker compared with the traditional languages (Meakins 2011).

19.5 Conclusions The chapter shows that there are many ways in which a language can become ergative morphologically, consistent with the synchronic diversity in ergative case marking (Gildea 2004). Many of these processes of grammaticalization are not cognitively motivated in terms of transference from one conceptual domain to another, as is often presumed. Contrary to Gildea (2004), however, it does not seem that the evidence supports the conclusion that grammaticalization of ergative case marking is a mechanical process involving constructions that just happen to have oblique case marked Agent NPs. Oblique sources account for a proper subset of ergative case markers. The discussion of a number of instances of grammaticalization of and from ergative markers in this chapter reveals the crucial role of constructions (see e.g. Hopper and Traugott 2003: 1–​2; Himmelmann 2004: 31). We have seen, however, that caution is required: evidence is needed for constructions just as much as for markers. It is not

 

464   William B. McGregor sufficient to hypothesize grammaticalization within constructions that are presumed on the basis of well-​known languages. The range of constructions in which ergative case marking may have emerged is as wide as the range of sources of the case markers themselves. Contra Queixalós and Gildea (2010: 13) biclausal constructions (such as nominalizations) and marked voice constructions (like passives) are not the only constructions in which ergative case marking can emerge. A number of grammaticalization scenarios show connections with focus: this is fundamental to the development of some ergative case markers from indexical items (via clefts); it appears to also be a possible consideration in the development of some ablative case markers to ergative markers, and in the case of Jingulu it seems that ergative markers can become markers of focus. Synchronically a strong connection between ergative case marking and focus is sometimes apparent—​e.g. McGregor and Verstraete (2010); Rumsey et al. (2013). I have not found convincing cases in which a focus marker serves as the historical source for an ergative case marker.9 This is, however, a possible origin for the ergative markers of Bunuban languages (McGregor 2008: 313) and Kuikuro (Southern Carib, Brazil) (McGregor 2012: 239; cf. Franchetto 2010). The discussion of sections 19.2.3 and 19.2.5 reveals that one needs to be cautious in drawing conclusions from ‘polysemies’ of case markers. While abstract and more grammatical uses of case markers sometimes develop from more concrete uses, we cannot presume this in every instance. It is also possible that such a situation emerged historically through independent development of another marker. And in some languages (e.g. Wangkumara) concrete senses appear to be secondary in diachrony.

Acknowledgments I am grateful to Barry Blake and Gerrit Dimmendaal for useful comments on an earlier draft of this chapter, and to Alan Rumsey and Lila San Roque for providing me with unpublished material.

Abbreviations 1, first person; 2, second person; and 3, third person; ACC, accusative; AG, agent; C, completive; DEF, definite marker; DEM, demonstrative; EMP, emphatic; ERG, ergative; FUT, future; GEN, genitive; HAB, habitual; INS, instrumental; LOC, locative; M, masculine; NP, noun phrase; PAS, preferred argument structure; PASS, passive; PFV, perfective; PL, plural; POT, potential mood; PRS, present; PRT, partitive; PST, past tense; SG, singular; SUB, subordinate.

9  Kilian-​Hatz (2008: 55, 2013: 376–​377) and König (2008b: 276) argue that the postposition à in Khwe (Khoe-​Kwadi, Botswana) developed from a copula into a focus marker, and thence to an optional marker of accusative case. (See however McGregor forthcoming for an alternative scenario in which the accusative marker developed from a previous attention-​directing indexical item.)

 

Chapter 20

Dec onstru c t i ng Iranian erg at i v i t y Geoffrey Haig

20.1 Introduction The Iranian languages constitute one of the major branches of Indo-​European. They are most closely related to the Indo-​Aryan branch, from which they separated some 4,000 years ago. The oldest reliably datable attestations of Iranian, the Old Persian inscriptions, are approximately 2,500 years old, but the earliest Avestan texts represent an Iranian language probably spoken several centuries earlier, though the dating remains contentious because they were committed to writing at a later date (Skjærvø 2009: 43). Today, Iranian languages are spoken across a vast swathe of Asia, from the Pamir mountains in the border region of Tajikistan and China, to central Anatolia in today’s Turkey. Traditionally they are genetically classified into two main branches, East Iranian and West Iranian, and the latter is further divided into north-​west and south-​west Iranian. However, many details of the subgroupings remain controversial; see Windfuhr (2009) for recent discussion. The following typological features characterize most modern Iranian languages:

• OV word order; • differential object marking, though absent in Kurdish and Zazaki; • a very high frequency of complex predicates, based on a small set of light verbs; • a tense-​based alignment split, affecting transitive verbs,1 and involving some kind of ergativity in those clauses based on verb stems etymologically derived from a participle.

1  Generally, tense-​based alignment splits in Iranian are only relevant for transitive verbs, but there are two counter-​examples: Wakhi (East Iranian, Tajikistan, and Afghanistan), where in the Hunza dialect intransitive subjects may occur in variant case forms (apparently restricted to first/​second person

 

466   Geoffrey Haig This chapter focuses on the last feature. Though absent in some modern Iranian languages, most notably Persian, most Iranian languages have, at some point in their history, passed through a stage with an alignment split, involving ergative structures in the past tense. In the case of Persian, the relevant structures are well attested for the Middle Persian period (Jügel 2015), but Persian has since ironed out this wrinkle and returned to a unified accusative alignment in all tenses. Elsewhere in Iranian, however, the characteristic split is still evident in various aspects of the morphosyntax. Given the extent of the Iranian languages, this chapter can only consider a selection of languages and themes related to ergativity. A common thread running through the presentation is the massive variation in what is loosely termed ergativity in Iranian (see section 20.2 on terminology). In the literature, a number of additional holistic labels for alignment types have been coined with the aim of capturing the variation (double oblique, tripartite, neutral, etc.). I will suggest that such labels, while useful for taxonomic purposes, actually obscure more important generalizations and commonalities cross-​cutting supposedly distinct alignment types. A more insightful account can be framed in terms of distinct sub-​aspects of alignment, hence the ‘deconstruction’ in the chapter’s title. From its inception, the literature on ergativity in Iranian has focused on the two interrelated issues of micro-​variation, and origins, the former pioneered in Payne’s (1980) study of alignment variation in Pamir languages, and the latter in Bynon’s (1980) paper on the diachronics of ergativity (see section 20.3). Synchronic analyses from the perspective of formal syntax models are less prominent, possibly due to the ubiquitous issue of variation, but see Dorleijn (1996) for an attempt to confront formal approaches to ergativity with dialectal variation. The current chapter continues the tradition of relating micro-​variation to a diachronic scenario. The chapter is built around three case studies of dialect clusters, all from the West Iranian branch of the family: Kurdish, Balochi, and Taleshi. For East Iranian languages, the respective chapters of Windfuhr (ed.) may be consulted. Kurdish, Balochi and Taleshi have been chosen because each exhibit a rich range of micro-​variation with regard to ergativity, they are geographically divergent, and the sources are reasonably accessible. Figure 20.1 shows the approximate locations of the three dialect clusters (note that Balochi is also spoken by around 500,000 speakers on the other side of the Persian Gulf in Oman). In section 20.2, ergativity in the Iranian context is introduced and terminological conventions laid down, while section 20.3 looks at the origins of ergativity. Sections 20.4, 20.5, and 20.6 contain surveys of Kurdish, Balochi and Taleshi respectively, while

pronouns, cf. Bashir (2009: 842–​843). The second exception appears to be the Mutki dialect of Zazaki (West Iranian, East Anatolia), though the analysis is still preliminary (Öpengin & Anuk 2015). Outside of these marginal cases, the restriction to transitive verbs is stable, and constitutes a major difference between the ergative structures of Iranian, and those of the neighboring East Aramaic languages, where (in some dialects at least) the special marking of the A is shared by the S arguments of certain unergative intransitive verbs (see Doron & Khan 2012; Khan, Chapter 36, this volume). The difference is noteworthy given that the East Aramaic brand of ergativity is generally considered to have developed under Iranian influence.

 

Key: NK=Northern Kurdish; CK=Central Kurdish; SK=Southern Kurdish; SB=Southern Balochi; WB=Western Balochi; EB=Eastern Balochi. Sources: Kurdish: Haig & Öpengin (2014); Taleshi: Paul (2011); Balochi: Jahani & Korn (2009)

Deconstructing Iranian ergativity    467

Figure 20.1  The approximate locations of the Kurdish, Balochi, and Taleshi dialect clusters

 

468   Geoffrey Haig section 20.7 summarizes the main points and proposes some more general principles behind the observations of the preceding sections.

20.2  Conceptual and Terminological Issues Throughout this chapter I will adopt the following abbreviations: S = core argument of an intransitive verb, bearing the least-​marked morphological case form A = core argument of a transitive verb, semantically exhibiting the most (potential) control P = core argument of a transitive verb, semantically exhibiting the least control

The classic definition of ergativity is a grammatical system in which ‘S is marked in the same way as O[=P] and differently from A’ (Dixon 1994: 16), abbreviated here as A≠S, S=P. Overwhelmingly, A≠S involves additional marking of the A in comparison to the S, rather than the other way round; I am unaware of any language in which S is consistently more morphologically marked than A. We can restrict the notion of ‘marked in the same way’ to the following two domains of morphology (i) case marking, e.g. inflectional affixes or adpositions that indicate predicate/​argument relations, and are located on the relevant NP; (ii) agreement (Dixon’s cross-​referencing), which involves the replication of features such as gender or person of a particular argument on another constituent. In addition to the above morphological criteria, we can gauge the similarities between S, A, and P in terms of their implication in syntactic rules, such as control of reflexives, or coreferential deletion across coordinated clauses. The latter kinds of similarities, generally termed ‘syntactic ergativity’, turn out to be largely irrelevant in the Iranian context. There is no reliably attested example of an Iranian language which consistently groups S and P in terms of syntactic rules, while there are many well-​documented cases of the grouping of S and A. Thus, most of Iranian syntax appears to work on a S/​A-​pivot, in Dixon’s (1994) terms, regardless of the alignment of its morphology (see e.g. Haig 1998 on Northern Kurdish).

20.2.1 Deconstructing Ergativity It is well known that ergativity generally characterizes only subdomains of a grammar, where it coexists with non-​ergative alignments of other parts (see the extensive literature on various kinds of ‘split’). As a consequence, the holistic view of ergativity as a deep and fundamental trait characterizing an entire grammar has largely been abandoned in favour of more contingent approaches, which focus on individual constructions and their interplay (Haig 2010). As Bickel (2010: 442) puts it, ‘once popular expressions like ‘ergative language’ are simply senseless’. This, then, is one way in which the concept of

 

Deconstructing Iranian ergativity    469 ergativity can be deconstructed. But there is a second, and more fundamental, fault line running through the concept of ergativity. Recall that the classic definition of ergativity involves two independent statements: A≠S, and S=P. As I hope to show in this chapter, there are sound empirical reasons for treating the two sub-​aspects distinctly. Turning first to A≠S, one way of conceiving it is in terms of a special marking, or singling out, of transitive subjects as a morphologically marked category among subjects generally. This most typically occurs in the case system, but there are often differences in the agreement system between S and A as well. Diachronically, marked A, or A≠S, appears to be the most persistent aspect of ergativity in Iranian, and is relatively stable across the micro-​variation surveyed here. The second component of ergativity, S=P, is, in its pure form (i.e. full identity in case and agreement) exceedingly rare in Iranian. While in case marking, many languages do unite S and P, agreement systems almost always make a distinction. In a sense, then, agreement is the weak link in the ergativity chain, a fact that echoes the typological generalization that agreement systems favour accusative alignment, while case systems are more prone to ergativity (cf. Haig (2008: 301–​304) for discussion of Iranian, and Bickel (2010: 442) for a typological perspective). But the failure of S and P to unite is not restricted to agreement. In many varieties of both Balochi and Taleshi, the case marking of P also differs from that of S. But crucially, these differences in case marking between S and P are generally independent of the presence or absence of A≠S. It can thus be demonstrated that the two aspects of ergativity, A≠S and S=P, are in principle independent of one another. For the purposes of this chapter, I will refer to any grammatical subsystem with A≠S, i.e. where transitive subjects are marked relative to intransitive subjects, as ‘ergative’, but the reader should be aware that this is merely a terminological convention, and in most cases construction-​specific details diverge from the standard definition of ergativity cited at the outset of this section.

20.3  The Origins of Iranian Ergativity, and Its Principal Manifestations In Old Iranian, transitive clauses with all finite verb forms had a unified accusative alignment, regardless of the tense of the verb: the A was in the nominative case, the verb agreed with A, and the P was marked by the accusative case. This is demonstrated with a transitive clause from Old Avestan (Haig 2008: 25, transcription and glosses adapted): (1)

aṯ zī θwā fšuyantaē=cā and indeed 2sg.acc cattle.breeder.dat=and θwōrəšta      tatašā fashioner.nom  has.created.3sg

vāstrāi=cā herdsman.dat=and

‘And indeed the Fashioner(A) has created you(P) for the benefit of the cattle-​breeder and the herdsman’ (Old Avestan, Yasna 29,6).

 

470   Geoffrey Haig Alongside the system of finite verbs, Old Iranian also had a set of participles, generally involving a final -​ta. These participles were originally ‘verbal adjectives, with a resultative sense’ (Haig 2008: 41), in their semantics comparable to English participles such as broken, or fallen. They were already widely attested in Old Avestan; examples include vista-​ ‘known’, mərəta-​ ‘died, dead’, or bərəta-​‘carried, brought’ (Jügel 2015: 101–104). These participles occurred both in attributive, and predicative function, and agreed in gender, case, and number with the noun they were associated with. Participles could be derived from both transitive and intransitive verbs, but in Old Iranian, the attested examples of participles with intransitive verbs appear to be restricted to change-of-state predicates (Jügel 2015: 101–104). The following Young Avestan example illustrates a participle used predicatively (and here without the copula), and negated via a verbal negation particle: (2) āaṯ yaṯ =hē zasta nōiṯ frasnāta then when=clpro.3sg.gen hands not washed.ptcpl ‘For if his hands are not washed […]’2 (Young Avestan, Jügel 2015: 66, glosses added, transcription simplified). In Old Persian, much of the Indo-​European system of aspectual distinctions for verbs had disappeared; for past tense reference in narratives, the default verb form was the imperfect, while the Old Iranian Aorist is only marginally attested, and presumably already largely defunct. Clauses with a ta-​participle as predicate became increasingly integrated into the system of aspectual distinctions, and according to Jügel (2015) they contributed a resultative, perfective aspectual nuance. The following example contains two clauses headed by the ta-​participle, and a third with imperfective. Jügel (2015: 70) suggests that the first two clauses express the speaker’s deeds from the perspective of completed events (‘deeds’), while the third clause, containing the imperfect form akunavam, focuses on the process itself, which involved accompaniment by the will of Ahuramazda. (3) a. taya that.which.n.sg

manā 1sg.gen/​dat

b. taya=mai that.which=clpro.1sg.gen/​dat c. awa  wisam wašna that all will.instr

kərtam do.ptcpl.n.sg

apataram kərtam, afar do.ptcpl.n.sg

auramazdāhā akunav-​am Ahuramazd.gen do.ipfv-​1sg

a. ‘That which I did here, b. and that which I did afar, c. all that I did through the will of Ahuramazda’ (Old Persian, Jügel 2015: 70, glosses added) 2 

idā utā here and

[participle, non-​finite] [participle, non-​finite] [imperfective, finite]

The translation here follows Jügel’s German translation ‘Denn wenn seine Hände nicht gewaschen (sind) […]’. An alternative interpretation is taken up below.

 

Deconstructing Iranian ergativity    471 Aside from the nuances of aspectual interpretation, (3) also serves to introduce the key syntactic issues surrounding ergativity in Iranian: first of all, the predicate in (3a) and (3b) is the participle kərtam, which agrees in gender and case with its underlying object (‘that which’). In the first clause, the doer is expressed through the independent first person singular pronoun manā. It is in a case that is historically derived from the old Iranian Genitive, but which had absorbed all the old functions of the Dative (the Dative no longer existed as a distinct formative in Old Persian). In the second clause, the doer is expressed by the clitic version of this pronoun, =mai, which is hosted by the first element of its clause in compliance with well-​known principles of second-​position clitic placement (Wackernagel’s law). By the Middle Iranian period, loosely covering the first millennium ad, constructions such as (3a–​b) had become the sole means of expressing past time reference. The verb system had broken down to a basic binary opposition between two stems: what I will term a ‘present stem’, and the reflex of the resultative participle in -​ta, termed in this study the ‘past stem’. The syntax of transitive clauses based on the past stem in Middle Iranian basically continued what we have illustrated for Old Persian, but with a crucial difference. Case morphology had largely been lost, so S, A, and P appeared in the same unmarked case (with the exception of the first person singular and certain kinship terms in Parthian and early Inscriptional Middle Persian texts). The A, therefore, appeared in the common, unmarked case form. But more frequently, the A was expressed through a clitic pronoun in a Wackernagel position, the etymological continuation of pronouns such as =mai in (3b). The Middle Iranian (Middle Persian and Parthian) forms for the clitics are shown in Table 20.1; they are still recognizable in the modern languages discussed below (though the third person singular has a variant form in some languages): Table 20.1 Pronominal clitics in Middle Iranian Singular

Plural

First person

=m

=mān

Second person

=t/​d

=tān/​=dān

=š

=šān

Third person

Source: Jügel (2015: 222), simplified

The verb itself (often an auxiliary) generally agreed with the P, quite in keeping with its participial origins. The following Middle Persian examples illustrate the clitic pronoun expressing a past A, as well as the agreement of the verb (here an auxiliary) with the P: (4) u=š druyist [fn. omitted, GH] kird hēd and=clpro.3sg complete do.pst(3sg) aux.2pl ‘and he created you(pl)’ (Middle Persian, Jügel 2015: 373, glossed added)

 

472   Geoffrey Haig The use of the pronominal clitic to express the A  generally only occurred when the A was not overtly present in the clause. Thus in the first part of (5) no pronominal clitic occurs, but it does in the second: (5) ud kū šāhān and when king.pl

šāh ēn paykar dīd king dem statue see.pst(3sg)

u=š dād … and=clpro.3sg give.pst(3sg) ‘And when the King of Kings saw this statue, and gave […]’ (Middle Persian, Jügel 2015: 534, glosses added) In an intransitive past clause, on the other hand, the S is never expressed through a pronominal clitic,3 and the S controls agreement on the verb. The same applies to the A of a present transitive clause (not shown here): (6) āmad hēnd gāwiraz-​ān come.pst aux.3pl throne.preparer-​pl ‘Those who prepare the throne have come’ (Middle Persian, Parthian, Jügel 2015: 260, glosses added) The Middle Iranian construction has generally been analysed as ‘ergative’, due to the agreement of the verb with the P, and the fact that A has distinct pronominalization patterns (and in the earliest texts, a distinct case) when compared to S. Thus the uniform accusative alignment that characterized finite verbs in Old Iranian (cf. (1)) had evolved into a system with split alignment in Middle Iranian: accusative alignment in the present, and ergative alignment in the past. The question that has been debated for decades concerns the nature of the mechanisms behind the transition from the Old Iranian accusative alignment to the Middle Iranian ergative alignment. Two main lines of explanation have been proposed. The first assumes that the Old Persian construction with a Genitive/​Dative A was basically an agented passive, i.e. passive verb form+by-​phrase, yielding for (3a) ‘that which was done by me’. Over time, this was reanalysed into an active construction, and the syntactically non-​core argument (the by-​phrase) develops into a fully fledged subject. Following Bynon (1980), this approach gained currency in diachronic syntax generally, and the Iranian case was added to the list of languages apparently supporting the ‘well-​attested

3  The corpus of Middle Iranian texts investigated by Jügel actually contains isolated examples of pronominal clitics with S (13 cases among 3,522 intransitive subjects), but Jügel (2015: 328) considers them most likely to be scribal errors. In the later languages, pronominal clitics are occasionally attested with intransitives (e.g. Balochi of Turkmenistan, Axenov 2006: 108–​109), but this appears to be a marginal extension, with no obvious systematic regularities. As a general rule holding across Iranian, wherever indexing of subjects with pronominal clitics is attested, it is restricted to past tense A’s.

 

Deconstructing Iranian ergativity    473 […] reanalysis of earlier passive constructions’ as a source of ergativity (Lightfoot 1999: 140). Haig (2008: Ch. 2) notes that the account in terms of origins from an agented passive construction suffers from an empirical flaw, namely the striking lack of evidence for any of the assumed intermediate stages. He develops an alternative account, according to which the Old Persian construction was the extension of already existing Non-​ Canonical Subject constructions, involving a Genitive/​Dative marked Benefactive, Experiencer, or Possessor. These already exhibited semantic and pragmatic features typical of subjecthood (topicality, high animacy), thus the necessity for assuming wholesale restructuring of the construction is reduced, because the Benefactives, etc. already shared crucial properties with transitive subjects. The link between possession/​benefaction and agency has been pointed out many times; the Young Avestan example (2), repeated here as (7), provides a neat illustration of the constructional polysemy that was already available in Old Iranian, and which can be considered pivotal in the extension of this kind of construction to clauses involving agency: (7)

āaṯ yaṯ =hē zasta nōiṯ frasnāta then when=clpro.3sg.gen hands not washed.ptcpl ‘For if his hands are not washed […]’

The translation given here follows Jügel (2015). But it is not hard to see how one could read: ‘For if he has not washed his hands’, with the clitic pronoun expressing not only the possessor of ‘hands’, but the A of ‘washed’. The following example from Old Persian likewise illustrates the overlap of possession and agency (Haig 2008: 95): (8)

avaθā=šām thus=clpro.3pl.gen

hamaranam kartam battle do.ptcpl

While Kent (1953) still translates with a passive plus by-​phrase ‘thus by them battle was done’, one could equally translate ‘their battle was fought’, or more naturally in the actual context of this example, ‘they engaged in battle’. On this account, the main change is not one of reanalysis or internal restructuring of the construction, but of frequency of occurrence, and markedness within the entirety of the tense system. Thus while in Old Persian, clauses headed by the ta-​participle occupied a relatively minor aspectual niche (see (3)), structurally they were already largely identical with their Middle Iranian ‘ergative’ descendants. The difference is simply that by Middle Iranian, the ta-​participle verb forms were the only available alternative for clauses with past time reference, hence the unmarked choice for past transitive clauses. The ‘change’ thus involved the extension of an existing construction into new contexts, brought about by the disappearance of the competitors through loss of finite verb forms such as the Aorist. A similar conclusion is recently reached by Jügel (2015), who goes as far as to classify the Old Persian m.k. construction with an overt agent as ‘ergative’, and likewise refers to the extension of a construction with a Benefactive to cover constructions with Agents

 

474   Geoffrey Haig (2012a: 464).4 For more detailed coverage, see Haig (2008: ch. 2) and two recent in-​depth studies on Middle and Old Iranian, Jügel (2015) and Bavant (2014).5 The reduction of the verb system to two stems, one of which was in origin a resultative participle, had far-​reaching consequences for the syntax of the daughter languages. These participles, like their counterparts in English, such as broken, fallen, etc., were unable to assign accusative case, and this defect was inherited by all the daughter languages. For participles from transitive verbs, the default assignment of nominative case was to the underlying object, with which the participle would agree, though agreement with the P is not consistent, even in Middle Iranian. With the nominative already assigned, and an accusative blocked, the expression of the A had to follow a different pattern. As we have seen, the case used for this purpose in Old Iranian was the Genitive/​ Dative case, the case otherwise used for Benefactives, Possessors and Recipients.6 In the daughter languages, the reflex of this case continues to mark the A of the ergative. Furthermore, many of the later languages maintained the system of indexing an A via a pronominal clitic, already illustrated in (4) and (5). Thus the most salient aspects of ergativity in Iranian were already clearly discernible in the syntax of Old Iranian, and most later developments can be plausibly reconstructed back to the earlier constructions. In sum, three structural components that characterized ergativity throughout Iranian languages can clearly be traced back to developments that had their seeds in Old Iranian: (i) the binary opposition in the verb stems, with the past member based on the old participle; (ii) an inherited case system, generally reduced to a two-​way opposition between an unmarked ‘Direct’ case, and a marked ‘Oblique’ case, the latter the continuation of the old Genitive/​Dative. In some languages, the latter has disappeared entirely (Persian, Central Kurdish), and in some, additional cases based on grammaticalized adpositions have developed. (iii) The remarkably persistent use of clitic pronouns to index the A of a past tense transitive clause, already evident in Old Persian (3b), and still found in many modern West Iranian languages, where they exhibit forms cognate with the Middle Iranian paradigm shown in Table 20.1. 4  A reviewer points out that while examples such as (7) and (8) illustrate the overlap of Benefactive/​ Possessor readings with Agency, it is not readily apparent how the much-​discussed Old Persian examples such as (3), with the phrase ‘that which I did’, can be reconciled with Benefactive/​Possessor readings. There are two responses to this. The first would be that the extension to general agency had already occurred in the Old Persian texts, hence the participial construction was already an available option with any transitive verb. The second is to note that even these examples can be viewed from a possessive/​ resultative perspective, to be read as ‘these are my deeds’. Ultimately, the restricted size of the Old Iranian corpus means that reliable reconstruction of all the assumed intermediate stages cannot be achieved. 5  Bavant (2014) and Jügel (2015) provide detailed philological analysis and develop more articulated accounts of the Old Iranian verb system, incorporating additional data from Avestan. Notably, neither scholar advocates a return to the agented-​passive interpretation. Bavant concludes that the question of the origins of Iranian ergativity is ultimately unanswerable, given the restricted nature of the Old Iranian corpus, while Jügel’s analysis largely confirms Haig (2008), though differing in details and terminology. 6  The two principal ideas behind this analysis, namely that ergativity in Iranian (i) primarily results from the intrinsic inability of the participles to assign accusative case; and (ii) exhibits both structural and diachronic parallels to constructions with ‘Dative’, or ‘Non-​Canonical Subjects’, were around for decades in Indo-​European linguistics (Benveniste 1952/​1966), albeit in rather different terminologies, before being explicitly developed in Haig (2008). It is interesting to note that more recently, and quite independently, the same ideas are espoused in Minimalist approaches to ergativity; see Mendívil-​Giro (2012).

 

Deconstructing Iranian ergativity    475

20.4 Kurdish Kurdish is a cover term for a group of closely related north-​west Iranian languages (or dialects) spoken across a large region of Eastern Turkey, North Iraq, Syria, Armenia, and west Iran. The main dialects are Northern Kurdish, also termed Kurmanji, and Central Kurdish (also called Sorani). These two dialects differ with regard to ergativity very strikingly. Most of this section is devoted to Northern Kurdish, which exhibits a rather clear case of ergativity, while in 20.4.2 I will briefly illustrate the relevant constructions in Central Kurdish.

20.4.1 Northern Kurdish: Canonical Ergativity Northern Kurdish is not the official language of any nation state, but in three parts of the Northern Kurdish speech zone, different versions of a more or less standard written language have developed: in the former Soviet sphere, mainly in Armenia, Kurmanji was written in the Cyrillic script; the Kurds of Turkey and Syria adopted a Roman-​based script for their version of the standard language, while in North Iraq, a modified version of the Arabic script is used for the Kurmanji dialects of that region. The Roman-​ based alphabet has become the most widely used standard for print and internet media, and examples here are based on that standard. Estimates of the number of Kurmanji speakers are unreliable, and vary between 12–​20 million, most of whom live in (or have recently migrated from) today’s Turkey. Dialectal variation across Kurmanji is discussed in Öpengin & Haig (2014). In Kurmanji generally, pronominal clitics are entirely absent, though transitional dialects between Northern and Central Kurdish exhibit them in some contexts. Ergativity, then, where it is found, is solely manifested in case morphology, and in verbal agreement morphology. Nouns and pronouns occur in two forms, the unmarked Direct, and the marked (or suppletive) Oblique. The paradigm for case and number on nouns is given in Table 20.2, and for pronouns in Table 20.3, which give the forms of the Roman-​alphabet based standard, with some dialectal variants in brackets. Table 20.2 Case and number on nouns in Northern Kurdish Singular, masculine and feminine Masculine singular ‘village’ Definite

Feminine singular ‘woman’

Indefinite

Dir.

Obl.

Dir.

Obl.

gund

gund-​î

gund-​ek

gund-​ek-​î

Definite

Indefinite

Dir.

Obl.

Dir.

Obl.

jin

jin-​ê

jin-​ek

jin-​ek-​ê

 

476   Geoffrey Haig

Plural ‘women’ Definite Dir jin

Indefinite

Obl.

Dir.

jin-​an

jin-​in

Obl. jin-​(in-​)an

Table 20.3 Case and number on pronouns in Northern Kurdish Singular

Plural

Direct

Oblique

Direct

Oblique

1

ez

min

em

me

2

tu

te

hûn

we

3

ew

(e)wî (m.)/​(e)wê (f.)

ew

wan

With regard to the structure of the NP, determiners and quantifiers precede the head, while lexical modifiers (adjectives and possessors) follow the head and are linked to it by means of a vocalic particle traditionally termed the Ezafe (ez). Ezafe constructions are well known from Persian, but unlike Persian, the Ezafe particle in Kurmanji is sensitive to the gender and number of the head noun. Possessors in Ezafe constructions take the Oblique case, while adjectives do not inflect for case:    (9) bajar-​ek-​î mezin ‘a big town’ town-​indf-​ez.m big (10) mal-​a me house-​ez.f 1pl.obl

‘our house’

The southern and south-​eastern dialects of Kurmanji exhibit a fairly consistent brand of ergativity, which I  will refer to as canonical ergativity. In canonical ergativity, S and P display identical features with respect to both case marking and agreement, while A is distinct in both respects. Canonical ergativity is restricted to past tense transitives. In intransitive clauses, the S is in the direct case, and the verb agrees with it; this alignment is impervious to a change in tense, as shown in (11): (11)

her sê t-​ê-​n-​e /​ hat-​in-​e mal-​ê each three indic-​come.prs-​pl-​drct come.pst-​pl-​drct  house-​obl.f ‘All three are coming/​came home’ (Southern Kurmanji, dialect of Midyat)

 

Deconstructing Iranian ergativity    477 In transitive clauses, however, alignment is sensitive to tense. In the present tense (12a), the A is in the Direct and agrees with the verb, while the P is in the Oblique. In (12b), the A is in the Oblique, while the P is in the direct, and agrees with the verb: (12)

a. ez 1sg.dir

wî 3sg.obl.m

na-​bîn-​im neg-​see.prs-​1sg

b. min ew ne-​dît 1sg.obl 3sg.dir neg-​see.pst(3sg) ‘I (a) don’t/​(b) didn’t see him’ Example (13) has a plural A and a singular P. Number indexing on the verb targets the A in the present tense (13a), and the P in the past tense (13b): (13) a. gundî tişt-​ek-​î na-​bêj-​in villager(dir.pl) thing-​indf-​obl neg-​say.prs-​pl b. gundî-​yan tişt-​ek ne-​got villager-​obl.pl thing-​indf(dir) neg-​say.pst(3sg) ‘The villagers (a) aren’t saying/​(b) didn’t say anything.’ The pragmatically neutral word order, however, remains APV, regardless of the morphology. In terms of case marking and agreement, then, the canonical ergative construction of Northern Kurdish in the past tenses involves S=P, S≠A. The system is summarized in Table 20.4. Table 20.4 Canonical ergativity, Northern Kurdish Type of verb

Case of core arguments

Present transitive

A-​Direct

P-​Oblique

With A

Past transitive

A-​Oblique

P-​Direct

With P

Intransitive, all tenses

S-​Direct

Verbal agreement

With S

Analytical passives can be formed from transitive verbs, using the verb hatin ‘come’ as auxiliary coupled with the infinitive, which is based on the past stem, for example hat kuştin(-​ê) ‘lit. (he/​she) came (to) killing (-​obl)’ = was killed’. Passives can be formed from transitives in both the past and the present tenses, with no apparent restrictions applying to past transitives.

20.4.1.1 Deviations from Canonical Ergativity The dialects of Northern Kurdish diverge from the scheme shown in Table 20.4 in various ways. The most widespread is the behaviour of verbal agreement with respect to

 

478   Geoffrey Haig plurality. There is a strong tendency for a plural A, particular if not overtly present in the clause, to trigger plural agreement on the verb—​even when the P (which would normally control agreement) is singular. Plural agreement with a past A is actually found in all varieties of Kurmanji, spoken or written, when the A is not overtly present, and the preceding clause is an intransitive clause with the same subject. An example from the written language (the poem Ji Biçukan re, by Cegerxwîn), is typical: (14) herdu çû-​n-​e cem rovî both go.pst-​3pl-​drct to fox _​_​_​ doz-​a xwe jê ra got-​in ø case-​ez.f refl adp.3sg.obl adp say.pst-​3pl ‘The two of them went to the fox (and) put their case before him’ (lit. told him self ’s case) The past transitive verb form gotin has a plural agreement marker, although its direct object is singular (doz ‘case’, identified as singular by the singular form of the feminine ezafe). The plural agreement here thus reflects the plural number of the A herdu, which is not overtly expressed in the second clause. In fact, in this example (and similar ones), it would be impossible for the verb to agree with its singular object doza xwe ‘their case’. Thus under the conditions of co-​referential deletion of a plural A, verbal agreement generally switches its alignment to the A, rather than the P. The same pattern may extend to clauses with an overt, Oblique marked A. The following has an overt A, and an explicitly singular P, yet the verb agrees in number with the A:7 (15)

pîrek-​a(n) jî hirç-​ik dî-​n woman-​pl.obl add bear-​indf see.pst-​pl ‘The women too saw a bear’ (Haig 2008: 234; note the non-​standard form of the indefinite suffix)

This pattern is particularly common with those transitive verbs that generally lack a referential, nominal object, such as gotin ‘say’ (the ‘object’ of this verb is usually a clausal complement, usually direct speech), see Haig (2008: 231–​242) for discussion of other relevant factors. The second well-​documented deviation from canonical ergativity is the so-​called double-​oblique construction, familiar from Payne’s (1980) account of East Iranian Pamir languages. In this construction, both and A and P are in the Oblique case, and the verb agrees with neither, generally taking the default third person singular form. Under 7 

The plural agreement with the A in this example would not be acceptable in all dialects of Northern Kurdish, for example Shemzinan (Ergin Öpengin, p.c.).

 

Deconstructing Iranian ergativity    479 the loose definition of ergativity given at the outset of this chapter, Double Obliques still qualify as a kind of ergativity because A≠S (A is Oblique). But it is not canonical ergativity, because S and P have different case and agreement properties. Double Oblique constructions are typical of north-​eastern dialects of the Northern Kurmanji cluster, in particular the dialect of Muş, and is also found in the writings of authors from this region (examples from Haig 2008: 226, glosses slightly modified): (16) gundî-​yan wan bizor ji hev kir villager-​pl.obl 3pl.obl by.force from recipr do.pst(3sg) ‘The villagers separated them by force’ (lit. did them from each other) (17) îşev min keç-​ek-​ê last.night 1sg.obl girl-​indf-​obl.f di xewn-​a xwe da di-​dît adp sleep-​ez.f refl adp ipfv-​see.pst(3sg) ‘Last night I was seeing a girl in my dream’ A final deviation from canonical ergativity occurs in tight sequences of same-​subject clauses, where the first verb is intransitive and the second transitive. The first verb is typically a motion verb çûn ‘go’, or hatin ‘come’. In such clause sequences, the subject of the intransitive verb may appear in the Oblique case, thus anticipating the case assignment of the second verb. Agreement on the first verb is usually default 3sg, though there is some variation here. This pattern only seems to be possible when the two clauses concerned are not linked by any overt coordinator (e.g. û ‘and’). An example is the following: (18) min çû cot-​ek sol li bazar-​ê kirî 1sg.obl go.pst(3sg) pair-​indf shoe at market-​obl buy.pst(3sg) ‘I went and bought a pair of shoes at the market’ (Haig 2008: 250) A very similar pattern also occurs in Balochi (Korn 2009: 68). One way of interpreting these sequences is as a kind of verb serialization, with the second verb determining the case assignment of the entire sequence, but there has been little research on these constructions (see Haig (2008: 248–​252) for discussion).

20.4.1.2 Non-​Canonical Subjects In the south-​ eastern dialects, predicates expressing physical and mental sensations, desire, and possession, may have their experiencer/​possessor argument in the Oblique case. The use of Oblique experiencers is semantically not fully predictable, but is connected to certain predicates rather idiosyncratically, cf. (19) with an Oblique experiencer, and (20) with the experiencer in the Direct case and agreement with the predicate:

 

480   Geoffrey Haig (19) min sar e 1sg.obl cold cop.3sg

‘I am cold.’

(20) ez birsî me 1sg.dir hungry cop.1sg ‘I am hungry.’ In the south-​eastern dialects, expressions of desire use the verb viyan ‘be necessary, desirable’. An overtly expressed ‘wanter’ is clause initial, and in the Oblique case. If there is an overt nominal ‘wanted’ in the clause, it is in the Direct case, and agrees with the verb (21); otherwise the verb takes the default third person singular ending, shown in (22): (21)

te ez na-​vê-​m 2sg.obl 1sg.dir neg-​be.necessary.prs-​1sg ‘You do not want me’ (lit. to.you I not.desirable.am’) (MacKenzie 1961a: 192, glosses supplied, transcription adapted)

(22) min di-​vê-​t [bi-​ç-​im] 1sg.obl indic-​be.necessary.prs-​3sg [irr-​go.prs-​1sg] ‘I want [to go]’ (lit. ‘to.me is.desirable I go’) Predicative expressions of possession show the same pattern: (23) min du bira he-​ne 1sg.obl two brother existant-​cop.3pl ‘I have two brothers.’ The fronted Obliques in (21)–​(23) exhibit syntactic pivot properties, and can thus reasonably be considered to be subjects. In this respect, these constructions resemble the ergative constructions discussed in the preceding section. However, unlike the ergative construction, they are not restricted to past tenses, i.e. there is no tense-​sensitivity. As a general tendency, the frequency of such non-​canonical subjects decreases as one heads northwards and westwards from the south-​eastern periphery of the Kurmanji speech zone. Thus in Northern Kurmanji, expressions of desire use the transitive verb xwastin ‘want, request’, and the wanter appears in the direct case in the present tenses (see Haig (2006) on the areal distribution of non-​canonical subjects in Northern Kurdish).

20.4.1.3 Summary of Northern Kurdish (Kurmanji) Clauses based on past tense verbs in Northern Kurdish shows a fairly clear case of canonical ergativity, with S=P in case marking and agreement, and distinct from A. However, dialectally a number of deviations from this scheme can be discerned: plural agreement triggered by a plural A, and Double Oblique constructions, lacking agreement with A or P. Both of these deviations affect agreement on the verb, and the latter also affects the

 

Deconstructing Iranian ergativity    481 case marking of P. The Oblique case of the A, however, is remarkably consistent across all dialects. Among early first-​language learners (up to 6 or 7 years old), constructions with a Direct-​marked A can be observed, and in the dialect of Gilan, a dialect of north-​ west Iran isolated from the rest of Northern Kurdish, this appears to be the community norm (Shojai 2005). Whether these phenomena are portentous of a general drift back towards accusativity is a moot point. Among those mature speakers who acquired Northern Kurdish in the social context of an intact speech community, particularly in the southeast part of the Northern Kurdish speech zone, the ergative construction is alive and well. However, the political unrest of the last decades, accompanied by mass migration from the region, means the number of persons currently acquiring Northern Kurdish in an intact speech community is declining rapidly (Çağlayan 2014), and many young speakers now show uncertainties and variation in the relevant constructions.

20.4.2 Central Kurdish: the Role of Pronominal Clitics Central Kurdish, spoken to the south of Northern Kurdish in North Iraq and West Iran, differs in its system of alignment quite strikingly from Northern Kurdish. In the Suleimaniye dialect, the best-​described variety (MacKenzie 1961, 1962), all nouns and pronouns lack overt case marking, thus S, A, and P are not distinguished at all by case marking. The morphosyntax of past transitive clauses nevertheless still differs from that of present transitive clauses. The principal difference is that in past transitive clauses, the A is obligatorily indexed via a mobile pronominal clitic, while in present transitive clauses, A is obligatorily indexed via a verbal suffix, in just the same manner as an S. Intransitive verbs, all tenses: (24) min a-​rrō-​m /​ rrōišt-​im 1sg indic-​go.prs-​1sg /​ go.pst-​1sg ‘I go, am going/​I went’ (MacKenzie 1961a: 106, glosses and transcription modified) Transitive verbs, present tense: (25) min sag-​aka na-​kuž-​im 1sg dog-​def neg-​kill.prs-​1sg ‘I am not killing the dog’ In past transitives, on the other hand, the A is obligatorily indexed via a mobile pronominal clitic, etymologically the descendants of the Middle Iranian clitics from Table 20.1. Unlike the Middle Iranian clitics, though, these do not gravitate to the Wackernagel position, but are hosted by the first available constituent of the VP (see Öpengin (2013) for a more detailed account). Different possibilities are illustrated in (26)–​(28). In (26), the clitic

 

482   Geoffrey Haig is hosted by the P, undoubtedly the preferred option, but in the absence of the P, the clitic drifts rightward to the first stressed element of the predicate, in (27) the negation prefix. In the absence of any other hosts, the clitic will land on the verb stem itself, as in (28). (26) min sag-​aka=m 1sg dog-​def=clpro.1sg ‘I didn’t kill the dog’

na-​kušt neg-​kill.pst(3sg)

(27) min na=m=kušt 1sg neg=clpro.1sg=kill.pst(3sg) ‘I didn’t kill (it)’ (28) min kušt=im 1sg kill.pst=clpro.1sg ‘I killed (it)’ Note that ‘every single past transitive construction requires an A-​past clitic’, regardless of the presence or absence of an overt A constituent in the same clause (Haig 2008: 288). In other words, despite the evidently clitic nature of the marker itself, functionally, it is an agreement marker (see Samvelian 2007 for this position). The agreement marker for S, as in (24), or for a present A as in (25), on the other hand, is a verbal suffix, and cannot be hosted by any other constituent. Thus the present tense version of (25), using a mobile clitic to index the A, is hopelessly ungrammatical for the meaning provided: (29) *min sag-​aka=m na-​kuž-​ē 1sg dog-​def=clpro.1sg neg-​kill.prs-​3sg Intended: ‘I am not killing the dog’ The question of whether the past transitive constructions of Central Kurdish qualify as ‘ergative’ cannot readily be resolved. As mentioned, case marking is irrelevant, as there is no overt case morphology. With respect to agreement, one could argue for A=S (i.e. accusative), because both involve obligatory agreement. On the other hand, it could be argued that A≠S, because the nature of the markers used for agreement with A and S are different (mobile pronominal clitics versus verbal suffixes). As there is no principled way to resolve the matter, the choice of taxonomic label is of no theoretical import (Haig 2008: 305). Turning now to agreement, the basic system is that a verbal suffix, of the same paradigm as those used to index an S, indexes a P, but only under the condition that an overt P is not present in the clause. Thus P-​indexing is not obligatory agreement, in the sense that A-​indexing is, but is alternating, or conditioned indexing. The difference is shown by the following, taken from the Mukri dialect of Central Kurdish (Öpengin 2013, glosses and transcription slightly modified): (30) a. dena de=y-​kušt-​in otherwise ipfv=clpro.3sg-​kill.pst-​3pl

 

Deconstructing Iranian ergativity    483 b. dena ewāni=ī otherwise them=clpro.3sg ‘otherwise he would kill them’

de-​kušt *-​ini ipfv-​kill.pst *-​3pl

However, a plural agreement suffix may appear on the verb if the P is nominal, rather than pronominal, though the details remain somewhat obscure (Mukri Kurdish, Öpengin 2013; note the vestiges of the Oblique case in this dialect in (32), kuř-​eke-​ī): (31) estēre-​k-​āni=mān de-​bižārt-​ini star-​def-​pl=clpro.1pl ipfv-​count.pst-​3pl ‘(we would sit at night and) count the stars.’ (32) nāme-​k-​āni=ī dā-​ni be kuř-​eke-​ī letter-​def-​pl=clpro.3sg give.pst-​3pl to boy-​def-​obl ‘he gave the letters to the boy.’ A further complication with agreement in past transitive clauses is that the set of verbal suffixes that would normally index a P may be co-​opted to index a non-​core argument. The non-​core argument concerned is most typically a recipient, benefactor, or external possessor (33), but is often an adpositional complement, in which case the adposition itself remains ‘stranded’, while its complement is expressed through the verbal suffix, cf. (34) and (35): (33) bač-​ka-​kān=ī a-​xward-​im child-​def-​pl=clpro.3sg ipfv-​eat.pst-​1sg ‘It used to eat my children’ (MacKenzie 1961: 115, glosses and transcription modified) (34) ew beserhāt-​e=ī bo gērā-​m-​ewe dem adventure-​def-​clpro.3sg to narrate.pst-​1sg-​asp ‘(he) narrated this adventure to me …’ (Mukri dialect, Öpengin 2013) (35) feqet qise=m lē de-​pirsī-​y only issue=clpro.1sg from ipfv-​ask.pst-​2sg ‘I would only ask council of you.’ (Mukri dialect, Öpengin 2013) This construction, solidly attested in Middle Iranian (Jügel 2015: 378), is one of the most intensely discussed features of Central Kurdish syntax; see MacKenzie (1961), Samvelian (2007), Haig (2008), and Öpengin (2012, 2013). Co-​opting an agreement suffix for non-​ core arguments is not possible in present tense transitive clauses, where verbal suffixes are restricted to agreement with an A. I would interpret this as additional evidence for the fundamental instability of P-​agreement in past transitive clauses: it is either missing entirely, or is facultative, or it is co-​opted for other constituents.

 

484   Geoffrey Haig

20.5  Balochi: the Interplay of Inherited and Innovated Case Balochi is a cover term for a bundle of related north-​west Iranian languages spoken in southeast Iran (Sistan and Balochistan province), and in the neighbouring regions of Afghanistan and Pakistan, including a sizeable speech community in Karachi (see Figure 20.1). Baloch have also migrated further afield, with speech communities in Turkmenistan, the Gulf states, East Africa, and more recently to North America, Europe, and Australia. Balochi is divided roughly into three dialect groups: Western Balochi, Southern Balochi, and Eastern Balochi. Like Kurdish, Balochi is not the official language of any particular state, and competing versions of more or less standardized written forms coexist (see Jahani & Korn (2009) for a more detailed overview). Balochi exhibits a clear case of tense-​based alignment split, with the typical concentration of variation in the realm of past transitive clauses, while the morphosyntax of present-​stem clauses exhibits the unbroken monotony of accusative alignment. This section largely draws on the surveys by Korn (2008, 2009) and Jahani (2015), though the terminology and mode of presentation differs in some respects from the sources, and in the interests of brevity a number of complexities have been omitted; readers should consult the originals in case of doubt.

20.5.1 Case Marking in Balochi With regard to the case system, there are three crucial differences between Balochi and Kurdish. First, all dialects of Balochi have developed additional innovated structural cases; second, the case marking of first and second person pronouns often differs from third person nominals; third, the marking of P is often mediated by DOM. Taken together, these factors yield systems of some complexity, and among scholars of Balochi, there is no agreement regarding ‘the number of cases and what they should be called’ (Jahani & Korn 2009: 651). Table 20.5 shows the system which Korn (2009: 46) assumes to be the common underlying case system for nouns (I have omitted the vocative case, and the footnotes; there is no grammatical gender in Balochi). Table 20.5 Case and number marking on nouns in Balochi Direct Singular Plural

-​ø

Oblique

Objective

Genitive

-​ā

-​ārā

-​ay

-​ān

-​ā(n)rā

-​ānī

 

Deconstructing Iranian ergativity    485 Pronouns of the first and second person show considerable cross-​dialect variation, and according to (Korn 2009:  47), a common system cannot be reconstructed for them. It is not clear whether the distinction between Direct and Oblique, which is assumed in Table  20.5 for nouns, should be also be maintained for first and second person pronouns.8 For illustrative purposes, Table 20.6 provides the Southern Balochi forms only, and does not assume a Direct/​Oblique distinction (based on Jahani & Korn (2009: 653–​654), but ignoring the locative). Table 20.6 Case and number on first and second person pronouns, Southern Balochi Direct/​Oblique

Objective

Genitive

1p sg

man

m(a)nā, manārā

m(a)nī

1p pl

mā

mār(ā)

maē, mē

2p sg.

taw, tō

t(a)rā, tarārā

taī

2p pl

šumā

šumārā

šumē

The Objective case is a clear example of an innovated object marker (Haig 2008: Ch. 3), as opposed to the inherited Oblique case illustrated in 20.4.1 for Northern Kurdish. The Objective displays an obvious phonological resemblance to the Persian object clitic =rā, which is known to have grammaticalized from a postposition with basically benefactive senses. Whether the Balochi Objective case is the result of an independent, but parallel grammaticalization, or was borrowed into Balochi from Persian, or both processes worked in parallel, is not clear. The Balochi case system is rendered more complex by the parallel existence of a dedicated ‘Genitive’ case, used for prenominal possessors, though never for S, A, or P. With the exception of some dialects of Iran and Turkmenistan (see below), most dialects of Balochi show some version of A≠S in their case marking alignment in past tense clauses. However, no dialect of Balochi has canonical ergativity in the sense outlined in 20.4.1 for Northern Kurdish. In Balochi, S and P are distinct in their agreement systems (see below) and frequently also differ with regard to case marking. The A in the past tense is generally Oblique/​Objective marked across Balochi, with the exception of some dialects of Iran and Turkmenistan (see below). The P, if third person, goes into the Direct case. Together this yields ‘model ergative’ case marking (Korn 2008: 252). With first and second person pronouns, the case system works differently

8  Table 4 of Korn (2008) and table 0.3 of Korn (2009) imply that a three-​way distinction Direct vs. Oblique vs. Objective is relevant in the paradigm, at least for some dialects, while Jahani & Korn (2009) collapse Direct and Oblique to a single column, the convention followed here.

 

486   Geoffrey Haig (see below). (36) and (37) illustrate third person arguments, while (38) illustrates the use of a clitic pronoun with the A: (36) sābir-​ā ē hawāl uškit Sabir-​obl dem news(dir) hear.pst(3sg) ‘Sabir heard this news’

(Korn 2008: 252, W. Balochi)

(37) āy-​ā gōk kušt dem-​obl cow(dir) kill.pst(3sg) ‘He/​she killed the cow

(Korn 2008: 252, S. Balochi)

(38) pīālā=ȭ bowl(dir)=clpro.1sg ‘I have taken the bowl’

(Korn 2008: 254, S. Balochi)

zurt-​a seize-​perf(3sg)

20.5.2 Variation in the Marking of P: the Impact of DOM All dialects of Balochi exhibit DOM in their present tenses, with discourse identifiability (definiteness) as the main triggering factor. Examples (39) and (40) contrast an overtly marked (Objective) P with an unmarked P (Bohnacker & Mohammadi 2012: 67–​68, transcription follows the source, glosses modified): (39) man wtī dōst-​ā har rōč a gend-​on 1sg.dir refl.gen friend-​obl every day ipfv see.pres-​1sg ‘I see my friend every day’ (40) man ya davār-​ē gend-​on 1sg.dir one house-​indf see.pres-​1sg ‘I (can) see a house’ Bohnacker & Mohammadi (2012: 69) note that DOM in Balochi ‘has not been studied systematically’, but suggests that the nature of the determining factors are very similar to those of Persian, which undoubtedly involve some notion of discourse recoverability and identifiability. Obviously, a first or second person P will always be marked, as it is inherently definite and identifiable. The system can be summed up in the following hierarchy, a variant of the Silverstein, or Animacy, Hierarchy discussed in numerous publications: (41) Hierarchy for DOM in Balochi: decreasing probability of overt marking of P 1st/​2nd person pronouns >third person definite >third person indefinite Here ‘third person’ refers to all nouns, including pronouns, demonstratives, etc. The question that we need to address concerns the impact of DOM on the P in an ergative construction, which as noted earlier, is generally in the Direct case. In fact, DOM does

 

Deconstructing Iranian ergativity    487 impact on the marking of P in both the ergative and the accusative constructions, but its effects occur at different points on the hierarchy given in (41). The system is summed up in Table 20.7, showing case marking of direct objects in present and past tenses, with the shaded regions contrasting the effects of DOM in the tenses. Table 20.7 The respective range of DOM in past and present transitive constructions Position on hierarchy present tense past tense

1st/​2nd person

3rd person definite

P is Objective P is Objective

3rd person indefinite P is Direct

P is Direct/​Obj /​Obl

For the present tense, we observe what was discussed earlier, namely that a definite P, regardless of person, will be overtly marked with the Objective case. In past transitives, we see that while first and second person P are always marked with the Objective, and indefinite third person are never marked with the Objective, for definite third person P, there is variation, indicated by pale shading. Essentially what is happening is that DOM operates in both tenses, but with different cut-​off points. All first and second person P’s are overtly marked, but in the past tenses, overt P-​marking has not diffused down the hierarchy to consistently affect definite third person P. What Table 20.7 does is to identify the ‘variation hot-​spot’ in the case marking of P, namely third person definites, in past tenses. These points are illustrated in the following. First, a first or second person pronominal P in the past tense is always in the Objective case, across all dialects: (42) bādšāh-​ā man-​ã: khušth-​a king-​obl 1sg-obj kill-​perf(3sg) ‘The king has killed me’ (Korn 2008: 263, Eastern Balochi; the object is glossed ‘obl’ in the original but I have adapted the gloss in line with Table 20.6) (43) rāh-​ā mn-​ā tunn-​ā jat-​a way-​obl 1sg-​obj thirst-​obl strike-​perf(3sg) ‘On the way, thirst has struck me’ (Korn 2008: 263, Western Balochi) For a third person P in the past tense, on the other hand, there is considerable variation. First, it may occur in the Direct case, and this appears to be the normal option in Southern Balochi outside the Karachi dialect (Carina Jahani, p.c.): (44) mard-​ā bačakk-​Ø dīst-​ant man-​obl child(pl) see.pst-​pl ‘The man saw the children’

(Jahani 2015)

It may also occur in the Oblique, though this is apparently ‘not very common’ (Korn 2008:  261), and only possible when the P is definite. Finally, the P may be in the Objective case: Korn (2009: 263) suggests that ‘in Balochi of all major groups’, a P that is

 

488   Geoffrey Haig human may be in the Objective case if definite, though this has since been questioned for Southern Balochi (Carina Jahani, p.c.). The factors determining the variation in the case marking of a definite third person P are not fully understood. The important point for this overview presentation is that we identify the underlying commonalities across all dialects, namely that DOM is operative in all tenses, but it ‘lags behind’ in the past tenses, where it only consistently affects first and second person objects. Third person definites are the realm of variation.

20.5.3 Verbal Agreement As mentioned, verbal agreement in past transitives works differently from intransitives: only plural number appears to be regularly reflected in verbal agreement, while the category of person is not involved (Jahani and Korn 2009: 663). Jahani (2015), qualifies this statement, pointing to Southern Balochi examples where person marking indexing a P is possible. However, this only occurs in the absence of an overt P in the clause, suggesting that this kind of P-​indexing is pronominal rather than agreement. These examples do not change the fundamental fact that in the past tenses, verbal agreement with P works differently to verbal agreement with S, the latter being always obligatory, across all dialects. If a P is plural, then plural number may be indexed on the verb, though this seems to be most frequent under the condition that overt plural marking (via a plural suffix) of the P is absent. Lack of overt coding of plurality can arise under two conditions: first, the object itself is not expressed at all (because its reference is contextually recoverable), as in (45). Second, the object is overtly expressed, but is in the direct case and hence carries no overt morphological signal of plurality (cf. Table 20.5), as in (46):9 (45) nũ: gwāt čanḍ-​ēnt-​ē now wind swing-​caus.pst-​3pl ‘Now the wind swung them (=the clothes)’

(Korn 2008: 256, Iranian Balochi)

(46) ã: hī-​ā kull-​ẽ: bandī yala kuθ-​aɣ-​ant dem-​obl all-​adj prisoner(dir.pl) free do-​perf-​3pl ‘He has freed all the prisoners’ (Korn 2008: 253, E. Balochi) If the plurality of the P is overtly expressed, plural agreement on the verb is not necessary: (47) mā mard-​ãnrā jaθ-​a 1sg.obl man-​obj.pl strike-​perf(3sg) ‘I struck the men’

(Korn 2008: 261, East Balochi)

9  Carina Jahani (p.c.) points out that plurality of the P can be indicated in other ways, for example through quantifiers, and the verb may still index plural number of the P. The relevant condition thus seems to be lack of the plural suffix on the P itself, not general lack of contextual indication of plurality.

 

Deconstructing Iranian ergativity    489 The above-​mentioned condition for number agreement with the P is a necessary, but not a sufficient one. Even when plurality is morphologically expressed on the P, plural agreement may still appear on the verb, though this appears to be unusual, and subject to dialectal variation. In sum, verbal agreement with a P is (i) restricted to the feature of number (excluding the pronominal indexing of Southern Balochi (Jahani 2015); and (ii) is in part dependent on the presence or absence of overt plural marking on the P itself. It is thus not the kind of obligatory agreement in person and number that characterizes verbal agreement with S in all dialects of Balochi. This again confirms the findings from Kurdish, that agreement with the P is the least stable and consistent aspect of ergativity.

20.5.4 Iranian Balochi: Loss of Ergativity The Western Balochi dialects spoken in Iran have alignment systems that differ from the ergative systems just outlined. Two types are illustrated here: the Sarawani dialect of southeast Iran, described in Baranzehi (2003), and the Sistan dialect from the southeast periphery of Iran (Delforooz 2010). Turning first to the case system for nouns in Sarawani, we find a two-​way distinction between Nominative, and Objective, shown in Table 20.8 (Korn 2008: 255). Table 20.8 Case system for Sarawani dialect of Iranian Balochi Nominative

Objective

Singular

-​Ø

-​ā(rā)

Plural

-​ān

-​ānā

In these varieties, the past A is indexed via a pronominal clitic, though the clitic does not co-​occur with a co-​referential free pronoun in the same clause (48). It is typically hosted by the first lexical category of the VP, as in (49) and (50). (48) mō dāt-​ē ramazān-​ā […] 1sg(nom) give.pst-​3pl Ramazan-​obj ‘I gave (them) to Ramadan’ (Sarawani dialect, Baranzehi 2003: 83) (49) zekk-​ī tālān kort … goat.skin=clpro.3sg pouring do.pst(3sg) ‘She poured (oil from) the goatskin […]’ (Sarawani dialect, Baranzehi 2003: 83) Example (48) also illustrates the ‘anaphoric’ usage of plural agreement with the P, discussed earlier, which also characterizes this dialect. In the past tenses, a third person P is apparently consistently unmarked even when definite (50), while first and second

 

490   Geoffrey Haig person P, in line with the generalization summarized in Table 20.7, are Objective marked (not illustrated here). (50) Alī Hasan=ī zat Ali(nom) Hasan=clpro.3sg strike.pst(3sg) ‘Ali struck Hasan’ (Korn 2008: 258, Iranian Balochi, Lashari dialect) In the present tenses, however, a definite third person P is marked with the Objective case: (51)

[…] dān-​ān-​t-​a be-​bār … grain-​pl-​poss2sg-​obj imper-​take(2sg) ‘(come and) take your grain(s)!’

(Baranzehi 2003: 82)

Thus with regard to case marking of P, the Sarawani dialect complies with the system outlined in Table 20.7: we find DOM in both tenses, but the cut-​off point in the past is higher, with only first and second person objects overtly marked. The most remarkable aspect of the Sarawani dialect is the lack of overt marking of A, already illustrated. Here then, we have S=A in case marking, though not in agreement. To what extent this is ergative is again a matter for (futile) debate. The system shows considerable parallels to that of Central Kurdish, except that a case distinction is still available, and used to mark P in the present tense. In the past tense, however, we have the same lack of case marking, and frequent use of pronominal clitics to index the A. The final pattern to be considered is the complete absence of ergativity, illustrated with the Sistan dialect of Balochi (Delforooz 2010). Delforooz provides no sketch grammar or case paradigms, but the basics of the system are readily apparent from the abundant text material. A very similar system, at least with regard to alignment, is also documented for Balochi of Turkmenistan (Axenov 2006). In these varieties, we find A=S in all tenses: both are nominative, and both control verbal agreement. The marking of P is consistent across all tenses, and mediated by DOM: somewhat simplified, the rule is that if definite, P is marked with the Objective case, whereas an indefinite P is unmarked for case. Examples for the direct-​marked A, and an indefinite, unmarked P is (52); an example for an Objective-​marked, definite P is (53): (52) tajjār p=ēšān gwarag=ē kušt merchant for=dem.pl.obl lamb=indf kill.pst(3sg) ‘The merchant slaughtered a lamb for them’ (Delforooz 2010: 344) (53) gwālag-​ā purr=ē kurt=u sack-​obj full=clpro.3sg do.pst(3sg)=and ‘(he) filled the sack and …’

(Delforooz 2010: 165)

The occasional use of a third person pronominal clitic to express a past A is illustrated by purr=ē ‘full=clpro.3sg’ in (53). With the exception of the latter phenomenon, alignment is basically identical to that of Persian, with unified accusative case marking in both tenses, unified S/​A agreement, and DOM mediated by definiteness.

 

Deconstructing Iranian ergativity    491

20.5.5 Summary of Balochi The Balochi system illustrates the interplay of the inherited case system and an innovated object marker. Common to all of Balochi is the possibility of innovated object markers to mark a definite P, regardless of tense, though what counts as ‘definite’ varies from dialect to dialect, and according to tense. All varieties case-​mark a first and second person P in all tenses; most mark a definite third person P in the present tense, but there is variation across the dialects with regard to case marking of third person, definite P in past tenses. In no dialect is the cut-​off point on the Animacy Hierarchy for marking the P higher in the present tenses than in the past tenses. The overt marking of first and second person P can be related to the origins of the Objective case, typically a grammaticalized postposition that marked recipients. Recipients are very frequently first and second persons, thus the link to first/​second person direct objects (textually actually rare) is quite obvious, and reflected in similar systems, e.g. DOM in Spanish. From the first/​second person, Objective marking encroaches downwards to third person definite objects, at dialectally differing paces, and differing according to the tense of the verb. Interestingly, the addition of an innovated object marker to a basically ergative case-​marking system does not automatically lead to breakdown and replacement of the earlier ergative system. Instead, we find systems combining overt marking of A (inherited Oblique case) with Objective marking of the P (with an innovated object marker). However, the Sarawani dialect shows a development otherwise unattested: the complete abandonment of the overt marking of A, yet the maintenance of the zero-​marking of third person P.10 Notably, this dialect makes widespread use of pronominal clitics to index the A, exhibiting a striking similarity to Central Kurdish (20.4.2). But the nature of the case system in Sarawani is undoubtedly unusual in the Iranian context.

20.6  Ergativity in Iranian Taleshi Taleshi refers to the north-​west Iranian language spoken by a group of speech communities along an approx. 100 km-​long strip of the south-​west Caspian coast, from just north of Rasht in Iran and extending into Azerbaijan (see Figure 20.1). The following data stem from Paul (2011), who focuses on the three main dialect groups of Taleshi in Iran: the northern region (Anbarâne Ardabil, here abbreviated Anb.), central (Asâlam, abbreviated Asâ.), and southern (Mâsâl Šânderman, abbrev. Mâs.). Of these, the dialect of Mâsâl is considered the ‘purest’ representative of the group by native speakers (Paul 2011: 18). Taleshi exhibits the two-​stem verb system, familiar from the preceding discussion, and in the interests of comparability, I will continue to employ the labels ‘present’ and ‘past’. However, these labels are less appropriate for Taleshi, for the following reasons. First, in the northernmost dialect, Anbarani, the stem distinction has been lost on all but seven 10  Geoffrey Khan (p.c.) points out that some North-​Eastern Neo-​Aramaic dialects show a parallel development in their perfect verb forms.

 

492   Geoffrey Haig high-​frequency verbs, yielding a single stem for all tense and aspect forms (in most cases a reflex of the old past stem, cf. Paul 2011: p. 104, table 23). In cases of uncertainty, the labels past/​present are omitted from the glosses. Second, all dialects have verb forms expressing past imperfective aspect, but based on the present stem of the verb. In Anb. a preverbal augment precedes the stem (54), while in Mâs. no such augment occurs (55): (54) gândəm devan a-​k-​im wheat scythe aug-​do.pres-​1sg ‘I was scything the wheat.’ (55)

(Anb. dialect, Paul 2011: 129)

zua-​te-​i=na vâ-​in pisakula boy-​dim-​indf=loc say.pres-​3pl baldy ‘(She had a little boy.) They used to call the little boy baldy.’ (Mâs. dialect, Paul 2011: 131)

In accordance with the general rule for Iranian (Haig 2008: 10), alignment with such verb forms is dictated by the verb stem (the present stem), rather than the actual time reference of the entire predicate (past). Constructions such as (54) have thus predictably accusative alignment. Taleshi has also developed periphrastic progressive forms, used in both past and present contexts. Either a clitic form of ‘be’ is used, or a frozen form of kârd ‘do’. Alignment in such clauses is also accusative, regardless of actual time reference (Paul 2011: 103, table 22).

20.6.1 Case Marking in Taleshi The case marking of nouns is provided in Table 20.9; it shows obvious parallels to that of Northern Kurdish and some dialects of Balochi. The marking of plural differs, however: unlike Northern Kurdish and Balochi, we find distinct forms for singular Direct (zero), and plural direct (-​e). The exception is Anb. dialect, which has gone the same way as Central Kurdish, generalizing the Oblique plural form to the Direct case, hence neutralizing the case distinction in plural number. Table 20.9 Case and number on nouns in Taleshi

Singular

Plural

Direct

Oblique

Anb.

-​ø

-​ə

Asâ.

-​ø

-​i

Mâs.

-​ø

-​i

Anb.

-​un/​-​ün/​-​ân

Asâ.

-​e

-​un/​-​mun

Mâs.

-​e

-​ân

Source: Adapted from Paul 2011: table 15

 

Deconstructing Iranian ergativity    493 Table 20.10 (adapted from Paul 2011: 81) provides the paradigms for first and second person pronouns. It will be noted that the Oblique/​Direct distinction is only overtly realized in the first person singular. A second important difference to the case system with the third person (see Table 20.9) concerns the functional range of the Oblique case. With nouns, the Oblique is also the case for prenominal possessors (i.e. a ‘Genitive’). With first/​second pronouns, however, an additional form is used for the Genitive, which is generally assumed to derive from the fusion of the Oblique form of the pronoun with a preposition related to Old Iranian *haca ‘from’, for example Anb. cəmân ‘1sg.gen’. The same kind of formation is also found with demonstrative pronouns. In the interests of brevity, the special possessives are not treated further here; see Paul (2011: 84) for discussion. Finally, an additional ‘Accusative’ form is entered in the table for the Oblique of Anb. This point is taken up below. Table 20.10 Case and number on first and second person pronouns in Taleshi Direct 1S

Oblique

Anb.

Asâ.

Mâs.

Anb.

Asâ.

Mâs.

âz

az

az

mâ(n)/​mânə (Acc.)

mən

mə(n)

2S

tə

tə

tə

tə

tə

tə

1P

ama

ama

ama

ama

ama

ama

2P

šəma

šəma

šəma

šəma

šəma

šəma

In Taleshi, ergative case marking is restricted to transitive clauses in the simple past (based on the past stem), and the perfect tenses (based on a participle, secondarily derived from the past stem). In all other environments, accusative alignment is found. Accusative alignment shows no variation; it is ‘the same in all three dialects’ (Paul 2011: 92). The A is unmarked (Direct), P is marked (Oblique). The marking of P is mediated by DOM: only definite direct objects (in the sense of ‘identifiable’, as defined in Lambrecht 1994: 77) are overtly case marked. Paul (2011: 69) refers to ‘specificity’ as a triggering factor, but his examples contain specific, indefinite objects which are not case-​marked, as in: (56) əm əšta=râ ka=i sâz-​ə dem self=for house=indf build-​3sg ‘This one builds a house for herself.’ (Paul 2011: 71, Mâs. dialect, glosses adapted) In view of these and other examples, I will continue to refer to ‘definiteness’, rather than specificity as the relevant factor behind DOM in Taleshi. Obviously, pronouns of the first and second person are definite, and are overtly case marked when in object function. Ergative patterns, found with simple past, and perfect tenses, show the by now familiar variability. The main dialect division lies between Anb. and Asâ. dialects on the one hand, and Mâs. on the other, and I treat these two groupings separately before drawing more general conclusions.

 

494   Geoffrey Haig

20.6.2 Anbarane and Asâlem Dialects: Case Marking in Past Tense Transitives The A is consistently in the Oblique case, while P is consistently in the Direct. The effects of DOM are restricted, and only affect a P that is first or second person. The case marking of a third person P, on the other hand, is invariably Direct, regardless of definiteness. Consider the following: (57) əm pis-​i əštan kis-​e iâ nâ=n dem baldy-​obl self ’s bag-​pl.dir here put=tr.pl camun kis-​e=əš ž=in əštan asb-​i poss.3pl bag-​pl.dir=clpro.3sg load=tr.pl self ’s horse-​obl ‘This baldy put his own bags here and loaded their bags onto his horse.’ (Paul 2011: 93, Asâ. dialect, glosses adapted) Both the P’s in this example are definite, but are in the Direct case (plural -​e). Note further that they trigger number agreement with the verb. This example also illustrates the use of a clitic pronoun to express the A in the second clause (see 20.6.3). Even a pronominal P (third person) remains in the Direct case, cf. av ‘him’ in (58):11 (58) səpa užnan av gat=e dog(obl) again 3sg bit=tr ‘The dog bit him again.’

(Paul 2011: 93, Anb. dialect, glosses adapted)

Anb. and Asâ. dialects thus show fairly consistent Oblique marking of the A, in all persons, Direct marking of the P (if third person), and number agreement with P. As in Balochi, a P that is first or second person, always goes into the Oblique case (only actually visible for the first person singular, cf. Table 20.10): (59) užna=š=an again=clpro.3sg=also ‘He struck me again.’

mân 1sg.obl

ža hit.tr (Paul 2011: 98, Anb. dialect, glosses modified).

(60) av-​ə mân nəfin kârd=e […] 3g-​obl 1sg.obl curse did=tr ‘He cursed me so [...].’ (Paul 2011: 99, Anb. dialect, glosses modified). Past transitive verbs do not agree with a first/​second person (singular) P. It would be interesting to see whether such verbs agree with a first/​second person plural P, at least in 11 

Overt expression of the Oblique singular appears to be systematically suppressed on nouns ending in -​a, hence the absence of the expected Oblique case on səpa ‘dog’ in (58).

 

Deconstructing Iranian ergativity    495 number. Unfortunately, no clear examples of such clauses were found, so this question remains unanswered. In sum, the case marking of A is consistently Oblique,12 so we have consistent A≠S. The case marking of P in the past tenses exhibits the same tendency identified for Balochi in Table 20.7. The difference is that in Taleshi, overt marking of a P in the past tenses is completely restricted to first and second persons. The system is illustrated in Table 20.11. Table 20.11 Case marking of P and the interaction of DOM and tense in Taleshi, Anb., and Asâ. dialects Position on hierarchy present tense past tense

1st/​2nd person

3rd person definite

3rd person indefinite

P is Oblique P is Oblique

P is Direct P is Direct

20.6.3 Ergativity in the Mâsal Dialect In the Mâs. dialect, there are two quite distinct constructions available for past transitive clauses. The first is identical to that illustrated above for Anb. and Asâ. dialects, and need not be discussed further (see Paul 2011: 95). The second construction is very different and involves, quite remarkably, fully accusative case marking. The A takes the Direct case, the P (if definite) the Oblique. This construction is apparently contingent on the presence of a pronominal clitic, indexing the A, and hosted by the verb (note that the pronominal clitic occurs inside the so-​called transitivity marker (tr) on the verb). Thus we have: A: Direct P: Oblique, if definite Pronominal clitic indexing the A, and hosted by the verb (61) a əštan tüng-​i avi â-​kard=əš=a 3sg: dir self ’s jug-​obl loss pvb-​caused=clpro.3sg=tr ‘He lost his jug.’ (Paul 2011: 95, Mâs. dialect, glosses modified) (62) az kuf-​i tâ kard=əm=a 1sg: dir felt-​obl fold did=clpro.1sg=tr ‘I folded the felt.’ (Paul 2011: 101, Mâs. dialect, glosses modified)

12 

The ‘Accusative’ form of the Anb. first person singular is also used to mark the P in present-​stem clauses (Paul 2011: 98). I have no explanation for the distribution of this form.

 

496   Geoffrey Haig These constructions are almost identical to those of the Sarawani dialect of Balochi, as discussed. But here, it appears that the Direct case of the A is contingent on the pronominal clitic being hosted by the verb. Why the presence of the pronominal clitic on the verb should have such a drastic impact on the case marking is not immediately obvious, though a correlation between pronominal clitics and case marking is discernible across Iranian, and is taken up in the summary (20.6.4). One can of course note that when the pronominal clitic is hosted by the verb, as in (62), the clause has an undeniable surface similarity to a transitive clause in the present tense, where the A is also indexed on the verb, via agreement suffixes.13 Indeed, in many parts of the paradigm, the pronominal clitics and the corresponding agreement suffixes are near-​identical in form, thus heightening the parallels. One might conjecture, then, that surface similarity in the agreement system left the present and past constructions so close that only a minor change (a change in case marking of A, and of definite P) was required to bring the past construction in line with that of the present. The change can be seen as resulting from the pressure towards cross-​system harmony (Haig 2008: 192–​198).

20.6.4 Summary of Taleshi Like Balochi, the Taleshi dialects surveyed here show a range of case-​marking strategies in their past transitives, from clearly ergative to fully accusative. And in both groups, DOM mediates the marking of P.  Likewise, both groups have preserved, to varying degrees, the system of indexing an A through a pronominal clitic. The differences stem from the fact that there is no innovated object marker in Taleshi, while in Balochi, marking of objects is largely (perhaps exclusively) effected via an innovated object marker. Although the overall range of variation in the Taleshi dialects surveyed here appears less than that of Balochi, that is probably an artefact of the smaller number of dialects sampled and the more restricted scope of the material available.

20.7  Conclusions and Outlook Ergativity emerged during the transition from Old to Middle Iranian. I have suggested that the main mechanism involved the extension of constructions already existing in Old Iranian, to become the sole means for expressing past-​time reference in the daughter languages. The main mechanism is thus extension, rather than restructuring or reanalysis. 13  The available material in Paul (2011) does not permit a satisfactory analysis of this construction. Furthermore, example (243) on p. 101 has the clitic on the verb, but the A in the Oblique case. It seems possible that some of the A’s could be seen as examples of left-​dislocated topics, which would go into the Direct case, and are then resumed by the pronominal clitic on the verb.

 

Deconstructing Iranian ergativity    497 Although ergativity in the sense defined here has been a remarkably persistent trait of much of Iranian, there is nevertheless an undeniable sense of instability when one surveys the family. On the one hand we have straightforward accusative alignments, with little variation in case and agreement, throughout the present tenses. This relative stability contrasts sharply with the situation obtaining in past transitive constructions, where we are confronted with a range of distinct variants, even for relatively small dialect clusters like Taleshi. Why should the Iranian ergative constructions so consistently exhibit this degree of inter-​and intra-​language variability, and tendency towards diachronic change? One explanation is in terms of pressure towards cross-​system harmony (Haig 2008). On this view, the instability of the ergative construction arises not from anything inherently instable in ergativity itself, but from the fact that speakers must essentially master two grammars for case and agreement alignment for transitive clauses, one applying to the present tenses and the other to the past. Now there is no obvious semantic or functional connection between tense and case marking: Case marking is concerned with predicate/​ argument relations, while tense is concerned with temporal reference. In other words, the presence of a tense-​based alignment split introduces unmotivated and opaque structural complexity into the grammar. All other things being equal, we might imagine that over time, minimal changes accrue that successively iron out the differences, ultimately returning to a unified alignment across all tenses. This is in fact what has generally happened: all the attested changes in the ergative construction of the past transitives can be viewed as bringing them closer to the corresponding present tense constructions. A shift in the other direction would of course have been a logical possibility, that is, we might have expected present tense constructions to adapt towards the ergative structures of the past. To the best of my knowledge, this has not happened anywhere in Iranian. The variation we have witnessed in the preceding sections nevertheless allows for certain generalizations regarding the possible pathways of change. First, we can fairly reliably reconstruct a common constructional schema behind all the variants of ergativity discussed here, as in Table 20.12 (see Haig (2004, 2008: 100) for justification).

Table 20.12 The proto-​ergative construction of early Middle Iranian A

P

Verb

Oblique case, expressed through:

Direct:

agrees with P in

*N-​closed, unrounded vowel (sg)

*N-​ø (sg)

person and number

*N-​ān (pl)

*N-​ø (pl)

alternatively, the A is expressed via a pronominal clitic, cf. Table 20.1

(N=nominal host of case marking)

 

498   Geoffrey Haig The A was overtly case-​marked, or was expressed through a pronominal clitic, P was unmarked, and the verb agreed with P. If we turn to the various systems we have considered, most can be relatively simply derived from the proto-​ergative construction in Table 20.12 via fairly straightforward processes of morphosyntactic change, which have ready parallels outside of ergativity: (i) Loss of the pronominal clitics The ‘canonical ergative’ construction of Northern Kurdish has precisely the case-​ marking pattern shown in Table 20.12, but it has completely abandoned the pronominal clitics. Complete loss of clitics is also found in Zazaki (not discussed here). The Balochi of Karachi also seems to have largely lost the clitics, at least none are visible in Farrell’s (2003) material. Persian, on the other hand, has retained pronominal clitics, but not in the function of indexing an A past. Taleshi has retained them, but not in the function of adnominal possessors. Thus changes in the clitic system, including complete loss, suggest that they are a relatively autonomous subsystem, subject to various changes which may produce epiphenomenally variants of the ergative construction. (ii) Weakening of agreement with P Northern Kurdish dialects show several deviations from canonical ergativity, most notably affecting agreement with the P, which is either lost, or is controlled by the A. In Balochi, agreement with the P is largely reduced to the feature of number, and is not obligatory. Similarly, agreement with a P only seems to reflect number in Taleshi (though the material is insufficient to pass final judgement on this); agreement with a first/​second person P is certainly not possible. In Central Kurdish too, we find weakening of verbal agreement with P: with very few exceptions (e.g. (31)) it is anaphoric, rather than agreement, and the relevant suffixes may be co-​opted for indexing other arguments. (iii) Case marking of the P Consistent direct marking of the P is still found in Northern Kurdish, but in all other varieties surveyed, there are exceptions. The commonest occur in connection with DOM:  the highest positions on the Animacy Hierarchy (41) are generally overtly marked, rather than Direct. The common pattern for Taleshi and Balochi, and quite possibly for other languages with DOM and ergativity in the past tenses, is shown in Table 20.13. Table 20.13 Common scheme for case marking of P in Balochi and Taleshi Position on hierarchy present tense past tense

1st/​2nd person

3rd person definite

P is overtly marked P is overtly marked

Variation

3rd person indefinite P is zero-​marked P is zero-​marked

 

Deconstructing Iranian ergativity    499 (iv) Case marking of the A The Oblique marking of the A is surprisingly stable, and is found in all varieties surveyed, with the exception of Central Kurdish, which has lost all case marking. From these facts, some more general principles can be formulated. First of all, it is agreement with the P that is the least stable component, synchronically and diachronically, in Iranian ergativity. Relatively stable P-​agreement occurs in East Iranian Pashto, and in some varieties of West Iranian Zazaki, but in both languages it also involves gender agreement, and there is good evidence that gender and person agreement have distinct characteristics. Consistent P-​agreement in the feature of person is unknown to me in any coherent dialect group within Iranian. Consistent person agreement with S, or present tense A, on the other hand, is the norm. The second least stable aspect of ergativity in Iranian is the case marking of the P. In many cases, variation can be attributed to the impact of DOM (cf. Table 20.13), but it is not only the effects of DOM. In the double-​oblique construction of Northern Kurdish, a P is Oblique marked regardless of definiteness, etc. Synchronically and diachronically more stable than the Direct marking of the P is the Oblique marking of the A.14 Throughout this chapter, I  have attempted to analyse the ergative construction in Iranian not in terms of changes from one alignment type to another (e.g. ergative to accusative). Rather, I see the changes in terms of the partially independent workings of distinct subsystems. If we look at the data in this manner, the variation finds a natural explanation: variant constructions are epiphenomena, labels attached to the many contingent combinations that arise through the changes in the respective subsystems. This is not to claim that the individual subsystems are fully independent. On the contrary, there is a loose interdependency that makes itself felt in certain correlations. For example, it is probably no accident that the only language to have completely grammaticalized the clitic pronouns into agreement markers for the A, Central Kurdish, has also completely lost case marking. But outside of this extreme case, several different combinations of clitic pronouns and case marking are found (e.g. in Balochi), so the correlation is weak at best. Whether ergativity itself can be considered in any sense a privileged constellation within Iranian appears doubtful.

Acknowledgments I am indebted to Carina Jahani, Ergin Öpengin, and an anonymous reviewer of this volume for numerous insightful comments and corrections on earlier versions of this chapter. None of these people bear any responsibility for the remaining errors.

14  Speculatively this fact might be related to the high frequency of zero anaphora realizations of A in actual discourse (Haig & Schnell, 2016). Thus, in actual usage, case marking of the A is seldom overtly expressed, hence contributes comparatively little to the differences between the grammars of present and past tense transitive clauses.

 

500   Geoffrey Haig

Abbreviations ACC, Accusative case; ADD, Additive particle; ADJ, Adjective; AUG, Augment (verbal); CLPRO, Clitic pronoun; COP, Copula; DEF, Definite; DEM, Demonstrative; DIM, Diminutive; DIR, Direct case (traditional term for unmarked case in two-​term Iranian case system); DOM, Differential Object Marking; DRCT, Directional particle; EB, Eastern Balochi; EZ, Ezafe particle (adnominal linking element); F, Feminine; GEN, Genitive; IMPER, Imperative; INDF, Indefinite; INDIC, Indicative; IPFV, Imperfective; M, Masculine; N, Neuter; NK, Northern Kurdish; NOM, Nominative; OBJ, Objective case; OBL, Oblique (inherited marked case form in two-​term Iranian case systems); POSS, Possessive; PTCPL, Participle; SB, Southern Balochi; SK, Southern Kurdish; TR, Transitive; WB, Western Balochi.

 

Chapter 21

In transitivit y a nd t h e deve l opment of e rg at i v e alignme nt Edith Aldridge

21.1 Introduction Ergatively aligned languages pose a challenge to the generative approach to argument licensing and case marking. According to Chomsky (2001 and subsequent works), the functional heads finite T and transitive v are each merged with an unvalued ɸ-​feature. D(P)s enter the derivation with valued ɸ-​features and an unvalued case feature. (1)

Accusative language vTr: [uɸ] vIntr: No [uɸ] TFin: [uɸ]

The unvalued ɸ-​feature on T or v acts as a probe and seeks a matching counterpart in its c-​command domain. As soon as it finds an appropriate goal, i.e. a valued ɸ-​feature set on a DP, the ɸ-​feature on T is valued, and the DP supplying the valued ɸ-​features is valued for case. Consequently, transitive v values accusative case on the structurally most prominent VP-​internal DP (i.e. the object), while T values nominative case on the highest DP in the clause, e.g. the subject. (2)

a. She[NOM] walks. b. She[NOM] sees him[ACC].

 

502   Edith Aldridge c.

TP

T [u

vP

]

DP[

v’

]

v[u

VP

]

V

DP[

]

The challenge presented by ergative languages is the fact that nominative case does not always appear on the DP structurally closest to T. Specifically, the object has nominative case in a transitive clause rather than the subject. Seediq (Aldridge 2004: 78) (3) a. Wada kudurjak ka qedin=na. pst flee nom wife=3sg.gen ‘His wife ran away.’ b. Wada bube-​un na Pihu ka dangi=na. pst hit-​tr gen Pihu nom friend=3sg.gen ‘Pihu hit his friend.’ A common approach to this locality problem is to propose that subjects in transitive clauses are assigned some type of inherent case (genitive in Seediq) and consequently do not require nominative case from T. This allows T to ignore the subject and value nominative case on the direct object (Bok-​Bennema 1991; Bittner and Hale 1996a; Woolford 1997; Ura 2000; Alexiadou 2001, Chapter 15, this volume; and others). (4) 

TP

T [u ]

vP v’

DP [ , INH] v

VP V

DP [ , NOM]

A further consequence is that v cannot have its own set of unvalued ɸ-​features, since there is no VP-​internal DP to value them. This in turn predicts that ergative alignment arises diachronically in an accusative language as the result of two parameter settings: (1) v in a semantically transitive clause lacks the ability to license structural case;

 

Intransitivity and the development of ergative alignment    503 and (2) this v is able to assign inherent case to the external argument in its specifier. It is important to point out that each of these properties is independent of the other. In differential object marking languages like Finnish, objects receive structural accusative case in bounded events but not in unbounded events. Note, however, that the subject continues to surface with structural nominative case. Finnish (Kiparsky 1998a: 3) (5) etsi-​n karhu-​a/​#karhu-​n seek-​1.sg bear-​part/​bear-​acc ‘I’m looking for the (a) bear.’ A shift from accusative to ergative alignment will only be observed when the lack of accusative case for an internal argument is accompanied by the availability of inherent case from v for the external argument. The necessity of the convergence of these two changes also suggests a reason for the relative rarity of ergatively aligned languages, as opposed to the relative commonality of differential object marking in general. In this overview, I summarize how an analysis of this sort has been or can be made to account for accusative-​to-​ergative change in a variety of languages or language families.

21.2  Passive-​to-​E rgative Hypothesis A classic approach to the origin of ergative alignment is positing a passive construction as the source. In a passive clause, the external argument—​if it surfaces at all—​is marked by an adposition rather than structural case, while an internal argument—​typically the theme or patient—​has nominative case. This bears superficial resemblance to a transitive clause in an ergative language, having an obliquely marked external argument and nominative internal argument. EA

IA

(6) a. DPOBL DPNOM V

(passive clause)

b. DPINH DPNOM V

(ergative clause)

Anderson (1977) has proposed just such an origin for ergative clauses in the perfective aspect in Indo-​Aryan languages. Many modern Indo-​Aryan languages exhibit a split-​ergative alignment, whereby imperfective aspect follows an accusative pattern, and perfective aspect is ergatively aligned. Note the ergative suffix on the subject in the Hindi example in (7b). Other DPs appear without a case marker. Note further that the verb shows agreement with the nominative subject in (7a) and the nominative object in (7b).

 

504   Edith Aldridge Hindi (Mahajan 1990: 72–​73) (7) a. raam roTii khaataa thaa. Ram(m).nom bread(f) eat.ipfv.m was.m ‘Ram (habitually) ate bread.’ b. raam-​ne roTii khaayii thii. Ram(m).erg bread(f).nom eat.pfv.f was.f ‘Ram ate bread.’ The ergative pattern in the perfective is generally traced to a construction in Sanskrit built on the participle -​ta (Proto-​Indo-​European *-​to), exemplified in (8b). Note the case on the external argument, glossed as “instrumental.” Classical Sanskrit (Klaiman 1978: 205) (8) a. naro vedān pat̥hati man.nom.sg Veda.acc.pl.m recites.3sg ‘The man recites the vedas.’ b. narena vedāh̥ pat̥hi-​tāh man.ins.sg Veda.nom.pl.m recite-​TA.pl.m ‘The man recited the vedas.’ Anderson suggests that the reanalysis of passive to ergative was motivated by the loss of the inflected perfect and its replacement by the participle, on the basis of the semantic similarity between perfect and passive in that both present a state resulting from a completed action (Anderson 1977: 336). The passive-​to-​ergative analysis has been adopted by Pray (1976), Bubenik (1989), Hook (1991), and others for Indo-​Aryan. Cardona (1970), Payne (1980), Bubenik (1989), and others have made similar claims for the related Iranian languages, which manifest ergative alignment in the past tense. However, questions have also been raised regarding the empirical basis for positing a passive source. Specifically, the source constructions for both the Indo-​Aryan and Iranian ergative clause types do not appear to have the characteristics of a canonical passive. Pray (1976) points out that the nominative object in Sanskrit remains in its base position between the agent and verb rather than moving to clause-​initial subject position, as can be seen in (8b). Furthermore, the agent functions as a subject in being able to control into a gerundive embedded clause, as in (9a), and serving as the addressee of certain types of imperatives, as in (9b, c). Additionally, an intransitive verb can be inflected with participle -​ta, as in (9c). Sanskrit (9) a. tapah̩ kr̩tvā mayā devo ārādhitah̩ austerity having.done by.me god.nom was.propitiated ‘Having performed austerities, I propitiated the god.’ (Pray 1976: 202)

 

Intransitivity and the development of ergative alignment    505 b. mañjūs̩ām ānāyya pr̩chyantām devatās tvayā chest.acc having.caused.to.bring let.them.be god.nom.pl you.ins ‘Having sent for your chest, inquire of the gods.’ (Pray 1976: 203) c. tvayā gamyatām you.ins let.it.be.gone ‘Please go.’

(Pray 1976: 203)

Haig (2008, 2010, Chapter 20, this volume) shows that the corresponding construction in Old Persian likewise did not have the properties of a passive. Here, too, the genitive agent behaves syntactically as a subject and not as an adjunct by-​phrase. Not only is the agent frequently highly topical and animate, this argument can also be the pivot for clausal coordination. Old Persian (Kent 1953; DB III, 47–​49; Haig 2008: 52) (10) avaθā=šām hamaranam kar-​tam utā thus=3pl.gen battle do-​ptcp and avam Vahyazdātam agarbāya utā that Vahyazdata take.prisoner.pst.3pl and ‘They fought battle and (they) took that Vahyazdata prisoner and ….’ In short, there is a lack of evidence that the external argument was ever a demoted adjunct. Facts of this nature prompt Klaiman (1978, 1987) to propose that clauses built on -​ta participles were already ergative in Sanskrit times. However, this view raises the question of where the ergative construction came from, a topic which I turn to in the next section. Before concluding this section, I briefly summarize one more well-​known passive-​ to-​ergative proposal. Hale (1968), Hohepa (1969), and Chung (1978) have proposed that ergative alignment in Polynesian languages like Tongan and Samoan also arose through the reanalysis of a passive construction. Maori as an accusative Polynesian language with a passive formed by adding a suffix “-​Cia,” where “C” is a lexically determined consonant.1 Maori (Chung 1978: 170) (11) a. Ka inu te tangata i te wai. uns drink the man acc the water ‘The man drinks the water.’ 1  Historically, this consonant was part of the base to which the suffix -​ia attached. Final consonants were lost in Proto-​Central Pacific, a subgroup of Oceanic containing Polynesian, Fijian, and Rotuman, so this consonant surfaced only when followed by a suffix. Eventually, the consonant was reanalyzed as being part of the suffix (Pawley 2001: 196).

 

506   Edith Aldridge b. Ka inu-​mia te wai e te tangata. uns drink the water obl the man ‘The water is drunk by the man.’ In contrast, Tongan and Samoan have transitive ergative constructions which very closely resemble the Maori passive. Note the “e” marker on the external argument. (12)

a. Na’e taa’i ‘e Mele ‘a Sione. pst hit erg Mary abs John ‘Mary hit John.’

(Tongan; Chung 1978: 53)

b. Sā ‘āmata-​ina e lātou le pese. pst begin-​tr erg they the song ‘They began the song.’

(Samoan; Chung 1978: 55)

Pawley (2001) provides support for the passive analysis of -​Cia by showing that a wide range of Oceanic languages both within and outside of the Polynesian subgroup have a suffix -​a, which attaches to a transitive verb to derive an intransitive, stative verb. The following examples from the Southeast Solomonic language Arosi show a verb transitivized with the suffix -​Ci and then made stative by the further addition of -​a. Arosi (Pawley 2001: 200–​201; from Fox 1970) (13)

a. age age-​ri age-​ri-​a ‘thatch’ (V) ‘thatch something’ ‘thatched’ b. hunu hunu-​‘i hunu-​’i-​a ‘kill’ ‘kill something’ ‘dead’

However, it is less clear that a diachronic connection can be made between passive -​Cia in languages like Maori and transitive morphology in Tongan and Samoan. The transitivizing suffix employed in Tongan is not -​Cia but -​‘i. Tongan does have -​Cia, but this suffix creates intransitives which are often stative, consistent with Pawley’s (2001) findings sketched in (13). Tongan (Pawley 2001: 204) (14) a. Na’a ku tanu-​‘i (‘a e kappa). pst 1sg bury-​tr abs the can ‘I buried it (the can).’ b. Na’a ku tanu-​mia. pst 1sg bury-​cia ‘I was buried.’

 

Intransitivity and the development of ergative alignment    507 Likewise, in Samoan -​Cia is not generally used as a transitivizing suffix. Transitive verbs are either bare or take the suffix -​ina, as shown in (12b). As in Tongan, when -​Cia appears, it creates stative verbs. Samoan (Pawley 2001: 206) (15) a. alu alu-​mia ‘go, go out’ ‘be in great demand, sell quickly’ b. fuli fuli-​sia ‘turn over, roll over’ ‘turned over’ In short, the intransitive, non-​agentive *-​Cia seems to be retained broadly in this function in modern Eastern Oceanic languages, including the Polynesian languages, regardless of whether they have ergative or accusative alignment, and it is difficult to see a direct connection between this suffix and transitivizing morphology in the ergative Polynesian languages. It may be more reasonable to posit ergative alignment as conservative with Clark (1973, 1976), Kikusawa (2002, Chapter 23, this volume), and Otsuka (2011a), which may help to reconcile Proto-​Polynesian alignment with the non-​ accusative type of alignment found in higher-​order subgroups of the Austronesian family of the Philippines, Taiwan, and elsewhere. To summarize this section, passive is one intransitive source which has been claimed to give rise to ergative alignment. However, empirical evidence does not favor this claim for Indo-​Iranian or for Polynesian languages. In section 21.3, I discuss an alternative analysis for Indo-​Aryan and Iranian ergative clauses as arising from an active intransitive clause type, specifically a possessive construction.

21.3 Possessive Origin Like their Indo-​Aryan relatives, Iranian languages also have split-​ergative alignment in which ergative clauses are employed in the past tense, as in Modern Kurdish. Transitive subjects take oblique marking, while intransitive subjects and direct objects are nominative. Modern Northern Kurdish (Haig 2010: 258) (16) a. min tu dît-​î 1sg.obl 2sg see.pst.2sg ‘I saw you.’ b. te ez dît-​îm 2sg.obl 1sg see.pst.1sg ‘You saw me.’

 

508   Edith Aldridge c. ez zarok bû-​m 1sg child cop.pst.1sg ‘I was a child.’ Old Persian, on the other hand, was a language with uniform accusative alignment. Subjects in both transitive and intransitive clauses are nominative. Old Persian (Haig 2008: 25) (17)

a. pasāva adam kāram frāišayam Bābirum thereupon 1sg.nom army.acc send.pst.1sg to.Babylon ‘Thereupon, I sent an army to Babylon.’ (Kent 1953; DB III, 84) b. yaθā Dārayavahauš xšāyaθiyā abava when Darius.nom king become.pst.3sg ‘when Darius became king’ (Kent 1953; XPf, 36–​37)

As in the case of Indo-​Aryan, the source for the ergative clause type in the Middle Iranian past tense was the participle ending -​ta, from PIE *-​to. In this construction, the object had nominative case, and the subject was marked with genitive case. Old Persian (Kent 1953; DB I, 28–​29; Haig 2008: 26) (18)

ima tya manā kar-​tam that which.nom 1sg.gen do-​ptcp pasāva yaθā xšāyaθiyā abavam after when king become.pst.1sg ‘This (is) that (which) was done by me after I became king.’

Benveniste (1952) proposes that this participle construction was originally a possessive construction expressing the perfect, similar to the English “I have eaten.” Like Benveniste, Haig (2008, Chapter 20, this volume) likewise analyzes this construction as expressing (external) possession. This construction was reanalyzed as finite and transitive as a result of the loss of past tense verbal inflection, the participle being co-​opted in order to express the past. Old Persian (Kent 1953; DB IV, 75; Haig 2008: 62) (19) utā=taiy tauhmā vasity biyā and.also=2sg.gen seed much may.be ‘and may you have much seed (offspring)’ (lit: ‘and may to you/​for you much seed be’) Bynon (2005) has proposed a similar analysis for Indo-​Aryan. She claims that the agent was a raised possessor in an anticausative construction that served as an evidential in Vedic. Bynon argues that the subject in the older construction had genitive

 

Intransitivity and the development of ergative alignment    509 case and that the later Sanskrit construction employing instrumental agents represents an innovation. Vedic (Bynon 2005: 56; MS 1.4.13:62.10; Kulikov 2001: 310) (20) átha yásya kapā́lam bhidyéta and who.gen pot.nom break.3sg.prs.opt tát sám dadhyāt it.nom pv put.3sg.prs.opt ‘And if someone’s dish would break, he should mend it.’ See Butt and Deo (Chapter 22, this volume) for a proposal that the Indo-​Aryan ergative construction developed from a result-​stative construction. Regarding the structure of the possessive construction, it was clearly intransitive in certain respects. The Proto-​Indo-​European resultative participle *-​to had the distributional properties of an adjective (Haig 2008: 41–​42), from which it can be inferred in both Indo-​Iranian and Indo-​Aryan, only one structural case (nominative) would have been available, this going to the internal argument in semantically transitive constructions, while the external argument was expressed as a possessor with genitive case. One possible formal analysis of this construction is that proposed by Mahajan (1997), as it relates ergativity in languages like Hindi to the HAVE–​BE alternation in possessive constructions. In transitive perfect (perfective) constructions in Hindi, the subject is marked with an oblique case, labeled ergative, as already discussed. In contrast, perfect constructions in French (and English) have nominative subjects and accusative objects. Note, too, the difference in the auxiliary: BE in Hindi and HAVE in French. (21) a. Rɑɑm-​ne bhinɖiiyɑ̃ɑ̃ pɑkɑɑyii hɛ̃ Ram-​erg.masc okra.fem.pl cook.prv.fem.pl be.pres.fem.pl ‘Ram has cooked okra.’ (Hindi; Mahajan 1997: 40) b. Jean a cuit les tomates. ‘Jean has cooked the tomatoes.’

(French; Mahajan 1997: 39)

Following Freeze (1992) and Kayne (1993), Mahajan proposes that possessor subjects in all languages are underlyingly PPs, and the auxiliary is universally BE. HAVE is the result of incorporation of the preposition introducing the subject into BE. But oblique subjects will surface if incorporation fails to take place, and the auxiliary will remain BE. (22)

IP VP PP

   

BE VP

V

DP

 

510   Edith Aldridge Mahajan suggests that incorporation is blocked in languages with verb-​peripheral word order like SOV, assuming that both government and adjacency are necessary for incorporation. In SVO languages, in contrast, the VP-​internal subject occupies the specifier of the VP selected by the auxiliary in Infl. Given the head-​initial word order, the auxiliary both governs and precedes the subject, so incorporation can take place. An obvious shortcoming of this approach is that it relies on directionality rather than structural relations alone. The unaccusative structure in (22) also raises some questions in light of the subject-​like behavior of the external argument observed in the past two sections. As an alternative, Whitman and Yanagida (2012) opt instead for a semantically transitive structure along the lines of that proposed for modern Hindi by Anand and Nevins (2006). Following Woolford (1997) and Ura (2000), they analyze the external argument as a DP assigned inherent case in the specifier of v. This DP then moves to [Spec, TP] to satisfy the EPP property of T. But nominative case is valued on the object. This structure accounts easily for the subject properties of ergative arguments. Except for the movement of the ergative subject, this proposal is essentially identical to the analysis of ergative alignment in (4). (23) 

TP DP[ , ERG]

T’ vP

T [u ]

tERG

v’

v

VP

V

DP [ , NOM]

21.4  Instrumental to NP Split-​E rgativity Garrett (1990) proposes an analysis of the origin of NP split-​ergativity in Anatolian. The Australian language Dyirbal is an example of a language with NP split-​ergativity. Third-​ person nominals are marked according to an ergative–​absolutive alignment, as shown in (24a, b). Intransitive subjects and transitive objects have no overt case-​marking, while transitive subjects are marked with a suffix. In contrast to this, first-​and second-​ person pronouns in Dyirbal are marked according to a nominative-​accusative pattern. The transitive subject in (24c) is bare, while the object takes a suffix. See also Coon and

 

Intransitivity and the development of ergative alignment    511 Preminger, Woolford, and Peterson (Chapters 10, 9, and 41, respectively, this volume) for more illustration of person splits in ergative alignment. Dyirbal (Dixon 1994: 161) (24) a. yabu banaga-​nyu mother.nom return-​nfut ‘Mother returned.’ b. nguma yabu-​nggu bura-​n father.nom mother-​erg see-​nfut ‘Mother saw father.’ c. nyurra ngana-​na bura-​n you.pl.nom we-​acc see-​nfut ‘You all saw us.’ Silverstein (1976) observes that the Dyirbal facts are part of a larger cross-​linguistic pattern. First and 2nd person pronouns and 3rd person animate nominals are more likely to be case-​marked nominative-​accusative, while those marked ergative–​absolutive are more often found at the other end of this animacy hierarchy. (25) NP Hierarchy (Silverstein 1976; revised by Dixon 1994: 85) 1st/​2nd Person Pronoun Dem/​3rd Person Pronoun Proper N  Common N Nom/​Acc marking ←============================→ Erg/​Abs marking (26) is a Hittite transitive clause with an ergative subject (marked with the ablative marker -​anza) and a nominative object. According to Garrett (1990), NP split-​ergativity manifests itself in Anatolian in the sense that the ergative suffix attaches only to neuter singular nouns. Hittite (KUB 14.4+rev 23’ [Hymnes 196]; Garrett 1990: 266) (26) nu KUR URU Ḫatti=ya apāš išḫan-​anza arḫa namma ptcl land Ḫatti=and that.nom.sg blood-​erg.sg pv moreover ‘and that murder moreover ended the land of Ḫatti’ For the source construction, Garrett (1990: 277) posits a transitive clause with an instrumental adjunct and no overt subject. Hittite (Garrett 1990: 277) (27) n=at witenanza parkunuzi ptcl=3sg.acc water.abl.sg pure.caus.prs.3sg ‘S/​he purifies it with water.’

 

512   Edith Aldridge Garrett proposes that an instrumental adjunct was reinterpreted as an agent when the subject did not overtly appear. Another key ingredient of the reanalysis is that it had to have taken place in transitive clauses. This is because of the functional overlap between instruments and agents. Garrett (1990: 265) notes that it is rare—​if even possible—​for a theme to be packaged as an instrument. In English, for example, instruments can function as subjects only in transitive clauses but not in intransitive clauses. (28) a. b. c. d.

John opened the door with the key. The key opened the door. John walks with a cane. *A cane walks.

Consequently, if an instrument is reinterpreted as a subject, it will always be a transitive subject. In short, this analysis conforms to the type of change sketched in section 21.1. It involves a reduction in the number of structural cases available, since a nominative subject is replaced by a PP external argument. And projecting the external argument as a PP would have allowed T to value nominative case on the object, resulting in ergative alignment.

21.5 Nominalization Source Synchronic or diachronic connections between ergative clauses and nominalizations have been proposed for a wide variety of languages. This is unsurprising, given that nominalizations clearly have the formal properties illustrated in (4), most notably the assignment of inherent—​specifically, genitive—​case to the external argument. For additional examples, see Alexiadou, Chapter 15, and Salanova, Chapter 43, in this volume)

21.5.1 Inuit Johns (1992) proposes that transitive clauses in the Inuit language Inuktitut are derived synchronically from nominalizations. First note that possessors are marked with the same case as transitive subjects, glossed as “relative.” Inuktitut (29) a. anguti-​up nanuq kapi-​ja-​a-​0 man-​rel bear.abs stab-​pass-​ptcp-​3sg/​3sg ‘The man stabbed the bear.’

(Johns 1992: 61)

 

Intransitivity and the development of ergative alignment    513 b. anguti-​up qimmi-​a man-​rel dog-​3sg ‘the man’s dog’

(Johns 1992: 69)

Johns proposes that transitive verbs combine with a passive participle, -​ja in (29a), which nominalizes the verb root. The external argument is merged within the nominal projection and assigned genitive case, while the internal argument is base generated outside the nominalization in subject position. The absolutive argument is assigned both case and Ɵ-​role as a result of verb movement to AgrV. (30)   

AgrPV (=IP) NP nanuq ‘bear’

Agrv’ AgrPN

Agrv -0

AgrN’ NP anguti-up AgrN ‘man-Rel’ N kapi-ja -a ‘stabbed one’

Since ergative NPs function as the subject of the clause and generally precede absolutives in linear order, Johns proposes that this argument moves to a position above the absolutive NP. The motivation for this movement is case, since its case-​assigner AgrN has moved to AgrV, and case is assigned in a spec–​head configuration. (31)    

AgrPv (=IP)

NP2 anguti-up

NP1 nanuq

AgrPV Agrv’ AgrPN AgrN’

N

tkapi-ja

Agrv kapi-ja-a-0 AgrN

tkapi-ja-a

Aside from the outdated theoretical assumption that case must be assigned in a spec–​head configuration, another disadvantage of this proposal is its requirement that the ergative NP move past the absolutive NP, which violates Relativized Minimality (Rizzi 1990b), since both of these NPs are located in case-​licensing A-​positions. Another question is whether a synchronic derivation involving nominalization is truly

 

514   Edith Aldridge warranted, an issue which I take up again in relation to Tagalog ergative alignment in section 21.5.3.

21.5.2 Old Japanese Active Alignment Modern Standard Japanese is a canonical accusative language: nominative case appears on subjects in both transitive and intransitive clauses, while objects in transitive clauses have accusative case. Modern Japanese (32) a. Taroo-​ga hasit-​ta. Taro-​nom run-​pst ‘Taro ran.’ b. Taroo-​ga ringo-​o tabe-​ta. Taro-​nom apple-​acc eat-​pst ‘Taro ate an/​the apple.’ According to Yanagida (2012) and Whitman and Yanagida (2012), Old Japanese of the eighth century likewise showed accusative alignment in finite root clauses. At this time in the history of the language, nominative and accusative cases were not morphologically marked. Old Japanese (Yanagida 2012) (33) a. 我期大王國所知良之 [Wa-​ga opo-​kimi] kuni siras-​u rasi. I-​gen great-​lord country rule-​conc seem ‘My great lord rules seems to rule the country.’ b. 烏梅能波奈伊麻佐加利奈利 [Ume-​no pana] ima sakari nar-​i. plum-​gen blossom now at.peak be-​conc ‘The plum blossoms are now at their peak.’

   (Manyoshu 933)

(Manyoshu 933)

In contrast to this, case-​marking of subjects in nominalized clauses exhibited an active alignment (see also Khan, Malchukov, Woolford, Chapters 36, 11, and 9, respectively, this volume, for other examples of active Split-​S alignment). All external arguments in nominalized clauses appear with genitive case, while internal arguments—​including subjects of unaccusative predicates—​are bare. The genitive case marker in (34a) is ga, which in modern Japanese has been reanalyzed as nominative. However, its genitive use in this period is clearly in evidence in (33a) and (34a) marking the possessor wa-​ga ‘my.’

 

Intransitivity and the development of ergative alignment    515 Old Japanese (34) a. 我背子之求流乳母尒 [wa-​ga seko-​ga motomu]-​ru omo-​ni I-​gen lord-​gen ask-​adn nurse-​dat ‘as the wet nurse that my lord asks for’ b. 久木生留清河原尒 [pisaki opu]-​ru kiyo-​ki kapara-​ni catalpa grow-​adn clear-​adn river.bank-​on ‘on the banks of the clear river where catalpas grow’

(Manyoshu 2926)

(Manyoshu 925)

In addition to the assignment of inherent genitive case to external argument subjects, Whitman and Yanagida further propose that the nominalized v lacks the ability to structurally license the object. Yanagida (2006), Yanagida and Whitman (2009), Yanagida (2012), and Whitman and Yanagida (2012) have shown that there was an asymmetry between null case-​marked objects in nominalized clauses and those taking the particle wo, which is the historical precursor of the modern accusative particle o: wo-​marked (accusative) objects are interpreted as specific and are required to precede a genitive subject, while bare objects remain in their base positions immediately preceding the verb. Old Japanese (35) a. 佐欲比賣能故何比列布利斯夜麻    (Manyoshu 868) [vP Sayopimye=no kwo=ga [VP pire puri]]-​si yama Sayohime=gen child=gen scarf wave-​pst.adn hill ‘the hill where the girl Sayohime waved her scarf ’ b. 蜻野叫人之懸者 (Manyoshu 1405) [AspP Akidu nwo=wo [vP pito=no [VP tObj kakure-​ba]]] Akizu field=acc man=gen speak.of-​when ‘When a man speaks of the moorland of Akizu …’ Yanagida and Whitman (2009) and Whitman and Yanagida (2012) propose that the v in the nominalized clause assigns inherent genitive case to its specifier but is unable to value structural case on an internal argument. A nonspecific object remains in its base position immediately preceding the verb. They point out that these objects are generally N0-​level categories, and they analyze them are incorporated to the verb. But when the object is specific, it raises from its base position to the specifier of an aspect projection dominating vP, where it values structural accusative case. As for bare internal argument subjects in unaccusatives, which are potentially phrasal, Whitman and Yanagida (2009) propose that T values nominative case on these DPs, given that vP is defective in unaccusatives. In this way, Whitman and Yanagida (2012) implement the proposal that non-​ accusative alignment emerges in an otherwise accusative language when the external

 

516   Edith Aldridge argument in a transitive clause is assigned inherent case, while the object is denied structural licensing by this same v head.

21.5.3 Austronesian Tagalog has been claimed by Payne (1982), De Guzman (1988), Aldridge (2004, 2012b), and others to be a language with ergative alignment. This can be seen in the contrast between (36a) and (36b), in which the object in a transitive clause takes the same ang case-​marker as the subject in an intransitive clause. Example (36c) is analyzed as an antipassive, a semantically transitive but syntactically intransitive clause type (see also Polinsky, Chapter 13, this volume, on the characteristics of antipassives). Like the simple intransitive in (36b), the subject (external argument) receives ang marking in the antipassive, while the object appears with inherent genitive case. Tagalog (36) a. Bili ng babae ang isda. buy gen woman nom fish ‘The woman bought the fish.’ b. Dating ang babae. arrive nom woman ‘The woman arrived.’ c. Bili ang babae ng isda. buy nom woman gen fish ‘The woman bought a fish.’

21.5.3.1 Synchronic Approach Kaufman (2009a, Chapter 24, this volume) offers an explanation for the use of genitive case to mark DP arguments that are not absolutives by analyzing Tagalog clauses as nominalizations embedded in a copula construction. Kaufman proposes that Tagalog lacks a v functional category. Consequently, lexical roots merge with n and project a nominal predicate. The external argument of a transitive clause is treated as a possessor merged in [Spec, n] and assigned genitive case. The nominalized predicate is a relative clause with a null operator in the specifier of PredP. T is treated as a null copula, which mediates a predication relation between PredP in its specifier and its DP complement. Tagalog (37) a. Bili ng babae ang isda. buy gen woman nom fish ‘The woman bought the fish.’ (lit. ‘The fish is what the woman bought.’)

 

Intransitivity and the development of ergative alignment    517 b.

TP PredP

OPi

T’ Pred’

Pred

T DP

binili ng babae bought gen woman

DPi ang isda nom fish

This analysis offers an account for the constraint in Tagalog and many other Austronesian languages (as well as syntactically ergative languages generally) that all DPs other than the absolutive are ineligible to undergo A’-​movement, since genitive DPs are all contained within the relative clause DP, an island to extraction. Example (38a) shows that a relative clause can be formed on the absolutive object in a transitive clause. However, the ergative subject cannot be extracted in this way, as shown in (38b). Tagalog (38) a.

isda-​ng2 bili ng babae fish-​lk buy gen woman ‘fish that the woman bought’

b. *babae-​ng bili ang isda woman-​lk buy nom fish ‘woman who bought the fish’ However, this biclausal analysis suffers from a number of shortcomings. Aside from the stipulation that all finite clauses are copula constructions, the structure in (37b) makes incorrect predictions about word order in the language, as pointed out by Aldridge (2009). For example, this structure does not allow the absolutive DP to intervene between predicate-​internal constituents. Example (39), however, shows an absolutive appearing between the verb and a genitive object (39a) and a goal PP (39b). Since these constituents are arguments of the verb, I assume they would be base generated in the PredP on Kaufman’s analysis. Consequently, their dislocation to clause-​final position should invoke an island violation along the lines of (38b), contrary to fact.

2  The linker appearing between the head NP and the modifying clause is spelled the same way as the genitive case marker. But the two differ in pronunciation, the linker pronounced as the velar nasal, and are not the same morpheme.

 

518   Edith Aldridge Tagalog (Aldridge 2009: 53–​54) (39) a.

[PredP Bili tDP ] ang babae ng bahay. buy nom woman gen woman ‘The woman bought a house.’

b. [PredP I-​bi-​bigay=ko tPP ] ang bulaklak kay Maria. appl-​red-​give=1sg. gen nom flower to Maria ‘I will give the flowers to Maria.’ In the next subsection, I propose that PP extraction from a predicate nominal is actually possible in certain Austronesian languages. However, there is a crucial structural difference between my approach and Kaufman (2009a). In Kaufman’s approach, extraction takes place from a relative clause in a specifier position, which should result in a violation of Huang’s (1982) Condition on Extraction Domain. In the structure I propose, the predicate nominal is the complement of a Pred head. I draw a parallel between this type of extraction and the lack of opacity effects in subextraction from indefinite objects in English.

21.5.3.2 Diachronic Approach Starosta et al. (1982/2009) take a diachronic approach to a possible connection between nominalizations and the type of ergative morphosyntax observed in most Philippine and Formosan3 languages. Specifically, they propose that many affixes appearing on transitive verbs in the modern languages were nominalizers in Proto-​Austronesian (PAn). The nominalizer exemplified in (40) is *-​an,4 which attaches to a verb and projects a relative clause predicated of the matrix subject in a copula construction. (40)

S NP N ‘climb’-an

NP ‘John’ (GEN)

NP ‘mountain’ (NOM)

‘The place where John climbed is the mountain.’   (Starosta et al. 1982/2009: 313) In time, biclausal copula constructions like (40) were reanalyzed as transitive verbal clauses like (41). Nominalizers like *-​an were consequently became verbal affixes. A reflex of *-​an continues to function solely as a nominalizer in Puyuma and Rukai but 3  The term “Formosan languages” refers collectively to the Austronesian languages spoken in Taiwan, but does not indicate a subgroup. 4  Starosta et al. reconstruct the morpheme as *-​ana, but *-​an is the commonly accepted reconstruction in more recent work by Austronesian historical linguists.

 

Intransitivity and the development of ergative alignment    519 is employed as a locative applicative attaching to verbs in many other Formosan and Philippine languages.5 (41)

S V ‘climb’-an

NP ‘John’ (GEN)

NP ‘mountain’ (NOM)

‘John climbed the mountain.’

(Starosta et al. 1982/2009: 314)

An advantage over Kaufman (2009a) is that the resulting structure is monoclausal and does not require that all root clauses be analyzed as copula constructions. However, the detailed steps in the reanalysis are not spelled out. Nor is a clear motivation for the change identified. In what follows, I develop the insight of Starosta et al. (1982/2009), Kaufman (2009a), and others that there is a connection between nominalization and ergative clauses in Philippine and most Formosan languages and propose an explicit account of how a nominalized relative clause in a copula construction could have been reanalyzed as a finite ergative clause. I also adopt Ross’ (2009) proposal that the reanalysis of nominalizations to root clauses took place in what Ross terms the Nuclear Austronesian subgroup and should not be attributed to Proto-​ Austronesian or Pre-​Austronesian. This subgroup encompasses all Austronesian languages except for Rukai, Puyuma, and Tsou, the last three still spoken in Taiwan. (42)

(based on Ross 2009)

Austronesian Rukai

Tsou

Puyuma

Nuclear An (NAn)

Much of Ross’ evidence for this grouping comes from the fact that affixes forming finite ergative verbs in NAn languages bear resemblance to morphemes only employed in nominalizations in extra-​NAn languages. Since the reanalysis of nominalizations as finite verbs has not taken place in the extra-​NAn languages, this change can be regarded as an innovation defining the NAn subgroup. As an example, the Puyuma finite transitive verb in (43a) takes the suffix -​aw, an affix which never appears on a verb in a relative clause. Verbs in relative clauses must be nominalized, as in (43b), where the verb takes perfective aspect marker and the nominalizer -​an. Puyuma (43) a. tu=trakaw-​aw na 3.gen=steal-​tr def.nom ‘Isaw stole the money.’ 5 

paisu kan  isaw money sg.obl Isaw

(Teng 2008: 147)

See Gildea (1998) for a similar account of the reanalysis of nominalizations as ergative clauses in Cariban languages.

 

520   Edith Aldridge b. ala amuna sadru [[tu=trekelr-​an] maybe because many 3=drink-​nmlz ‘Maybe because the milk he drank is a lot.’

na asi] def.nom milk (Teng 2008: 105)

These morphemes surface only on nominalized verbs in Puyuma but appear on finite verbs in NAn languages like Tagalog. Note that -​an has been reanalyzed as a locative applicative in Tagalog. Tagalog (44) Bilh-​an ng babae ng buy-​appl gen woman gen ‘The woman bought a/​the fish at my store.’

isda fish

ang nom

tindahan=ko. store=1.sg.gen

For the reanalysis of nominalizations as verbal clauses, I assume with Starosta et al. (1982/2009) that the input structure was a nominal predication. In Aldridge (2004, 2016), I  presented evidence for the following structural analysis of nominal predication in Tagalog. The predicate-​initial word order is derived by moving the predicate NP to a position above the subject. The trace position of the NP is indicated with angled brackets.6 Tagalog (45) a. Importante-​ng miyembro ng Sizzlers si Gilbert. • important-​lk member gen Sizzlers nom Gilbert ‘Gilbert is an important member of the Sizzlers.’ b.

CP

C

TP T

PredP NP

Pred’ Gilbert

Pred’ Pred

importante miyembro 6 

N’ Sizzlers

See Massam (2000, 2001, 2003), Rackowski and Travis (2000), Pearson (2001), Aldridge (2004), Cole and Hermon (2008), and others for derivations of Austronesian predicate-​initial word order by means of phrasal fronting of all or part of the predicate.

 

Intransitivity and the development of ergative alignment    521 This analysis can also accommodate cases in which part of the nominal predicate has been stranded before fronting. In the following example, the complement PP vacates the predicate to be fronted and is stranded inside the PredP. Cole and Hermon (2008) propose a very similar analysis of stranding accompanying predicate-​fronting in Toba Batak. Below, I suggest an important role for stranding in the reanalysis of nominal predications as verbal clauses. Tagalog (46) a. Importante-​ng miyembro si Gilbert ng Sizzlers. Important-​lk member nom Gilbert gen Sizzlers ‘Gilbert is an important member of the Sizzlers.’ b.

CP

C

TP T

PredP NP

Pred’ Gilbert

Pred’ Pred

XP sizzlers

importante miyembro

N’ tsizzlers

I propose that the input to the reanalysis in question was a nominal predication in which the predicate was a nominalized relative clause, specifically the type Krause (2001) terms “reduced relatives with genitive subjects.” The structure contains a position for aspect but not tense. As Chen (2008: 96) shows that the nominalizer -​ane (cognate with the nominalizer -​an in Puyuma) attaches closer to the root than the perfective marker -​nga, I place AspP outside nP. The external argument and nominalizing affix are merged in the nP layer. Internal arguments are merged together with the root.

 

522   Edith Aldridge Budai Rukai (Chen 2008: 84)7 (47) a. Ta-​badh-​ane ki tina-​ini ki lalake-​ini nonfut-​give-​nmlz gen mother-​3sg.gen obl child-​3sg.gen ka laimai. nom clothes ‘The clothes are what the mother gave her child.’ b.

CP

C

TP T

PredP DP

Pred’ laimai

Pred’ Pred

D

AspP

ta-badh-ane

nP

tina-ini

n’ √P

Interestingly, stranding is also possible from within this type of predicate nominal. In (48), the dative argument has been moved out of the relative clause and appears in a position following the matrix subject. I assume that the PP moves from complement position in the root phrase and exits the DP via the edge this phase (and possibly also the nP) in order to avoid violating the Phase Impenetrability Condition of Chomsky (2001 and subsequent works). Budai Rukai (Chen 2008:82) (48) a. Ta-​badh-​ane ki tina-​ini ka nonfut-​give-​nmlz gen mother-​3sg.gen nom

laimai clothes

ki lalake-​ini. obl child-​3sg.gen ‘The clothes are what the mother gave her child.’ 7  Chen (2008) does not analyze this construction as a nominalization but rather as “object voice.” However, he admits that affixes like -​ane are clearly nominalizers in the language. He also attributes stative semantics to the construction. So it is difficult for me to understand the rationale for not analyzing this construction as a nominalization.

 

Intransitivity and the development of ergative alignment    523 CP

b. C

TP T

PredP DP

Pred’ laimai

Pred’ Pred

XP PP



D' D

AspP

ta-badh-ane

nP n'

tina-ini n

It may seem surprising here that the predicate nominal is not an island to extraction. However, two points are worthy of note here. First, extraction takes place from complement position. This is a crucial distinction between the current analysis and Kaufman’s (2009a) proposal in (37b), in which the predicate is in a specifier position. Second, extraction from DP is in fact possible if certain conditions are met. For example, indefinite DPs in object position in English allow subextraction, but definite DPs do not. Regarding the copula constructions in Rukai, since the DP in question is predicational and not referential, it cannot be definite.8 (49) a. Who did you meet [some friends of ]? b. *Who did you meet [those friends of ]? Returning to the diachronic reanalysis of copula constructions as verbal clauses, I suggest here that the possibility of remnant predicate fronting may have played a key role. Specifically, when material from the relative clause is stranded to the right of the matrix subject, the utterance comes to strongly resemble a monoclausal verbal construction in its word order, as shown in (50b). Example (50a) repeats the stranding example in (48a). 8  Note further that reduced relative clauses in Tagalog existential constructions allow subextraction. See Aldridge (2012a) for discussion.

√P

 

524   Edith Aldridge Budai Rukai (50) a. Ta-​badh-​ane ki tina-​ini ka laimai nonfut-​give-​nmlz gen mother-​3sg.gen nom clothes ki lalake-​ini. obl child-​3sg.gen ‘The clothes are what the mother gave her child.’

(Chen 2008: 82)

b. Wa-​bai ku laimai ka kineple ki cegau. nfut-​give acc clothes nom Kineple obl Cegau ‘Kineple gave clothes to Cegau.’

(Chen 2008: 40)

The reanalysis is straightforward. The mechanisms involved are relabeling and pruning, as proposed by Whitman (2000). Because of the resemblance to monoclausal constructions, the nominalized verb is interpreted as the main verb. As a result, the nP in the relative clause is relabeled as vP. (51)

CP C

TP T

PredP DP

Pred’ laimai

Pred’ Pred

D

AspP

ta-badh-ane

nP > vP

tina-ini

n’ > v’ √P

Since the construction as a whole has been parsed as verbal and monoclausal, there is no longer any evidence for the DP and PredP layers, so these are pruned away, with the result that the AspP dominating vP will be directly selected by T. This clause will exhibit an ergative case-​marking pattern, because the genitive-​assigning functional head in the erstwhile relative clause is now the matrix v. As a former nominal category, this v also lacks the ability to license structural accusative case on the object. Consequently, the object will enter into an Agree relation with T and value nominative case in order to be

 

Intransitivity and the development of ergative alignment    525 licensed. In this way, the ergative v emerged in Proto-​Nuclear Austronesian with precisely the properties expected in the analysis in (4). (52)

CP C

TP T

AspP Asp

vP v’

DP[GEN] v

√P √

DP[NOM]

Identifying a nominalized relative clause structure as the historical antecedent of the ergative clause type has the added advantage of providing an explanation for the extraction restriction observed above. Recall that the internal argument in a transitive clause can be the head of a relative clause, as in (53a), but not the external argument, as shown in (53b). A subject can only be extracted from an intransitive clause. Example (53c) is an antipassive, in which intransitive morphology appears on the verb and the object receives inherent genitive case rather than structural nominative. On the other hand, extraction of non-​DPs is not subject to this constraint, as shown in (53d). Tagalog (53)

a.

isda-​ng bili ng fish-​lk buy gen ‘fish that the woman bought’

babae woman

b. *babae-​ng bili ang isda woman-​lk buy nom fish ‘woman who bought the fish’ c. babae-​ng bili ng isda woman-​lk buy gen fish ‘woman who bought a/​the fish’ d. Saan bili ng babae ang isda. where buy gen woman nom fish ‘Where did the woman buy the fish?’

 

526   Edith Aldridge I propose that this locality restriction is a direct consequence of the properties of relativization in Austronesian languages. First consider Rukai, which is an accusative language9 and has not undergone the reanalysis of nominalizations as verbal clauses. In Rukai, relative clauses formed on subject position are finite CPs. This is evidenced by the fact that they contain the same tense markers as finite clauses. Tanan Rukai10 (54) a. luða ay-​kɨla ku tina=li tomorrow fut-​come nom mother=1.sg.gen ‘My mom will come tomorrow.’ b. [kuaDa ay-​suwasuwaw] ka muka-​baru-​barua dem fut-​clean top girl ‘The one who will clean is the girl.’ On the other hand, object relatives are nominalized with the suffix -​anɨ, transcribed as -​ane for the Budai dialect in the preceding discussion. But they do not carry tense marking. The prefix a-​expresses imperfective aspect. Tanan Rukai (55) w-​aga=su sa aga sa [a-​kanɨ-​anɨ=ta ki maum] past-​cook=2.sg indef food indef imprv-​eat-​nmlz=1.pl.inc p night ‘Did you cook dinner (the food that we will eat tonight)?’ The analysis I propose here accounts for the restrictions in both Tagalog and Rukai. The strict locality between DPs observed in the Tagalog examples in (53) suggests that the feature driving movement in the relative clause is a feature which specifically attracts DPs. In Aldridge (2004), I proposed that Austronesian languages do not have A-​movement to [Spec, TP]. Rather, the EPP feature that appears on T in other languages is carried instead by C in Austronesian languages. This accounts for the extraction restriction, since movement to [Spec, CP] in Austronesian languages is subject to the same category sensitive locality observed in subject movement to [Spec, TP] in other languages.

9  Interestingly, Puyuma and Tsou are ergative languages. Aldridge (2015, 2016) proposes a revision to Ross’ (2009) subgrouping hypothesis by positing an Ergative Austronesian subgroup as sister to Rukai and parent to Puyuma, Tsou, and Nuclear Austronesian. Rukai retains the accusative alignment of Proto-​Austronesian. Ergativity was first innovated in irrealis clauses in Proto-​Ergative Austronesian. This alignment is retained in irrealis clauses in the Nuclear-​Austronesian languages. These languages have further reanalyzed nominalizations as ergative clauses in the realis paradigm. 10  Data for the Tanan dialect of Rukai were collected during fieldwork in Taiwan in 2013 and 2014 with support from the Chiang Ching-​Kuo Foundation.

 

Intransitivity and the development of ergative alignment    527 In this chapter, I restate this analysis within Chomsky’s (2005, 2008) proposal of C-​T inheritance. Specifically, the lack of A-​movement in languages like Tagalog and Rukai is accounted for by proposing that C-​T inheritance does not take place, so all movement targets [Spec, CP]. Another consequence of the lack of C-​T inheritance is that the unvalued ɸ-​feature which licenses the subject remains on C. Since this feature will probe for a set of valued ɸ-​features, which are only visible on DPs, category sensitive locality between DPs will be observed in relative clause formation.11 In a finite CP, the subject enters into an Agree relation with the ɸ-​probe on the C/​T complex and value its case feature. Since the languages under discussion are verb-​initial, I assume that the subject does not generally raise to [Spec, CP]. However, in cases of A’-​movement, the subject does move. I assume that this movement takes place directly from the subject’s base position in vP, as per Ouali’s (2006) adaptation of C-​T inheritance. Tense features are also retained on the C head. This accounts for the fact that subject relatives in Rukai can be formed on finite clauses and need not be nominalized. Extraction of the external argument in antipassives in Tagalog takes place in exactly the same way.

(56)

DP D

CP

DP[CASE:NOM] C’ vP

C

v’

v

VP

Relativization is not possible on object position in a full CP. If the object were to move to the edge of vP and consequently become visible to the ɸ-​probe on the C/​T complex, the derivation would not result in an object relative clause. There are two possible reasons for this. If the object’s ɸ-​features entered into an Agree relation with C/​T, the object would move to [Spec, CP]. However, this would leave the subject’s case feature unvalued, and the derivation would crash. Alternatively, if we assume the Activity Condition of Chomsky (2001), then the object would not be visible to the probe on C/​T, since its case feature has already been valued. The derivation would not crash, but the subject would still be the DP to move to [Spec, CP].

See Carstens (2005) and Henderson (2006, 2011) for other proposals that ɸ-​features can drive relative clause formation. 11 

 

528   Edith Aldridge (57)

CP C

TP vP

T

DP[CASE:ACC]

v’ v’

DP[CASE: ] v

√P √

In order to extract an object over the subject, the competition for ɸ-​feature valuing must be eliminated. This is accomplished in a nominalization, since the subject is given inherent genitive case, and there is also no C/​T layer. Rather, it is the nominalizer itself which creates the gap to form the relative clause. This probe undergoes Agree with the object and raises it to the edge of nP.12 This accounts for the fact that object extraction in Rukai requires a nominalization. Nuclear Austronesian languages like Tagalog have inherited this relativizing n as the transitive (ergative) v, accounting for why internal arguments are extracted in transitive clauses.13 DP

(58)

D

AspP Asp

nP DPOBJ

n’ n’

DP[GEN] n[

√P

:]

√

12  I assume that the moving constituent in a relative clause is a null operator, which will not be spelled out with phonetic content, and that movement to the highest strong phase edge suffices to create a structure which can be interpreted as a lambda abstraction at the interface. But minimal revision could also accommodate a head raising approach by further moving the DP to a peripheral position. 13  This proposal entails that transitive v always raises an object to its outer specifier, which is a welcome consequence of this analysis. It is generally assumed that absolutive objects in Tagalog raise to the vP phase edge when the verb bears transitive morphology (Rackowski 2002; Aldridge 2004; Rackowski and Richards 2005) but is spelled out in its base position if it does not undergo further movement to [Spec, CP]. Raising of the object accounts not only for its ability to undergo A’-​movement but also for its presuppositional, wide scope interpretation.

 

Intransitivity and the development of ergative alignment    529 Since the probe driving the movement can only agree with a DP, non-​DPs can dislocate freely, presumably motivated by focus. It should now be clear how nominalization enables object relativization by eliminating the competition to value ɸ-​features on the relativizing functional head. Interestingly, this DP locality restriction on A’-​movement is widely observed in syntactically ergative languages, particularly those in which morphological ergativity arguably resulted from the reanalysis of a clausal nominalization. Recall from section 21.5.1 that Inuit languages exhibit a syncretism between ergative and genitive case, and ergative clauses have been argued to be derivationally related to nominalizations. Mayan is another language family which has both the ergative/​genitive case syncretism and the DP extraction restriction. The preceding proposal speaks to this correlation by tracing the source of ergativity to a nominalized relative clause in which the inherent case-​marking on the transitive subject is a strategy for allowing the object to enter into an Agree relation with a higher functional head without intervention by the subject. See Aissen, Otsuka, and Erlewine et al. (Chapters 30, 40, and 16, respectively, this volume) for other approaches to the extraction restriction in ergative languages.

21.6 Conclusion Ergative alignment can be characterized as involving one structural case, which appears on intransitive subjects and transitive objects, and one inherent case, which is assigned to transitive subjects. In this chapter, I have summarized various accounts of the emergence of ergative alignment and shown that they are compatible with an analysis which traces the source of ergativity to an intransitive v which assigns inherent case to its specifier and does not structurally license an internal argument. I have further shown that the strict locality exhibited between DPs in syntactically ergative languages results when the diachronic source of ergativity is a nominalizing v which facilitates movement of an internal argument over an external argument in relative clause formation.

Abbreviations 1, first person; 2, second person; 3, third person; abl, ablative; abs, absolutive; acc, accusative; adn, adnominal; appl, applicative; caus, causative; conc, conclusive; cop, copula; dat, dative; def, definite; dem, demonstrative; erg, ergative; f, feminine; fut, future; gen, genitive; indef, indefinite; inc, inclusive; inh, inherent; ins, instrumental; intr, intransitive; ipfv, imperfective; lk, linker; m, masculine; N, noun; nmlz, nominalizer; nom, nominative; nfut, nonfuture; obl, oblique; opt, optative; P, preposition; part, partitive; pass, passive; pfv, perfective; pl, plural; prs, present; pst, past; ptcl, particle; ptcp, participle; pv, preverb; red, reduplication; rel, relative case; sg, singular; top, topic; tr, transitive; uns, unspecified tense/​ aspect/​mood; V, verb.

 

Chapter 22

Devel opments i nto an d ou t of e rg at i v i t y: Ind o-​A ryan diac h rony Miriam Butt and Ashwini Deo

22.1 Introduction This contribution provides an overview of what is known about the diachrony of ergativity in Indo-​Aryan. The Indo-​Aryan language family provides an ideal situation for a study of case and ergativity. Old Indo-​Aryan (Vedic, Sanskrit) did not have an ergative case, but many of the New Indo-​Aryan languages do. Some of the modern languages further show evidence of having lost ergative alignment. The written record for Indo-​Aryan goes back to about 1900 bc and is the longest documented continuous historical record available to us in historical linguistics. Table 22.1 provides a time line as relevant for the discussions in this chapter.1 The standard modern languages tend to be spoken by millions of people and have a rich literary record. They are thus in principle easily accessible for linguistic study and show an interesting variety of Table 22.1 Indo-​Aryan chronology

1 

timeline

stage

language

1900 BC–​1100 BC

I

Early Old Indo-​Aryan

1000 BC –​200 BC

I

Later Old Indo-​Aryan

300 BC –​700 AD

II

Middle Indo-​Aryan

1100 AD–​present

III

New Indo-​Aryan

Approximate dates are based on Alsdorf (1936); Witzel (1999); Jamison and Witzel (2002).

 

Indo-Aryan diachrony   531 patterns in the domain of ergativity. Although much has been written about the patterns found in Indo-​Aryan, a complete understanding of the diachrony of ergativity in Indo-​Aryan remains to be attained. The issues surrounding ergativity are complex. In addition to the mere absence or presence of an ergative case, ergativity has been discussed in terms of alignment patterns (cf. Aldridge (Chapter 21, this volume)), which tend to prominently involve not only case marking but also agreement and other linguistic phenomena that are sensitive to alignment issues. It is still a matter of current debate whether Old Indo-​Aryan (OIA) was already ergatively aligned, indeed there are proponents of an ergative analysis of Proto Indo-​European (cf. Bauer 2000). In addition to the inherent complexity attached with an understanding of ergativity, our understanding of the diachronic phenomena has also partly been hampered by too ready an acceptance of an overly simplistic scenario in some of the mainstream literature (cf. section 22.2.1). Another complicating factor is the comparative lack of historical evidence for the crucial time in which, following the erosion of the older case system, new case markers began to be innovated, namely the early New Indo-​Aryan (NIA) period from about 1100 ad–​1300 ad. A final obstacle to a definitive understanding is the comparative dearth of modern work on NIA languages. In this chapter, we first lay out some general background on the diachrony of ergativity in Indo-​Aryan (section 22.2). This discussion functions as a backdrop to the more detailed presentation in the following sections on the diachronic trajectory as far as it is known (section 22.3). We include a section thematizing further perspectives on ergativity, including proposals regarding markedness, the role of clausal and lexical semantics, language contact, and competition among case markers (section 22.4.4). The chapter ends with a discussion and an outlook onto where we believe future efforts at understanding ergativity in the history of Indo-​Aryan should be concentrated (section 22.5).

22.2  General Background Two central linguistic sources have been implicated cross-​linguistically in the rise of ergativity: passive and possessive constructions (Benveniste 1952; Anderson 1977; Plank 1979; Garrett 1990; Dixon 1994; Harris and Campbell 1995; Bynon 2005; Haig 2008). We discuss both here because both have been proposed for an understanding of the Indo-​ Aryan situation.

22.2.1 Passive-​to-​Ergative In the passive-​to-​ergative reanalysis scenario, ergative alignment is taken to arise when a passive participial construction that expresses the resultative or perfect aspect is reanalyzed as an active, ergative clause with perfective aspectual reference (e.g. Dixon 1994;

 

532    Miriam Butt and Ashwini Deo Harris and Campbell 1995). As part of this reanalysis, the former oblique agent adjunct of the participle (generally an instrumental) is reanalyzed as an ergative subject. The basic idea with respect to a reanalysis of former passives is illustrated in (1) with the constructed example from Sanskrit (based on Garrett 1990, 263). (1) a. ahi-​r indr-​eṇa ha-​ta-​ḥ serpent-​nom.m.sg Indra-​ inst.sg kill-​ptcpl-​nom.m.sg ‘The serpent has been killed by Indra.’ Actually: The serpent is one killed by Indra.’

Sanskrit

b. Reanalyzed as: serpent-​nom.m.sg Indra-​erg.sg kill-​perf-​m.sg ‘Indra has killed the serpent.’ As illustrated in (1), the participial implicated in the development of an ergative is the ta marked adjectival passive participial form. It assigns nominative case to the patient argument of the verb while the agent appears with instrumental case marking. The participle, which has resultative semantics, allows reference to states that hold of objects as a result of the occurrence of some event. With transitive unaccusative verbs as in (2), such a participle predicates a result state of the sole argument while with transitive change-​of-​state verbs, it predicates a result state of the patient argument. It is this resultative semantics that generates the ergative-​like configuration in which objects of transitive verbs and subjects of intransitive verbs come to be marked identically. Over time, this passive-​like stative construction comes to be reanalyzed as active and the former instrumental is reinterpreted as an ergative. (2) ahaṃ sarayūtīra-m ā-​ga-​ta-​ḥ I.nom Sarayu.bank-acc pv-​go-​ptcpl-​nom.m.sg ‘I came to the bank of the Sarayu.’

(Rāmāyaṇa 2.58.12) (from Bynon 2005, 11)

There are many indications that the ta participle already had an active interpretation in Epic Sanskrit. In (3), we provide an example taken from Bynon (2005, 13), who shows that an active analysis not only does justice to the narrative structure of the text, but also makes syntactic sense given that the instrumental agent is controlling the subject of the gerund clause. The example is taken from the Rāmāyana, a text dated to ∼200 BC, and written in what is known as Epic Sanskrit. For further examples, see Bynon (2005). (3) śabda-​m ālakṣ-​ya mayā … visṛṣ-​ṭo nārāca-​ḥ sound-​m.acc aim-​ger I.inst … release-​ptcpl.nom.m.sg arrow-​nom.m.sg ‘Aiming at the sound I released an arrow.’ (Rāmāyaṇa 2.58.15) Along with evidence for an active interpretation of the participle, there is also good reason to believe that the ta participle was integrated into the finite verbal system of the

 

Indo-Aryan diachrony   533 languages—​being used robustly to describe past, culminated events. An example from the Mahābhārata, another Epic Sanskrit text, shows that the participial form may be used as the main verbal predicate in the clause with past-​referring frame adverbials (Deo 2012, 10). (4)

purā devayug-​e ca eva dṛṣ-​ṭaṃ sarvaṃ formerly god.age-​loc.sg and ptcl see-​perf.n.sg everything mayā    vibho I-​inst.sg  lord-​voc.sg ‘Lord, formerly, in the age of the Deva (Gods), I saw everything.’ (Mahābhārata 3.92.6a)

Bynon (2005, 11) provides further evidence from the Rāmāyana in support of this use of the participle. Here, a story is being told and the same action is alternatively described using the aorist or the ta participle. The aorist was used to express resultative perfect and recent past meanings (Kiparsky 1998b; Condoravdi and Deo 2014). A reasonable conclusion to draw from the existence of alternations such as (5) is that the ta participle could already be used to describe events in the recent past in Sanskrit. For more examples, again see Bynon (2005). (5)

a. aśrauṣam … ghoṣam hear.1.sg.aor … noise.acc ‘I heard a noise.’ b. ṣru-​to hear-​ptcpl.nom.m ‘I heard a sound.’

mayā I.inst.sg

(Rāmāyana 2.57.16) śabdo sound.nom.m.sg

(Rāmāyana 2.58.13)

Bynon considers register as a possible conditioning factor in the use of the ta participle vs. the aorist form. Be that as it may, the diachronic record clearly shows that the ta participle replaced the older inflecting tense forms over time. Bynon cites Bloch (1906, 48, 58) as counting about 150 main clause past participle predicates vs. 1,033 finite verbs at the time of the Mahābhārata (Epic Sanskrit). In a text from a later period this proportion is reversed: 790 finite verbs vs. 1,750 morphologically non-​finite expressions. The reanalysis of the ta participial marking into active inflectional past/​perfective morphology is undisputed. The modern NIA languages tend to show split-​ergativity in that the ergative case predominantly appears in conjunction with this past/​perfective morphology (see discussion in section 22.3.3). This type of split-​ergativity is part of a wider cross-​linguistic pattern (Trask 1979), though one that is still not wholly understood. From the perspective of the passive-​to-​ergative hypothesis, the historical explanation is straightforward:  the ergative is naturally only associated with the past/​perfective morphology because it is a direct continuation of the old agentive instrumental adjunct. The adjunct gets reanalyzed as the subject of the clause with the original instrumental case being reanalyzed as ergative. This hypothesis is supported by the fact that cross-​linguistically ergatives are often form-​identical with instrumentals. However, the

 

534    Miriam Butt and Ashwini Deo hypothesis is problematic within the Indo-​Aryan context because it also tends to assume an instrumental-​to-​ergative change. The instrumental-​to-​ergative hypothesis was first proposed by Trumpp (1872, 113) for Indo-​Aryan. Despite being debunked immediately and roundly by Beames (1872–​79), it has continued to persist, being presented as textbook knowledge in works such as Dixon (1994) and Harris and Campbell (1995). What Beames (1872–​79) objects to is Trumpp’s claim that the OIA instrumental case marking was the direct ancestor of the modern ergative case forms. This scenario is improbable because of the erosion of the OIA case system as attested in Indo-​Aryan diachrony. In fact, at least some of the Indo-​Aryan languages went through a period in which there was no overt marking of ergative case but an ergative alignment of clauses (see section 22.3.2). The instrumental-​to-​ergative hypothesis historically has been closely associated with the passive-​to-​ergative hypothesis. The former involves the case marker per se, the latter notions of how clauses are structurally aligned (see Butt 2006b for background and discussion). It is therefore worth mentioning that while the ta participle is often referred to as “passive” in the literature, it must be distinguished from the actual passive found in Sanskrit. As illustrated in (6), this passive was formed with the morpheme -​ya. It was an integral part of the language, but crucially appears to have absolutely nothing to do with the realization of either an ergative case or an ergative structural alignment (Klaiman 1978). (6) devadattena kaṭaḥ kri-​ya-​te/​kri-​ya-​nte Devadatta.inst.sg mat.nom do-​pass-​3.sg/​do-​pass-​3.pl ‘by Devadatta a mat/​mats is/​are made’ Sanskrit (adapted from Hock 1986, 16) Given the idea that the rise of ergativity is tied to a reanalysis of a passive in passive-​to-​ ergative hypothesis, why did the Sanskrit -​ya-​ passive not give rise to the ergative pattern? One answer to this question may be the frequency of the appearance of the instrumental agent. Like the modern NIA languages, Sanskrit was able to drop arguments freely. A comparison of the instrumental agents occurring with the deverbal adjectival -​ta participle and the passive -​ya-​ shows that for both Sanskrit (Gonda 1951, 22) and the later Pāli (Peterson 1998), the instrumental agents in passives were rarely expressed. With the deverbal participle -​ta construction, on the other hand, the instrumental agent was almost always expressed overtly. The -​ta participle was also part of the overall tense/​aspect paradigm in Sanskrit (see earlier discussion in this section) and is the ancestor of inflectional tense/​aspect morphology in NIA. The passive -​ya-​ did not enter this system and has generally been lost in NIA. Implicating passive, that is the voice system, as a conditioning factor for ergativity is thus misleading, at least in the Indo-​Aryan context. Rather, stative expressions that are being drawn into the tense/​aspect system of a language appear to furnish the starting point of an ergative system. These stative expressions begin by being nominal/​adjectival and therefore have a “demoted” agent. As they are reanalyzed as verbal, the status of the agent argument is concomitantly reanalyzed and it is syntactically realized as the subject of a verbal predication. Research with a strong South Asian focus or background has tended to acknowledge this complex diachronic scenario and has also sought to understand the origin of the

 

Indo-Aryan diachrony   535 modern ergative markers in nuanced detail. The ergative is homophonous with the instrumental in some NIA languages—​the semantic connection here is obvious since agents and instrumentals share many semantic characteristics. However, we see that the ergative case clitic in languages like Urdu/​Hindi and Punjabi is also cognate to the dative case clitic in other languages like Gujarati and Rajasthani (Tessitori 1913, 1914; Montaut 2003, 2006; Butt 2006a; Butt and Ahmed 2011), a development that does not have an obvious explanation. See section 22.4.4 for further discussion.

22.2.2 Possessives Bynon (2005) also sees difficulties for the passive-​to-​ergative hypothesis in the explanation of how the instrumental/​agent came to occupy a clause initial position in the default word order (vs. the clause internal position illustrated in (1b)) and in the appearance of the agent/​instrumental on intransitives. She proposes an alternative hypothesis which looks to developments in Indo-​Iranian and which invokes a possessive construction as the ancestor of ergativity in Indo-​Aryan. The diachronic scenario generally assumed for Iranian is that there was an accusative aligned ancestral language that shifted to ergative alignment. This then shifted back to an accusative language, namely modern Persian, which is morphologically and syntactically accusative (but see Haig (2008) for a more differentiated discussion and Haig (Chapter 20, this volume)). The shift in alignment in Iranian is associated with a possessive construction exemplified by a sentence involving manā kartam ‘done of (by) me’ (paraphrasable as ‘my done thing’). The participle involved is the same -​ta participle as the one implicated in the OIA system, the common Proto Indo-​European ancestor has been reconstructed as *to/​no. The posited reanalysis is illustrated in (7) for Old Persian (based on Kent (1953)). (7)

a. ima tya manā kartam pasāva yaθā xšāyaθiya that which 1.s.g.gen do.ptcpl after when king abavam become.past.1.sg           Old Persian ‘This (is) that (which) was done by me after I became king.’ Reanalyzed as: ‘I did that after I became king.’ b. avaθ=šām hamaranam kartam thus=3.pl.gen battle do. ptcpl ‘Thus by them battle was done.’ Reanalyzed as: ‘Thus they did battle.’

  Old Persian

In the manā kartam construction, the agent is realized as a genitive possessor. As with the passive-​to-​ergative hypothesis, the overall idea is that the original genitive-​marked possessor argument was reanalyzed as an ergative-​marked agent, and that the stative/​

 

536    Miriam Butt and Ashwini Deo passive participle was reanalyzed as an active verb that was part of the finite verbal paradigm of the language. Bynon (2005) argues that the ergative pattern in Indo-​Aryan similarly arose out of a possessive construction via a mechanism of possessor raising. In particular, she argues that the source of the ergative construction in Indic and Iranian was anticausative, but not passive. Understanding the construction in terms of anticausativity has the advantage that the appearance of the ergative with intransitives can be explained readily. Furthermore, she argues that the construction functioned as an evidential, signaling that the event in question was inferred or reported rather than witnessed directly. This idea has the advantage of bringing in a semantic component to the understanding of the ergative case, potentially being able to explain modal and other effects associated with the ergative in languages such as Urdu/​Hindi and Nepali (see section 22.4.4). Bynon also makes the point, already argued for by many others, that the OIA -​ta participle was not passive, but already reflected an ergative alignment in Sanskrit, albeit in the absence of an ergative case. This argument has been made by Hock (1986) for Sanskrit, who talks about the -​ta participle as a “P(atient)-​oriented” construction. Similarly, Peterson (1998) argues that MIA showed ergative alignment, but in the absence of an ergative case (cf. Klaiman 1978; Bubenik 1996). The real innovation in NIA then, is the development of an overt ergative marker. Why and how this should have happened is not quite understood. Indeed, while Bynon’s argumentation is very persuasive and suggestive, she also does not address the questions of how or why the NIA languages innovated an ergative marker. She does suggest that the instrumental marking associated with the -​ta participle in Sanskrit is an innovation that replaced an older Persian-​type manā kartam pattern. The advantage of this account is that an immediate parallelism is drawn to the closely related Iranian branch. However, as far as we are aware, there is no evidence for such a change in the historical record.

22.3  The Diachronic Trajectory The previous section served to set the scene and discussed the two major linguistic sources that have been adduced as relevant to the development of ergativity in Indo-​ Aryan: passives and possessives. This section provides more information on the diachronic trajectory. After having thus delved into considerable detail, we return to a more high level discussion of other factors that have been implicated in understanding Indo-​ Aryan ergativity in the following sections.

22.3.1 Old Indo-​Aryan The affix -​ta (allomorph -​na) described in the previous section, inherited from Indo-​ European, is attested at all stages of Old and Middle Indo-​Aryan. It attaches directly

 

Indo-Aryan diachrony   537 to the root, creating an adjectival stem, and inflects for number and gender like any other adjectival form. As already mentioned, this affix has been reconstructed for Indo-​ European as *to/​-​no. In the oldest Vedic texts, the -​ta based form of the verb serves to describe a result-​state brought about by a preceding event when it is used predicatively in an adjectival passive construction. The -​ta forms (bold-​faced) in (8a) agree with the nominative patient while the agent remains unexpressed. In (8b), the agents and instruments are overtly expressed in the instrumental case. (8)

a. stīr-​ṇáṃ te barhíḥ strew-​perf.n.sg you.dat.sg Barhis.nom.n.sg su-​tá          indra      sóma-​ ḥ          kṛ-​tā́ press-​perf.m.sg  Indra.voc.sg  Soma-​nom.m.sg  do-​perf.m.pl dhānā́           át-​ tave te       hā́ri-​bhyāṁ barley.nom.m.pl  eat-​inf you.gen.sg  horse-​dat.sg ‘The Barhis has been strewn for thee, O Indra; the Soma has been pressed (into an extract). The barley grains have been prepared for thy two bay-​horses to eat.’ (Ṛgveda 3.35.7) b. nṛ-​bhir dhū-​táḥ su-​tó áśna-​iḥ man-​inst.pl wash-​perf.m.sg press-​perf.m.sg stone-​inst.pl áv-​ yo    vā́ra-​iḥ       páripū-​taḥ wool-​gen.sg  filter-​inst.pl  strain-​perf.m.sg ‘It (the Soma) has been washed by men, pressed with the help of stones, strained with wool-​filters.’ (Ṛgveda 8.2.2)

As shown in (9), the -​ta form agrees with the sole (nominative) argument of intransitive verbs. This results in a difference in the marking of the subject arguments of transitive and intransitive verbs. In (8) the verb does not agree with the instrumental agentive arguments. In (9), in contrast, the verb śri-​taḥ has a nominative subject soma and agrees with it in number and gender. (9)

div-​i somo adhi heaven.loc.sg soma.nom.m.sg on Soma rests (is supported) in the heaven.

śri-​taḥ rest-​perf.m.sg

(Ṛgveda 10.85.1)

This periphrastic resultative -​ta construction is the source of the ergative pattern observed in the perfective aspect in the later languages. In later stages of OIA, the construction was extended to marking the perfect aspect and it exhibited existential as well as universal perfect readings (Condoravdi and Deo 2014). By the time of Epic Sanskrit (late stage of OIA), the -​ta construction became a frequently used device for marking past perfective reference. The agent argument in these cases is most frequently overt and marked with instrumental case. The examples in (10) are from (Deo 2012, 10). Past eventive reference is indicated by the presence of past referring frame adverbials like purā ‘formerly’ and tadā ‘then’.

 

538    Miriam Butt and Ashwini Deo (10) a. purā devayug-​e ca eve dṛṣ-​ṭaṃ sarvaṃ formerly god.age-​loc.sg and ptcl see-​perf.n.sg. everything mayā   vibho I-​inst.sg  lord-​voc.sg ‘Lord, formerly, in the age of the Deva (Gods), I saw everything.’ (Mbh. 3.92.6a) b. hṛ-​tā gau-​ḥ sā tadā steal-​perf.f.sg cow-​nom.f.sg that-​nom.f.sg then t-​ ena       prapāta-​ s   tu    na   tark-​itaḥ he-​inst.3.sg  fall-​nom.m.sg  ptcl  neg  consider-​perf.m.sg ‘Then he stole that cow, but did not consider the fall (consequences).’ (Mbh. 1.93.27e)

22.3.2 Middle Indo-​Aryan The main change between Epic Sanskrit and the later MIA stage of the language concerns the erosion and simplification of the rich tense-​aspect system (Pischel 1900). Inflectional past referring forms such as the aorist, the inflectional perfect, and the imperfect disappeared from the language, leaving the -​ta construction as the only past referring device.2 This loss of the inflectional system has often been cited as a reason for the increase in the frequency and scope of the participial construction, which in turn led to the unmarking of the stative nature of the construction, and resulted in an active, ergative clause in late MIA (Hock 1986; Bubenik 1998). The examples that follow from an archaic MIA Mahāraṣṭrī text Vasudevahiṃḍī (ca. 500 ad) shows this ergative alignment. The verb agrees with the nominative subject in (11a). In (11b) the verb agrees with the nominative marked object while the agentive argument (that running one) appears in the instrumental. (11)

a. pat-​to ya seṇiyo reach-​perf.m.sg and S.nom.m.sg ta-​m        paesa-​m that-​acc.sg  place-​acc.sg ‘And King Seniya reached that place.’

rāyā king.nom.m.sg

b. t-​eṇa palāyamāṇ-​eṇa purāṇakuv-​o that-​inst.sg running-​inst.sg old.well-​nom.m.sg taṇadabbhaparichinn-​o     diṭ-​ṭho grass.covered-​nom.m.sg  notice-​perf.m.sg ‘That running one noticed an old well covered with grass.’ 

(VH.KH. 17.1)

(VH.KH. 8.6)

2  Traditional grammarians do provide instances of the inflectional perfect and the aorist during this period, but they only occur as isolated, unanalyzed forms for a few verbs like āha-​‘say-​aor’ and akāshi -​‘do-​aor’.

 

Indo-Aryan diachrony   539 Evidence that the agentive argument is indeed the subject of the clause comes from control in gerundial clauses. In (12), the controller of the gerundial clause having bought the buffalo is the argument that appears in the instrumental/​ergative case ṇ-​eṇa ‘by him’ in the main clause. The morphology on the main verb ‘kill’ is a cognate of the ta participial. (12)

tamm-​i ya sama-​e... so mahiso that-​loc.sg and time-​loc.sg that buffalo.nom.sg kiṇe-​uṇa  mār-​io buy-​ger   kill-​perf.m.sg ‘And, at that time, having bought that buffalo, he killed it.’

ṇ-​eṇa he-​inst.sg (VH:KH 14:21)

Another critical change between OIA and MIA is the restructuring of the case system—​ particularly the loss of morphological contrast between nominative and accusative as well as between the genitive and the dative cases. The syncretized paradigm is given in Table 22.2. Table 22.2 Syncretized case paradigm in MIA (Masica 1991, 231) Singular

Plural

-​u, a, aṁ -​eṁ, iṁ, he, hi

-​a, aĩ -​e(h)ĩ, ehi, ahĩ

-​hu, ahu, aho

-​hũ, ahũ

Genitive/​Dative

-​ho, aho, ha, su, ssu

-​na, hã

Locative

-​i, hi, hiṁ

-​hĩ

Nominative/​Accusative Instrumental Ablative

With respect to ergative alignment, syncretism is also observed between the nominative and instrumental forms of the first and second person plural pronouns. This is shown in Table 22.3. Table 22.3 Syncretisms in the MIA Pronominal System aspect

person

Non-​perf Perf

number singular

plural

1 1

haũ maĩ

amhaĩ/​amhẽ amhaĩ/​amhẽ

Non-​perf

2

tuhũ

tumhaĩ

Perf

2

taĩ

tumhaĩ

Non-​perf

3

so

te

Perf

3

tẽ, teṇẽ

tehĩ

 

540    Miriam Butt and Ashwini Deo Despite this syncretism, agreement is uniformly with the nominative argument—​with the nominative object in constructions based on the -​ta form and with the nominative subject elsewhere. The examples in (13) illustrate this pattern with the first person plural pronoun amhẽ. These are taken from the Paumacariu, a Jaina rendition of the epic Rāmāyana, ca. eighth century ad). (13a) contains the syncretized pronoun amhẽ which triggers agreement in the imperfective aspect while the same form fails to trigger agreement in (13b). In (13c) the second person plural syncretic form similarly fails to trigger verb agreement in the perfective. (13) a. amhẽ jāe-​va vaṇavāsa-​ho we.syncr go-​impf.1.pl forest.dwelling-​dat.sg We are going to our forest-​exile b. ki-​u amhẽ do do-​perf.m.sg we.syncr what What crime have we done?

avarāh-​o crime-​nom.m.sg

c. tumhẽ jaṃ cint-​iu you.syncr what.rel.m.sg think-​perf.m.sg taṃ              hū-​a that.correl.m.sg  happen-​perf.m.sg That, which you thought (would happen), happened.

(PC 2.23.14.3)

(PC 1.2.13.9)

(PC 3.47.9.6)

The proto-​ergative system that crystallizes in Late MIA thus has the following properties: (14) a. It exhibits an aspect-​based split in ergative alignment. Changes in the semantics of the –​ta form (which starts out as a marker of result-​stative aspect) render it a marker of perfective aspect (Condoravdi and Deo 2014), which contrasts with imperfective forms in the language. b. It exhibits partial syncretism within the instrumental marking paradigm and syncretism in nominative/​accusative case, making agreement the overt indicator of ergative alignment in many clauses.

22.3.3 New Indo-​Aryan The NIA languages display a variety of patterns and a comprehensive discussion of these can be found in Deo and Sharma (2006) and Verbeke (2013a). We identify three major patterns: in one set of languages, the ergative is one of the several new case markers innovated (e.g. in Hindi and Nepali); in another set of languages the loss of ergative case leads to the loss of ergative agreement patterns (e.g. in Bengali and Oriya), while for a third set of languages, the original pattern persists despite changes in morphosyntactic marking (e.g. Marathi, Gujarati). We review each of these patterns in the following sections.

 

Indo-Aryan diachrony   541

22.3.3.1 Renewal of Overt Ergative Marking Data from Early Hindi suggests that the original instrumental marking observed on transitive subjects for the MIA ergative system is lost in NIA. Examples here are from the opus of Kabir, a Bhakti poet from the fifteenth century ad. Note that the transitive subject arguments in the perfective clauses in (15a) and (15c) carry no overt marking but agreement is uniformly with the feminine object argument (explicit or unpronounced) chādar ‘sheet’. (15)

a. jo chādar sura-​nara-​muni which sheet.nom.f.sg gods-​men-​sages.∅erg Gods, men, and sage, all wore this sheet

oḍh-​i wrap-​perf.f.sg

b. oḍhi-​ke mail-​i kin-​i chadariyā wrap-​ger dirty-​f.sg do-​perf.f.sg sheet.nom.f.sg Having worn it, they (invariably) made it dirty (defiled it). c. dās    kabir jatan=se oḍh-​i servant Kabir.∅erg care=with wrap-​perf.f.sg (Your) servant Kabir wore it with great care d. jyon ki tyon dhar deen-​i just as it was hold-​ger give-​perf.f.sg He has given it back (to you) just as it was.

chadariyā sheet.nom.f.sg

In the later language, the subject of a transitive perfective clause uniformly receives overt ergative marking across all persons and numbers via an invariant case-​clitic ne. This is an innovation that is observed in some languages of the Western and Central subgroups of Indo-​Aryan, including Modern Standard Hindi. The modern Hindi paradigm and agreement pattern (uniformly with nominative object) is shown in Table 22.4. An example is provided in (16).

Table 22.4 Modern Hindi pronominal paradigm and agreement pattern aspect

person

Non-​perf Perf

number singular

plural

1 1

maĩ mai-​ne

ham ham-​ne

Non-​perf

2

tum

āp

Perf

2

tum-​ne

āp-​ne

Non-​perf

3

vah, yah

ve, ye

Perf

3

is-​ne, us-​ne

inho-​ne, unho-​ne

 

542    Miriam Butt and Ashwini Deo (16) a. sītā rām-​ko sītā.f.nom Rām.m=acc ‘Sita hits Ram.’ b. mai=ne Pron.1.sg=erg ‘I saw a sparrow.’

pīṭ-​tī hit-​impf.f.sg

chiḍiyā bird.f.nom

c. ram=ne chiḍiyā Ram.m=erg bird.f.nom ‘Ram saw a sparrow.’

hai be.pres.3.sg

dekh-​ī see-​perf.f.sg dekh-​ī see-​perf.f.sg

The path of change in this case is thus one of strengthening of the morphosyntactically weakened ergative alignment pattern of Older Hindi via the innovation of explicit marking. Nepali, a language that shares this part of the change with Hindi, innovates further a pattern of agreement in which the verb uniformly agrees with the subject—​regardless of whether it is marked ergative or nominative (Deo and Sharma 2006).

22.3.3.2 Loss of Ergative Marking and Agreement Pattern Bengali and Oriya, which are languages of the Eastern subgroup of Indo-​Aryan, show a contrast to the strengthening of ergative marking observed in Hindi and Nepali. Bengali and Oriya exhibit a trajectory that has led to a complete loss of ergativity. As Chatterji (1926, 742–​743) notes, Old Bengali texts show a minimal distinction between nominative and ergative (instrumental) case-​marked arguments, but the agreement pattern is clearly ergative. The only difference between the nominative and ergative case marking is the presence of nasalization on the ergative case marker.3 The verb agrees with the nominative subject in non-​perfective clauses and with the nominative object in perfective clauses. (17) a. kānhē pothī kānha.m.sg.nom book.f.sg.nom Kānha reads the book. b. kānhẽ pothī kānha.m.sg.erg book.f.sg.nom Kānha read the book.

paḍh-​ai read-​pres.m.sg paḍh-​ilī read-​perf.f.sg

Old Bengali

Old Bengali

The complete syncretism of nominative and ergative case marking over time went hand in hand with the loss of the ergative agreement pattern. Modern Bengali (and also Oriya) have lost all traces of ergative marking on the subject and the verb uniformly agrees with the subject in perfective (18a-​b) and non-​perfective (18c) clauses. The agreement marking has been innovated through auxiliary incorporation. 3 

Chatterji constructs the minimal pair in (17) to illustrate the Old Bengali pattern but also gives several examples that are directly taken from the textual corpus.

 

Indo-Aryan diachrony   543 (18)

a. āmī boi I.sg.nom book.f.sg.nom ‘I saw a book.’

dekh-​lām see-​perf.1.sg

b. rām boi rām.sg.nom book.f.sg.sg ‘Rām saw a book.’

dekh-​lo see-​perf.3.sg

c. āmī boi I.sg.nom book.f.sg.nom ‘I am looking at a book.’

dekh-​chi see-​prog.1.sg

Bengali

Bengali

Bengali

The loss of ergative marking and the corresponding agreement pattern is also attested in Indo-​Aryan texts from a different period and provenance. Jamison (2000) describes the intriguing pattern of ergative, perfective clauses in the Niya documents, a collection of texts found in the early twentieth century around Niya on the Southern Silk Route, datable to the third century ad. The perfective paradigm based on -​ta in this linguistic system innovates a set of new endings (through incorporated auxiliaries) that obligatorily agree in person and number with the clausal subject. In many cases, the subject argument of a transitive perfective clause is nominative (Jamison calls this the absolutive) as in (19a). However, in some cases, as in (19b), the subject argument exhibits the older instrumental/​ergative case. (19)

a. ahu sumiṃna I.nom.sg dream.nom.sg ‘I saw a dream.’

triṭh-​emi see-​perf.1.sg

(Niya 157, Jamison 2000, 71)

b. tade supi-​yehi aǵasit-​aṃti there.abl Supi-​inst.pl carry-​perf.3.pl ‘The Supis carried (them = mares) off from there.’ (Niya 212, Jamison 2000, 76). The data from the Niya documents that Jamison presents shows that it is a complex system with variation between nominative and ergative marking on the subject being sensitive to the animacy of both arguments (agentive and patientive) of the clause. Overt instrumental/​ergative marking is “essentially obligatory” when both the agent and the patient/​theme are human (Jamison 2000, 73) but may be optional elsewhere. In fact, there are attested examples where subjects in non-​perfective clauses receive overt marking. (20) y-​ena vṛchā who.inst.sg tree.nom.pl Whoever cuts down trees

chiṃn-​ati cut-​pres.3.sg

(Niya 482, Jamison 2000, 76)

 

544    Miriam Butt and Ashwini Deo This picture of the MIA ergative system points to an agentivity, animacy driven use of ergative case, i.e. a move toward semantic rather than purely structural determinants of ergative alignment. This aspect of the use of the ergative is elaborated upon in section 22.4.

22.3.3.3 Persistence of ergative agreement In Marathi, a language of the Southern subgroup of Indo-​Aryan, nominative and ergative case-​marked pronouns in the first and second person are entirely syncretized (an analogical extension of the syncretism in the first and second plural observed in MIA—​ see Table 22.5). Table 22.5 Marathi pronominal paradigm aspect

person

Non-​perf Perf

number singular

plural

1

mī

āmhī

1

mī

āmhī

Non-​perf

2

tū

tumhī

Perf

2

tū

tumhī

Non-​perf

3

to/​tī/​te

te

Perf

3

tyā-​ne, ti-​ne

tyā-​nī

The agreement pattern, however, persists—​the verb fails to agree with the subject of a transitive perfective clause despite lack of overt morphosyntactic evidence for an ergative pattern. In (21a), a non-​perfective clause, the verb agrees in number and gender with the subject (assuming the speaker in the utterance context is male) while in (21b), a perfective clause, the verb agrees with the object. (21) a. mī ek chimṇī I.sg.nom one sparrow.f.sg.nom ‘I am watching a sparrow.’ b. mī ek I.sg.erg one ‘I saw a sparrow.’

chimṇī sparrow.f.sg.nom

bagha-​toy see-​prog.pres.m.sg baghit-​lī see-​perf.f.sg

Gujarati (Central Indo-​Aryan) exhibits yet another pattern of agreement in ergative clauses. In understanding this pattern, it is necessary to take into consideration a diachronic change in NIA that has only recently received attention in discussions of ergativity from a diachronic perspective. The NIA languages innovated differential object

 

Indo-Aryan diachrony   545 marking (DOM) —​a pattern in which objects receive overt marking in a subset of cases, typically to mark animacy, definiteness, and/​or specificity properties of the object denotation (Bossong 1985). DOM is observed in both perfective and imperfective transitive clauses. The examples in (22) are from Urdu/​Hindi, which employs the marker ko to mark specificity on objects as seen in (22b) (Butt 1993b). The ko marker is also obligatory with animate human denoting objects (22c). (22) a. ram gari Ram.m.sg.nom car.f.sg.nom ‘Ram will buy a/​the car.’

xarid-​e-​ga buy-​3.sg-​fut-​m.sg

b. ram gari=ko xarid-​e-​g-​a Ram.m.sg.nom car.f.sg=acc buy-​3.sg-​fut-​m.sg ‘Ram will buy the car (a specific car).’ c. ram sita=ko Ram.m.sg.nom sita.f.sg=acc ‘Ram will see Sita.’

dekh-​e-​g-​a see-​3.sg-​fut-​m.sg

Urdu/​Hindi

Urdu/​Hindi

Urdu/​Hindi

In perfective transitive clauses, the presence of DOM together with ergative case marking has an effect on the ergative agreement pattern. In most NIA languages, in clauses with ergative subjects and accusative marked objects, the verb agrees with neither argument exhibiting default (masculine or neuter singular) agreement instead. In the perfective sentence in (23), we see that the verb carries default masculine agreement and thus fails to agree with the overtly marked accusative object. (23)

gita=ne gita.f.sg=erg ‘Gita saw Sita.’

sita=ko sita.f.sg=acc

dekh-​a see-​perf.m.sg

Urdu/​Hindi

Gujarati provides an example of an interesting variation of this pattern. Gujarati also has animacy/​specificity determined DOM marking. However, unlike Urdu/​Hindi, it exhibits a persistence of ergative agreement even in clauses with overtly marked accusative objects. The examples in (24) show that the verb agrees with the feminine, accusative marked object argument Sita. (24) a. rām-​e rām.m.sg-​erg ‘Ram saw Sita.

sitā=ne sita.f.sg=acc

b. sītā-​e rām=ne sita.f.sg-​erg ram.m.sg=acc ‘Sita saw Ram.’

jo-​i see-​perf.f.sg jo-​yo see-​perf.m.sg

Gujarati

Gujarati

 

546    Miriam Butt and Ashwini Deo

22.3.4 Summary The ergative pattern in Indo-​Aryan languages thus emerges as a consequence of a result-​ stative construction becoming the default construction for describing past completed eventualities. Over time, there is reduction in overt marking of ergative subjects, resulting in either renewal of ergative clitics (as in Hindi) or the loss of ergativity in its entirety (as in Bengali). The Niya documents, representing a distinct late MIA system, exhibit yet another system in which subject agreement is innovated in the perfective paradigm with semantically conditioned variability in the ergative marking on transitive subjects. In other NIA languages, ergative agreement is retained despite syncretism between the nominative and ergative cases in parts of the paradigm (Marathi) and the innovation of differential object marking (Gujarati). The focus in this section has been on the structural aspects of the development of ergative patterns in Indo-​Aryan. In the next section, we discuss further factors that have been adduced with respect to ergative case marking in the Indo-​Aryan context, namely markedness and clausal and lexical semantic factors.

22.4  Further Factors The picture that emerges from the overall literature on the history of the ergative in Indo-​Aryan is a complex one. The previous sections have focused mainly on structural factors such as changes of alignment and agreement. However, even while attempting to adhere to a strictly structural perspective, discussions with respect to Differential Case Marking (DCM) have crept in. The DCM involved semantic factors: animacy in the case of the Differential Subject Marking (DSM) found with the Niya documents and the typical DOM involving specificity that was needed to understand the NIA Gujarati patterns. In this section, we elaborate on such further perspectives and suggest that they hold a key part of the answer to understanding the diachrony of ergativity in Indo-​Aryan.

22.4.1 Markedness Malchukov and de Swart (2009) and de Hoop (2009) provide comprehensive surveys of the state of the art with respect to DCM of arguments and notions of markedness. One of the main issues this work addresses is the distribution of case marking across core arguments of verbs. The underlying question is: why do languages exhibit nominative-​ accusative systems vs. ergative systems vs. mixed systems? The literature on DCM and markedness suggests that the emergence of the various observed patterns is rooted in the maximization of distinctions between the core arguments in a clause.

 

Indo-Aryan diachrony   547 DSM in particular is taken to imply the presence of two different considerations, which may sometimes conflict with one another (see Malchukov (Chapter 11, this volume) for a related discussion). 1. Distinguishing strategy: in order to distinguish subjects from objects, mark non-​ prototypical subjects (i.e., subjects which could be mistaken for objects). 2. Indexing strategy: identify proto-​typical subjects (agents) and mark this particular semantic role. Ergative languages are taken to be one typical result of these two basic strategies. The split-​ergative pattern in Indo-​Aryan can be seen as a result of the indexing strategy, whereby the proto-​typical subject is marked, though only in the presence of perfective morphology. As far as we are aware, the question of why split-​ergativity tends to go hand in hand with perfect/​perfective or past morphology has not yet been fully resolved. We suspect, along with Trask (1979), that the solution to the puzzle lies in the fact that modern perfect/​perfective/​past forms in ergative clauses tend to come from old participles and that these participles expressed stative situations. The reinterpretation of the stative participles as eventive verbal predications thus may also necessitate a special marking of the “unexpected” or “marked” agents (since stative situations are not expected to have agents). As such, the employment of ergative markers in these situations may also alternatively be due to the distinguishing strategy whereby non-​prototypical subjects, i.e., agentive subjects of originally stative situations, are marked.

22.4.2 Complex Case Marking Patterns Invoking markedness as a conditioning factor for ergativity has the potential of contributing significantly to the overall understanding of ergative marking and ergative alignment. However, Indo-​Aryan exhibits patterns that are unexpected from the markedness perspective. Consider the examples in (23) and (24) presented in the previous section. An ergative marks the agent/​subject, but additionally the object also carries an overt case marker. For purposes of indexing or distinction of arguments, having two overt case markers is a situation of overkill. Furthermore, one finds examples in NIA in which the same case form marks both subject and object. From the perspective of indexing and distinction of arguments, this is again unexpected as disambiguation of arguments is not being achieved when both arguments carry the same case forms. Consider the examples from Haryani and Kherwada Wagdi in (25) and (26), which involve polysemy between ergative and dative/​ accusative.4

4 

In Kherwarda Wagdi, ne and ṇe are allophonic variants of one another (Phillips 2013).

 

548    Miriam Butt and Ashwini Deo (25) man=ne sahab=ne Pron.1.sg=acc/​dat Sahib.m.sg=erg ‘The Sahib hit me.’    (Shirani 1987)

mar-​a hit-​perf.m.sg

(26) va-​ṇe ve-​ne dekh-​yu Pron.3.sg=erg Pron.3.sg=acc see-​perf-​n.sg ‘He/​she saw him/​her.’  (Phillips 2013)

Haryani

Kherwada Wagdi

Both examples (25) and (26) can only be disambiguated by context. The question thus arises —​why mark both of the core participants and yet not distinguish? Examples like this are unexpected not only from the markedness perspective, but also from the perspective of the passive-​to-​ergative and possessive-​to-​ergative alignment theories. In particular, while ergative/​instrumental or ergative/​genitive polysemy is expected, polysemy between ergative and dative/​accusative is not. The next two sections offer further perspectives that allow for an understanding of examples of this type.

22.4.3 Differential Case Marking One crucial piece of the overall picture involves semantically motivated case alternations. Section 22.3 included examples of semantically motivated DCM, namely animacy of both agents and patients conditioning the appearance of the ergative marker in the MIA Niya documents and specificity conditioning DOM in the NIA languages Hindi/​Urdu and Gujarati. Indeed, examples of DCM on subjects, objects, and obliques abound in NIA. Butt and Ahmed (2011) observe that DOM is an old part of the language and argue that the strategy of using case marking to express semantic/​pragmatic information cannot be ignored in developing an understanding of how new case markers, including the ergative, developed in the NIA languages. A typical example of DOM involving partitivity in OIA is provided in (27). Other examples include accusative vs. instrumental marking to signal degrees of affectedness on causees or dative vs. accusative to signal abstract vs. concrete movement (Butt and Ahmed 2011). (27) a. pibā soma-​m drink.imp soma-​acc ‘Drink soma. (all of the quantity)’ b. pibā soma-​sya drink.imp soma-​gen ‘Drink (of) soma.’

Vedic (Ṛgveda VIII.36.1, from Jamison 1976)

Vedic (Ṛgveda VIII.37.1, from Jamison 1976)

 

Indo-Aryan diachrony   549 As far as we know, OIA did not contain DSM and the nominative was the only case used to mark subjects. This situation changed in MIA, where DSM is attested. One example is constituted by the dialect recorded in the Niya documents. Another example involves a genitive/​instrumental alternation on agents documented by Andersen (1986) for Aśokan inscriptions. Andersen finds that the genitive is rarer and can only apply when the agent is animate. No such restriction applies to the instrumental. The instrumental thus appears to function as the default case, while the genitive is used to mark animacy of the agent. Bynon (2005, 17) points to work by Jamison (1979a,b) to show that Vedic, the oldest attested state of Indo-​Aryan, already allowed for variable agent marking: (1) compounded in together with the verb; (2)  instrumental; (3)  genitive. The instrumental realization is by far the most frequent, but as part of her overall possessive-​to-​ergative view, Bynon argues that the genitive is originally the default marker (contra the received wisdom by Cardona (1970) and Jamison (1979a,b)) and draws a connection to evidentiality. The Sanskrit grammarian Pāṇini considered this to be part of the verbal system, with the aorist and imperfect encoding the reporting of events the speaker can attest to, while the perfect encoded events not witnessed by the speaker. Bynon (2005) thus sees the perfect as a modally marked form and the ergative ultimately as encoding a modally marked evidential. This understanding of the ergative is interesting in the light of modern patterns by which DSM is used to signal differences in modality. (28) provides an example from Bengali, (29) from Urdu/​Hindi (see Butt and King (2004) for detailed discussion). (28) a. ami toma=ke cai I.nom you=acc wants ‘I want you.’ (Klaiman 1980, 279)

Bengali

b. amar toma=ke cai I.gen you=acc wants ‘I need you.’ (Klaiman 1980, 279)

Bengali

(29) a. nadya=ne dakhane Nadya.f.sg = erg post office.m.sg.obl ‘Nadya wants to go to the zoo.’

ja-​na hε go-​inf.m.sg be.pres.3.sg          Urdu

     b. nadya=ko dakhane Nadya.f.sg=dat post office.m.sg.obl ‘Nadya has/​wants to go to the zoo.’

ja-​na go-​inf.m.sg

hε be.pres.3.sg        Urdu

In Bengali there is no ergative. The DSM is between an unmarked form (glossed as nominative) and the genitive. In Urdu the contrast is between ergative and dative. Also note that the ergative here appears without being licensed by perfective morphology. This use of the ergative appears to be an innovation (Bashir 1999; Montaut 2003, 2006, 2009), but the use of case morphology in conjunction with non-​finites to express

 

550    Miriam Butt and Ashwini Deo modality is an old part of Indo-​Aryan (Bynon 2005) (and Indo-​European as a whole), see (30). (30) samprati gan-​tavyā puri now go-​ger city.nom.f.sg ‘now I want to go to the city of Benares’

vārāṇasī Benares.nom.f.sg

mayā I.inst

Sanskrit (from Speijer 1886: section 41)

The diachronic and synchronic evidence in Indo-​Aryan thus points to a long tradition of DCM. Every instance of DCM that has been documented expresses a semantic distinction. The semantic parameters range from fairly well known ones like partitivity, specificity, and animacy to the expression of modality. Since the ergative case is often implicated in DCM, this observation about Indo-​Aryan needs to be taken into account for an understanding of ergativity.

22.4.4 Lexical Semantics, Borrowing, and Competition Butt (2001, 2006a) and Butt and Ahmed (2011) set out to understand the development of the NIA case system primarily from a lexical semantic perspective of case. The new NIA case markers have primarily been drawn from a set of originally spatial terms, e.g., ‘with, at, from, in, on’ (Beames 1872–79; Kellogg 1893; Chatterji 1926; Tessitori 1913, 1914; Montaut 2003, 2006; Hewson and Bubenik 2006; 2009, Reinöhl 2015). In the hundreds of NIA languages and dialects, the same handful of case markers crops up again and again. Particularly common for the marking of core arguments are k-​forms (e.g. ko, ke, khe, ku), n-​forms (e.g. ne, nai, ni, nũ), and l-​forms (e.g. lai, le, la). The forms distribute across the case paradigms and can take on different roles in different languages. In some languages, case polysemy can be observed. In Marathi the ne form furnishes both the ergative and the instrumental. In Haryani (cf. (25)), Kherwada Wagdi (cf. (26)) and Rajasthani (Allen 1960) ne expresses both the ergative and the dative/​accusative. The overlap in forms is one of polysemy and not syncretism since distinctions have not been lost over time. Rather, the originally spatial term changes and expands its lexical semantic space over time. Understanding exactly how this happens remains a general problem to be solved within historical linguistics.5 Butt and Ahmed take a Lexical-​Functional Grammar (LFG) perspective under which they treat the case clitics as instantiating individual lexical items (see also the discussion in Butt and King (2003, 2004) and the computational treatment of Urdu, Butt and King (2007)). These lexical entries are associated with semantic information and, as with any lexical item, both

5  Cf. McGregor (Chapter 19, this volume) who suggests that identifying direct lexical sources for ergative markers is difficult cross-​linguistically. If correct, this has interesting theoretical implications as it suggests that ergatives are derived from other case markers

 

Indo-Aryan diachrony   551 semantic and grammatical information may be polysemous, ambiguous, or underspecified and is subject to change over time. Butt and Ahmed see semantically based DCM as a major motivating factor in drawing new case markers into a grammatical system. Hewson and Bubenik (2006) observe that languages which develop an article system tend not to redevelop case markers (e.g. English, Romance). Case marking clearly fulfills a function beyond indexing and marking and where this function is not redistributed onto other grammatical means, new case markers are drawn into the system. In Indo-​Aryan this involves semantically based DCM. Butt and Ahmed concentrate on Urdu/​Hindi and on the following case markers: dative/​accusative ko, ergative ne and instrumental se. ko appeared in the language some two hundred years before ne did. Early on, ko was involved in DCM involving the (non)-​attainment of goals. ne appears to have entered the language through borrowing via language contact (Beames 1872–​79). The agent was generally not marked explicitly in Old Urdu/​Hindi, but it was marked indirectly via agreement patterns (see discussion in section 22.3). In addition, the language engaged in systematic alternations where agency/​control contrasted with involuntariness (cf. Montaut 2003, 2006, 2009). When the language came into contact with another language which explicitly marked agents, it was natural to borrow that marker, thus bringing ne into the language. However, only the agency reading of ne was borrowed since the other semantic domains were already taken up by existing case markers. A possible ablative meaning of ne, found in sister languages, was expressed by se and the dative/​accusative meaning was already expressed by ko. Butt and Ahmed suggest that ne is found with a more restricted usage in Urdu/​ Hindi than in languages like Rajasthani, Haryani, or Kherwada Wagdi because the other potentially available meanings were blocked by existing established case markers. Thus, the range of semantic functions a given case marker is taken to be determined jointly by the meaning of the original spatial term, the constraints imposed by the existing case system and the existing case alternations. If case markers already exist in the language that block part of the available meaning space, then the full range of possible semantic functions of new, incoming case form are not expressed. Butt and Ahmed paint a complex picture of the development of case marking systems. The picture involves factoring in lexical and clausal semantic information expressed by case marking, in particular, by DCM. It also involves an understanding of blocking and competition among case markers in a language. While Butt and Ahmed place their argumentative emphasis on semantic factors, the structural agreement and alignment factors discussed in sections 22.2 and 22.3 also play a central role, as does the fact that ergatives tend to be associated with the reanalysis of stative participial predications.

22.5  Conclusion and Discussion In conclusion, much more work on the effects of historical change on case systems needs to be done. The Indo-​Aryan situation is a particularly promising one for achieving a

 

552    Miriam Butt and Ashwini Deo deeper understanding of the diachrony of ergativity. It is characterized by the longest historical record available for any one language family. Synchronically, the large modern languages tend to be spoken by millions, the smaller ones by thousands rather than by hundreds. The modern languages and dialects show interesting variation that can be exploited for comparative analyses. Original fieldwork by Deo on the central Indian language Bhili and the dialectal variation found in Pawri, Dehawali, Ahirani, and Kokana has shown that these retain MIA patterns with respect to tense/​aspect, there is thus also the possibility of synchronically investigating patterns attested in the historical record. Beyond doing more original historical and field research, we see the most promising future research on the topic as being of the type that acknowledges the complex interplay between structural and semantic concerns and seeks to investigate this further.

Acknowledgements Ashwini Deo gratefully acknowledges the support of the National Science Foundation (NSF BCS-1255547/BCS-1660959). Miriam Butt gratefully acknowledges Opus Magnum funding from the Volkswagenstiftung.

Abbreviations ‘-​’, stands for a morpheme boundary; ‘=’, for a clitic boundary; abl, ablative; acc, accusative; aor, aorist; dat, dative; erg, ergative; f, feminine; fut, future; gen, genitive; ger, gerund; Imp, imperative; IMPF, imperfective; inf, infinitive; inst, instrumental; loc, locative; m, masculine; n, neutrum; neg, negation; nom, nominative; pass, passive; perf, perfective; plural, plural; ptcl, discourse particle; ptcpl, participle; prog, progressive; pv, verb particle; sg, singular; syncr, syncretic; voc, vocative.

 

Chapter 23

E rgativit y an d l a ng uag e change in Au st rone sia n l anguag e s Ritsuko Kikusawa

23.1 Introduction The focus of this chapter is change that takes place in the case-​alignment patterns found in pronominal systems in Austronesian languages. I will discuss how ergativity related systems are described and relevant phenomena can be analyzed in the context of syntactic reconstruction and language change. Although the notion “ergativity” has been applied to capture patterns observed in various linguistic aspects (Dixon 1994), the focus of this chapter will be on changes in case-​alignment patterns of nominals, or the coding of case, that is relevant to ergativity in Austronesian languages. In particular, how pronominal arguments are aligned to express S, A, and P of sentences will be examined.1 In Figure 23.1, proto-​typical ergative and accusative systems are presented. An ergative system is one where S and P are formally marked alike. The case form by which they are marked is commonly labeled as absolutive or nominative. In this system, A is marked differently and is commonly labeled as ergative, but also as instrument, genitive, etc., depending on other functions that the form carries in the language. On the other hand, the accusative system is where S and A are formally marked alike and the case is labeled as nominative. It is P that is marked differently in this system and the case form is labeled as accusative. 1 

The following abbreviations are used to indicate syntactic roles of the arguments of transitive and intransitive sentences, following Comrie (1978) and Dixon (2000): A=Agent of transitive verbs P= Patient of transitive verbs S=Subject (actor/​undergoer) of intransitive verbs E=Extended argument of intransitive verbs

 

554   Ritsuko Kikusawa

Intransitive

Transitive

S

S

A

P

ERG

NOM

Ergative

A NOM

P ACC

Accusative

Figure 23.1  Ergative and accusative alignments

Case-​marking strategies in Austronesian languages are typically found in paradigmatic pronominal sets, and/​or sets of typically short morphemes sometimes referred to as case-​marking determiners, specifiers, or adpositions that introduce lexical NPs. Another strategy is simply word order. The languages show a variety of alignment patterns, including ergative, accusative, and various other types that could be referred to as some kind of “split.”2 There are several geographical areas where ergative languages are commonly found in the Austronesian family; namely Taiwan, island Southeast Asia, including the Philippines and northern Sulawesi, Madagascar, and Polynesia. In the geographical areas between, a wide range of variation in alignment patterns is found, some of which will be discussed in this chapter. The geographical positions of the languages referred to in this section are shown in Figure 23.2, and their genetic relationships are presented in Figure 23.3. Two different case alignment patterns, namely ergative and accusative, involving pronominal systems are presented as examples in what follows. The first example provides sentences from Betsimisaraka Malagasy spoken in Madagascar (a western Malayo-​ Polynesian language); (1a) is an intransitive sentence, and (1b) is a transitive sentence. Here, S and P are expressed by a nominative pronoun, while A is expressed by a genitive pronoun, thus showing an ergative pattern. (1) Betsimisaraka Malagasy sentence examples (1)    a. Intransitive Mandry [izy]S. be.asleep 3sg.nom ‘S/​he is asleep.’

2 

(Kikusawa 2008: 49)

No language is ever fully ergative or accusative, for various kinds of lexical and/​or grammatical splits also exist within cases that are characterized as one type or another, and also alignment systems are not confined to the relations between what are referred to as S, A, and P. In this sense, it is not accurate to include “split system,” as though the term carries equal status to ergative and accusative (Bickel and Nichols 2009).

Figure 23.2  Geographical positions of the languages referred to in this chapter

 

Ergativity and language change in Austronesian languages    555

 

556   Ritsuko Kikusawa

Austronesian Eastern Formosan Malayo-Polynesian (Extra-Formosan) (Western Malayo-Polynesian)* Nuclear Malayo-Polynesian Ibaloy Tagalog Sunda-Sulawesi (Central-Eastern) Mamuju Pendau Indonesian Balinese Central Eastern Malayo-Polynesian Malayo-Polynesian Tetun South-Halmahera/ Oceanic West New Guinea Taba Admiralty Western Central-Eastern Oceanic Oceanic Southeast Solomonic

Eastern (Remote) Oceanic Central Pacific Polynesian

Fijian Rotuman Tongan West Futuna-Aniwan

East Futunan, East Uvean

Samoan Eastern Polynesian

Māori

Figure  23.3  Languages referred to in this chapter (in italic and boxed) and their genetic relationships

 

Ergativity and language change in Austronesian languages    557 b. Transitive Tia[=ko]A [izy]P like=1sg.gen 3sg.nom ‘I like him/​her/​it.’

(Kikusawa 2008: 49)

In (1a), it should be noted that the sole argument is labeled as nominative, rather than absolutive, since the term “nominative” is defined here as the case form that marks the sole argument of a basic intransitive sentence, regardless of whether the language shows an ergative or an accusative pattern. This definition is based solely upon distributional features within a language. This is unlike the traditional labels for case forms in European languages in that it does not imply any historical or etymological origin. In (1b), the form expressing A is labeled as genitive rather than ergative, since it is common in western Austronesian languages that the forms marking the agent of transitive sentences are identical to those marking the possessor of a noun, that are typically referred to as “genitive.” Austronesian languages that show an ergative pattern often have another dyadic sentence structure in addition to the transitive one, such as the sentence shown in (2). The general meaning of the sentence (‘I am fond of him/​her/​it’) is the same as that of (1b) (‘I like him/​her/​it’), however, the case alignment of the sentence is different. The agent ‘I’ is expressed by nominative instead of ergative, and the patient ‘him/​her/​it’ is expressed by oblique instead of nominative. (2)

Betsimisaraka Malagasy sentence Extended intransitive Tia [ananjy]E [zaho] like 3sg.obl 1sg.nom ‘I am fond of him/​her/​it.’

(Kikusawa 2008: 49)

A non-​transitive dyadic sentence such as (2)  is labeled in this chapter as “extended intransitive” and the non-​nominative argument in this structure is labeled as “E,” following Dixon 1998. This structure has been analyzed as forming part of a voice alternation system and has been referred to as “anti-​passive” in some literature (Payne 1982; de Guzman 1988; Gerdts 1988b; Manning 1996; Aldridge 2012b; and others). However, to avoid dealing with the fundamental question of whether the structure should be analyzed as such, and to keep structural comparisons free from the different implications that are associated with this term, it will not be used in this chapter, unless it is part of the terminology used in the original source. The typical ergative case alignment pattern of pronominal systems is summarized in Table 23.1.

 

558   Ritsuko Kikusawa Table 23.1 Typical case alignment patterns in Austronesian ergative pronominal systems actor Intransitive

undergoer S (nom)

Extended intransitive

S (nom)

E (obl/​loc)

Transitive

A (gen)

P (nom)

Accusative systems are commonly found in Oceanic languages. Example (3) shows one such system from Rotuman (spoken on Rotuman Island in the north of the Republic of Fiji). Here, unlike the Malagasy example, there is a single set of pronouns and the case is marked by the relative position to the verb. The form for 3pl is iris, which expresses S or A when preceding the verb (3a–​c), and P when it follows the verb and is not preceded by an adposition (3c). A common view of case-​marking is to include marking by word order, and following this view, Rotuman can be said to show an accusative pattern. (3) Rotuman sentence examples a. Intransitive [Iris]S ‘ā 3pl eat ‘They ate. /​They were eaten.’ b. Intransitive [ŋou]S pæ 1sg sit ‘I sat down.’

sio down

c. Transitive [ŋou]A fesiʔen 1sg hate ‘I hate them.’

[iris]P 3pl

(Churchward 1940: 123)

(Vamarasi 2002: 38)

(Churchward 1940: 22)

Just as in languages showing an ergative pattern, Rotuman also has a dyadic intransitive sentence structure, an example of which is presented in (4). Here, the form for 3pl is irisa, which is preceded by a locative preposition (se) and expresses E. There are two sets of pronouns in Rotuman, namely, short forms (or “incomplete phase”) and long forms (or “complete phase”). Short-​form pronouns occur in nominative and accusative case positions, while long-​form pronouns occur when preceded by a preposition (Churchward 1940). (4) Rotuman intransitive dyadic sentence example c. Intransitive + PP [ŋou]S fesiaʔ [se irisa]E 1sg hate loc 3pl ‘I hate (toward) them.’

(Churchward 1940: 22)

 

Ergativity and language change in Austronesian languages    559 Argument structure of Rotuman sentence structures can be summarized as in Table 23.2. Figure 23.4 schematically shows ergative and accusative alignments including extended intransitive sentence structures. Table 23.2 Typical case alignment patterns in Austronesian accusative pronominal systems actor Intransitive

undergoer S (nom)

Extended intransitive

S (nom)

E (loc)

Transitive

A (nom)

P (acc)

It should be noted that these are just two of a wide variety of ergative and accusative patterns found in Austronesian pronominal systems.3 Clarifying the ways in which so many typologically different systems emerged from a single proto-​system makes a fascinating scientific endeavor. The rest of this chapter is organized as follows. In section 23.2, methodology, terminology, and related issues are discussed. In section 23.3, three different sets of changes that resulted in a change in the case-​alignment of Austronesian languages will be presented. For each set of changes, those that affected the interpretation of the case system are discussed, and how the results may have influenced other grammatical features are considered. Section 23.4 presents concluding remarks.

Monadic

Intransitive

S

Extended Intransitive

S

S E

E

S

Diadic Transitive A

P

ERG

NOM

Ergative

LOC/ OBL

A

P

NOM

ACC

LOC

Accusative

Figure 23.4  Ergative and accusative alignments including the extended argument 3 

For specific examples of various alignment systems and some discussion about their differences from a historical perspective, see Kikusawa (2015b).

 

560   Ritsuko Kikusawa

23.2 Methodology Syntactic comparison and reconstruction was long considered to be a hazardous if not impossible endeavor. However, increasing interest is seen today in finding ways to apply the principles of the Comparative Method, and promising results are being published (Harris 2008, Gildea 2000, and others; see Barðdal 2015 for an overview). In such attempts, the following appears to be the consensus among researchers. First, the comparanda between languages need to be based on surface structures, where changes take place and are directly observed. The comparanda in this chapter are abstracted sentence patterns describing the argument structures of selected languages. Second, syntactic changes take place as a result of a cascade of changes (Roberts 2007; de Smet 2015). As Fried (2008: 47) states, “the gradualness of change [of grammatical phenomena] consists in discrete partial changes that involve specific features or aspects of a larger pattern before they affect the full pattern completely.” Thus, there is no holistic change such as a language switching from accusative to ergative, or ergative to accusative overnight. Approaches in this chapter follow this view. Since changes that take place in each form and category are discrete from other changes, in conducting historical examination, linguistic features forming part of the examined system have to be decomposed and analyzed separately. In dealing with case-​alignment systems, changes in personal pronouns are examined separately from changes in the case-​marking forms on lexical NPs. Moreover, word order change is perceived as a separate change, with the recognition that its results may interact with those of case-​marking systems. Verb morphology is by definition not part of a case-​alignment system, and comments on verb morphology will be restricted to those cases where an identified change in the coding of case affects or is affected by the verb morphology. On the other hand, person and number agreement markers, which in many languages compose part of the verbal morphology, often develop from pronominal forms that marked case and need to be examined in this context. Such an approach may appear to obscure the typological characteristics of each language and their ancestral languages, however, any system a language exhibits today is the end result of a collection of changes that took place in various grammatical features over time, changes which exist in layers. Having the above in mind and understanding that syntactic reconstruction is the reconstruction of patterns and not the reconstruction of actual sentences used in a proto-​language (cf. Harris 2008), the following two aspects are considered to be particularly important in this work: (i) identifying cognacy and sentence pattern correspondences; (ii) identifying the direction of change (Barðdal 2015; Luján, Barðdal et al. to appear). In this chapter, cognacy among sentence structures is identified based on morpho-​functional correspondence; more specifically, the occurrence of the ergative/​ genitive pronouns associated with the A or actor of syntactically transitive sentences. It was mentioned earlier that in Austronesian ergative languages, it is often found that A is expressed by a genitive form (cf. (1)). Even when not, it is often found that a trace of what was formerly a genitive pronominal set is identifiable, as will be seen in section 23.3,

 

Ergativity and language change in Austronesian languages    561 and these are invaluable in identifying cognate structures. This is a method inductively developed, rather than theory oriented, based on the fact that such pronouns are found to be more stable than other relevant components. It is not clear what the reason is for this stability; it may be because of the salient nature of the referents of the phrase marked as ergative, or simply because this was the only set of pronouns that carried clitic status in their shared proto-​language. Whatever the reason, tracing the occurrence of these pronominal forms in various sentence structures enables us to trace sentence structure changes and identify the directionality of change. Analyses and glossing follow the source description of each language, unless otherwise specified. It should be noted that case is usually determined according to typological criteria. The form or marking on S is by definition nominative. If A receives a different marking from that on S, it will be referred to as ergative, while if it is P that receives a different marking, it will be referred to as accusative. As has been mentioned earlier, the marking on A and the form of the associated pronouns is often shared in Austronesian languages with that of the possessor of a noun in noun phrases, and is consequently labeled as genitive (rather than ergative) in such cases. Typologically defined terms do not necessarily reflect etymological relationships, and the functional change of each case needs to be traced, based on formal correspondences. Finally, a few words are made here regarding macro-​and micro-​comparison. Languages compared in this chapter are all daughter languages of Proto-​Malayo-​Polynesian, the parent of the subgroup of Austronesian languages comprising all those that are geographically located outside of mainland Taiwan. Those that are directly compared with each other are often distantly related. The advantages of conducting such a macro-​comparison is that it makes it easier to identify direct inheritance. Historical examination of closely related languages (micro-​comparison) is often complicated by having to deal with a mixture of direct and indirect inheritance (that is borrowing from closely related languages), as well as sporadic local innovations, where earlier features are obscured by layers of change that took place subsequent to the split of the languages. It should be remembered that the comparison and reconstruction of lexical items and sound systems, which is today conducted applying “bottom-​up” methodology, was initially done by macro-​comparison, which set the basis for detailed bottom-​up micro-​comparison (cf. Blust 1990: 137-​138). Needless to say, follow-​up modification of any proposed hypothesis is always necessary, based on new data and the results of micro-​comparison. I consider that macro-​comparison is a reasonable way to start the investigation.

23.3  Ergativity Related Change in Austronesian Languages In this section, three sets of changes that resulted in a shift from an ergative to a different alignment system are described. In 23.3.1, a case of change from an ergative to a

 

562   Ritsuko Kikusawa bi-​transitive system via an inverse one will be presented. This was probably triggered by a word order change, which eventually led to a system where word order alone marks grammatical relations. Among the sequence of relevant changes, it will be argued that it was the loss of the contrast between two distinct dyadic sentence structures that resulted in a change in the case-​alignment of the language. In 23.3.2, a case of change from an ergative to an accusative system as a result of a merger of two pronominal sets will be provided. It is argued that this change resulted from the generalization of earlier genitive pronouns marking A to mark S, and the genitive pronominal forms subsequently began to be replaced with forms from other pronominal sets. In 23.3.3, an ergative to accusative change as a result of change in the distribution of morphological forms will be summarized. Languages in which the same set of case markers on lexical NPs can be interpreted as either ergative or accusative will be examined. For each set of changes discussed, the mechanisms by which the changes took place and their preconditions are described. Since the methodology for morphosyntactic comparison and reconstruction is not yet well established, I will explain how the changes described relate to the general principles of comparative (historical) linguistics. The combination of the results of the discussion in 23.3.1 through 23.3.3 gives a general overview of some of the case-​alignment changes in the Austronesian language family.

23.3.1 Change from a Morphologically Marked Ergative System to a Word Order Marked Bi-​transitive System Basic argument structures compared here are those in Ibaloy spoken in Luzon, the Philippines, based on Ruffolo (2004), and Pendau spoken in Sulawesi, based on Quick (2007). This is a comparison of an ergative pattern pronominal system and what has been described as an inverse system respectively. It will be shown that a historical comparison and reconstruction reveals that the shared proto-​system was an ergative pattern and that the inverse system developed as a result of word order change. It will be also shown that the Balinese system shares different stages of the same line of development. Ibaloy example sentences are presented in (5). Sentence (5a) is an intransitive sentence with the S is expressed with a nominative clitic pronoun. Sentence (5b) is an extended intransitive sentence where the S is expressed by the absence of marking implying a third person singular nominative (as is the case in many Philippine languages) and the E is expressed with an oblique PP. Sentence (5c) is a transitive sentence, where the A is expressed by a genitive pronoun and the patient by a nominative pronoun.4 The system as a whole thus shows an ergative pattern contrasting S and P vs. A, following the pattern described in Table 23.1. 4 

The sentences are cited from Ruffolo (2004), where all example sentences are extracts from texts and therefore the sentences do not show matching pronouns to show the case marking contrast.

 

Ergativity and language change in Austronesian languages    563 (5)

Ibaloy sentence examples with pronominal arguments a. imakadkami ali ʔakad[=kami]S ʔali go=1+/​nom toward ‘We went back home.’ b. on’aseba ʔon-​ʔasǝwa[Ø]S ActV/​ipf-​marry=3sg/​nom ‘She will get married to him.’

son [so=n obl=gen/​pers

(Ruffolo 2004: 175) si’kato siʔgato]E 3/​ind (Ruffolo 2004: 150)

c. tinodonganchaka toloŋ-​an[=da]A[=ka]P help-​lv=3pl/​gen=2sg/​nom ‘They helped you.’

(Ruffolo 2004: 418)

Pendau sentence examples are presented in (6). Here, there are two sets of pronouns, which are labeled as “absolute” and “genitive.” Absolute pronouns occur marking S in intransitive sentences as in (6a), the actor (preceding the verb) and the undergoer (following the verb) as in (6b), and the undergoer (preceding the verb) as in (6c, d). The genitive pronoun expressing the actor occurs following the verb in (6c, d). As can be seen in the examples, there are two dyadic constructions in Pendau, one in which the actor precedes the verb and the undergoer follows the verb (6b), and the other in which the order is reversed and the undergoer precedes the verb and the actor follows the verb (6c, d). (6) Pendau sentence examples with pronominal arguments5 a. Sampanyo [jimo]S neosa sampanyo jimo N-​pe-​osa after.that 3pl.abs r-​dy-​rest ‘After that they rested.’ b. [Io]Aav neng-​ebiling 3sg.abs av/​r-​leave ‘He left me.’

5 

[’a’u]Pav 1sg.abs

c. [’A’u]Piv ni-​ebiling 1sg.abs iv/​r-​leave ‘He left me.’

[-​onyo]Aiv -​3sg.gen

d. [Io]Piv ni-​ebiling 3sg.abs iv/​r-​leave ‘I left him.’

[-​o’u]Aiv -​1sg.gen

(Quick 2007: 235)

(Quick 1997: 467)

(Quick 1997: 467)

(Quick 1997: 467)

In Quick (1997), the absolute and genitive cases are labeled as proximate and obviate respectively. The glossing here follows that in Quick (2007).

 

564   Ritsuko Kikusawa Quick (2007: in particular 360–​387) analyzes the language as showing an inverse system based on an “emic analysis.” Both of the dyadic structures in this language are analyzed as syntactically transitive, and the structure where the actor precedes the verb and the undergoer follows the verb such as (6b) is referred to as “active voice,” while the one where the undergoer precedes the verb and the actor follows the verb such as (6c, d) is referred to as “inverse voice.” The arguments of each structure are indicated as Aav ‘A of active voice’, Pav ‘P of active voice’, Aiv ‘A of inverse voice’ and Piv ‘P of inverse voice’. The case marking system of Pendau is summarized in Table 23.3. Table 23.3 Case alignment pattern in the Pendau inverse system actor Intransitive

undergoer S (abs)

transitive (actor voice)

A (abs)

P (abs)

Transitive (inverse voice)

P (abs)

A (gen)

In what follows, to order to compare and reconstruct the sentence structures of the two languages described above, abstracted sentence structures are first presented, then cognate structures are identified and the changes that took place in their pronominal systems will be discussed. The changes presented here are considered to be generally shared between pronominal and other NP marking systems, but in cases where they are not, the differences will be indicated. Abstracted argument structures of Ibaloy are presented in (7). Personal pronouns are represented by the case names in each position. Genitive pronouns mark the A of a transitive construction, and nominative pronouns mark S in an intransitive construction and P in a transitive one. The NP expressing E in structure (7b) is marked by an oblique case-​marking form so. The pronominal P may be expressed by a clitic pronoun (7c) or an independent pronoun (7d). (7) Abstracted argument structures in Ibaloy (1) Personal pronouns (without Aux) a. Vi[=nom]S b. Vi[=nom]S [so-​n ind]E c. Vt[=gen]A[=nom]P d. Vt[=gen]A [nom/​ind]P Abstracted argument structures of Pendau are presented in (8), showing that in this language there is one intransitive structure (8a), and two transitive structures (8b, c). Sentence structure (8b) is referred to as “actor voice (av),” and the initial argument is A, and the argument following the verb is P. Sentence structure (8c) is referred as “inverse voice,” where the initial argument is a P, and the argument following the verb is A. In this

 

Ergativity and language change in Austronesian languages    565 language, the A of the inverse voice (Aiv) is expressed by genitive forms, while others (S, Aav, Pav, Piv) are expressed by “absolute” (equivalent of nominative) forms. (8)

Abstracted argument structures in Pendau (1) Personal pronouns a. Vintr [abs]S b. [abs]Aav Vav [abs]Pav c. [abs]Piv Viv[-​gen]Aiv

The etymological correspondences among the sentence structures of Ibaloy and Pendau are shown in Figure 23.5. In the figure, cognate structures are boxed together with dotted lines. Structures (7a) and (8a), structures (7b) and (8b), and structures (7c, d) and (8c) are identified to have developed from the same proto-​structure and thus are cognate structures. These cognate structures are first identified here by the occurrence or non-​ occurrence of genitive pronominal forms associated with the marking of A and are then confirmed by the distribution of other grammatical forms (Kikusawa 2003, Kikusawa to appear b). In Ibaloy sentence (7c) and its alternate (7d), and Pendau sentence (8c), a genitive pronoun expresses the A and these are thus identified as cognate structures. The monadic sentence structures (7a) and (8a) are assumed to have developed from the same intransitive structure, and sentence structures (7b) and (8b) are also assumed to be cognate structures based on the fact that they are dyadic with no genitive pronoun marking A occurring in the structure. These sentence correspondences are supported by various types of evidence, in particular by the distribution of various verbal affixes. For example, the distribution of the verb prefix ʔon-​‘imperfective, actor verb’ (5b) in Ibaloy, and neng-​‘realis, actor voice marker’ (6b) in Pendau are restricted to the intransitive or actor voice structures (7a, b) and (8b). This can be associated with the syntactic distribution of the commonly found meN-​/n ​ eN-​verb alternation in western Austronesian languages (e.g., məN-​/​nəN-​verbs in Ibaloy and mong-​/​nong-​ verbs in Pendau),6 although space does not allow me to go into details here. Likewise, the Ibaloy

Figure 23.5  Cognacy among sentence structures in Ibaloy and Pendau 6 

Vowels in the prefixes listed here for the two languages alternates depending on the phonological environment.

 

566   Ritsuko Kikusawa infix ‘perfective, locational voice’ (5c), and the Pendau prefix ni-​‘realis, inverse voice marker’ (6c, d), which are probably related, occur only in transitive structures such as (7c, d) and (8c). Based on the identified cognate relationships, it becomes possible to reconstruct how the different systems developed. For the sake of convenience, I will refer to the two systems as systems (7) and (8). Here I claim that the general direction of the change is as from (7) to (8). This is based on the fact that this direction of change is readily explained while the reverse is not, as summarized in what follows.7 The existence of the two systems is explained as the result of word order change and subsequent changes where morphological contrast was lost. The sentence structures shown in (9) represent the assumed proto-​system, which resembles the Ibaloy system (7). From this system, (8)  is considered to have developed with the nominative NP acquiring pre-​main-​verb position. The new word order is a result of fronting, which was a topicalized position in the proto-​system, becoming an unmarked position for nominative NPs.8 The process of the change from the proto-​system (9) to the Pendau system (8) is schematically shown in Figure 23.6.    (9) Reconstructed shared proto-​system of Ibaloy and Pendau a. Vi[=nom]S b. Vi[=nom]S [ni ind]E c. Vt[=gen]A [nom]P In the proto-​system, there are two dyadic constructions but with the nominative argument having a different macro-​role in each. The nominative NP in (9b) is actor while in (9c) it is undergoer. Thus, the fronting of nominative NPs yielded two sentences where the positions of the actor and the undergoer are reversed, such as the one described in (6b, c–​d) and repeated here as (10a, b). (10) Two dyadic sentence examples in Pendau (repeated from (6)) a. [Io]Aav neng-​ebiling [’a’u]Pav (=(6b)) 3sg.abs av/​r-​leave 1sg.abs actor undergoer ‘He left me.’ b. [’A’u]Piv ni-​ebiling 1sg.abs iv/​r-​leave undergoer ‘He left me.’

7 

[-​onyo]Aiv (=(6c)) -​3sg.gen actor

(Quick 1997: 467)

(Quick 1997: 467)

See Kikusawa (2003a, to appear b) for details of the relevant discussion. The same process is proposed by Aldridge (2004, 2008b, 2010) but from a different theoretical perspective. 8 

 

Ergativity and language change in Austronesian languages    567

Figure 23.6  Change from a shared proto-system to the pre-Pendau system

To illustrate the claimed change, Ibaloy extended intransitive sentences are presented in (11) and transitive sentences in (12), where nominative NPs are fronted to be topicalized. When the nominative NPs are fronted, the semantic actor (S) precedes the verb in extended intransitive sentences (11), and the semantic undergoer (P) precedes the verb in transitive sentences (12), showing the two word orders as in (10). According to Ruffolo (2004:  470), “When the topicalized constituent corresponds to the Nominative complement of a verb, the clause usually contains a resumptive Nominative pronoun which is co-​referential with the topic. The pronoun is a bound form. When the topic refers to a third person singular entity no resumptive pronoun surfaces in the clause.” A  resumptive pronoun occurs in examples (11a) and (12a) where the topicalized constituent is 3pl and a resumptive clitic pronoun is ʔida. In examples (11b) and (12b), the topicalized constituent is 3sg and there is no pronominal form occurring in the main clause. (11)

Ibaloy intransitive sentences with a topicalized nominative NP a. sota enganak tan sota nankaama, [sota ʔəN-​ʔanak tan sota nanka-​ʔama]TS nom/​rec ActV/​pft-​sponsor and nom/​rec StaPatV/​pl-​old.man menginom     ira     ni      tapey. məN-​ʔinom      [ʔida]S     [ni2     tapəy]E ActV/​ipf-​drink  3+/​nom  gen  rice.wine ‘as for the sponsors and the old men, they drink rice wine’ (Ruffolo 2004: 471) b. nem   sota      embalangan      aso, [nəm sota      ʔən-​balaŋa=n    ʔaso]TS but      nom/​rec  StaV/​en-​red=lk  dog timiyed chi toktok ni chontog tijəd[=Ø]S [di toktok ni dontog]E climb loc head summit gen mountain ‘but as for the red dog, it climbed to the top of the mountain.’ (Ruffolo 2004: 471)

 

568   Ritsuko Kikusawa (12)

Ibaloy transitive sentences with a topicalized nominative NP a. say bebaknang ket in’obdaanto ira [saj cv-​baknaŋ]TP kət ʔin-​ʔobla-​an[=to]A [ʔida]P top pl-​rich TpLk BnfV/​ipf-​wok-​BnfV=3/​gen 3+/​nom ‘as for the rich people, he works for them’ (Ruffolo 2004: 473) b. emin ya kanenmi [ʔəmin ja kan-​ən=mi]TP all lk eat-​PatV/​ipf=1+/​gen ‘all our food, he eats’

ket kət TpLk

kanento kan-​ən[=to]A eat-​PatV/​ipf=3/​gen (Ruffolo 2004: 473)

When the topicalized constituent is pronominal, it is the independent pronoun which occurs in the fronted position, with a corresponding resumptive pronouns. In (13), for example, the nominative pronoun si-​Kam occurs in the fronted position with the nominative clitic pronoun (=2sg) occurring as a resumption pronoun. The sentence structures with topicalized nominative NPs are presented in (14). It can be hypothesized that from this system, the Pendau system must have developed via the one where the resumptive pronouns are optional as in (15) and (16). (13) Ibaloy sentence with a topicalized pronominal NP [Si-​Kam,]TS daw[=ka]S [shi Bagiw]E 2sg.ind go=2sg.nom loc Baguio ‘You, you go to Baguio.’

(Ballard 2011: 796)

(14) Sentence structures with topicalized S and P in Ibaloy a. [ind]S Vi[=nom]S b. [ind]S Vi[=nom]S [ni ind]E c. [ind]P Vt[=gen]A (15)

Pre-​Pendau system (1) a. [ind]S Vi(=nom) b. [ind]S Vi(=nom) c. [ind]P Vt[=gen]A

(16) Pre-​Pendau system (2) a. [ind]S Vi b. [ind]S Vi c. [ind]P Vt[=gen]A

[ni ind]E

[ni ind]E

Note that this change is naturally motivated if we assume that the proto-​system was not only morphologically ergative but also syntactically ergative. In the Philippine-​type languages today, including Ibaloy and Tagalog, nominative NPs are the only core NPs that can be fronted for topicalizing (or, “extracted,” cf. Kaufman, Chapter 24, this volume). The fronted position is marked for these languages with either an intonational break, or a

 

Ergativity and language change in Austronesian languages    569 formal signal following the topicalized phrase labeled as a “topic linker,” while in Pendau and other languages in Indonesia, pre-​verbal position is default for these NPs. Thus, the word order change between the two systems is readily explained by the originally allowed fronting of nominative arguments becoming a fixed position. If we assume the direction of the change was the reverse, it would be difficult to explain both motivations and the actual procedure. We would have to assume that, first, the sentence initial NPs, the macro-​role of which (actor or undergoer) had been marked by word order and somehow started to follow the verb at a certain stage. It would have to be assumed that in this process, the nominative NP expressing the actor in the actor voice sentence had acquired post-​verbal position, the position that marked the other NP as expressing the undergoer. The pre-​verbal position would have had to be retained to become the marked position for topicalization. All these processes are much more difficult to explain and support than the proposed hypothesis. The proposed change appears to apply to lexical NPs as well as pronominal NPs as described in what follows. In Ibaloy, the case marking of lexical NPs shows the same case alignment as the pronominal system. Abstracted sentence structures are presented in (17). The Ibaloy forms si and ʔi (=y) expressing S and P, mark the difference between personal and common nouns respectively, and so are nen and ni expressing A. Two dyadic sentence examples are given in (18), where nominative NPs S and P are underlined. Thus, in Ibaloy, both personal pronouns and NP markers show an ergative pattern. As Kaufman (Chapter 24, this volume) states, “according to the criteria of case and transitivity, Tagalog, and the vast majority of Philippine languages, are clearly ergative,” and Ibaloy is not an exception.9 The relative position of the two lexical NP arguments is free, except that the genitive NP expressing A  normally precedes the nominative complement (Ruffolo 2004: 417). (17)

9 

Abstracted argument structures in Ibaloy (2) Lexical NPs (personal/​non-​personal)10 a. Vi [si/​ʔi NP]S b. Vi [si/​ʔi NP]S [so=ni2, di NP]E c. Vt [nǝn/​ni1 NP]A [si/​ʔi NP]P

In the description of Philippine languages, attempts have been made to simultaneously capture the system combining two different grammatical features, namely, case-​marking patterns and the alternation of verb forms. Results of such attempts include symmetrical voice analyses, focus analyses, among others. Although an ergative analysis and a voice analysis are not mutually exclusive, those who take an either-​or position, rejecting the former, often do so questioning the typological acceptability of the status of what are analyzed in this chapter as “extended intransitive,” as syntactically intransitive. See Kaufman (Chapter 24, this volume) for further discussion. 10  Ruffolo (2004: 471) glosses both ni and ni as genitive. However, in this chapter, because of the 1 2 functional difference and also semantic difference (an NP marked by ni1 is always definite while one marked by ni2 is always indefinite), the two forms are differentiated and the former is analyzed as genitive, while the latter as oblique. Incidentally, the NP marked with ni2 (oblique) does not alternate with a genitive pronoun, as ni1 (genitive) does.

 

570   Ritsuko Kikusawa (18)

Ibaloy sentence examples with lexical NPs a. engoney i aki ni ʔǝN-​ʔonǝj [ʔi ʔaki]S [ni2 ActV/​pft-​see nom monkey obl ‘the monkey saw a mouse’

otot ʔotot]E mouse

(Ruffolo 2004: 238)

b. naon’an ni dedaki sota bibiid Batan na-​ʔonǝj-​an [ni1 cv-​laki]A [sota cv-​biʔi=d batan]P Pot.LocV/​pft-​see-​LocV gen pl-​man nom/​rec pl-​woman=loc Batan ‘the men happen[ed] to see the women of Batan’ (Ruffolo 2004: 306) The case-​alignment pattern is shared by pronouns and lexical NPs in Pendau as well. Schematic Pendau sentence structures with lexical NP arguments are presented in (19), followed by two sentence examples illustrating (19b, c) in (20). The forms marking lexical NPs show the same pattern as pronouns, and carry the same case, with si and Ø marking respectively personal and common nouns when absolutive/​nominative. The forms ni and nu marking respectively personal and common nouns when genitive (19). The post verbal positions of (19b, c) are fixed. The positioning of other NPs (including the sole NP in (19a)) is flexible and may precede the V or occur in sentence final position following the other NP (Quick 2007: 366–​369). (19) Abstracted argument structures in Pendau (2) Lexical NPs (proper/​common) a.          Vintr [si/​Ø NP]S b. [si/​Ø NP]Aav Vav [si/​Ø NP]Pav c. [si/​Ø NP]Piv Viv [ni/​nu NP]Aiv (20) Pendau actor and inverse voice sentence examples with lexical NP arguments a. [Si kai]A neng-​ita-​i [si be’e]P abs/​pnm grandfather av/​r-​see-​loc abs/​pnm grandmother ‘The grandfather saw the grandmother.’ (Quick 1997: 466) b. [Si be’e]P ni-​ita-​i abs/​pnm grandmother iv/​r-​see-​loc ‘The grandfather saw the grandmother.’

[ni gen/​pnm

kai]A grandfather (Quick 1997: 466)

By comparing the pre-​Pendau system in (21), which is extracted from Figure 23.6, and the current Pendau system repeated here in (22), the following changes are inferred: (i) nominative NP started to occur in the sentence-​initial position relatively freely; (ii) the position of the non-​fronted NP in the extended intransitive structure (E in (21b) was fixed to the post-​verb position; (iii) the sentence-​initial position became the (semi-​)default position for nominative NPs, and (iv) the syntactic distinction between extended intransitive and transitive ((21b) and (21c)) was lost. Of these, the last point is discussed in some detail in what follows.

 

Ergativity and language change in Austronesian languages    571 (21)

Pre-​Pendau system a. [nom]S Vi b. [nom]S Vi [ni ind]E c. [nom]P Vt[=gen]A

(22) Argument structures in Pendau with pronouns (=(8)) a. Vintr [abs]S b. [abs]Aav Vav [abs]Pav c. [abs]Piv Viv[-​gen]Aiv (23)  Argument structures in Pendau with lexical NPs (=(19))    a. Vintr [si/​Ø NP]S     b. [si/​Ø NP]Aav Vav [si/​Ø NP]Pav    c. [si/​Ø NP]Piv Viv [ni/​nu NP]Aiv As has been shown, both Ibaloy and Pendau have two kinds of dyadic sentence structure. The difference in the case alignment analysis of the two languages resides in the fact that, in Ibaloy, only one of the two dyadic structures is considered to be a canonical transitive, while in Pendau, both of them are analyzed as transitive. The Ibaloy analysis depends primarily on the morphology of the verb. Non-canonical transitive, verbs typically match the morphological structure of monadic intransitive verbs, hence their analysis as extended intransitive constructions. For example, verb affixes on-​ and məN-​and their alternating forms occur in clear intransitive and extended intransitive sentences as in (24) and (25). It can be seen that the prefix ʔon-​both in a monadic (24a) and dyadic (24b) sentences, while the prefix maN-​occurs in a monadic sentence (25a) and its alternate naN-​in a dyadic sentence (25b).11 (24) Examples of Ibaloy prefix on-​in monadic and dyadic intransitive sentences a. onchakchak [i chanom]S ʔon-​dakdak ʔi danom ActV/​ipf-​boil nom water ‘the water will boil’ b. onbono [sota too]S [ni ʔon-​bono sota toʔo ni ActV/​ipf-​kill nom/​rec person gen ‘the person will kill a snake (or snakes)’

11 

oleg]E ʔoləg snake

(Ruffolo 2004: 216)

(Ruffolo 2004: 218)

The vowels in Ibaloy maN-​/m ​ əN-​ and naN-​/n ​ əN-​alternate depending on word stress.

 

572   Ritsuko Kikusawa (25) Examples of Ibaloy prefix məN-​in monadic and dyadic intransitive sentences a. mantejaw [ira]S ni pigen adew maN-​tajaw ʔida ni piga=n ʔakəw ActV/​ipf-​traditional.dance 3+/​nom gen several=lk day ‘they will dance for several days’            (Ruffolo 2004: 243) b. nangda naN-​ʔala[=Ø]S ActV/​pft-​get=3/​nom   ‘he got some sugar cane’

ni ni gen

onas ʔonas sugarcane

(Ruffolo 2004: 242)

At the same time, the referent of the oblique NP of such constructions has to be indefinite or partitive, while the second NP of canonical transitive constructions are always interpreted as definite. In Pendau, however, Pav, the argument that historically corresponds to the oblique in Ibaloy, is “referentially definite or indefinite, but when it takes a demonstrative…it must be definite” (Quick 2007: 366). Quick (1997: 462) concludes that the two dyadic sentence structures such as (20) should both be analyzed as two primary transitive clauses in this language. The selection of which depends on contextual and structural constraints, which can be explained using functional/​pragmatic parameters. Unlike Ibaloy where there is a semantic difference between two corresponding dyadic structures, the difference between the two structures has apparently been lost in Pendau. The applicative verb morphology occurring on verbs in both active and inverse voice structures is probably the result of an innovation subsequent to the loss of this distinction (see Kikusawa to appear a for details). The change is schematically shown in (26), which applies to both pronominal and lexical NPs.12 From the earlier intransitive verb (Vi), the Pendau actor voice verb develops (Vav) and from the earlier transitive verb (Vt), the Pendau inverse voice verb (Viv) develops. (26) Illustration of word order change from system (9) to system (8) proto-​system nominative fronted inverse system Vi S E  → S Vi E → A Vav P Vt A P  → P Vt A  → P Viv A The generalization of the change that has been discussed above is from a tripartite case contrast system (Ibaloy and other Philippine languages, with nominative, genitive and oblique cases marking core arguments) to a bipartite one (Pendau, with genitive and 12 

The reconstruction of case-​marking forms is commonly more difficult than general lexical reconstruction for various reasons, and although relevant to the current discussion, I take it as out of scope of this chapter. Case-​marking forms are typically short (commonly monosyllabic) and tend to undergo sporadic sound changes. Such paradigmatic forms are susceptible to replacement, and their functional interpretation changes when the system changes, obscuring cognate relationships (cf. Reid 1978).

 

Ergativity and language change in Austronesian languages    573 nominative cases marking core arguments), with neutralization of the contrast between earlier nominative case and oblique case forms and the functional difference between them being marked by word order. The bi-​partite case-​marking system is found in many languages in Indonesia today, in which only residues of the earlier genitive pronoun are found within a single pronominal system. In such a system, word order is the major means of case marking. Example sentences from Balinese are presented in (27). It can be seen that the same pronoun occurs expressing the A of AVP pattern dyadic sentence (27b) and the A of PVA pattern dyadic sentence (27c), the latter of which would be expressed by a genitive pronoun in languages like Pendau. (27)

Sentence example from Balinese (Wechsler and Arka 1998: 388) a. [Ia]S pules (s)he sleep (S)he is sleeping. b. [Tiang]A numbas I av.buy ‘I bought the pig.’ c. [Bawi-​ne punika]P pig-​def that ‘I bought the pig.’

[bawi-​ne pig-​def

punika]P (high register) that

tumbas ov.buy

[tiang]A (high register) I

The sequence of changes is in line with the knowledge that in verb-​medial languages, “dispreference for [morphological] case marking” is commonly observed and explanations for this tendency have been offered (Siewierska and Bakker 2009; cf. Hawkins 2004 for possible explanations). If in fact word order change took place and languages became verb-​medial, the languages can then be said to have followed this path toward less case marking. Whatever the motivations are, with no semantic difference between two dyadic sentences, and the system being typically described as showing a “voice system” with actor/​agent voice and patient voice sentence structures, the system can no longer be analyzed as showing ergative. The change presented in this section is from a morphologically marked ergative system to a word order marked bi-​transitive system. The discussion is summarized in Figure 23.7.

23.3.2 A Development of an Accusative Pattern Clitic Pronoun System from an Earlier Ergative System In this section, data will be presented from Tongan and Samoan to demonstrate the development of an accusative clitic pronoun system from an earlier ergative system. In section 23.3.1, a shared proto-​system of Ibaloy and Pendau was reconstructed as having developed from an ergative system as in (28).

 

574   Ritsuko Kikusawa

Figure 23.7  Changes from an ergative system that resulted in a bi-transitive system

(28) Shared proto-​system of Ibaloy and Pendau (=(9)) a. Vi [nom]S b. Vi [nom]S [ni ind]E c. Vt[=gen]A [nom]P From this system, it was shown how an inverse system with two pronominal contrasts in Pendau and a bi-​transitive system with a single-​pronominal system in Balinese developed. In this section, it will be shown that the development of accusative pattern pronominal systems are also accounted for by assuming a similar proto-​system, but with the following preconditions: (i) nominative pronouns expressing S were clitic pronouns; (ii) second-​position, or Wackernagel clitics had acquired the pre-main verb position as their default position. This is illustrated by Ibaloy examples in (29). In (29a), the clitic pronoun =mo ‘2sg.gen’ expressing the A follows the main verb kalat-​ən ‘to bite s.o.’ In (29b), where an auxiliary verb ʔəg occurs, the genitive pronoun expressing the A (the form =da in (29b)) “climbs up” to the second position, to follow the auxiliary. Example (29c) shows a case where both the genitive clitic expressing A (=to) and the nominative clitic expressing P (=ka) occurring in the second position. (29) Ibaloy sentence examples with auxiliary verbs a. nem kedatenmo ira nəm kalat-​ən[=mo]A [ʔida]P but bite-​PatV/​ipf=2/​gen 3+/​nom ‘but you will bite them.’

(Ruffolo 2004: 179)

 

Ergativity and language change in Austronesian languages    575 b. egcha kedaten ʔəg[=da]A kalat-​ən neg=3+/​gen bite-​PatV/​ipf ‘they did not bite them.’

ira [ʔida]P 3+/​nom

c. egtoka kegtinan ʔəg[=to]A[=ka]P gətin-​an neg=3/​gen=2/​nom ipf-​step-​LocV/​ipf ‘he will not step on you.’

(Ruffolo 2004: 178)

(Ruffolo 2004: 179)

It should also be noted that a nominative pronoun in Ibaloy can be replaced by an independent pronoun (as ʔida in (29a, b). In such a case, an independent pronoun “carries a more emphatic or contrastive meaning” (Ruffolo 2004: 174). An abstracted Ibaloy pronominal system illustrating constructions with an auxiliary verb is presented in (30). It should be noted that the pronominal forms that occur as clitic pronouns are limited to nominative and genitive. (30) Abstracted argument structures in Ibaloy with clitic/​independent pronoun alternation    Clitic pronouns Independent pronouns    a. Vaux[=nom]S Vi e. Vaux Vi [ind]S     b. Vaux[=nom]S Vi [son ind]E f. Vaux Vi [ind]S [son ind]E    c. Vaux[=gen]A[=nom]P Vt     d. Vaux[=gen]A Vt [nom]P g. Vaux[=gen]A Vt [ind]P Languages compared with this system in this subsection are Tongan and Samoan, which retain the earlier sentence structures with auxiliary verbs, the ones that are shown in (30). Tongan example sentences are presented in (31) and (32). Sentences in (31) show the clitic pronoun system. Here, S and A are expressed by the form ku, which precedes the main verb, while P is expressed by the form au following the main verb. Thus the clitic pronoun system shows an accusative pattern. (31)

Tongan sentence examples with clitic pronouns (Otsuka, Chapter 40, this volume) a. Na‘a [ku]S kata PST 1SG laugh ‘I laughed.’ b. Na‘a [ku]A ma‘u PST 1SG get ‘I caught a fish.’

[‘a ABS

e SPEC

ika]P fish

 

576   Ritsuko Kikusawa c. Na‘e taa‘i [au]P PST hit 1SG ‘John hit me.’

[‘e ERG

Sione]A John

Sentences in (32) present example sentences with independent pronouns, where pronominal arguments are underlined. When an argument is expressed with an independent pronoun, the pronoun is preceded by a case marking form, kiate ‘dative’ (32a), ‘e ‘ergative’ (32b, c), or ‘a ‘absolutive’ (32b). (32) Tongan sentence examples with independent pronoun arguments a. ‘E tokoni [‘a Sione]S [kiate koe]E fut help  abs   Sione     to-​person 2sg ‘Sione will help you.’ (Otsuka 2000: 258) b. Na‘e taa‘i [‘e pst hit erg ‘Sione hit you.’

Sione]A John

[‘a abs

c. Na‘e tala mai [‘e ia]A pst tell dir.1 erg 3.sg ‘He told me (that) it was correct.’

koe]P 2sg (Otsuka, Chapter 40, this volume) ‘oku tonu prs correct (Otsuka, Chapter 40, this volume)

Sentence structures with Tongan personal pronouns are presented in (33). Tongan personal pronouns occur in two different patterns:  (i)  a set of clitic pronouns marking both S and A, with an independent pronoun marking P of a transitive clause, and thus occurring in an accusative case-​alignment pattern as in (31); (ii) independent pronouns receiving the same case marking as personal lexical NPs, showing an ergative pattern as in (32). According to Otsuka (Chapter 40, this volume), the use of independent pronouns in lexical NP slots is, like in Ibaloy, “marked and has an effect of emphasis.” (33) Tongan pronominal system Clitic pronouns a. Vaux [cltc]S Vi b. Vaux [cltc]S Vi … c. Vaux [cltc]A Vt [cltc]P

Independent pronouns e. Vaux Vi [ʔa ind]S f. Vaux Vi [ʔa ind]S [kiate ind]E g. Vaux Vt [ʔe ind]A [ʔa ind]P

The Tongan system as described is here compared with the Ibaloy sentence structures with clitic and independent pronouns, which are repeated in (34). (34) Abstracted argument structures in Ibaloy with clitic/​independent pronoun alternation (=(30)) Clitic pronouns Independent pronouns a. Vaux[=nom]S Vi e. Vaux Vi [ind]S b. Vaux[=nom]S Vi [son ind]E f. Vaux Vi [ind]S [son ind]E

 

Ergativity and language change in Austronesian languages    577 c. Vaux[=gen]A[=nom]P Vt d. Vaux[=gen]A Vt [nom]P

g. Vaux[=gen]A

Vt [ind]P

Comparing the structures in (30) and (33), the following differences can be pointed out: (i) clitic pronouns expressing S and A form a single set in Tongan, while in Ibaloy, S is expressed with nominative and A with genitive; (ii) A can be expressed by an independent pronoun marked by ʔe in Tongan, while in Ibaloy, A is always expressed by a clitic pronoun and never by an independent pronoun, which alternates with a lexical NP; (iii) in Tongan, independent pronouns have to be morphologically case marked when occurring in an argument position. Accounting for the development of accusative pattern clitic pronoun sets (expressing S and A) in Oceanic languages, such as the one in Tongan, it has been hypothesized that the clitic pronouns showing an accusative pattern system are the result of a merger of the earlier genitive set (marking A) and nominative set (marking S) (Kikusawa 2002). The outline of this hypothesis is summarized in (35). (35) Development of a split pronominal system from an Ibaloy-type ergative pronominal system Clitic pronouns

Independent pronouns

a. Vaux[=nom]S

Vi

a. Vaux

Vi

[ind]S

b. Vaux[=nom]S

Vi [son ind]E

b. Vaux

Vi

[ind]S

c. Vaux[=gen]A

Vt [nom]P

c. Vaux [=gen]A

a. Vaux[=cltc]S

Vi

a. Vaux

Vi

[ind]S

b. Vaux[=cltc]S

Vi [son ind]E

b. Vaux

Vi

[ind]S

[son ind]E

c. Vaux[=cltc]S

Vt [nom]P

c. Vaux

Vt

[ind]A

[ind]P

[son ind]E Vt [ind]P

The original pronominal system is assumed to have been one similar to that in Ibaloy (and other Philippine-​type and Formosan languages) with sentence structures containing auxiliary verbs. At this stage, clitic pronouns and independent pronouns must have alternated depending on various syntactic and pragmatic factors. When A was expressed by a pronoun, it had to be realized by a genitive clitic pronoun and no independent pronoun could be used. Starting from this system, a clitic pronoun set with a new function (covering both S and A) developed by merging the earlier nominative and genitive sets. The precondition of this change was probably the position of the second order clitic pronouns that became fixed in the pre-​main-​verb position. Then, S and A contrasted E and P by their relative position to the main verb, and because the S and A share the semantic role actor, while E and P share the semantic role undergoer, morphological marking was no longer necessary (Kikusawa 2003b, 2015).

 

578   Ritsuko Kikusawa It should be noted that merger is unidirectional and thus implies strong directionality. Observation of the forms of clitic pronouns supports this hypothesis. For example, Lynch, Ross and Crowley (2002: 68) reconstructed three sets of “subject proclitics” for Proto Oceanic, which are labeled as Set I, II, and III. Regarding the existence of so many sets for a single function, they state “[a]‌lthough subject proclitics (or prefixes) occur in many well distributed Oceanic languages and we can infer their presence in P[roto-​] Oc[eanic], their forms vary considerably and a number of competing reconstructions can be made. We organize these into three sets … we see that Sets I and II respectively reflect the P[roto-​]M[alayo-​]P[olynesian] nominative and genitive clitics …” Meanwhile, genitive clitic pronouns expressing A  started to alternate with the corresponding independent pronouns marked as lexical NPs in situ. The general change is from an ergative system to a system with accusative pattern clitic pronouns and independent pronouns marked in the same way as lexical NPs. The pronominal sets changed from a system with genitive clitics (expressing A), nominative clitics (expressing S and P), and independent pronouns (expressing S and P) to a system with a clitic pronoun set (expressing the actor), an independent set (expressing S, A, and P), and a genitive pronoun set, which no longer occur on verbs but only on nouns expressing possessors. In such a system, actual pronominal forms are often shared between different sets. The results in Samoan have been analyzed as an ergative language with an accusative pronominal pattern (Mosel and Hovdhaugen 1992), or an ergative language with an accusative “raising” pattern (Chung 1978), for example. A possible scenario with specific developmental stages to account for these changes is the following. First, some of the genitive pronouns that originally only expressed A in transitive constructions, were generalized also to express S in intransitive constructions. Genitive pronouns were then gradually replaced by corresponding forms from (an)other pronominal set(s). The end result of this sequence of changes varies depending on the language. In some languages, most if not all the clitic pronoun forms are the retention of forms in the original genitive set, while in some languages, clitic pronoun forms show a mixed etymology, and yet in other languages, all the forms have been replaced with independent pronominal forms as will be seen with Tongan and Samoan data. Table 23.4 presents the forms of Tongan pronouns. There is often more than a single form for each function. When we closely examine them, it can be seen that clitic and genitive pronouns share the same in common. This reflects the split of the earlier genitive set splitting to a clitic set occurring on the verb and a genitive set occurring on the noun. It can be also seen, however, that in non-singular pronouns, the longer genitive form and part of the corresponding independent form are identical. This appears to indicate the spread of the function of independent forms. The independent form spreading to cover the function of the other pronoun sets is more obvious in Samoan (Table 23.5). Like Tongan, Samoan personal pronouns are commonly described as consisting of three sets, namely, pre-​main-​verb pronouns

 

Ergativity and language change in Austronesian languages    579 Table 23.4 Tongan personal pronounsa Clitic 1sg

Genitive

Independent

ou, u ku

ku au

2sg

ke o, u koe

3sg

ne

ne no, na ia

1dl.incl

ta

ta taua

1dl.excl

ma

ma maua

2dl

mo na tau mau mou nau

(ki)mautolu

mou moutolu

3pl

(ki)tautolu

mau mautolu

2pl

(ki)naua

tau tautolu

1pl.excl

(ki)moua

na naua

1pl.incl

(ki)maua

mo moua

3dl

(ki)taua

(ki)moutolu

nau nautolu

(ki)nautolu

a. The longer possessive forms are used for emphasis.

(corresponding to “clitic pronouns”), independent pronouns, and possessive pronouns. Here, it can be seen that in some plural forms, a long form, or the earlier independent form, occurs to cover all the three functions. Like Tongan, Samoan shows a split system as shown in (36) followed by sentence examples in (37). Sentence (37a) exemplifies the intransitive structure (36b), while sentences (37b–​d) exemplify the transitive structure (36)c.

 

580   Ritsuko Kikusawa Table 23.5 Samoan personal pronouns (based on Mosel and Hovdhaugen 1992: 121–​124) 1sg.gnrl

preverbal

possessive

=ʔu

=ʔu

independent

ʔou, oʔu aʔu 1sg.em

ta

=ta, =tā taʔita

2sg

ʔe, e, ʔē, ē =u ʔoe

3sg

na

=na

ia 1dl.excl

tā

ia tā tāʔua

1dl.incl

mā

mā māʔua

2dl

lua

(ʔi)tāʔua

(ʔi)māʔua

lua =ulua

3dl

ʔoulua

ʔoulua

lā

lā

ʔoulua

lāʔua

(ʔi)lāʔua

1pl.excl

tātou

tātou

(ʔi)tātou

1pl.incl

mātou

mātou

(ʔi)mātou

2pl

tou

tou =utou

3pl

ʔoutou

ʔoutou

ʔoutou

lātou

lātou

(ʔi)lātou

(36) Samoan pronominal system Pre-​main-​verb pronouns a. Vaux [cltc]S Vi b. Vaux [cltc]S Vi … c. Vaux [cltc]A Vt …

Independent pronouns a. Vaux Vi b. Vaux Vi c. Vaux Vt

[ind]S [ind]S [ind]P

[(ʔ)i(ā) ind]E [e ind]A

 

Ergativity and language change in Austronesian languages    581 (37)

Samoan sentence examples with pronominal arguments a. olo [oʔu]S galue t 1sg work ‘I am working.’ b. ʔua [na]A fasi-​a perf 3sg hit-​tr ‘He hit the girl.’

[le the

teine]P girl

c. Na [ia]A opo-​ina past 3sg hug-​tr ‘He hugged the girl.’

[le the

teine]P girl

d. Na opo [e ia]A past hug erg 3sg ‘He hugged the girl.’

[le the

teine]P girl

(Clark 1976: 91)

(Mosel 1987: 461, my analysis)

(Cook 1991: 82, my analysis)

(Cook 1991: 82, my analysis)

The 1sg and 2sg forms in Samoan show similar patterns to those that are found in Tongan, with forms (ʔou, oʔu and ʔe, e) exclusively occurring in the clitic pronoun set, and others shared with the genitive set, or with both the genitive and independent sets. By comparing with Tongan, it can be seen that independent pronouns have extended their distribution to clitic pronoun position. What is particularly interesting in the given context is that the use of ia in a pre-​main-​verb position is apparently a recent innovation. The 3sg pre-​main-​verb pronouns =na and ia go back etymologically to earlier genitive forms (PAn *=ni-​á, POc *=ña (ACD)) and nominative forms (PAn *=ia (ACD), POc *=ia (Ross 1988)), respectively. This is supported by the fact that in Tongan (a closely related language), the form ia is restricted to the independent set.13 According to Mosel (1987: 461), in Samoan, the 3sg clitic pronoun =na occurs only to express A, and reflects the original distribution of POc *=ña. In addition, she claims that the occurrence of the 3dl and 3pl forms is restricted to A of a transitive construction and only to the S of an intransitive construction that is an actor, but not to the S that is an undergoer.14 This appears to be a transitional stage prior to the point where the same set of clitic pronouns is used for A and all of the exponents of S. If the earlier genitive pronouns were in fact generalized from marking A to also mark S, as I claim, we would expect to see this reflected in languages where a cross-​referencing system developed, and this is in fact what we find in Tetun (East Timor) and Taba (a language of southern Halmahera in North Maluku province of Indonesia). Taba, for example, has actor cross-​referencing proclitics (showing an accusative pattern), the forms of four of 13 

Note that in Samoan this 3rd person pronoun is the only one that is shared exclusively between the pre-​main-​verb and independent sets, and not with the genitive set. 14  Mosel and Hovdhaugen (1992: 697–​698) comment, however, that they found “counter-​examples for all proposed constraints” regarding the occurrence of third person pre-​main-​verb pronouns expressing S.

 

582   Ritsuko Kikusawa which (the three singular forms and 1pl.incl) are shared with the endings of the corresponding possessive ligatures (Table 23.6). Example sentences are presented in (38). Table 23.6 Taba agreement forms and possessive pronominal endings (Bowden 2001: 188–​189)a Actor cross-​ referencing proclitics

possessive ligatures

independent pronouns

1sg

k=

nik

yak

2sg

m=

nim

au

3sg

n=

ni

i

1pl.incl

t=

nit

tit

1pl.excl

a=

amam

am

2pl

h=

memeu/​mmeu

meu

3pl

l=

nidi

si

a. Whether the cross-​referencing clitic forms for 1pl.excl and 2pl etymologically

relate to the corresponding possessive ligature and independent pronoun requires investigation. The form for 3pl appears to have developed by the same process as the other forms did, namely, retaining the part following ni of the possessive ligature.

(38) Taba sentence examples i. Mapot i be.heavy 3sg ‘It’s heavy.’

(Bowden 2001: 102)

ii. Yak kwom yak k=wom 1sg 1sg.cr=come ‘I’ve come.’

(Bowden 2001: 187)

iii. Yanti ncung um yanti n=sung um Yanti 3sg.cr=enter house ‘Yanti entered the house.’

(li)15 (li) (loc)

(Bowden 2001: 102)

15  I could not find a sentence example with all the arguments expressed with a pronoun for this construction. An example of a locative complement phrase expressed with a pronoun (yak li) is as follows:

Malusa nim wlo maduga m=ha-​lusa nim wlo m=ha-​duga 2sg=caus-​say 2sg.poss liver 2sg=caus-​only ‘You said your heart was only for me.’

yak yak 1sg

li. li loc (Bowden 2001: 323)

 

Ergativity and language change in Austronesian languages    583 iv. Am aamsi am a=am=si 1ex.pl 1ex.pl=see=3pl ‘We already saw them.’

do do real

(Bowden 2001: 35)

Cross-​referencing marking is obligatory for the A  and the S of “actor intransitive verbs,” while optional for the S of “undergoer intransitive verbs” (Bowden 2001: 118–​ 119, 188), thus showing a transitional stage where the verb agreement marking originally marked A is extending its distribution to cover part of the intransitive structure. Bowden states that although there are times when either actor intransitives or undergoer intransitives can be used to refer to the same events, “actor intransitives are more commonly associated with animate Actors and Undergoer intransitive more commonly associated with inanimates” (Bowden 2001: 199–​200). Note that among Samoan clitic pronouns, reflexes of 1st and 2nd person pronouns occur expressing both S and A, while the occurrence of 3dl and 3pl is restricted to S that expresses an undergoer and A, and 3sg only to S. Both the Samoan and Taba data could be related to the semantic saliency hierarchy.

23.3.3 Ergative and Accusative Interpretation of Polynesian Sentence Structures Changes discussed in the previous two subsections both involved structural changes that resulted in a new interpretation of the case-​alignment system. The change that is going to be briefly discussed in this subsection, is one that took place in the marking of lexical NPs in Polynesian languages, where it is believed that the earlier ergative system changed to an accusative system. However, this change does not involve changes in surface-​structures. Instead, according to Otsuka (2011a), it is the combination of the distribution of certain verb forms and sentence types that resulted in the change in the interpretation of their sentence structures. It has been commonly known that languages that are analyzed as ergative (such as Tongan) and those as accusative (such as Māori) share the same set of sentence structures shown in (39).16 The judgment as to the pattern of the case-​alignment system of these languages has depended on which of the dyadic sentence structures serves as the canonical transitive. More specifically, when the structures (39b) and (39c) are analyzed as the canonical transitive structure, the language is analyzed as accusative. On the other hand, when the structure (39d) is analyzed as the canonical transitive 16 

Whether the Proto-​Polynesian system was ergative and changed to accusative in some languages, or whether it was accusative and changed to ergative in some languages has produced extensive debate in Austronesian linguistics (see Ball 2007: 130–​132 and Pucilowski 2006 for a summary).

 

584   Ritsuko Kikusawa structure, the system is analyzed as ergative. These are illustrated with specific examples in what follows. (39) Abstracted argument structures of lexical NPs in Polynesian languages a. Vaux Vi [NP]S(actor/​undergoer) b. Vaux V [NP]actor [i NP]undergoer c. Vaux V [NP]actor [ki NP]undergoer d. Vaux V [e NP]actor [NP]undergoer Tongan and Samoan systems are analyzed as ergative, as shown in (40). In these languages, structure (39b) carries a partitive or “less affected” (Otsuka 2011a) reading, showing semantically lower transitivity. Thus, structure (40c) is analyzed as the canonical transitive and structure (40b) as an extended intransitive. This analysis is parallel to that of Ibaloy and other Philippine languages where the extended intransitive construction also carries partitive and less affected senses. It should be noted that the analysis is also parallel to that of Rotuman, an accusative language ((3) and (4)). (40) Abstracted argument structures of lexical NPs in Samoan a. Vaux Vi [NP]S b. Vaux Vie [NP]S [i/​ʔi NP]E c. Vaux Vt [e NP]A [NP]P (41) Samoan sentence examples with lexical NPs a. ‘ua ‘ata [le tama]S perf laugh the boy ‘The boy laughed.’ b. Sa fa’afetai [le teine]S [i pst thank the girl obl ‘The girl thanked her mother’

lona her

c. ‘olo’o fafao [e le tama]A prog pack erg the boy ‘The boy is packing the banana-​case.’

[le the

(Milner 1976: 26) tina]E mother pusafa’i]P banana-​case

(Chung 1978: 217)

(Milner 1976: 59)

On the other hand, Māori and other Eastern Polynesian languages are commonly analyzed as accusative, and in such analyses, it is structure (39b). that is analyzed as the canonical transitive as shown in (42). According to Otsuka (2011a), even in the languages which have been analyzed as accusative, the i marked NP in a structure such as (42b) is less affected, just as in Tongan and Samoan, one of the facts that Gibson and Starosta used to claim that Māori is an ergative language (Gibson and Starosta 1990).

 

Ergativity and language change in Austronesian languages    585 They claim that if consistent criteria had been applied exclusively to transitivity and the case-​marking of lexical NPs, it is likely that all Polynesian languages would be analyzed the same—​showing an ergative system. (42) Abstracted argument structures of lexical NPs in Māori a. Vaux Vi [NP]S b. Vaux Vt [NP]A [i NP]P c. Vaux Vpass [NP]S [e NP]agent (43) Māori sentence examples with lexical NPs a. kua haere [a Hone]S tam go pers Hone ‘Hone has gone.’

(Pucilowski 2006: 13)

b. e kai ana [ngā tamariki]A [i ngā āporo]P tam eat tam the.pl child acc the.pl apple ‘the children are eating the apples’           (Bauer 1997: 40) c. i patua [te kurī]S [e tam hit the dog by ‘the dog was hit by the child’

te the

tamaiti]agent child

(Bauer 1997: 41)

The two systems presented in (40) and (42) are compared in Figure 23.8. What is it then that make Tongan and Samoan more readily analyzed as ergative and Māori and other Eastern Polynesian languages as accusative? Otsuka (2011a) argues that in Tongan and Samoan, verbs with highly transitive meanings do not occur (or are very limited) in sentences with the pattern shown in(39b) and (39c), but in (39d), while in Māori and other Eastern Polynesian languages, any verb could occur in structures (39b) and (39c). Thus, with proto-​typical transitive verbs, some analysts find the (39d), the sole structure in which such verbs occur, as the canonical transitive, while (39b) and (39c) as derived from (39d). As a result, Tongan and Samoan are more likely to be analyzed as ergative. Otsuka demonstrates the details of the developmental paths of Māori and other Eastern Polynesian systems from the clearly ergative system found in Tongan and Samoan. This includes the generalization (extension) of the distribution of the *-​Cia suffix (which I claim originated from the earlier canonical transitive structure but appears to occur in both dyadic sentences), and the affectedness alternation associated with the existence or non-​ existence of the suffix. Although I have a reservation in accepting her statement that “any dyadic verbs [sic] may occur in at least two types of constructions freely” in Philippine-​type languages, the developmental paths that she proposes is in line with what is expected based on the cognacy of sentence structures.

 

586   Ritsuko Kikusawa

Figure 23.8  A comparison of ergative and accusative analyses

 

Ergativity and language change in Austronesian languages    587

23.4  Concluding Remarks In this chapter, I have shown three sets of changes each of which resulted in a change in the case-​alignment system. The first case was a change in the alignment system which is claimed to have started with word order change. The resulting structures allowed the reading of both of the two dyadic sentence structures as transitive, and the system is no longer analyzed as ergative, but a “voice system,” an Austronesian linguistic term for languages with two transitive structures where the positions of the two core arguments are reversed. In the second case, a merger of earlier nominative pronouns marking S and genitive pronouns marking A resulted in a change from an ergative-​to an accusative-​pattern pronominal system. In the third case, proto-​typical transitive verb forms extended their distribution from what was earlier a canonical transitive structure to dyadic intransitive structures. This resulted in the change in the interpretation of the case-​alignment system. The comparison and reconstruction was mainly based on a macro-​comparison between a Philippine-​type ergative language and geographically distant Malayo-​Polynesian languages. The generally assumed working hypothesis, that the Philippine-​type language, Ibaloy, is structurally more conservative and that other languages belonging to lower branches in the family tree show innovations, is supported by the results obtained in this study. The changes referred to in this chapter all involve what appear to be a change in case-​ alignment from an ergative system to an accusative one. Whether languages in the higher-​order subgroups of Austronesian exhibit an ergative system or a voice system is not an issue here, for whichever of the possible analyses one follows, a sequence of the changes by which other languages developed needs to be clarified and explained, and if the proposed hypothesis is correct, it should be translatable to the framework which yields the other analysis. The approach followed in this chapter is to examine case-​marking systems and the verb systems separately, considering the independent and gradual nature of syntactic change. However, by combining the results of the comparison and reconstruction of the case system with that of the verb system, we should be able to capture how some 1,200 languages that are typologically so diverse developed in the Austronesian language family. Ball (2007: 140–​142) lists a few Oceanic languages where a change from accusative to ergative appears to have taken place. However, each of these is obviously a relatively recent independent innovation as Ball claims, and they must have taken place subsequent to the general flow of ergative to accusative change that took place in Austronesian languages, after the Oceanic languages had developed new accusative systems. Such changes all involve developments in the marking of lexical NPs, and what is seen is the susceptibility of case marking on lexical NPs, sporadic or borrowed, to change or reanalysis contrasting with the relative stability of pronominal sets that were originally distinguished by case, in particular the genitive set. An anonymous reviewer questions the validity of determining structural cognacy by the occurrence or partial occurrence of the genitive pronouns associated with the actor, since the realis construction in Proto-Austronesian was historically a nominalization, while the irrealis paradigm is claimed to have a different source (Aldridge, Chapter 21,

 

588   Ritsuko Kikusawa this volume), so that genitive pronouns could be sourced to different structures. While the details of the reconstruction of Proto-​Austronesian morphosyntax is still highly controversial and the subject of much ongoing research, this chapter has dealt mainly with case-​alignment systems, and the syntactic reconstruction of patterns within Malayo-​Polynesian. The typological characteristics that each language exhibits is a sum of the results of a large number of accumulated changes. By clarifying developmental paths of components consisting of languages, we come to learn explanations as to how existing systems, typologically common or unusual, developed. Thus, in the context of diachronic examination, the focus is not on whether a system changed from ergative to accusative, or vice versa, but how each of these changes contributed to the analyses that linguists propose.

Abbreviations -​, affix boundary; /​, may alternate with; [ ]‌A, expressing A; [ ]E, expressing E; [ ]P, expressing P; [ ]S, expressing S; =, (clitic boundary); A, agent of transitive verb; abs, absolute, absolutive; ActV, actor verb, Aux, auxiliary verb; av, av, actor voice; BnfV, beneficiary-​oriented verb; cltc, clitic pronoun; dy, dynamic verb; E, extended argument of intransitive verbs; erg, ergative; gen, genitive; excl, exclusive; genr, general; incl, inclusive; ind, independent pronoun; ipf, imperfective aspect; iv, inverse voice; lk, linker; loc, locative; lv, locational voice; nom, nominative; NP, noun phrase; obl, oblique; ov, object voice; P, patient of transitive verbs; pass, passive; PatV, patient-​oriented verb; pft, perfective; pfv, perfective; pl, plural; pnm, personal; pst, past tense; pot, potential; r, realis; rec, recognitional demonstrative; reciprocal marker; S, subject (actor/​ undergoer) of intransitive verbs; sg, singular; spec, specific; ta, tense & aspect; tp, topicalized P; TpLk, topic linker; ts, topicalized S; Vaux, auxiliary verb; Vav, agent voice or actor verb; Vi, Vintr, intransitive verb; Viv, inverse voice verb; Vpv, patient verb; Vt, transitive verb.

 

Chapter 24

Lexical cat e g ory and alignm e nt i n Austron e sia n Daniel Kaufman

24.1  The Alignment of Philippine-Type Languages Philippine-​type languages are often cited as exemplifying a cross-​linguistically unique voice system, in which verb morphology can select not only an agent or patient, but also locative, instrumental, and other adjunct type relations as subject.1 Current syntactic treatments characterize this phenomenon alternatively as a rich voice system, a rich applicative system, “case agreement” or thematic nominalization. Relatedly, there is disagreement as to whether Philippine languages are best analyzed as ergative, accusative, active, or symmetrical. The uncomfortable position of Philippine languages in regard to more common alignment systems has been a long-​standing topic in the typological literature (some earlier treatments include DeWolf 1988; Shibatani 1988; Himmelmann 1991).2 Here, I present a critical review of three current syntactic analyses of Tagalog: the ergative analysis (Gerdts 1988b; De Guzman 1988; Liao 2004; Aldridge 2004), the case-​agreement approach (Richards 2000; Rackowski 2002; Rackowski 1 

I use the term “Philippine-​type language” here to refer to a subset of Austronesian languages spoken in the Philippines as well as northern Sulawesi and Borneo. Proto-​Malayo-​Polynesian, the ancestor of all Austronesian languages spoken outside of Taiwan is reconstructed as a Philippine-​type language although some Austronesian languages of Taiwan may also be considered Philippine type (e.g. Amis, Paiwan, Seediq). See Himmelmann (2005) for discussion. 2  Richards (2013:fn.2), for instance, states, “I suspect that the debate about whether Tagalog is ergative will prove to be a terminological one; Tagalog resembles ergative languages in some respects, and differs from them in others, and the only question is how vague we want the technical term ‘ergative’ to be.”

 

590   Daniel Kaufman and Richards 2005)  and a nominalism analysis (Starosta et  al. 1982/2009; Kaufman 2009a,b), while advocating for the latter. I also present new data from Mamuju (South Sulawesi subgroup, Indonesia) to demonstrate that canonical ergative languages also exist within the Western Malayo-​Polynesian subgroup of Austronesian. I argue that this retreat to the canonical ergative type is a result of the historical re-​emergence of verbal predication. The typical Austronesian “voice” paradigm is exemplified in (1) and (2) with Tagalog.3 Note that the use of each morpheme correlates with the selection of a different argument or adjunct as the “ang phrase.” (1) a. Bili ng bulaklak ang bátà para um.beg-​buy ng flower ang child for ‘The child bought a flower for the prisoner.’

sa sa

b. Bili-​∅ ng bátà ang bulaklak para buy-​in ng Juan ang flower for ‘A/​the child bought the flower for the prisoner.’ c. I-​bili ng bátà ng bulaklak ang i-​give ng child ng flower ang ‘A/​the child bought the prisoner a flower.’

b. I-​bigay ng bátà sa bilanggô ang i-​give ng child sa prisoner ang ‘A/​the child gave the flower to the prisoner.’ 3 

sa bilanggô sa prisoner bilanggô prisoner

d. Bil-​han ng bátà ng bulaklak ang give-​an ng child ng flower ang ‘A/​the child bought a flower from the prisoner.’ (2) a. Nag-​bigay ng bulaklak sa bilanggô mag.beg-​give ng flower sa prisoner ‘The child gave a flower to the prisoner.’

bilanggô prisoner

bilanggô prisoner

ang ang

bátà child

bulaklak flower

The different verbal morphemes glossed act for actor voice/​agreement, pat for patient voice/​agreement, conv for conveyance voice/​agreement and loc for locative voice/​ agreement. The glosses should be self-​explanatory except for the conveyance voice. The types of arguments and adjuncts selected by the conveyance voice do not seem to form a natural semantic class. They include instruments, benefactives, and themes. I adopt the term conveyance voice from Wolff (1973) based on its selection of objects that are conveyed away from the agent. It is also referred to as instrumental, benefactive and circumstantial voice in the literature. Note that the actor voice is indicated by in (i) but by nag-​in (ii). The latter form is best analyzed as a combination of and pag-​, a transitivity related prefix, but this will not concern us here. Regarding spelling conventions, the Tagalog genitive case marker ng is an abbreviation for /​naŋ/. I follow the accentual conventions of (Wolff et al. 1991) where final stress is treated as the unmarked default and penultimate stress/​length is marked with an acute accent. The grave accent is used to indicate word-​final glottal stop. Thus, bátà indicates /​ˈbaːtaʔ/​.

 

Lexical category and alignment in Austronesian    591 c. Bigy-​an ng bátà ng bulaklak ang give-​an ng child ng flower ang ‘A/​the child gave the prisoner a flower.’

bilanggô prisoner

The proper analysis of the alternating morphology on the predicates (1) and (2) (glossed neutrally as , nag-​, i-​, -​an, etc.) as well as the markers introducing the arguments (ang, ng, sa) are at the crux of the debate on Philippine-​type languages. As emphasized by Foley (2008), what makes the system atypical is its morphologically symmetrical nature; note that in each of the examples (1) and (2), one and only one voice marker appears.4 Following the typologically oriented work of Himmelmann (1991) and Foley (1998), I trace this symmetry to the root level, specifically, the lack of a clear contrast between entity-​ denoting and event-​denoting roots. Crucially, when this contrast develops, as in many languages of Indonesia, more canonical alignment patterns emerge as well (Kaufman 2009b).5

24.1.1 The Ergative Analysis of Philippine Languages Under Aldridge’s (2004) ergative analysis of Tagalog, our example sentences in (1) would be analyzed as in (3), following the interpretation of the morphology presented in Table 24.1. (3)

a. Bili ng bulaklak ang bátà antipass-​buy obl flower abs child ‘The child bought a flower the prisoner.’

para for

b. Bili-​∅ ng bátà ang bulaklak buy-​tr erg child abs flower ‘A/​the child bought the flower for the prisoner.’

4 

sa prep

para for

sa prep

bilanggô prisoner bilanggô prisoner

c. I-​bili ng bátà ng bulaklak appl-​give erg child obl flower ‘A/​the child bought the prisoner a flower.’

ang abs

bilanggô prisoner

d. Bil-​han ng bátà ng bulaklak give-​appl erg child obl flower ‘A/​the child bought a flower from the prisoner.’

ang abs

bilanggô prisoner

Note that the symmetry referred to here is primarily morphological. The different “voices” are clearly not interchangeable in the majority of cases as they correspond to different possibilities for the definiteness of arguments. The term “symmetrical” has also been used in the sense of (Bresnan and Moshi 1990) to refer to the treatment of ditransitive objects. This is also unrelated to the morphological symmetry discussed here. 5  In later historical developments, this ergativity has also given way to a more accusative syntax, as it does in Indonesian and Malagasy. See Aldridge (2008a) for a formal treatment of accusative features in Austronesian.

 

592   Daniel Kaufman The sentence in (3b) contains a canonical transitive clause with the zero allomorph of the transitive suffix -​in.6 In this clause, the agent is marked with ergative case and the patient is marked with absolutive case. The sentence in (3a) must then be analyzed as an antipassive construction in which the agent is marked with absolutive and the patient is marked with oblique. In Tagalog, there is no formal difference between oblique and ergative case marking. Other Philippine languages, such as Ivatan (Reid 1966), distinguish these functions with separate case markers. Table 24.1 Ergative analysis of Tagalog morphology Argument marking

Predicate marking

ang ng1

absolutive ergative

ng2

oblique

sa

preposition

antipassive1

mag-​

antipassive2

-​in

transitive

-​an

applicative1 (directional)

i-​

applicative2 (instrumental, benefactive, etc.)

In support of the antipassive analysis of (3a), we can note the indefinite/​non-​specific interpretation of the object, a fact which has been commented upon in all descriptions and analyses of Tagalog (Bloomfield 1917; Schachter and Otanes 1982:  76; Wolff et  al. 1991; Kroeger 1993; Maclachlan and Nakamura 1997:  310; Richards 2000; Rackowski 2002; Kaufman 2005; as well as older descriptive works). As pointed out by proponents of the ergative analysis, this is a typical property of antipassive patients and is seen clearly in numerous Inuit and Mayan languages. In Tagalog, the objects of putative antipassive verbs are robustly indefinite (except when the agent is extracted, see Adams and Manaster-​ Ramer 1988). For instance, in (4a) we see that a definite demonstrative is not felicitous (without a partitive reading) in the object position of an antipassive verb. There is no such constraint on the clause in (4b), with what is a canonical transitive verb on the ergative analysis. (4) a. *?Káin nito ang bátà     eat obl.this abs child     (For, ‘The child ate this.’ OK for ‘The child ate from this.’) 6  The zero allomorph of -​in is used in conjunction with the aspectual infix . The infix also has a zero allomorph which is used in the prospective aspect. Because both of these voice markers have zero allomorphs in their aspectual paradigm, neither can be said to be less marked than the other.

 

Lexical category and alignment in Austronesian    593 b. Káin-​∅ ng eat-​tr erg ‘A/​the child ate this.’

bátà child

ito abs.this

According to the criteria of case and transitivity, Tagalog, and the vast majority of Philippine languages, are clearly ergative. The problem, as noted by Foley (1998, 2008) and Himmelmann (1991, 2005) among others, is that there are few if any ergative languages outside the Austronesian family that use the same morphology for putative antipassives as well as monadic (underlyingly intransitive) predicates. In Philippine languages, not only is used in antipassive contexts as in (5a), it is used in underlyingly intransitive predicates as in (5b). The same problem is noted by Paul and Travis (2006: 321) for Malagasy man-​and holds throughout Philippine-​type languages. (5)

a. Patay kill ‘Galvan killed a goat.’ b. Lákad walk ‘Galvan walked.’

ng obl si abs

kambing goat

si abs

Galvan Galvan

Galvan Galvan

Furthermore, is used on meteorological verbs, which differ from canonical intransitives in disallowing an overt absolutive argument, as shown in (6). In this case then it seems that simply functions to create an event-​denoting predicate rather than reducing valency or relating a particular thematic role to the ang phrase. (6) a. áraw sun ‘It was sunny.’

(*ang araw)    abs sun

b. ulan rain ‘It rained.’

(*ang ulan)   abs rain

c. Lindol earthquake ‘There was an earthquake.’

(*ang lindol)    abs earthquake

The question of the Tagalog antipassive is discussed at length by Aldridge (2012b), who argues that antipassives are not derived but rather combine with intransitive verbal morphology. Aldridge (2012b: 198–​200) claims that a demotion analysis of antipassives requires downward movement of the demoted object, a possibility that is excluded by Chomsky (1995, 2005). While this is easy to argue from a Philippine perspective, it

 

594   Daniel Kaufman remains to be explained why antipassive morphology only attaches to transitive verbs in other well-​known ergative languages.7 The two other verbal alternations involving i-​and -​an are treated as applicatives on the ergative approach (Aldridge 2004). The first promotes benefactives, instrumentals and conveyance objects to absolutive while the latter promotes directional and locative arguments. There are two difficulties with treating these morphemes as applicatives. The first is that, unlike traditional applicatives, they are always required to introduce arguments that correspond to their functions, whether selected by the predicate or not. For example, a predicate like bigay ‘give’ requires applicatives to promote both the theme and the recipient to absolutive position, as shown in (7). (7) a. Bigy-​an mo ng give-​appl 2s.erg obl ‘Give the child a mango.’

mangga mango

b. I-​bigay mo ang mango appl-​give 2s.erg abs mangga ‘Give the mango to the child.’

ang abs

bátà child

sa prep

bátà child

This is typologically unusual, as one of the two alternations is generally treated as basic, but in Tagalog and other Philippine languages all such arguments are derived as there is no unmarked verb. The second difficulty is that we expect applicative morphology to co-​ occur with -​in, which under the ergative analysis must be treated as a simple transitive marker. Yet no Philippine language employs combinations such as i-​STEM-​in or STEM-​ in-​an for these putative applicative constructions. On the other hand, a strong piece of supporting evidence for the ergative view is the well-​ known restriction on extraction in Philippine and other Austronesian languages (as first discussed in a generative context by Keenan 1972). Ng-​marked arguments in Tagalog cannot be topicalized, relativized, or clefted. This is demonstrated via ungrammatical topicalization of an ergative argument in (8) and an antipassive patient in (9). In both cases, topicalization of the corresponding absolutive argument is fully acceptable, as shown in the (a) sentences. (8) a. Ang bulaklak ay bili-​∅ abs flower top buy-​tr ‘The flower, the child bought.’ b. *Ng bátà ay bili-​∅ ang erg child top buy-​tr abs (For, ‘The child, bought the flower.’) 7 

ng erg

bátà child

bulaklak flower

The Chol intransitive marker -​i and transitive marker -​V may, however, be relevant here. The -​i suffix appears with underlying monadic predicates as well as derived intransitives, such as passives. Antipassives are formed with a transitive light verb and thus do not display -​i. See Coon (2012) for discussion.

 

Lexical category and alignment in Austronesian    595    (9)

a. Ang bátà ay bili abs child top buy ‘The child, bought a flower.’ b. *Ng bulaklak ay bili obl flower top buy (For, ‘A flower, the child bought.’)

ng obl ang abs

bulaklak flower bátà child

Similar restrictions are attested in Mayan and Inuit languages among others. It is these facts, together with the reduced transitivity of “actor voice” verbs marked with and mag-​, that form the strongest arguments for the ergativity analysis of Philippine-​ type languages. To review, the most basic arguments for the ergative view are shown in (10) and those against it, in (11). (10)

a. indefiniteness (and low scope) of “actor voice” objects b. restrictions on the extraction of “actor voice” objects and transitive agents c. case marking pattern (e.g. gen/​erg syncretism) common to other ergative languages

(11)

a. absence of unmarked transitive and intransitive predicates b. complementary distribution of “transitive” -​in, and “applicatives” i-​conv and -​an loc. c. absence of a true de-​transitivizing antipassive

24.1.2 The Case Agreement Analysis of Philippine Languages Chung (1998), Richards (2000), Rackowski (2002), and Pearson (2001) present variations of an agreement approach to the Austronesian verbal alternations seen above in which the morphology instantiates agreement with an argument in an A-​bar position (see also Erlewine, Levin, and van Urk, Chapter 16, this volume, for a similar approach employing an A-​position). I restrict my focus here to the analyses of Richards (2000), Rackowski (2002) and Rackowski and Richards (2005) as the languages investigated by Chung (1998) and Pearson (2001) (Chamorro and Malagasy, respectively) differ from Tagalog in some important respects. In case agreement approaches, all arguments have their case checked in their respective A-​positions after which object shift may occur. Object shift moves a DP to an outer specifier of vP above the merged position of the external argument. The verb then agrees with the highest argument, not for canonical agreement features, but rather for its case features. Rackowski (2002) proposes that complements of verbs receive accusative case and verbs register accusative agreement with -​in (or its zero allomorph). External arguments receive nominative case from T with the corresponding verbal agreement being

 

596   Daniel Kaufman or mag-​. Nominative agreement with the external argument only takes place when no other argument has moved to a higher position. This is the case when the direct object is indefinite and there are no applicatives present. On this analysis, Tagalog also has null high and low applicatives (cf. Pylkkänen 2002) which promote benefactives, instrumentals, locatives, and other types of arguments and adjuncts. The high applicative assigns oblique case and the verb displays oblique agreement with the prefix i-​. The low applicative assigns dative case and the verb displays dative agreement with the suffix -​an. The merged structure of a benefactive (high) applicative construction is shown in (12) and the result of subsequent movement in (13). The applicative argument ends up on the edge of vP and enters into an agreement relationship with the verb which is spelled out as i-​. (12)

vP v’

DPEA v

ApplP DPBen

Appl’ Appl

VP V

(13)

DPDO

vP vP

DPBen i

v’

DPEA v

ApplP ti

Appl’ Appl

VP V

DPDO

This analysis is significantly more abstract than that of the ergative analysis, in which voice morphology is interpreted directly as applicatives and transitivity markers. Here, the actual applicatives are only detectable indirectly through distinct agreement on the predicate. Note also that the cases with which the predicate agrees are abstract. As seen earlier, there are only three phonologically distinct argument markers in Tagalog but for the case agreement analysis to hold, there must be distinct oblique and dative cases underlyingly that the verb agrees with. The interpretation of Tagalog’s core functional morphology under this analysis is shown in Table 24.2.

 

Lexical category and alignment in Austronesian    597 Table 24.2 Case agreement analysis of Tagalog morphology Argument marking

Predicate marking

ang ng

agreement trigger default case

sa

preposition

nominative agreement1

mag-​

nominative agreement2

-​in

accusative agreement

-​an

dative agreement (directional)

i-​

oblique agreement (instrumental, benefactive, etc.)

The ang marker, interpreted as absolutive case on the ergative analysis, is treated as a topic marker of sorts under case agreement analyses.8 The ng-​case, which instantiates both ergative case and oblique case on the ergative analysis is simply glossed as case in the case-​agreement analysis. It is not clear what kind of case it is and the details of its assignment remain to be explained. As with the ergative analysis, there are also difficult typological and theoretical questions that arise with ostensive applicatives in the case agreement approach. First of all, applicative objects must somehow retain their underlying (oblique and dative) case for case agreement to take place but in familiar nominative-​accusative languages applicatives promote arguments to a position where they receive accusative case and behave as canonical direct objects.9 Second, it remains unclear why applicative objects must shift to the edge of vP and block agreement with the external argument. Without such movement, we expect to find structures like that in (14b), where the ang phrase

8 

However, the use of “topic” cannot be taken too literally, as the ang phrase has been shown conclusively to not be an actual discourse topic (Adams and Manaster-Ramer 1988; Kroeger 1993; Kaufman 2005). For instance, in the following mini-​dialogue, an ang phrase appears within the focus of a completely felicitous response. But see Richards (2000) for arguments that the ang phrase is similar to the topic position in Icelandic and other Germanic languages. (i) A: Ano ang gá∼gawà what ang imprf∼do ‘What is Jojo doing?’

ni case

    B: Hú∼hugás-​an niya imprf∼wash case.3s ‘He’s washing the dishes.’

ang ang

Jojo Jojo? mga pl

pinggan. dish

9  The literature on Bantu applicatives bears this out as well as studies of non-​accusative languages such as those of the Salish family. Kiyosawa and Gerdts (2010: 41), for instance, shows that applicative objects across Salish languages are extracted in exactly the same manner as direct objects, not as obliques, which require nominalization of the clause.

 

598   Daniel Kaufman benefactive fails to trigger case agreement on the verb. As pointed out by Rackowski and Richards (2005), such promotion without further movement is ungrammatical. They offer potential analogues from other syntactic domains but little if any cross-​linguistic support exists for applicative objects blocking agreement with an external argument in nominative-​accusative languages.10 (14) a. Ni-​lútò-​Ø ni Romeo ang adóbo para sa babáe beg-​cook-​Acc case Romeo ang adobo for prep woman ‘Romeo cooked the adobo for the woman.’ (Rackowski and Richards 2005: ex.13a) b. *Ni-​lútò-​Ø ni Romeo ng babáe ang adóbo beg-​cook-​Acc case Romeo case woman ang adobo ‘Romeo cooked the adobo for a woman.’ (Rackowski and Richards 2005: ex.11b) The theory makes crucial use of Pylkkanen’s (2002) distinction between low and high applicatives but the Tagalog facts do not accord well with the theory’s prediction. Low applicatives relate a promoted argument to a direct object and are crucially dependent on the presence of a direct object. Rackowski (2002) argues that this is the case based on the data in (15). (15)

a. Bigy-​an ko ang give-​dat 1s ang ‘I gave the father his child.’ b. *Bigy-​an ko ang give-​dat 1s ang ‘I gave (to) the father.’

ama father

ng case

anak. child

ama. father

However, the judgment reported for in (15b) seems better attributed to pragmatic rather than syntactic factors. There can be no act of giving without a theme and there is thus an expectation that the theme will be included in an out-​of-​the-​blue context. Similar examples could be produced without a theme but more conclusive is the data shown in (16), which demonstrates that monadic predicates of all types can take -​an, counter to the expectations of a low applicative in Pylkkanen’s framework.

10 

An analogy is made to the wager-​class of verbs in English, in which an argument must move from its underlying position with no obvious motivation. Collins (2002) and Epstein and Seely (2006:83–​5) question the judgments surrounding this phenomenon in English and it remains to be seen if the proposed link can be further substantiated. (i) a. *John wagered Mary to have won the race. b. Maryi was wagered ti to have won the race.

 

Lexical category and alignment in Austronesian    599 (16)

a. Ni-​lakár-​an ng bátà beg-​walk-​dat case child ‘The child walked the road.’

ang ang

daan road

b. Sigaw-​an ni shout-​dat case ‘Romeo shouted at Jojo.’

Romeo Romeo

si ang

Jojo Jojo

c. ubu-​han ni cough-​dat case ‘Romeo coughed at Jojo.’

Romeo Romeo

si ang

Jojo Jojo

d. Na-​ulan-​an si sta.beg-​rain-​dat ang ‘Jojo was rained on.’

Jojo Jojo

The role of specificity shift is crucial in both the ergative and case agreement approaches to Tagalog. Both predict, in their own ways, that a definite/​specific object should not be possible for a matrix clause verb marked with or mag-​. On the ergative analysis, these verbs are antipassives whose objects are assigned inherent oblique case from v and remain in VP. In VP, they are subject to existential closure (Diesing 1992) and receive an existential indefinite interpretation. In the case agreement analysis, specificity triggers object shift, which puts a shifted object above the external argument and thereby triggers accusative agreement rather than nominative agreement with or mag-​. Neither approach, however, seems to predict all the specificity facts of applicative constructions correctly. Rackowski (2002) and Rackowski and Richards (2005) explicitly claim that the underlying direct object in applicative constructions can be specific, offering the example in (17) as evidence. (17)

I-​pinaglútò ni Romeo ng adóbo ang babáe obl-​cook case Romeo case adobo ang woman ‘Romeo cooked (the) adobo for the woman.’ (Rackowski and Richards 2005: 570)

However, there is no independent evidence for the definite interpretation of the object in structures like (17). Nothing in (17) forces a specific interpretation and, in fact, evidence points to definite direct objects being just as marked in structures such as (17) as they are as object of verbs marked with or mag-​. This becomes clear by using pronominal arguments as diagnostics. As Rackowski and Richards (2005) show with (18), a pronominal object of an marked verb is ungrammatical. Unexpectedly, the same ungrammaticality appears with direct objects of -​an and i-​marked verbs, as in (19). (18)

*Sampal ko ang mandurukot. slap 1s.case ang pickpocket (For, ‘The pickpocket slapped me.’) (Rackowski and Richards 2005: 568, ex.4b)

 

600   Daniel Kaufman (19) Bigy-​an ni Maria (ng pera cook-​dat case Maria case money Juliette Juliette ‘Maria gave money/​*Romeo to Juliette.’

/​*ni Romeo) si /​ case Romeo ang

In a regular matrix clause with a definite object and an oblique, the object must surface with ang case and the oblique must surface as an oblique/​prepositional object, as shown in (20). (20) I-​bigay ni Maria obl-​cook case Maria ‘Maria gave Romeo to Juliette.’

si ang

Romeo Romeo

kay obl

Juliette Juliette

Rackowski and Richards (2005: 568) follow Chomsky (2001) in positing that: object shift occurs as the result of an EPP-​feature on v that is present only when it has an effect on semantic outcome. There is an effect on semantic outcome because the position at the edge of the vP is assigned a specific interpretation, while everything internal to vP is assigned a nonspecific interpretation.

But there is an issue of derivational look-​ahead in this account as well as an empirical problem. As ng-​marked objects cannot be specific/​definite in a regular matrix clauses, we must assume that they are barred from leaving vP. This means that the presence of an applicative object must arbitrarily preclude an EPP-​feature on v on this account. Barring any larger generalization which can be extracted from this coincidence, it merely describes the facts. Regarding ng-​marked agents, Rackowski & Richards recognize the need to account for both an existential indefinite as well as a specific/​definite reading for them. As seen in (21), a ng-​marked agent can felicitously be a proper name or an indefinite pronoun. (21) Hindi bigy-​an (ni neg give-​dat case pera money ‘Jose/​nobody gave Maria money.’

Jose/​ninuman) si Jose/​case.anyone ang

Maria Maria

ng case

The reasoning for this ambiguity is that only internal arguments have the opportunity to undergo object shift. For arguments that do not have such an opportunity, specificity is unpredictable. But as we have seen in (19), this empirical claim cannot be upheld. A separate treatment is required for ng-​marked objects, which cannot be definite/​specific, and ng-​marked agents, which can. This partially supports the ergative analysis in that ergative arguments are not known to show the kind of definiteness effects attested for antipassive patients. However, the restricted interpretation of direct objects of -​an and

 

Lexical category and alignment in Austronesian    601 i-​verbs is equally problematic. On the ergative analysis, non-​specific readings are only forced for antipassive patients but something extra must be said for -​an and i-​ verbs, which can only be analyzed as transitive. We find that the case-​agreement analysis captures several important truths about Philippine voice systems, enumerated in (22). On the other hand, there remain theoretical issues with the analysis, some of which are summarized in (23). (22) a. no unmarked verbs, (virtually) all inflected verbs show “voice morphology” b. “voice morphology” is only marked once per verb c. range of meanings associated with “voice morphology” resembles that of case marking (23) a. incorrect predictions regarding definiteness/​specificity b. unusual properties of (null) applicatives c. unclear what morphological case is and how it is assigned

24.1.3 The Nominalization Analysis of Philippine Languages Yet another interpretation of Philippine-​type verbal alternations views them as participant nominalizations. This was first proposed by Starosta et al. (1982/2009) on a historical basis and further developed by Ross (2002, 2009) and, in synchronic terms, by Kaufman (2009a).11 On this approach, each “voice” indicates a different participant nominalization, as shown in Table 24.3.12 The three argument markers are interpreted as case, as in traditional descriptions. The ang marker indicates subject case, which can be termed either nominative or absolutive. The ng-​case is first and foremost genitive case, that is, the case of possessors, but is recruited for other purposes due to the regular use of nominalized predicates. Finally, sa is an oblique case marker (and not a preposition as in the two approaches reviewed above).13 11 

This also builds on an important body of work beginning with Bloomfield’s (1917) Tagalog grammar and furthered by Capell (1964), Naylor (1980), Himmelmann (1987, 1991, 2008), DeWolf (1988), and Gil (1993, 1995, 2000) who propose precategorial analyses for Tagalog, or, in the case of Himmelmann (2008), a distinction that cross-​cuts traditional categories. The technical side of the analysis has a close analogue in Johns’ (1992) nominalization-​based approach to Inuktitut. See also Coon (2014) for a similarly v-​less analysis of partially overlapping facts in the morphosyntax of Chol, a Mayan language. 12  For reasons of space, I do not discuss the difference between and mag-​although I treat mag-​as containing an inner causative as suggested by Travis (2000). This is supported in Kaufman 2009c where I show comparative evidence for mag-​being composed historically, and potentially synchronically, of the morphemes pa-​caus-​. 13  For all locality based approaches to the extraction restriction, including case agreement and the ergative analysis, there must in fact only be two cases. The oblique case, marked by sa, is treated as a preposition by these approaches because sa phrases can typically be fronted and topicalized regardless of voice. In other words, oblique phrases are effectively invisible for calculations of locality under previous approaches. But there are many reasons to believe that sa is a case marker rather than a preposition. It is in complementary distribution with the other case markers and is selected by bona fide prepositions, e.g.

 

602   Daniel Kaufman Table 24.3 Nominalist analysis of Tagalog morphology Argument marking

Predicate marking

ang

nominative/​absolutive

ng

genitive

sa

oblique

actor nominalization

mag-​

inner causative + actor nominalization

-​in

patient nominalization

-​an

locative nominalization

i-​

circumstantial nominalization

This view differs from the previous two approaches in treating all predication in Philippine languages as inherently copular. The interpretation of the ang phrase is not derived through the use of applicatives but rather through copular identification with the predicate itself. To compare with Johns’ (1992) analysis of Inuktitut, she proposes that predications are formed compositionally in the manner shown in (24). (24) a. kapi-​jaq stab-​pass.part ‘the stabbed one’ b. anguti-​up kapi-​ja-​a man-​erg stab-​pass.part-​3s/​3s ‘the man’s stabbed one’ OR ‘the one that the man stabbed’ c. anguti-​up nanuq kapi-​ja-​a man-​erg polar bear.abs stab-​pass.part-​3s/​3s ‘The polar bear is the man’s stabbed one.’ OR ‘The man stabbed the polar bear.’ (Johns 1992) Applying this approach to Tagalog, we arrive at the following literal translations for the basic “voice” alternations: (25) a. Káin ng dagà eat gen rat ‘The cat was the eater of a rat.’

ang nom

púsà cat

gáling ‘from’, patúngo ‘towards.’ The oblique also displays a human/​non-​human distinction (sa for non-​ humans vs. kay for humans) which is a unique feature of the two other case markers and not found on any of the bona fide prepositions. These facts alone should dispel any notions that the oblique could be a preposition but see Kaufman (2009a: 40) and Gerassimova and Sells (2008: 196–​197) for further arguments.

 

Lexical category and alignment in Austronesian    603 b. Káin-​∅ ng púsà ang eat-​pat gen cat nom ‘The rat was the eaten one of the cat.’

dagà rat

c. Kaín-​an ng púsà ng dagà ang pinggan eat-​loc gen cat gen rat nom plate ‘The plate was the cat’s eating place of the rat.’ d. I-​káin ng púsà ng dagà ang conv-​eat gen cat gen rat nom ‘The dog was the cat’s “eating benefactor” of the rat.’

áso dog (Kaufman 2009a: 6)

This avoids the difficulties of theta-​linking approaches discussed by Rackowski (2002) since nominalization, like case, is well known to reflect thematic roles imprecisely. More importantly, nominalization also makes sense of the fact that Philippine predicate morphology does not distinguish between argument and adjunct. Barker (1998: 714) discusses at length the way in which English -​ee nominalizations select participants that are not part of the argument structure of the corresponding verb.14 This offers an excellent analogue to one of the more typologically difficult aspects of Philippine-​type voice. As argued in Kaufman (2009a), the nominal nature of event-​denoting predicates in Tagalog can also go a long way in explaining other curiosities of Philippine syntax, including: (26) a. Near identical syntactic distribution of canonical event-​denoting predicates (i.e. “verbs”) and entity-​denoting predicates (i.e. “nouns”). b. Identity between the case of ergative agents and possessors across Philippine languages. c. Ungrammaticality of extracting genitive marked arguments corresponds with the difficulties of extracting possessors and other nominal dependents cross-​linguistically. d. Symmetric nature of Philippine “voice” –​only one instance per predicate, as would be expected of nominalization morphology. 14   Barker utilizes the notion of  “episodic linking” rather than argument structure to account for how participant nominalizations in English identify their referent.

… the meaning of the verb amputate guarantees the existence of a person undergoing amputation, even though there is no syntactic argument that corresponds to this participant… the fact that the person undergoing amputation is a participant of every amputation event is sufficient to enable a set of amputation events to characterize the -​ee noun amputee: for each amputation event e, there exists an individual x which is a participant in e such that x is (becomes) an amputee. Thus amputee is episodically linked to the meaning of amputate despite the fact that there is no corresponding syntactic argument position. Similarly, Aronoff (1980) discusses the role of Gricean principles in certain morphosyntactic alternations.

 

604   Daniel Kaufman Event-​denoting predicates in Philippine-​type languages can always serve as the direct complement to case markers.15 There is no strong evidence from Philippine languages for null complementizers or headless relatives in (27b) and (c).16 (27) a. Ang áso nom dog ‘the dog’ b. Ang lú∼lutú-​in nom imprf∼cook-​pat ‘the thing to be cooked’ c. Ang mag lú∼lutò nom act-​imprf∼cook ‘the one who will cook’ The syntax of ng-​marked possessors and ng-​marked agents appears identical. That is, there is no reason to believe that the addition of the morphology in (28b) leads to a substantial difference in syntactic structure between (28a) and (b). (28) a. Súlat ni Juan write gen Juan ‘That is Juan’s letter.’ b. Súlat-​∅ write-​pat ‘Juan wrote that.’

ni gen

iyan that:nom Juan Juan

iyan that:nom

Johns (1992) makes the same claim for Inuktitut pairs such as those in (29). Inuktitut ja in (29b) functions similarly to the Tagalog patient nominalizer -​in/​∅. In both cases, a NP or DP predicate combines with a structural subject through a copular structure to yield a predication such as (28b).17 (29) a. anguti-​up qimmi-​a man-​gen dog-​3s/​3s ‘the man’s dog’

15 

Kaufman (2009a: 25) points out an important exception to this pattern with the descendants of the original (non-​nominalized) verbs in certain dialects of Tagalog. 16  But see Richards (2009) for some arguments to this effect and Kaufman (forthcoming) for a defense of the symmetry. 17  This is similar to Pearson’s 2005 proposal of base generating the equivalent of the ang-​phrase in Malagasy in an A-​bar position located in a high rightwards specifier and co-​indexed with a lower null operator in an A-​position.

 

Lexical category and alignment in Austronesian    605 b. anguti-​up kapi-​ja-​a man-​gen stab-​pass.part-​3s/​3s ‘the man’s stabbed one’ OR ‘the one that the man stabbed’ There are two potentially independent conditions conspiring against fronting genitive arguments. On one hand, extracting ergatives should bear similarity to sub-​extraction from NP in other languages, a highly constrained operation. Indeed, in Tagalog itself extraction of possessors is subject to precisely the same constraints as ergative extraction as shown by the illicit topicalizations in (30).18 (30) a. Bili-​∅ ni Maria buy-​pat gen Maria ‘Maria bought Juan’s car.’

ang nom

kótse car

b. *Ni Juan ay bili-​∅   gen Juan top buy-​pat (For, ‘Juan, Maria bought (his) car.’)

ni gen

ni    Maria GEN Maria

Juan. Juan ang kótse _​_​_​_​. nom car

c. *Ni Maria ay bili-​∅ _​_​_​_​ ang kótse   gen Maria top buy-​pat nom car (For, ‘Maria, (she) bought Juan’s car.’)

ni Juan. gen Juan

18 

Richards cites Cena’s (1979) examples in (i) of apparent possessor extraction as an argument against constraints on linking ergative extraction to possessor extraction. (i) a. Kasama ng doktor companion gen doctor ‘The child is with the doctor’

ang nom

anak child

    b. ang doktor [na kasama nom doctor lnk companion ‘the doctor that the child is with’

ang nom

anak] child (Cena 1979; Richards 2009)

As shown in (30), possessor extraction of the normal type is subject to exactly the same constraints as ergative extraction. A minimal pair for Cena’s example with an ostensibly verbal predicate, shown in (ii), demonstrates clearly that relativization of the ergative argument is no more marked than relativization of the possessor. I leave the analysis of this construction to further work noting only that Cena’s examples do not entail different treatments for possessors and ergative arguments. (ii) a. sa∼sama-​han ng imprf∼accompany-​loc gen ‘the doctor accompanies his/​her child’

doktor doctor

ang nom

    b. ang doktor [na sa∼sama-​han nom doctor lnk imprf∼accompany-​loc ‘the doctor that accompanies his/​her child’

anak child ang nom

anak] child

 

606   Daniel Kaufman Potentially separate from the issue of extraction from NP is the general ban on genitive predicates. This is independently necessary in Tagalog as shown in (31) and has clear cognates cross-​linguistically. Genitive phrases can be modifiers, as in (31a) but not predicates, as shown in (31b). As in Hungarian and many other languages (Szabolcsi 1983), predicate and extracted possessors must be expressed as obliques or datives, as in (32). (31) a. Ang súlat ni nom write gen ‘The letter of Juan’s’

Juan Juan

b. *Ni Juan ang súlat gen Juan nom write (For, ‘The letter is Juan’s.’) (32) Kay Juan ang obl Juan nom ‘The letter is Juan’s.’

súlat write

Argument questions in Philippine-​type languages have been widely analyzed with the interrogative phrase as the predicate of the clause (Paul 2001; Aldridge 2002; Oda 2002; Massam 2003; Potsdam 2006, 2009; Gerassimova and Sells 2008). As we have already seen that genitive predicates are banned, we also correctly predict that genitive case interrogatives should be ruled out, as seen in (33). Note that the genitive interrogative is allowable in an in-​situ post-​nominal position, as shown in (34) (see also Erlewine, Levin, and van Urk, this volume, Chapter 16, n.21). (33) a. Kaníno ang súlat? obl.who nom write ‘Whose is the letter?’ b. *Nino ang súlat? gen.who nom write (For, ‘Whose is the letter?’) (34) Ang súlat nino?!? nom write gen.who ‘The letter of who?!?’ Similarly, if event-​denoting predicates in Philippine-​type languages are nominalized, we can rule out interrogatives like those in (35b) and (c), which attempt to make a predicate out of a genitive agent. The ungrammaticality in (35b) ensues from the constraint against genitive predicates and (35c) is ruled out on the simple basis of case preservation. What should be a genitive marked agent, as in (35a), cannot be expressed as an interrogative in the nominative (ang) case. A copular predication has at most two nominative marked

 

Lexical category and alignment in Austronesian    607 arguments, one in the predicate position and one in the subject position. The nominal approach to Philippine-​type languages is thus uniquely able to unify the ungrammaticality behind (34) and (35), which is common to the vast majority of Philippine-​type languages. (35)

a. Sú∼sulát-​in ni imprf∼write-​pat gen ‘Jojo will write the book.’

Jojo Jojo

ang nom

b. *Nino ang sú∼sulát-​in   gen.who nom imprf∼write-​pat   (For, ‘Who will write the book?’) c. *Sino ang sú∼sulát-​in   nom.who nom imprf∼write-​pat   (For, ‘Who will write the book?’)

libro book ang nom

ang nom

libro? book libro? book

On this approach, genitive marked arguments in Philippine-​type languages bear a certain resemblance to English of-​phrases. Just like ng-​phrases, of -​phrases are highly restricted in predicate position (36), and also display restrictions on their extraction from NP, as shown in (37).19 (36) The team is (*of) Juan’s. (37)

a.   Juan was an employee of Rizal. b. *?Of whom was Juan an employee?

In (38), we see that these constraints do not hold for instrumental agents of passive verbs. (38)

a. Juan was employed by Rizal. b. By whom was Juan employed?

As pointed out in Kaufman 2009a, this could hold the key to why some ergative languages show a restriction on extracting the ergative argument while others do not (see Dixon 1994 and Manning 1996 for discussion). If the ergative argument is treated as a nominal dependent, we expect it to share a morphological case with possessors and resist extraction. If the ergative argument is a verbal dependent, we

19  See Davies and Dubinsky (2003) for a review of approaches to extraction from NP. Davies and Dubinsky rely on participant structure to derive the complex pattern of extractability from English NPs but they do not consider extraction from participant nominalizations. Other approaches have implicated the left branch condition (Ross 1967), the ECP (Chomsky 1981), the case filter (Huang 1982), movement of non-​constituents (Bošković 2005), and most recently, the treatment of nP or DP as a strong phase (Chomsky 2001; Svenonius 2004).

 

608   Daniel Kaufman expect it to share a morphological case with instrumentals or obliques and to allow extraction more freely. In particular, it seems that a genitive-​ergative syncretism in combination with a pseudo-​cleft strategy for interrogatives is what conspires to constrain extraction in ergative languages. In languages with the same syncretism but in-​situ interrogatives, as Inuktitut, constraints against ergative interrogatives are not attested, as evidenced by (39). Indeed, even in the rare contexts that Tagalog allows wh-​in-​situ, as in (40), genitive interrogatives become acceptable (compare above). (39) Kia Alaana who.gen Alana.nom ‘Who bit Alana?’

kii-​ja-​nga? bite-​pass.part-​3s/​3s

(40) Gawà-​∅ nino make-​pat gen.who ‘Who made those shoes?’

ang nom

sapatos shoe

(Yuan 2013) na iyon? lnk that (Schachter and Otanes 1982: 512)

The typological correlation is worth investigating on a larger scale but here we are most concerned with what happens when a bona fide v category develops from n and its consequences for ergativity. The crucial properties of the v versus n heads which determine lexical category are the following. (41) n properties: a. possessor is projected [Spec, n] b. genitive case to nP-​internal phrases c. strong island properties (42) v properties: a. agent is projected in [Spec, v] b. accusative case to object c. islandhood dependent on v features The v category projects an agent and more generally, v is associated argument structure that is arguably not present in low nominalizations. While apparent arguments can nonetheless be expressed in low nominalizations, they are not distinguished by a distinct object case nor are they obligatory. The general assignment of genitive case within nP can be seen in (43), where the agent, instrument and a manner adverbial are all marked with ng-​case. (43)

nP [Buks-​an

ko ng súsi ng maingay] open-​loc 1s.gen gen key gen loud ‘I opened the door loudly with the key.’

ang nom

pintuan door

 

Lexical category and alignment in Austronesian    609 Note also that the ergative argument is not obligatory nor must it be interpreted as pro when omitted in line with nominalizations cross-​linguistically (Abney 1987; Himmelmann 1991: 22; Alexiadou 2001). The sentence in (44), for instance, can be uttered in an out-​of-​ the-​blue context without any implication regarding an agent. Optionality is expected for a possessor but not for the external argument of a transitive verbal projection.20 (44) Mukhang kain-​∅ ang lahat seem:lnk eat-​pat nom all ‘It seems that everything has been eaten.’ Relatedly, there are also patient voice predicates that function as adversatives and cannot take agents of any sort. Two such examples are shown in (45) and (46).21 (45)

Ni-​langgam-​∅ ang asúkal beg-​ant-​pat nom sugar ‘The sugar was infested with ants.’

(46) In-​ú∼ubo-​∅ ang bátà beg-​cough-​pat nom child ‘The child is (effected by) coughing.’ As discussed earlier, the full range of definiteness/​specificity effects have presented difficulties to previous treatments of Tagalog. In the simplest case, which all theories have an account for, actor voice patients tend to be non-​specific and strictly disallow pronominals, as seen in (47). More difficult to account for is the parallel constraint on themes of locative and circumstantial voice predicates, as seen earlier in (19). (47) a. Nag-​pi∼pinta act.beg-​imprf∼paint ‘John paints faces.’ b. *Nag-​pi∼pinta act.beg-​imprf∼paint (For, ‘John paints me.’)

ng gen ko 1s.gen

mukha face

si nom

si nom

Juan Juan

Juan Juan

20  As Lawrence Reid points out (p.c.), this is not the normal state of affairs in the Cordilleran languages, spoken in North Luzon, Philippines. In many of these languages it seems that an ergative agent is obligatory with transitive predicates. I take this property to be a later historical development based on its narrow distribution but nonetheless very important with regard to the viability of a nominal analysis for the Cordilleran subgroup. 21  We can again draw a parallel to English patient nominalizations with -​ee in that they appear to operate outside of argument structure (Barker 1998), attaching to monadic bases (standee, escapee) as well as nominal bases (hoaxee).

 

610   Daniel Kaufman The nominalization approach suggests a parallel between these facts and similar constraints on definite of-​phrase objects of nominals in English, as shown in (48).22 (48) a.   John is a painter of portraits b. *John is a painter of me In English, this effect extends to themes of recipient nominalizations, as shown in (49). While (49a) shows that a definite theme is possible, a pronominal theme is completely unacceptable. Definite genitive case themes are thus possible in both English and Tagalog (under the right circumstances), but both languages strictly disallow pronominals in this position. A formalization of these facts will not be offered here but these parallels are clearly promising. (49) a.   John was the awardee of the nobel prize. b. *John was the awardee of it. In regard to the semantics of the nominative/​absolutive phrase, it seems far more felicitous to attribute its referential properties to the case marker itself as an operator rather than strictly to object shift (as originally advocated by Himmelmann 1991). Out-​of-​the-​ blue exclamations such as those in (50) support such an approach. It would make little sense to derive the definiteness of (50b) via object shift or agreement with T. On the analysis suggested here, ang contains both case and definiteness features.23 (50) a. Dagà! rat ‘A rat!’ b. Ang pangúlo! ang president ‘The president!’ In the next subsection, we look at root level phenomena to demonstrate that much of the syntax attributed to the voice/​agreement system is already present at the first phase of word building, further suggesting the absence of v in the functional inventory. 22 

Note that John is a painter of mine is acceptable on the possessive reading where I have a painter. According to several speakers I have consulted, this reading is also acceptable for the Tagalog constructions as in (47b) although I have not found naturally occurring examples of an aspect-​inflected actor voice predicate with a possessor. 23  This approach is further supported by other Central Philippine languages which distinguish [±specific] variants for each case marker (see McFarland 1974 and Zorc 1977 for examples). Of course, an analysis in which ang functions as a specific/​definite determiner is not necessarily incompatible with object shift. As Edith Aldridge points out (p.c.), Austronesian languages which still show the same specificity/​definiteness pattern without overt case markers suggest that a structural approach might still be necessary.

 

Lexical category and alignment in Austronesian    611

24.1.4 Ergativity on the root level Foley (1998) argues that Tagalog roots are fundamentally different from those of English: The lexeme give in English is a verb, with a corresponding argument structure . The Tagalog root bigay ‘give’ however is precategorial; it lacks a true argument structure, but does have a precategorial semantic structure like ‘the giving of something to someone by someone.’

On this view, it is the voice morphology which imbues words with argument structure rather than anything in the underlying semantic representation. This explains why notional valency has little consequence for the voice marking potential of roots in Philippine-​type languages. There are, however, problems with the claim that roots lack category and argument structure altogether. In contrast to early claims made about Tagalog roots, they appear regularly in their bare form as both predicates and arguments. Unlike the predictions of the precategorial analysis, bare roots are restricted to particular readings and do not display flexibility in their distribution of thematic roles to genitive and nominative marked arguments. As discussed by Himmelmann (1991: 40) and Kaufman (2009a), the reading of a bare root is consistently that of a patient-​oriented (or “proto-​object”) entity. In Table 24.4, we find a list of roots and their glosses in the first two columns. In the following columns we see the event denoting predicates formed from these roots through the use of voice morphology. Table 24.4 Tagalog root meanings Root

Gloss

Actor voice form

Gloss

WALK EAT

lákad káin

‘a walk, an errand’ ‘eating, a meal’

lákad káin

‘to walk’ ‘to eat’

THINK

ísip

‘thought, thinking’

mag-​isip

‘to think’

KILL

patay

‘corpse’

patay

‘to kill’

BREAK

básag

‘a break’

básag

‘to break’

TEACH

túrò

‘lesson, teaching’

mag-​túrò

‘to teach’

SAY

sábi

‘what is said’

mag-​sábi

‘to say’

BUY

bili

‘price bought for’

bili

‘to buy’

TAKE

kúha

‘taken object’

kúha

‘to take’

Most surprising is the interpretation of stems with the causative prefix, as exemplified in Table 24.5 (adapted from Schachter 1976: 105). Here, we find that even when embedded under a causative head, these roots maintain their patient-​ oriented interpretation.

 

612   Daniel Kaufman Table 24.5 Tagalog causative meanings Causative stem

Gloss

CAUS - HAND.OVER CAUS - BRING

pa-​ábot pa-​dala

‘something caused to be handed over’ ‘something caused to be brought’

CAUS - MAKE

pa-​gawà

‘something caused to be made’

CAUS - COOK

pa-​lútò

‘something caused to be cooked’

CAUS - KEEP

pa-​tágò

‘something caused to be kept’

The patient-​oriented nature of roots is projected to the clause level in predications, which is analyzed as essentially copular. Thus, even with a bare root predicate as in (51), the agentive argument surfaces with genitive case and the patient/​theme argument surfaces with nominative/​absolutive case. (51) Háwak ni Jojo ang hold  gen Jojo nom ‘Jojo holds the money.’

pera money

Just as with fully inflected event-​denoting predicates, extraction obeys the predicted pattern. The genitive argument is trapped within the domain of the nominal predicate and cannot be topicalized or otherwise extracted, as shown in (52). (52) a. Ang pera ay háwak nom money top hold ‘The money, Jojo holds.’ b. *Ni Jojo ay háwak    gen Jojo top hold    (For, ‘Jojo, holds the money.’)

ni gen ang nom

Jojo Jojo pera money

So what does nominalization add to an already nominal root? Primarily, it changes its reference to that of a potential participant. Again, we can compare English -​ee in a word like hoaxee where it similarly attaches to a nominal stem to select a notional participant. Second, it also allows the root to combine with aspect inflection, which bestows an agentive reading to the genitive/​ergative argument.24 This is highly reminiscent of

24  Aspect inflection is generally impossible without voice with one exception. Predicates inflected for the recent perfective show no voice marking and assign genitive case to all arguments. This is probably related to the exclamative function of the recent perfective in Tagalog, as discussed in Kaufman 2011.

 

Lexical category and alignment in Austronesian    613 what has been described by Davis and Demirdache (2000: 100) for St’át’imcets, a Salish language. In St’át’imcets, notionally bivalent roots surface as unaccusative predicates when used in their bare form. This is seen in (53), where the bare predicate qam’t ‘hit’ assigns the patient role to the subject ti sqáycw-​a ‘the man.’ Similar to Tagalog, an agent must be licensed by a voice-​related morpheme, as seen in the contrast between (54a) and (b). In (54b), the suffix -​en introduces an agent and creates a transitive predicate. For Davis and Demirdache (2000), roots like qam’t and máys as well as “classic” unaccusative roots contain an underlying causer (following Chierchia 2004) but no agent. This accords relatively well with Tagalog bare root predicates in that they license causers but not true agents.25 (53)

qam’t ti sqáycw-​a hit det man-​det ‘The man was hit (with something thrown).’

(54)

a. Mays ti tsítcw-​a built det house-​det ‘The house got built.’ b. Máys-​en ti tsítcw-​a built-​dir det house-​det ‘(They) built the house.’

Let us briefly consider one theory of how Tagalog predicates are built from the root up. The category determining head n introduces a referential index (in the sense of Baker 2003). When the inserted root is entity-​denoting, as in (55), this is entirely straightforward. In (55), the merged n+√ stem simply takes on a referential index associated with an instance of ⟦dog⟧. But determining how the reference is set with an event-​denoting root, as in (56), is more complex, as it is often a particular participant rather than the event which the bare stem denotes on the surface. (55)

nP Possessor n0

n’ √PEntity √dog

25   The lack of an agentive reading with bare root predicates can be gleaned from (i), where the adverb nang sindayà ‘intentionally’ is infelicitous.

(i) Bigay ko ang bulaklak (?*nang give 1s.gen nom flower gen ‘The flower is my gift (?*intentionally).’

sinadyà) intentionally

 

614   Daniel Kaufman (56)

nP Possessor

n’

n0

√PEvent Theme

√’ √give

Recipient

Despite lacking an argument structure, an event-​denoting root projects associated thematic roles as Davidsonian event arguments (or “participants,” in the terminology of Grimshaw 1990). In the case of GIVE this will include at least a theme and a recipient, as shown. Let us then posit that when √P combines with n, the root raises to n and n probes downwards to select the nearest thematic role, which determines the reference of the stem. This will naturally exclude an agentive interpretations of roots, as Agents are introduced by the higher functional projection, VoiceP (Kratzer 1996). It will also exclude possessor interpretations as the possessor role is projected to the specifier of nP and is thus outside the c-​command domain of n. We thus have a mechanism for deriving a phrase such as that in (57), where the root bigay GIVE is identified with the theme of the event, i.e. a gift. The possessor and the recipient are expressed as genitive and oblique phrases, respectively.26 (57) Bigay ni Nonoy kay Neneng give gen Nonoy obl Neneng ‘The flower is a gift of Nonoy to Neneng’

ang nom

bulaklak flower

The higher functional category VoiceP licenses an agent and hosts the voice marking in Voice0, which allows the predicate to be identified with one of several thematic roles via episodic linking (Barker 1998) as in participant nominalization. The AspP projection houses aspect morphology and certain aspectual adverbs. These two functional categories undergo morphological merger, as evidenced by several points of syncretism in the Voice/​Aspect paradigm and by the fact that voice marking is a prerequisite for aspect marking. We see a fully inflected analogue of (57) in (58).

26 

Note that this predicts patient-​orientation for all low nominalizations. This has been argued for independently by Salanova (2007) and is supported by the broad comparative study of Koptjevskaja-​ Tamm (1993). It may even be present in English bare nominalizations of bivalent roots (e.g. take as in ‘The take was $500’), which seem to only refer to the patient and never the agent.

 

Lexical category and alignment in Austronesian    615 (58)

I-​bigay ni Nonoy kay Neneng conv-​give gen Nonoy obl Neneng ‘The flower was given by Nonoy to Neneng’

ang nom

bulaklak flower

The peculiar meaning of causative stems seen earlier in Table 24.5 is also predicted here by the fact that the causative head lies beneath the voice head. This is corroborated by the order of prefixes: the conveyance voice prefix and actor voice prefix are always external to the causative marker: i-​pa-​ conv-​caus-​, mag-​pa-​ act-​caus-​. The referential index is set by n and is not affected by further movement to the causative head, pa-​. So while the causative projection introduces a causer, the causer is introduced too late to be identified with the meaning of the bare stem. CausP

(59) Causer

Caus’ pa-

nP Possessor

n’

n0

√PEvent Theme

√’ √give

Recipient

The structure in (59) underlies the predicate phrase in examples like (60). (60) Pa-​bigay ni Nonoy ang caus-​give gen Nonoy nom ‘The flowers are Nonoy’s caused gift’

bulaklak flower

This section has shown that ergativity in Tagalog is present on the root level even before the attachment of voice morphology. This poses an additional challenge to both the case agreement and canonical ergativity approach reviewed above. Under those approaches it is left to explain why object shift should occur in the absence of a VP and, if it occurs, why it is not registered as agreement or transitivity marking on the predicate. With this brief introduction to Tagalog word structure, we are now ready to examine Mamuju, an Indonesian language which has maintained an ergative pattern while developing a strong noun-​verb distinction.

 

616   Daniel Kaufman

24.2  Mamuju and the re-​emergence of v In the following, we observe the basic morphosyntactic alternations of Mamuju, an Austronesian language of the South Sulawesi subgroup, to show the consequences of v on the syntax.27 What differentiates Mamuju from Tagalog is the loss of the nominal properties of event denoting predicates and the development of a strong Noun/​Verb contrast (Himmelmann 2005: 128–​131). Two important consequences of this are that agreement and valency changing operations make reference to underlying argument structure and that constraints on ergative extraction are loosened. Mamuju displays clear differences from Tagalog on the level of root and word. One simple distinction between nouns and verbs in Mamuju is that only verbs can be the complement of the tense marker na future. As seen in (61b), nominal predicates cannot follow na. In Tagalog, there are no tense/​aspect markers whose distribution distinguishes root classes. (61) a. na menjari=aʔ guru fut become=1sg.nom teacher ‘I will be a teacher there at Udayana.’ b. (*na) guru=aʔ jao fut teacher=1sg.nom there ‘I am a teacher there at Udayana.’

jao there di prep

di prep

Udayana Udayana

Udayana Udayana

Conversely, verbal roots cannot combine directly with a possessor, unlike nouns, as shown in (62). An overt nominalizing head must combine with a verbal stem before prior to modification by a possessor. As seen earlier, this is not the case in Philippine-​ type languages. (62) a. *langi-​na    swim-​3.gen b. bau-​na fish-​3.gen ‘his/​her fish’ If the symmetric nature of Philippine-​type alignment systems are due to the nominal nature of the roots, we expect that the emergence of v will allow for bona fide 27 

Mamuju has so far only been described in a single article by Stromme (1994), who also published interlinearized text collections (Stromme 1991). That work has been supplemented by elicitation with native speaker Husni Husain over the course of a semester at the CUNY Graduate Center in 2009 as well as two weeks of field work in the Mamuju area itself. I thank Husni Husain for generously sharing his knowledge with us as well as my graduate students Eva Szymanski, Josh Gray, Ji Young Shim for their valuable input.

 

Lexical category and alignment in Austronesian    617 argument structure to be projected from the root and an absence of symmetry in the derivational morphology. Recall that in Tagalog, the notion of valency plays almost no role in determining a root’s morphological potential. This is seen clearly in Table 24.6, where a notionally monovalent root, langoy ‘swim’ is compared with a bivalent root patay ‘kill.’ Both roots can take the full range of voice morphology suggesting that they are essentially of the same type at the point where they merge with Voice. Table 24.6 Tagalog word classes √LANGOY ‘swim’

√PATAY ‘kill’

actor voice -​in patient voice

langoy languy-​in

‘nom to swim’ patay ‘to swim to nom’ patay-​in

‘nom to kill’ ‘to kill nom’

-​an locative voice

languy-​an

‘to swim in nom’ patay-​an

‘to kill from nom’

This contrasts starkly with Mamuju and other South Sulawesi languages. In Table 24.7, we see that the cognate Mamuju roots are in complementary distribution with regard to their morphosyntactic potential. An unergative verb like langi ‘swim,’ requires prefixation with mo to form an intransitive predicate. In contrast, a bivalent root like patei ‘kill,’ cannot take the intransitive prefix but rather must be prefixed with antipassive mang-​ to enter into a intransitive predication. Bivalent stems must be prefixed with ergative agreement when forming transitive verbs, as seen with ku-​patei ‘I kill (X).’ This agreement marking is impossible with monovalent roots like langi.

Table 24.7 Mamuju word classes √LANGI ‘swim’

√PATEI ‘kill’

mo-​active mang-​antipassive

mo-​langi *man-​langi

‘abs to swim’ —​

ku-​1sg.erg

*ku-​langi

—​

*mo-​patei mam-​patei ku-​patei

—​ ‘abs to kill’ ‘erg to kill abs’

This development can be understood as part of the emergence of a robust v category which converts the thematic structure of an event-​denoting root into actual argument structure. Note also that, unlike in Philippine-​type languages, we now find an unmarked class of intransitive and transitive verbs. Unaccusatives, like tama ‘enter’ in (63), do not require derivational morphology when surfacing as intransitive verbs. Similarly, bivalent roots are only prefixed with ergative agreement when functioning as transitive

 

618   Daniel Kaufman verbs, as seen in (64). All predicates can host second-​position pronominal clitics reflecting the absolutive argument. (63) tama=do=ʔ di enter=already=1.abs prep ‘I already entered the room’

songi room

(64) na-​kita=ko 3.erg-​see=2.abs ‘S/​he sees you.’ Note also the divergence in possessor and ergative agreement. The former is expressed via a set of suffixes and the latter through a set of verbal prefixes. In all Philippine-​type languages, these two functions are expressed with the same set of pronouns, the typically second-​position genitive clitics. Importantly, Mamuju has a robust antipassive, which only combines with transitive predicates and is used to introduce indefinite objects, as can be seen in the comparisons in (65) and (66).28 In the transitive clause, a missing patient argument would be interpreted as a null pronoun retrievable from discourse. The ergative argument typically follows the verb directly although scrambling may also occur to yield an ABS ERG order. (65) a. na-​kande i Husni 3.erg-​eat art Husni ‘Husni is eating the fish.’ b. mang-​kande (bau) i antipass-​eat fish art ‘Husni is eating (fish).’

bau fish Husni Husni

(66) a. mu-​kita=a’ 2.erg-​see=1.abs ‘You see me.’ b. mang-​kita=a’ antipass-​see=1.abs ‘I see a fish.’

bau fish

In addition to the anti-​passive, a limited number of verbs appear to take a “super-​ antipassive,” which has also been described for the neighboring Seko Padang language by 28  The antipassive mang-​prefix derives historically from a distributive/​pluractional infix * (Kaufman 2009c). The path from Proto-​Malayo-​Polynesian distributive marker to an antipassive marker in the South Sulawesi languages is straightforward. Just like antipassives, distributive verbs are bivalent but cannot take a specific or definite object. The Makassarese cognate aN(N)-​is discussed by Jukes (2013).

 

Lexical category and alignment in Austronesian    619 Payne and Laskowske (1997). The super-​antipassive also expresses the subject as an absolutive but does not allow for any objects, whether definite or indefinite, as shown in (67).29 (67) kande=ko eat=2.abs ‘You eat.’

(*bau) fish

Mamuju also has a passive, marked on the verb with the prefix ni-​, shown in (68b). Ergative agreement is obligatory on transitive verbs and thus the passive allows for backgrounded or impersonal agents.30 (68) a. Apa na-​kande todapa? what 3.erg-​eat people ‘What do the people eat?’ b. Apa ni-​kande (*todapa)? what pass-​eat people ‘What is being eaten?’ Mamuju has bona fide applicatives that promote benefactives and oblique arguments to absolutive. Furthermore, the division between low and high applicatives does not run into the difficulties noted earlier for Tagalog. In (69a), the applicative -​ang attaches low and transitivizes an intransitive verb by introducing a theme, thereby feeding ergative agreement. When -​ang merges as a high applicative, as in (69b), it attaches to a transitive stem (pa-​lamme) and introduces a benefactive. (69) a. ku-​lamme-​ang buku 1s.erg-fall-​appl book ‘I dropped a book.’

29 

The super-​antipassive is more useful than it appears at first sight. With the loss of case marking on DPs, the difference between antipassive and super-​antipassive is all that distinguishes a post-​verbal object from a post-​verbal subject in sentences like (i). (i) a. mangapa bongi itte ampe’ why/​when night dem conj ‘What time did the fish eat last night?’

kande eat

    b. mangapa bongi itte ampe’ why/​when night dem conj ‘What time did he eat fish last night?’

mangande antipass:eat

bau? fish

bau? fish

30  The existence of a passive in Philippine-​type languages is a somewhat difficult question. Reid and Liao (2004) and Tanangkingsing and Huang (2007) claim that ma-​verbs are essentially passives in Bontok and Cebuano, respectively, because of their inability to license agents. On the other hand, the ma-​prefix indicates non-​volitionality and is obligatory on certain unaccusative predicates like ‘to fall’,

 

620   Daniel Kaufman b. ku-​pa-​lamme-​ang=ko 1s.erg-​caus-​fall-​appl=2s.abs ‘I dropped a book for you.’

buku book

Although the data on double applicatives is still unclear for Mamuju, other languages of Sulawesi allow for promotion of multiple adjuncts to arguments through applicative stacking. Tukang Besi, for instance, allows the comitative and benefactive applicative combination shown in (70). Among the South Sulawesi languages, Selayarese allows double applicatives when both the theme and recipient of a ditransitive verb are definite, as shown in (71) and Sirk (1996: 82) discusses an identical construction for Bugis. Recall that no Philippine-​type language allows double applicatives of this type (on the analysis of the voice/​nominalization morphology as applicative marking).31 (70) No-​wila-​ngkene-​ako te ina-​no 3.rl-​go-​com-​appl core mother-​3poss ‘She went with Wa Ki’i for her mother.’ (71) Ku-​kiring-​i-​ang=ko 1s.erg-​send-​appl-​appl=2.abs ‘I sent you the money.’

te core

doe’-​injo money-​def

Wa Ki’i. Wa Ki’i (Donohue 1999: 248)

(Basri 1999: 313)

As shown in (72b), applicatives are incompatible with the antipassive. Aldridge (2012b) predicts this behavior via the requirement that applied objects require structural case, which is systematically lacking in antipassives.32 which makes it look quite different from a traditional passive. (See Kaufman 2012 for a discussion of the historical development of ma-​verbs). I thank Laurie Reid for bringing this point to my attention. 31   An apparent counterexample is found in the Cordilleran languages that employ the circumfix i-​ -​an for the benefactive. While these were historically two morphemes (*i-​conveyance and *-​an locative), they have been reanalyzed as one. This is clear from the syntax of benefactives in the relevant languages, in which only one argument, the benefactive, is “promoted” to the nominative/​absolutive. 32  Note that in Bajau, and potentially other Austronesian languages, antipassives can both host definite objects, as seen in (i), and freely allow combinations of actor voice and applicatives, as shown in (ii). The prefix must still be considered antipassive as it only attaches to transitive stems to make an absolutive argument from the agent.

(i)

Nga-​daka’ manu’ iru antipass-​catch chicken that ‘I caught that chicken for my father.’

aku 1s

(ii) Nga-​daka-​an uwa’-​ku manu’ antipass-​catch father-​1s.gen chicken ‘I caught that chicken for my father.’

pugay do.for iru that

uwa’-​ku. father-​1s.gen aku. 1s (Donohue 1996: 789)

Bajau thus represents an intermediate stage between Mamuju and modern Indonesian, where the cognate meng-​prefix has extended its domain to include many monadic verbs as well (e.g. Indonesian menangis /​meN-​tangis/​ ‘cry’, menyala /​meN-​nyala/​ ‘to be on/​alight’).

 

Lexical category and alignment in Austronesian    621 (72)

a. Mam-​baca=aʔ buku antipass-​read=1.abs book ‘I read a book for Husni.’ b. *Mam-​baca-​ang=aʔ    antipass-​read-​appl=1.abs    (For, ‘I read Husni a book.’)

bua for

Husni Husni

buku book

Husni Husni

Antipassives are, however, allowed to combine freely with applicatives when the agent is extracted, as in (73). This is the same condition under which definite patients are possible for actor voice predicates in Philippine-​type languages. (73)

Sema mam-​baca-​ang who antipass-​read-​appL ‘Who read Husni a book?’

Husni Husni

buku? book

The Mamuju data observed thus far suggests that a verb’s underlying argument structure largely determines its morphosyntactic potential. The v category combines with event-​ denoting roots and projects a basic argument structure from the thematic structure. This structure can then be adjusted by transitivizing and detransitivizing morphology. In Philippine-​type languages, to the extent that we can speak of a fixed argument structure, it is created by the nominalizing/​voice morphology rather than inherited from the root. I argued earlier that the simplest account of extraction constraints in Philippine-​type languages unifies the patterns found across event-​denoting and entity-​denoting predicates. In Mamuju, where event-​denoting predicates no longer have any nominal properties at all, we would expect a loosening of these extraction constraints. This expectation is borne out for topicalization, as shown in (74). Here, the ergative argument is extracted to a pre-​verbal topic position, which was shown earlier in (8) to be generally ungrammatical in Tagalog.33 (74)

Baco na-​patei Ali Baco 3.erg-​kill Ali ‘Ali killed Baco.’ OR ‘Baco killed Ali.’

However, relativization and question formation are still restricted in the usual way, as seen in (75), which only has a single interpretation. 33  Edith Aldridge (p.c.) points out that the extraction might be expected if na-​functions as a resumptive pronominal clitic. From a morphological perspective, ergative person marking in Mamuju appears to be standard agreement in that it is obligatory, attaches consistently to the left edge of the verbal stem and can co-​occur with a co-​referential pronominal argument. Nonetheless, it is possible that agreement plays a role in licensing topicalization, especially given that antipassive objects, which do not trigger agreement, are more highly constrained. See Kaufman (2008) for details.

 

622   Daniel Kaufman (75) Sema na-​kita Ali? who 3.erg-​see Ali ‘Who did Ali see?’ NOT ‘Who saw Ali?’ It would seem then that extraction constraints can, historically speaking, outlast the nominal features of the predicate. Mamuju would seem to be an ideal candidate for an ergative analysis along the lines of Aldridge (2004) and would not pose any of the empirical hurdles found with Philippine-​type languages enumerated in (11). The mang-​prefix would be a true antipassive while the formatives me-​, mo-​ and mu-​could be treated as instantiations of intransitive v. Unlike as in Philippine languages, polyvalent roots would give rise to unmarked transitive verbs and the suffixes -​i and -​ang behave like true applicatives attaching to an unmarked transitive verb and, with agent extraction, to an antipassive. Nonetheless, I would argue that the restrictions on extraction in Mamuju obtain the best explanation as historical residue from Philippine-​type morphosyntax. Recall that two potentially independent factors account for extraction restrictions in Philippine-​type languages: the constraints on extraction from NP and the ban on genitive predicates. As shown earlier in detail, Mamuju has developed a strong N/​V distinction and thus constraints on extraction from NP have become irrelevant for ergative extraction. This explains why ergatives can be freely topicalized to the preverbal position as in (74). The second factor, however, remains firmly in place as seen in (76), which shows the only way a would-​be genitive argument can be extracted. The “dummy” predicate ampunna ‘owner’ allows a possessor to be expressed as an absolutive argument. (76) Sema ampunna ku’bur itte di bao di who owner grave that prep down prep Timbu me-​ loda   batu? Timbu av.have-​roof  stone ‘Whose is that grave down in Timbu with the stone roof?’ (Stromme 1991:Maradika Lasalaga) Although there is no overt morphological case marking of DPs in any of the languages of the South Sulawesi subgroup, we find that the interrogative and relative marking elements themselves bear unambiguous traces of absolutive case. The interrogative pronoun sema ‘who,’ which is plausibly derived from Proto-​Austronesian *si-​ima (Blust et al. 2010), contains a reflex of the personal nominative/​absolutive case marker *si-​.34 34 

The presence of Proto-​Malayo-​Polynesian nominative/​absolutive *si can also be found in Indonesian/​Malay siapa ‘who,’ among many other Austronesian languages. See Ross (2006) for a detailed reconstruction of the relevant Austronesian case markers. An alternative source of Mamuju sema is a hypothetical sai-​ma ‘who-​relt.’ Blust et al. (2010) reconstructs *sai ‘who’ for Proto-​Malayo-​Polynesian and ma is found in Bajau and Kambera as a relativizer (although not attested for Mamuju). The nominative case element on sai is not as easy to isolate but given the reconstructions of the nominative singular proper name marker *si and its plural counterpart *sa, it is not far-​fetched to also relate the s(a) of sai to the nominative case function.

 

Lexical category and alignment in Austronesian    623 Crucially, no genitive/​ergative variant (which we would expect to come out as nema) exists in Mamuju (nor in any of the South Sulawesi languages). Recall that genitive marked interrogative pronouns do exist in Philippine-​type languages but can only be used with in-​situ interrogative phrases, as seen earlier in (40), repeated here in (77). (77) Gawà-​∅ nino make-​pat gen.who ‘Who made those shoes?’

ang nom

sapatos shoe

na iyon? lnk that (Schachter and Otanes 1982: 512)

Similarly, the Mamuju relative marker anu is clearly cognate with the Tagalog nominative interrogative ano (underlyingly also /​anu/​). As the interrogatives themselves show morphological traces of absolutive case, it follows that only absolutive arguments can be clefted given case preservation.35 The proper morphological analysis then of interrogative sentences such as (75) is shown in the glossing of (78), where the case marked interrogative constrains the types of arguments that can be extracted. Similarly, the grammatical voice of verbs in relative clauses is constrained by the case of the relativizer, as exemplified in (79). (78) Sema na-​kita Ali? abs.who 3.erg-​see Ali ‘Who did Ali see?’ NOT ‘Who saw Ali?’ (79) Ku-​pe-​ingarangng-​i=mo apa 1.erg-​tr-​remember-​app=emph abs.what sambongi. last.night ‘I remember all that I dreamt last night.’

anu abs.relt

ku-​so’na 1.erg-​dream

(Stromme 1991:Alibe Niso’na)

This analysis, although syntactically trivial, offers a morphologically well-​founded basis for the difference between relativizations and question formation on one hand, which both require case marked operators, and topicalization, which does not. As we have seen, the former operations are just as constrained in Mamuju as they are in Philippine-​ type languages while the latter operation is freer, as constraints against extraction from NP are no longer relevant for event-​denoting predicates.36 35  The morphological argument is less clear with ‘what’ than ‘who’ as nominative ‘what’ can be treated as unmarked. Compare Tagalog ano ‘what,’ ng ano ‘gen what,’ saan ‘obl.what/​where.’ However, the a-​initial in the Mamuju relative marker anu and interrogative apa ‘what’ is found in a wide range of nominative/​ absolutive interrogatives throughout Philippine-​type and Formosan languages. In Tagalog, we can compare the argument interrogatives a-​no, a-​lin ‘which’ and si-​no ‘nom.who’ which all begin with the putative nominative markers a-​or si-​(for proper names) to the adjunct interrogatives ilan ‘how many,’ kailan ‘when,’ bakit ‘why,’ magkano ‘how much’ none of which begin with a-​. Ross (2006) presents the comparative data but tentatively reconstructs the *a-​formative with demonstrative features rather than case features. 36  Note however that this approach requires a view of the grammar that attributes more power to the lexicon than is currently popular in generative theorizing. In particular, a lexical gap in the case paradigm for certain operators must be able to block syntactic movements. This would not be

 

624   Daniel Kaufman Note that this analysis extends beyond Mamuju and can account for a similar pattern in Indonesian, a language which is generally considered to have adopted a nominative-​ accusative alignment pattern (Chung 1976, 2008; Cole et  al. 2008; Aldridge 2008a). As can be seen in (80), actor voice objects can be topicalized but not relativized in Indonesian. (80) a. Orang itu, saya meng-​ajak ke sini. person that 1sg act-​invite to here ‘As for that person, I invited (him/​her) here.’ b. *Orang itu yang saya meng-​ajak  person that relt 1sg act-​invite (For, ‘That person whom I invited here.’)

ke sini. to here (Arka and Manning 2008: 53)

Again, we can plausibly attribute (80b) to the relativizer, which Adelaar (1992) reconstructs as ia-​ng 3s.nom-​lnk. Indeed, ia functions as a strictly nominative case pronoun in modern Indonesian as well. Compare in (81) the distribution of nominative ia with dia, a case neutral pronoun which can function as either subject or object of an actor voice clause (Musgrave 2001). (81) Dia/​ia 3s.neut/​3s.nom ‘He sees him.’

me-​lihat act-​see

dia/​*ia 3s.neut/​3s.nom

Were Indonesian to have in its functional inventory a case neutral relativizer (hypothetically diang), we might expect that extraction could take place from a wider array of syntactic positions. The prediction here is that Austronesian’s well-​known “subjects only” condition on extraction can easily dissipate once the nominal features of event-​denoting predication are lost, as it is then “relic” properties of functional items that keep the restriction in place rather than the strong island characteristics of Philippine-​type predicates. While constraints on ergative extraction are relatively tenacious among Indonesian languages, we do find several ergative languages that allow ergative extraction in a manner unknown among Philippine-​type languages. This is exemplified here by Sumbawa Besar (as described by Shiohara 2013) and Selayarese (as described by Basri and Finer 1987; Basri 1999).37 The transitive verbal clause in Sumbawa employs an unmarked verb countenanced in a framework such as Distributed Morphology where the lexicon is reduced to almost nothing and the putative lexical gap relied upon above would itself have to be derived synchronically in the syntax. On the other hand, it is a perfectly natural analysis in Combinatory Category Grammar (Steedman 2001) where Mamuju anu would be of the category: (N\N)/​(S/​NPABS), that is, a predicate which seeks a clause with a nominative argument gap to its right and a noun to its left to yield another noun via predicate modification. 37 

See also Erlewine, Levin, and van Urk, Chapter 16, this volume, who discuss Balinese pre-​verbal subjects in object voice.

 

Lexical category and alignment in Austronesian    625 stem with prefixal ergative agreement and introduces the ergative argument with the case marker léng, as seen in (82).38 (82)

ka=ku=inóm kawa past=1sg=drink coffee ‘I drank the coffee.’

nan that

léng by

aku 1sg

(Shiohara 2013: 132)

Antipassives are formed with the nasal prefix N-​and subject agreement (which is similar, but not identical to ergative agreement). As shown in (84), antipassives do not allow for objects. (83)

a. ka=ku=ng-​inóm past=1sg=antipass-​drink ‘I drink (something).’ b. *ku=ng-​inóm kawa 1sg=antipass-​drink coffee (For, ‘I will drink (the) coffee.’)

aku. 1sg (nan) (that)

(Shiohara 2013: 132)

Crucially, the ban on antipassive objects (much like Mamuju’s super-​antipassive in (67)) appears to be indefeasible and thus trumps the restriction against extraction of the ergative argument. The result is that agent extraction proceeds from a transitive clause as shown in (84a) in the presence of an object. The antipassive is used in subject questions only when no object is present, as in (84b). (84) a. sai    adè    ka=tumpan’ jangan=ta who  nom past=get fish=this ‘Who caught the fish?’ b. sai adè ka=n-​umpan’? who nom past=antipass-​get ‘Who already had a catch (in fishing)?’

(Shiohara 2013: 135)

Topicalization of the ergative argument to a preverbal position is also possible, as expected, although interestingly, Shiohara shows that the ergative marker must be omitted in this construction, as shown in (85) (see n. 38).

38 

The pattern of a leng-​phrase agent triggering prefixal agreement appears very similar on the surface to that found in Acehnese (Durie 1985; Legate 2012b, 2014b). However, while Sumbawa clearly follows an ergative pattern, Acehnese is typically described as having an active-​stative alignment type which Legate (2012b, 2014b) derives from properties of vP. Shiohara (2013) in fact glosses leng as ‘by’ but because leng seems obligatory on external arguments of transitive verbs, I re-​gloss leng as erg. Shibatani (2008) argues that there is a significant difference between the verb in agent extraction, as in (84a), and a canonical declarative like (82). Space does not allow a more in depth discussion of the complex Sumbawa facts.

 

626   Daniel Kaufman (85) (*leng) aku (ku=)inóm erg 1sg 1sg=drink ‘I drink the coffee.’

kawa coffee

nan. that

(Shiohara 2013: 137)

Selayarese presents a similar state of affairs. The alternation in (86) shows that person marking and alignment are identical to Mamuju in the simple case. A definite object requires a transitive clause with ergative agreement and a second-​position absolutive clitic. An indefinite object must be introduced with an antipassive/​intransitive verb. (86) a. Ku-​halli’=i sapo=ɲjo 1.erg-​buy=3.abs house=def ‘I bought the house.’ b. M-​mali=a sapo intr-​buy=1.abs house ‘I bought a house.’

(Mithun 1991a: 175)

As with Sumbawa and Mamuju, the external argument of both a transitive and intransitive predicate can be topicalized, as shown in (87). (87) a. I Baso’ la-​alle=i pm Baso 3.erg-​take=3.abs ‘Baso took the money.’ b. I Baso’ (a)ng-​alle(=i) pm Baso intr-​take=3.abs ‘Baso took some money.’

doe’=ɲjo money=def doe’ money

(Finer 1994: 159)

Like Sumbawa, but unlike Mamuju, this extends to cases of agent extraction in transitive clauses, as seen in (88), where the verb still takes ergative agreement rather than antipassive/​intransitive morphology. Just like in the dialect of Sumbawa described by Shiohara (2013), this is the only option when the object is definite.39 (88) Inai la-​sumbele=i who 3.erg-​slaughter=3.abs ‘Who slaughtered the buffalo?’

tedong=injo? buffalo=def

(Hasan Basri 2006 p.c.)

39 

Note that the extraction of antipassive objects in Selayarese is still restricted just as in more morphosyntactically conservative languages. Recall that in Indonesian, where the cognate prefix meN-​ marks transitive active voice verbs, topicalization (but not relativization) of the object is permissible, as shown earlier in (80). (i) *Doe (a)ng-​alle=i money intr-​take=3.abs (For, ‘Money, Baso took.’)

i pm

Baso’ Baso (Finer 1994)

 

Lexical category and alignment in Austronesian    627 This also extends to relative clauses, as seen in (89), which are optionally introduced by the relative marker nu. (89) a. asu (nu) n-​datalaʔ dog relt intr-​chase ‘the dog that chased a pig.’

bahi pig

b. asu (nu) la-​datala-​ɲjo=i dog relt 3.erg-​chase-​def=3.abs ‘the dog that chased the pig.’

bahi=ɲjo pig=def (adapted from Basri 1999: 292)

While this diachronic analysis of case marked relativizers and interrogative operators can extend to many Austronesian languages transparently there are inevitably exceptions. Bajau as described by Donohue (1996), for instance, shows the classic extraction restrictions but has a relativizer ma that alternates with ∅. Bajau ma is homophonous with the oblique case marker/​preposition rather than an absolutive/​nominative one and yet it is only the absolutive argument which can be relativized. The converse problem is found in Malagasy, where a focus marker which almost certainly also derives from PMP *anu allows for clefting of prepositional phrases. (Note also that the Selayarese relativizer in (89) derives from anu yet allows relativization of the ergative argument.) (90) T-​amin-​ny antsy no nanapaka ity pst-​with-​det knife foc pst.act.cut this ‘It was with the knife that Sahondra cut this tree.’

hazo tree

ity this

Sahondra Sahondra (Paul 2001)

Clearly, the etymology of relativizers and interrogatives is only suggestive of the solution proposed here. We cannot expect that the features and structures involved in questions and relative clauses will not diverge from their etymological roots. Nor should we expect that the features of a relative marker or interrogative may not be inherited by a lexical item that comes to replace it.

24.3 Conclusion I have argued here that lexical categories play a critical role in accounting for the difference between Philippine-​type language and related ergative languages of Indonesia. In particular, symmetrical voice systems in which every event-​denoting predicate is marked with voice morphology emerge from participant nominalizations. True argument structure with unmarked intransitive and transitive event-​denoting predicates are rooted in verbal categories. Mamuju was presented as a canonical ergative language with a highly developed N/​V contrast. As a result, it displays canonical antipassives and

 

628   Daniel Kaufman applicatives of a kind not found in Philippine-​type languages. When event-​denoting predicates lose their nominal properties and become verbalized, certain syntactic properties of ergative alignment (e.g. extraction constraints) are more likely to be lost as predicted by Manning’s (1996: 21) hypothesis that “syntactic ergativity” results from nominalization, in contrast to more surface-​oriented “morphological ergativity.” Further research in the comparative syntax of Indonesian languages is necessary to better understand the correlates between alignment-​type, lexical category and extractability. The bewildering assortment of agreement and argument marking patterns (see, for instance, Kikusawa (Chapter 23, this volume) as well as the papers in Adelaar and Himmelmann (2005), Wouk and Ross (2002), Arka and Ross (2005) and Adelaar (2013) for a sample) will likely require decades to fully sort out. One of the purposes of this chapter has been to suggest new potential correlations for further investigation as we expand our empirical scope to the many under-​described Austronesian languages of Indonesia and beyond.

Acknowledgements I thank Lisa Travis, Edith Aldridge, Laurie Reid and an anonymous reviewer for extensive comments on this chapter which led to considerable improvements. None of them should be held responsible for the views expressed herein, for which I am alone to blame.

Abbreviations ABS, absolutive; AV, actor voice; ACC, accusative; ACT, actor nominalization, active voice; ANTIPASS, antipassive; APPL, applicative; ART, article; BEG, begun aspect (a component of both the perfective and progressive); CAUS, causative; COM, comitative; CONJ, conjunction; CONV, conveyance nominalization; CORE, core argument; DAT, dative; DEF, definite; DEM, demonstrative; DET, determiner; DIR, directive transitivizer; EMPH, emphatic; GEN, genitive; ERG, ergative; FOC, focus; FUT, future; IMPRF, imperfective INTR, intransitive; LOC, locative nominalization; LNK, linker; NEG, negation; NEUT, neutral case; NOM, nominative; OBL, oblique; PASS, passive; PASS.PART, passive participle; PAT, patient nominalization; PL, plural; PM, personal marker; POSS, possessive; PREP, preposition; RELT, relativizer; RL, realis; STA, stative; SUPERANTIPASS, super anti-passive; TOP, topic; TR, transitive.

 

Aquisition

 

 

Chapter 25

The ac quisi t i on of e rgativit y: An ov e rv i ew Edith L. Bavin

25.1 Introduction An important focus for researchers of child language has been identifying properties that seem to facilitate the acquisition of structures relative to a specific language, structures which indicate events and who is doing what to whom. In order to acquire a language, and so become productive in using it, children extract patterns from the input and generalize to new instances. Crosslinguistic research is important in generalizing the factors that influence how children achieve this. Some research focuses on developmental paths, for example, when children start producing grammatical morphemes, when they produce them in appropriate contexts, and which ones appear early and which late. Other research focuses on cues in the input, for example, morphology and word order that might help in identifying and acquiring grammatical morphology. In this chapter I discuss some of the variation found in ergative alignment and properties that might impact on how readily a child acquires the ergative system and usage patterns of the ambient language. In addition, I summarize data reported for three ergative languages to illustrate language specific features and their influence on the acquisition of the ergative features of the language. Overall, overgeneralizing or undergeneralizing in the use of ergative morphology is not frequent in relation to the acquisition of other morphology (as pointed out by Pye 1990). Based on research from a number of ergative languages representing different language families (see Bavin and Stoll 2013), it is clear that children acquiring ergative languages acquire the system rapidly and by age 2;6 or 3 years have knowledge of the system.1

1 

Note: more information about the acquisition of ergative languages is presented by Austin, Chapter 26, this volume, and Pye and Pfeiler, Chapter 27, this volume.

 

632   Edith l. Bavin

25.2 Alignment and acquisition 25.2.1 Alignment A major distinction across languages is how events are encoded: whether the language uses nominative-​accusative alignment or ergative-​absolutive alignment. Ergativity is typically associated with transitivity. In ergative-​absolutive languages, transitive subjects (A) and intransitive subjects (S) are distinguished; S and objects (O) are treated similarly. In nominative-​accusative languages A and S are treated similarly. In the prototypical transitive clause, an agent fills the subject slot but other semantic roles may do so; in ergative languages the A is identified as ergative. Ergativity is generally shown morphologically rather than syntactically. The distinction between A and S is evident in case marking (as in Hindi; Narasimhan, 2005, 2013), verb agreement (as in Mayan languages; Brown et al. 2013; Pye, Pfeiler, and Mateo Pedro 2013) or both case and verb agreement (as in Basque; Austin 2013). Case marking is more common than verb morphology in marking ergativity. Ergative languages rarely show ergative alignment in their syntactic operations (for examples from Dyirbal, a syntactically ergative language, see Dixon 1979), so that the child learner needs to distinguish two systems: morphological and syntactic. Warlpiri, a Pama Nyungan language spoken in central Australia, is an ergative language in terms of its case marking (see also Laughren, Chapter 39, this volume). Ergative alignment is illustrated in (1) by the case marking on the A arguments), while the O and S arguments are unmarked (absolutive case). Note that the arguments can be interpreted as definite or indefinite and word order is variable. Case marking is used on free-​ standing pronouns, which register person and number, but it is not obligatory on first and second person singular pronouns. Clitics, which appear in second position in the clause, register the person and number of core arguments. However, unlike the case marking system, clitics representing the S and A arguments are treated alike and are distinguished from those representing O arguments. Except for dative third person, the third person singular clitics are null in form (see Hale 1982). Examples (2a and 2b) illustrate pala, the third person dual clitic for A and S. In (2a) the object clitic for the first singular object, ju, is distinguished from rna, the first person singular subject (S and A) form, which is used in (2c) and (2d). (1)

a. Luwa-​rnu wati-​ngki marlu shoot-​PST man-​ERG kangaroo ‘The/​a man shot the/​a kangaroo.’ b. Nyina-​mi ka wati sit-​N.PST IPFV man ‘The/​a man is sitting.’

(2) a. Nya-​nyi ka-​pala-​ju karnta-​jarra-​ngku See-​N.PST IPFV-​3DU.SUBJ-​1SG.OBJ woman-​DU-​ERG ‘Two women are looking at me.’

 

An overview of the acquisition of ergativity    633 b. Nyina-​mi ka-​pala sit-​NPST IPFV-​3DU.SUBJ ‘They (two) are sitting.’ c. Nya-​ngu-​rna-​ngku see-​PST-​1SG.SUBJ-​2SG.SUBJ ‘I saw you.’ d. Nyina-​mi ka-​rna sit-​NPST IPFV-​1SG.SUBJ ‘I am sitting.’ Syntactic ergativity is shown if syntactic operations, such as clause conjoining, follow ergative-​absolutive alignment (see Dixon 1979, 1994). However, in Warlpiri S and A arguments are treated similarly in that the same complementizer, karra, is used to indicate that the main clause S (shown in 3a) or A (shown in 3b) controls the subordinate subject (examples from Hale 1983), whereas for a main clause object as controller of the subordinate subject, the complementizer is kurra. (3) a. Karnta ka-​ju wanka-​mi yarla karla-​nja-​karra Woman IPFV-​1SG.DAT speak-​N.PST yam dig-​INF-​COMP ‘The woman is speaking to me while digging yams.’ b. Ngarrka-​ngku ka purlapa yunpa-​rni karli jarnti-​rninja-​karra-​rlu man-​ERG IPFV corroboree sing-​N.PST boomerang trim-​INF-​COMP-​ERG ‘The man is singing a corroboree while trimming the boomerang.’

25.2.2 Semantic Bootstrapping Pinker’s (1984, 1989) bootstrapping theory presents a possible explanation for how syntactic roles are acquired. Children are assumed to draw on basic (innate) knowledge of syntactic categories, semantic roles and linking rules (the expected links between semantic and syntactic functions). Initially, children would link an agent thematic role to the subject of a verb. Once that link is established they would extend subjecthood to include other thematic roles. That is, semantics is assumed to provide an entry into syntax. If the role of agent helps bootstrap children into acquiring syntactic structures, a prediction would be that children acquiring a language with ergative case marking would identify prototypical agents first and then extend the ergative marking to other semantic roles, irrespective of the verb’s transitivity, and so extend the case marking to S arguments (see also Pye 1990; Van Valin 1992; Siegel 2000; Van Valan 1992). Pinker did not perceive ergative languages as posing a problem; his stated view (1984: 372) was that children would either notice similarities in the encoding of the agent (A) in a transitive clause and actor (S) in an intransitive clause (nominative-​accusative languages), or similarities for encoding the intransitive actor (S) and transitive patient

 

634   Edith l. Bavin (O) (ergative–absolutive languages), suggesting that children would need to distinguish verb type and associated morphological marking for the arguments for each, that is identify distributional patterns in the input using the available language specific cues. This, however, simplifies the problem since the use of ergative alignment varies across and within languages. Ergative languages typically have split systems, with ergative-​ absolutive alignment used in some contexts but not all, and the use of ergative morphology may be variable in languages, (as discussed in section 25.3). If the input to young children includes a large proportion of highly transitive clauses, it might be expected that children would use a high proportion of transitive verbs in their early utterances. So rather than assuming that if children first use ergative marking with highly transitive verbs they are drawing on a link between agent and ergative marker before extending the marker to other A  arguments (as predicted from the semantic bootstrapping approach), an alternative explanation is that the input favours such verbs and children are acquiring the patterns they hear in the input. Children need to identify the specific linguistic contexts, and social contexts, for using case morphology or verb agreement associated with ergative-​absolutive alignment on the basis of patterns in the input. Rumsey, San Roque and Schieffelin (2013) reported on the acquisition of ergative marking in three Trans New Guinea languages. In conversations and monologues from Duna, one of the languages studied, ergative marking was used on only 54% of the A nominals; in Ku Waru, a related language, approximately 62% of A arguments in a sample of adult speech were marked with ergative marking. Clearly, when there is variability in whether an A argument is marked with ergative morphology or not, the opportunity is reduced for children to determine the contexts where ergative marking is applicable. In another study, on the acquisition of Samoan, Ochs (1982) found that the young children studied used ergative marking in less than 5% of obligatory contexts. The use of ergative case marking in Samoan is constrained by social factors, the social distance between speaker and listeners, and so is rarely used when communicating with family members. Thus the young children would have restricted opportunities to hear ergative marking and identify the conditions for its use.

25.3 Variability, Pragmatic Functions and Multifunctionality 25.3.1 Split Systems and Noncanonical Use of Ergative Marking A general problem in determining where ergative morphology applies is that ergative alignment does not generally apply overall in a language. The prototypical pattern for ergative alignment is for a single argument of a monovalent verb to be identified as absolutive and the primary argument of a bivalent verb to be identified as ergative. However, ergative marking varies within languages and split systems are common. Comrie (2013c) discusses some of the variability in the encoding of ergativity, questioning whether it represents syntactic or semantic alignment. Ergative morphology may also be found for the single argument of a monovalent verb in some languages.

 

An overview of the acquisition of ergativity    635 Split alignment systems are conditioned by various factors. For example, nouns and demonstratives may show ergative-​absolutive alignment while pronouns do not. Splits may also be restricted to certain pronouns based on person or number. In one such language, Arctic Quebec Inuktitut (Allen 2013), the split system is based on number; singular ergative and absolutive case forms on nouns and demonstratives are distinguished, but not the plural forms. Verb aspect is the basis for the split system in Hindi; ergative alignment is applied in clauses with verbs in perfective aspect. In Trans New Guinea languages (Papua New Guinea), ergative case marking depends on word order; it is used in OAV word order, but in AOV the A is marked ergative only if both the A and O arguments are proper names or kin terms. The specific contexts for the application of ergative marking need to be identified based on language specific distribution patterns. Split systems would be problematic if a child assumes ergative marking is associated with any A argument of a bivalent verb. Another potential problem for acquisition is that for some languages ergative marking is extended in specific contexts to S arguments, as in the Mayan languages discussed by Pye, Pfeifer and Mateo Pedro (2013). In Nam, the infrequent extension of ergativity to S arguments is restricted to complex clauses. In contrast, the extension of ergativity to S arguments in Yukatec is frequent, occurring after a progressive verb. In some contexts, ergative morphology can also crossreference the O argument of a transitive verb (see Pye et al. 2013 for details). There is no strong evidence that split ergative alignment systems within a language lead to delayed acquisition. For example, Narasimhan (2013) reports that children do not extend the ergative case to agents of transitive clauses with non-​perfective aspect marking. Nor are children found to extend ergative marking generally to S arguments if this is not a feature of the input language. The acquisition data support the view that children rely on distributional evidence in the input to identify where ergative marking applies, and its acquisition is relatively error free.

25.3.2 Pragmatic Functions Although the term ‘ergative language’ is typically used in reference to the grammatical or semantic conditioning of alignment patterns, ergative marking often serves pragmatic functions, indicating, for example, focus, contrast, or individuation. McGregor (2010) illustrates that there are two relevant features of the S (agentive and referential) in using ergative marking in intransitive clauses in Warrwa, a Nyulnyulan language of North West Australia. The referential use indicates that an ergative marked argument is unexpected in the particular context. In Hindi, the S argument of a small number of intransitive verbs, including verbs of ‘bodily emission’ and others (e.g. chĩĩk ‘sneeze’, ro ‘cry’, cillaa ‘shout’, nahaa ‘bathe’) can be marked as ergative in perfective contexts and when the S is viewed as being in control, that is, in volitional contexts (Narasimhan 2013). DeLancey (2011) discusses noncanonical ergativity in Tibeto-​Burman languages; the case marker is optional on A arguments and in specific contexts is used on some S arguments (also see LaPolla 1995a). He argues that while a language may show a split ergative pattern in elicited forms, in natural discourse, ergative marking is found only in some clauses, usually with pragmatic functions -​indicating emphasis or contrast.

 

636   Edith l. Bavin

25.3.3 Multifunctionality In addition to canonical and pragmatic functions, ergative case morphology may have other functions. It is frequently used for instrumental case, as in Darma (Sino Tibetan) and Warlpiri, as illustrated in (4). In Chintang (Sino-​Tibetan; Stoll and Bickel 2013) it marks instrumental case and also cause and source. In Kaluli, the ergative case marking is also used for instrumental case as well as genitive case, and in Mayan languages ergative morphology is also used on nominal possessors (Brown, Pfeiler, de León and Pye 2013). Marking nominals associated with the A argument in Warlpiri is another function for Warlpiri ergative case forms, as shown in (5b) in which wita ‘small’ is associated with wirriya ‘boy’. Even if the A argument is not overt in the clause, as in (5c), associated words will be marked as ergative. In this example, the ergative marker on the locative marked pirli ‘rock/​hill’ indicates that the A (ellipsed in this clause) was at that location. (4) Paka-​rnu kurlarda-​rlu hit-​PST spear-​INS ‘(Someone) hit (something) with a spear.’ (5) a. Wirriya nya-​ngu wita-​ngku boy see-​PST small-​ERG ‘The small one saw the boy.’ b. Wita-​ngku nya-​ngu wirriya-​rlu small-​ERG see-​PST boy-​ERG ‘The small boy saw something/​someone’ c. Marlu pantu-​rnu pirli-​ngka-​rlu kangaroo spear-​PST rock/​hill-​LOC-​ERG ‘(Someone) on a rock/​hill speared the kangaroo.’ In the data from ergative languages reported to date, instrument marking tends to be less frequent in utterances to very young children than ergative case, which means exposure to the instrumental function of ergative forms is reduced, and it is typically not reported to be observed in children’s speech until after it has been used on an A argument.

25.4  Determining when Ergative Marking is Acquired 25.4.1 What Counts as Acquisition? That children have acquired ergative morphology is evident if they map thematic roles onto syntactic roles using the language specific morphology or word order and in the

 

An overview of the acquisition of ergativity    637 expected contexts. Following Roger Brown (1973), a measure that is frequently used to determine acquisition is the percentage of use in obligatory contexts. Brown chose 90% as indicating acquisition in research on the acquisition of grammatical morphemes in English. When a single ergative form appears, or just a few forms appear, it is not strong evidence that the child has identified the function of ergative marking since the word + case marking could be extracted from the input as a whole unit. However, if the child also used the nominal without the case marker in an S or O role, or if the child produced an appropriate ergative morpheme with a number of different lexical items it does indicate some knowledge of its function, although not necessarily all. When a child marks ergative case but does not use the appropriate allomorph, it is indicative that the child has identified the function, but the distribution of the particular forms is not yet mastered. Although children may use ergative morphology in similar proportions to the speakers who provide input, suggesting mastery of ergative alignment, they may in their early productions use only a limited set of verbs, as reported by Stoll and Bickel (2013) in their study of the acquisition of ergative marking in Chintang. A  distribution of ergative markers with a larger set of verbs would provide stronger evidence of generalization, that is, acquisition. In a language that allows ellipsis of core arguments and a single unmarked nominal is included in a child’s utterance, there may be some uncertainty as to whether a child is failing to use an ergative case form on an A argument or appropriately representing an O argument since absolutive case typically has null marking. The potential ambiguity might be resolved from the context, but not necessarily. However, accounts of the acquisition of both absolutive and ergative morphology for languages that use overt absolutive forms in cross-​referencing arguments on verbs are available for Mayan languages (Brown et al. 2013) and Basque (Austin 2013).

25.4.2 Sampling Child Language Much of the research on the acquisition of ergative language is based on naturalistic data –​children are recorded in daily activities interacting with family members and others. One advantage of this type of data collection is that language input to the children is also available. This provides information about how others use language in the presence of children and which structures the children hear, rather than relying on a description of the mature system. Data from the same children collected over a period of time allows researchers to identify individual developmental patterns. A major problem in collection longitudinal data, however, is that the resources required are not readily available and when such research is conducted, data samples tend to be small, collected from a few children over a limited period. Then there may be few examples of ergative marking in the children’s productions, particularly if the language has frequent ellipsis of core arguments and the proportion of ergative marked nominal in the input is low. With just a few examples the researcher is limited in being able to draw conclusions about acquisition.

 

638   Edith l. Bavin Stoll and Bickel (2013) were able to collect a large data base in their research on Chintang, so enabling a meaningful comparison of the ratio of ergative use for adults and children. The ratios were found to be similar, as was the distribution in terms of agent versus other functions. However, some differences were evident in usage; for children ergative use was more item specific -​with fewer lexical items. Comparisons such as made by Stoll and Bickel are not reliable with only a few examples and since there is variability in language development between individual children, small data sets limit the generalizations that can be made. Pooling comparable data can help overcome the small sample issue. Alternatively, cross sectional data from children of different ages can provide an overview of developmental stages over a longer time period. Individuals do not produce all they know, which justifies the use of elicited data, including narratives, in acquisition research. As well as narratives from adults, they have been collected from children beyond the early stages of acquiring a language—​children who are still mastering linguistic properties of the language. Narratives are useful for documenting how reference is maintained in extended discourse, and identifying contexts where argument ellipsis is less likely or more likely to occur (see, for example, Allen 2000; Bavin 2000) and in identifying pragmatic contexts from ergative marking. Comparing acquisition patterns across closely related languages with structural similarities and differences and similar cultural norms allows the researcher more opportunity to focus on whether differences between the languages account for different development patterns. For example, Brown et al. (2013) report on the emergence of ergative marking for children acquiring four Mayan languages. The languages have many similarities -​for example they use verbal inflection to cross-​reference the ergative argument. In all languages the verb morphology is complex. While there are similarities there are also differences, including the form of the pre-​consonant ergative marker and its position in relation to other verb morphology. For two of the languages, the non-​syllabic ergative marking used with consonant initial roots was found to be produced later than the ergative marker in the other two languages. As discussed by the authors, the non-​syllabic form makes it less salient but, when they are acquired, aspect inflections form a syllable with the ergative marker so making it easier to perceive. Acquiring ergative morphology is not done in isolation; the data show that the acquisition of ergative forms may be combined with or influenced by other morphology in the language. Brown et al. found other language differences in the acquisition of case morphology and suggest these might result from difficulties in processing adjacent affixes, but experimental data rather than naturalistic data would be required to confirm this.

25.4.3 The Social Context While comparing research findings across typologically different languages is important for understanding the factors that influence language acquisition, it can be difficult

 

An overview of the acquisition of ergativity    639 because of different socialization practices and cultural expectations of the child. It is important to document the culturally specific input and interaction routines between caregivers and children. For any language, social context will influence how much opportunity there is for children to hear, process, and use specific language features. Different types of verbs are used with different activities, for example (see Hoff and Naigles 2002) and this will impact on which case markings are used. Similarly, the use of pronouns in conversations is influenced by the number of participants involved. The linguistic forms used by others in the presence of the child influence the patterns detected and acquired. Communication between child and caregiver(s) is often predominantly biased to first and second person agents. In a language system in which ergative case marking is not marked on these pronouns, there will be reduced opportunity for children to hear ergative case morphology. As an example, in Arctic Quebec Inuktitut, first and second person pronouns are only indicated by verb inflection, not free standing pronouns (Allen 2013), and so ergative case marking will not be available for the child to hear when these pronouns fill the argument slots. Schieffelin (1985) argued that the acquisition of ergative marking on focused pronouns in Kaluli is facilitated by pragmatic scaffolding. Through interactions which socialize children to appropriate behavior exchanges, such as giving and sharing, adults model appropriate communication which clearly distinguish the speaker and listener. In these modelling routines children are exposed to the obligatory contexts for ergative marking with focused pronouns. In Ku Waru, through modeling of appropriate language by adults in multi-​turn exchanges with young children and correction when the children omit required ergative markers (Rumsey, San Roque, and Schieffelin), children are exposed to appropriate ergative marking

25.5  Developmental Patterns for Three Ergative Languages: Kaluli, Arctic Quebec Inuktitut, and Warlpiri 25.5.1 Kaluli Kaluli has a split ergative system, dependent on word order and pragmatic factors (Schieffelin 1981, 1985; Rumsey, San Roque and Schieffelin 2013). Neutral (absolutive) case is used with AOV word order unless both the A and O are proper names or kin terms; then the A argument is marked with ergative case. Ergative marking is obligatory on nouns, deictics, and demonstrative A arguments in (O)AV word order. Ellipsis of a noun argument from three constituent clauses is common. A focused set of pronouns is used for ergative arguments of transitive verbs but in AV clauses ergative marking is not required; it is used where new information is provided by the A, as in responses to ‘who’ questions.

 

640   Edith l. Bavin Ergative marking used with genitive function appeared first in the naturalistic data from the three children studied, aged approximately two years at the start of data collection. Inconsistent marking for A arguments in AV clauses followed; later ergative marking was used on OAV clauses. Some errors were noted from one child. There were some overgeneralizations of A marking to AOV word order, but by age 2;8 the children were sensitive to the word order constraints of ergative marking (Scheiffelin 1985) although for AV clauses the children continued to omit the case marking in contexts where it was appropriate to use it. A trend for using ergative marking in AV order in clauses with past tense, highly transitive, and positive rather than negated verbs was noted. For focused pronouns, no errors were found for OAV clauses and no overgeneralizations were made by using focused pronouns in AOV order. Possible explanations provided by Schieffelin for the apparent ease of acquiring appropriate ergative marking with focused pronoun arguments compared to non-​pronominal arguments were: they appear consistently in preverbal position; they are independent, unlike case markers; and they have one function. In addition, first and second person arguments are frequent in interactions with young children, in which conversations are about the here and now and focus on the speaker and listener.

25.5.2 Arctic Quebec Inuktitut Ergative is associated with transitive constructions but ‘detransitivization’ processes can change the role of the arguments for clauses containing a bivalent verb, as in Arctic Quebec Inuktitut. Allen (2013) discusses two main influencing factors influencing the use of ergative marking by children acquiring the language: ellipsis of core arguments and detransitivizing processes. Another possible influencing factor is that ergative case marking is distinguished from absolutive case only in singular forms for nouns and demonstratives, not dual or plural, so the split system could potentially lead to overgeneralisation in the use of ergative case markers. Detransitivizing reduces the opportunities to use ergative case marking. Allen reports naturalistic data from four children aged between 2;0 and 2;10 at the start of the study, which continued over a period of nine months. Only 16 examples of transitive constructions were identified; of these, only seven had overt third person subjects. First and second person A arguments (which do not require ergative case marking) were more common than third person; third person A arguments were typically ellipsed. However, ergative case marking was present when required in the child productions. In contrast, intransitive clauses numbered 197, reflecting the high frequency of detransitivized constructions for bivalent verbs: passive, noun incorporation and antipassive constructions. In antipassives and noun incorporation constructions the agent takes on the S role; in passive constructions the O takes on the S role and the agent is treated as an adjunct. As with the child data, the input showed frequent use of detransitivized constructions with noun and demonstrative agents.

 

An overview of the acquisition of ergativity    641 In narratives elicited from older children (8–​9 and 15–​16 years) and adults, ergative case was present when required, but as with the naturalistic data, obligatory contexts were infrequent. Third person A arguments were more frequent in intransitive (detransitivized) constructions than transitive. The data from Inuktitut indicates that frequency of ergative case markers is not necessarily a requirement for acquisition. Frequency has not been found to be predictive of ergative use in other languages also. For example, Pye et al. (2013) found that frequency did not predict acquisition of extended ergativity in 2–​3-​year-​ old children. If ergative marking is used infrequently in the input, it may focus children’s attention to the contexts when it is used; this point was made by Stoll and Bickel (2013) in relation to their findings on the acquisition of ergative marking in Chintang. They argue that since argument ellipsis is frequent in that language, overt noun phrases will be salient.

25.5.3 Warlpiri 25.5.3.1 Some Features of the Language Warlpiri has variable word order and there is extensive ellipsis of core arguments, and thus case morphology is not always available to aid the child in identifying the alignment patterns of the language. However, as indicated previously, this may make ergative marking more salient when ergative marked nouns do appear. There are a number of factors that could potentially influence how readily children are able to acquire the distribution patterns for ergative marking. Since the cross-​referencing of core arguments with clitics follows a nominative-​accusative pattern, as shown in the examples in (2), they will not be of value in determining the case frame of the verb. In addition, there are allomorphs for ergative case and children need to identify that these forms mark the same function but in different environments. The allomorphs children heard in the input at the time of data collection were ngku, nkgi, rlu, and rli, with vowel harmony conditioning the use of rli or rlu and ngku or ngki. Words longer than two syllables had rlu/​rli and nkgu or ngki was used on two syllable words. Examples are wati-​ngki ‘man-​ ERG’, karnta-​ngku ‘woman-​ERG’, maliki-​rli ‘dog-​ERG’, and wirriya-​rlu ‘boy-​ERG’. Another potential influencing factor is multifunctionality. The forms of the ergative case marking are also used for instrumental case, as well as agreement between an A argument and nominal associated with it, as illustrated in example (4). There are also homophones for ngku and rli. The ergative form ngku is homophonous with the second person singular cross-​referencing clitic, which is illustrated in example (2c) and rli is the first person inclusive dual subject clitic. In conversations and stories there is lot of repetition, using different word orders. In this style of communication some of the information is repeated and more information is added. In the repetitions core arguments will sometimes be overt and sometimes not, which may help focus children’s attention on how the arguments, when they are overt, are marked when included with particular verbs. It also exposes them to the variability of including or not including core arguments.

 

642   Edith l. Bavin A feature of the language that might assist children in learning which verbs require an A argument, is the small number of verb roots. These are divided into five verb classes based on verb inflections for tense, mood, and finiteness. There is high correlation between valency and verb class; almost all verbs in class one require a single S argument (e.g. nyinami ‘sit’), and most in class 2 have an ergative-​absolutive case frame (e.g. luwarni ‘shoot’). The small number of verbs included in classes 3, 4, and 5 include bivalent and univalent verbs. A verb has a specified case frame, with the ergative-​absolutive case frame used for prototypical transitive verbs such as pakarni ‘hit’ but others also (e.g. nyanyi ‘see’. However, a verb’s case frame can be modified, resulting in a change of meaning. If the absolutive object of a bivalent verb is replaced by a dative object, the verb is interpreted as attempted action. In addition, a number of preverbs are used in the language and these modify the verb’s meaning and can also modify the case frame (see Nash 1982).

25.5.3.2 The Data Naturalistic cross sectional data  Bavin collected naturalistic data from young

Warlpiri children interacting with at least one other child and adults close by, and narrative data from children and adults (previously discussed in Bavin 1992, 2000, 2004, 2013). The naturalistic data showed no use of verbs before the age of 2;0. At age 2;0 locatives, deictics, nouns and some pronouns were frequent. From age 2 to 3 years, there was great variability in how many verbs were produced but the majority were imperatives. More verbs were used overall from age three, and these included a proportion of verbs with ergative-​absolutive case frames. A arguments of ergative verbs were not overt until around age three. From the data from the 3-​year-​olds (N=9) the contexts for use of ergative marking ranged from 0 to ​9. With the exception of one young 3-year-old who used no verbs, the number of verb tokens ranged from 10 to 59; for ergative-​absolutive case frame verbs the range was a 6–​30. That is, the majority of A arguments were not overt, either because the verb form was imperative or the argument was ellipsed. In all obligatory contexts, an ergative case marker was included. For the 4-​year-​olds (N=6) there were 18–​56 verb tokens (15–​33 ergative-​absolutive verbs) and the contexts for ergative marking on A arguments ranged from zero (for three of the children) to four, and there were also two instrumental contexts and one context for ergative agreement–​marking on a nominal associated with a non overt A (watiya-​ngku ‘with a stick’). All had ergative marking. The verbs in these clauses varied. They included pakarni ‘hit’, nyanyi ‘see’, ngarni ‘eat’ warrirni ‘look for’, kanyi ‘carry’, and others. Ergative case marking was not restricted to past tense. The child who used ergative forms for instrumental and agreement functions at age 4 had used no ergative forms in an earlier session at age 2;1. However, age is not a reliable way of predicting how much a child will use ergative case forms. One child produced nine ergative forms at age 3;6 but none six months later. Even though 20 ergative-​absolutive verbs were included in her utterances, there were no contexts for ergative case to be used. Although children used ergative case when required from about age 3, the allomorph used was not always appropriate. Vowel harmony caused no confusion but length of word as a conditioning factor for which allomorph to use did, for example, ngku was

 

An overview of the acquisition of ergativity    643 used on words longer than two syllables, even though previously a word had been used with the appropriate rlu case form. It could be that ngku/​ngki are heard more in the early input (e.g. on wati ‘man’ ngati ‘mother’ and kurdu ‘child’) and used as the default. In elicited narratives, discussed in the next section, case forms were found to be confused sometimes also by older children.

Elicited narratives  Narratives were collected using picture books. The first of two

books comprised eight pictures drawn to show people going hunting, shooting a kangaroo, coming home with the animal, and cooking it. In her narrative, one child (aged 5;2) used six ergative-​absolutive verbs; of these five had overt A arguments and all six had an overt object. However, a second child, aged 5;0, used eight verbs but no overt core arguments. Another child aged 6;4 used 12 verbs, including six ergative-​absolutive verbs, but only one had an overt A argument, with the eleventh verb she used. It was at a stage in the story when the hunters had returned home with a kangaroo and a woman was cooking the animal on the fire. So there was a new subject (agent), which may have motivated the inclusion of the argument. Other children, 6-​and 7-​year-​olds, also tended to use an overt A with ergative marking for this particular scene although they had not included an A argument elsewhere. Possibly because it was a highlight of the story, the overt A was used as a way of focusing on the event Book 2, with 12 coloured pictures, was part of a published book. Narratives were collected from 28 children aged 4; 8–​124 and six adults, parents of some of the children. The six adults used ergative forms (numbering between 4 and 12) with some of these used for instrument and agreement functions. There was a great deal of variation across speakers as to whether and where A arguments were included, from zero instances of seven possible contexts from a 5-​year-​old to 11 in 12 possible contexts for a 23-​year-​ old. All of the adults included some ergative arguments with appropriate case marking, and the older children used more than the younger children. From 9 years up all but one child used some ergative case markers; below that age, eight of 14 children used at least one. The number of verb tokens included in the narratives ranged from 5 to 41, so some stories were lengthy, even from the younger children. When ergative arguments were included, ergative case marking was used. A arguments were used sometimes when new characters appeared, and sometimes, it seems, to highlight part of the story; and possibly for emphasis or contrast. The data indicate that pragmatic factors need to be taken into account in documenting when children do include ergative arguments when argument ellipsis is common in the language.

25.6 Conclusion Available crosslinguistic research findings show that children with typical development become attuned to features of the input language from an early age. This is evident from

 

644   Edith l. Bavin data in a number of collected volumes (e.g. Slobin 1985, 1992, 1997; Berman and Slobin 1994; Bowerman and Brown 2008), and journal articles. Ergativity is not represented in the same way across all ergative languages and the acquisition of morphology associated with it, as with the acquisition of other case morphology and verb inflections, is influenced by a number of factors. However, overall, reports on children’s acquisition of ergative languages show clear evidence that by age 3;0 years, often earlier, children are attuned to patterns of ergative marking in the input language. In summary, overgeneralizations are not frequent, although if there are different allomorphs to indicate ergativity some overgeneralizing of one form may occur. Ergative marking may have other functions, but this does not appear to significantly delay acquisition. Ergative markings may also be homophonous with other morphemes, but children seem to keep them separate based on distribution. Frequency of specific constructions in the input can facilitate the detection and acquisition of usage patterns but low frequency of ergative marking in the input, the result of ellipsis or the use of detranisitivized constructions, does not necessarily delay the acquisition of ergative features of a language. Children do not seem to be significantly hindered by having to acquire two systems—​one morphological and one syntactic—​nor do split morphological systems and variable use of ergative marking seem to be problematic overall, although they may impact on developmental patterns, as shown in Kaluli where ergative marking depends on word order and is not always obligatory. More child language research on ergative languages will provide opportunities for greater understanding about the impact of the type of ergative marking, its position, the conditions for use, variability in use and input styles. One aspect that needs to be considered is that many ergative languages are undergoing rapid change. For example, in Kurmanji Kurdish (Mahalingappa 2013) a shift in the case marking system has resulted in inconsistent input to children. However, by age 2;6 children are reported to use the system in a similar way to adults. Language change occurs in contact situations; an interesting example of this is discussed by Meakins (2015). Gurindji Kriol, an optional ergative language, has emerged from contact between a Pama Nyungan ergative language, Gurindji, and Kriol, a nominative-​accusative language. The Gurindji ergative marker merged with the verb system of Kriol, along with other case markers, and its function underwent a series of changes from ergative marker to optional nominative marker. Children are a medium for change and acquisition research of ergative languages in contact and bilingual situations can add to our understanding of factors that influence changes in alignment and changes in the functions of ergative case marking.

 

An overview of the acquisition of ergativity    645

Abbreviations 1, first person; 2, second person; A, ergative-​ like subject of a transitive verb; COMP, Complementizer; DAT, dative; DU, Dual; ERG, ergative; INF, Infinitive; INS, instrumental; NPST, nonpast; OBJ, object; PST, past; SUBJ, Subject; S, Intransitive subject; SG, Singular; V, verb. In the chapter O is used for object (when referring to transitive versus intransitive sentences and Subject Verb Object word order).

 

Chapter 26

The role of de fau lts in the ac qui si t i on of Basque erg at i v e a nd dative morph ol o g y Jennifer Austin

The use of default agreement plays a key role in morphological theories from diverse perspectives, as well as in many analyses of child language acquisition. In this chapter, the development of ergative and dative agreement and case in 20 bilingual and 11 monolingual Basque-​speaking children between 2;00–​3;06 years old is examined. I propose that the most commonly produced errors in child Basque involve the substitution of unmarked absolutive forms for ergative and dative case and dative verbal morphemes; for independent reasons, the absolutive is considered to be unmarked inflection in adult Basque (Arregi and Nevins 2012). These errors suggest that in early stages of morphological acquisition, children learning Basque use default forms which encode a subset of the morphemes as a “best match” to support their developing language when they are unable to produce or retrieve target forms.

26.1 Introduction The notion of a default, or a linguistic form which surfaces as the result of a general rule in cases where a more specific rule cannot apply, has a long history in linguistic analysis. In Kiparsky’s (1973) analysis of disjunctive ordering in phonological rules, he termed this phenomenon the Elsewhere Condition. In many theoretical approaches to morphology, default inflection also plays a critical role. For example, in the Distributed Morphology (DM) framework, vocabulary items are underspecified and compete with

 

Acquisition of Basque ergative and dative morphology    647 each other for insertion. Default (elsewhere) morphological forms are selected when there is no exact match between morphosyntactic features and phonological forms (Harley & Noyer 1999). Defaults also figure prominently in the Network Morphology Framework (Brown & Hippisley 2012), which distinguishes between two types of morphological defaults: normal case defaults, which generally apply in the absence of a more specific rule, and exceptional case defaults, which are lexically specified and used to revert back to the default form from a specific rule which otherwise applies (Aronoff 2013). In several theoretical models of child morphological development, default inflection is used as a mechanism for building of morphological paradigms in language development rather than serving as elsewhere forms. According to Bybee (1985, 2007) and Albright (2002), inflectional paradigms usually have a default or base form that children use for creating other parts of the paradigm. While Bybee proposed that this basic or default form is the least inflected member of the paradigm, Albright suggested that the morphological form with the greatest information is selected to construct the rest of the paradigm. In a similar vein, researchers using a constructivist approach view children’s paradigm building as a means to organize and advance the learning of morphology in language acquisition (Dressler & Karpf 1995; Aguirre 2003, 2006; Bittner, Dressler, & Kilani-Schoch 2003; Dressler 2005). In this chapter I will argue that the use of default forms can serve another purpose in child grammars, namely as a kind of repair strategy while children are still in the process of acquiring inflectional morphology. I develop the analysis introduced in Austin (2009, 2010) and propose that children’s inflectional errors in Basque are default forms selected by the Subset Principle (Halle, 1997) as the closest match when a target form cannot be retrieved or produced. My claim is that this approach provides an account of patterns seen in the development of both nominal and verbal inflection. I will begin by presenting the morphological characteristics of Basque that are germane to this chapter as well as evidence for the use of default morphology in adult Basque speakers (Arregi & Nevis, 2012). This overview will be followed by a discussion of data from previous studies of child Basque and the details of my proposal regarding children’s use of default inflection in acquiring Basque.

26.2  Basque Verbal Morphology and Case Basque is a language with an ergative case and agreement system. In Basque, there are distinct agreement morphemes for the subjects of transitive verbs (ergative agreement and case), and for the subjects of intransitive verbs and direct objects (absolutive agreement and case). This pattern is shown in sentences in (1a–​b). While the subject of a

 

648   Jennifer Austin transitive verb such as entzun ‘to listen’ in (1a) is marked with the ergative morpheme –​k, the direct object receives absolutive case, which is zero-​marked (Ø). In contrast, the subject of an unaccusative verb such as ibili ‘to walk’ in example (1b) has the same absolutive case marker as the direct object of a transitive verb (Ø): (1)

a. Transitive verb: Ni-​k berri-​a-​Ø entzun du-​t I-​ERG news-​DET-​ABS hear   AUX.ABS.3SG-​ERG.1SG ‘I have heard the news’ b. Unaccusative verb: Ni-​Ø ibili naiz I-​ABS walk AUX.ABS.1SG ‘I have walked’

However, in Basque the subjects of unergative verbs receive ergative and not absolutive case, as seen in example (2). (2) Unergative verb:

Ni-​k korritu du-​t I-​ERG run AUX.ABS.3SG-​ERG1SG ‘I have run’

Verbs in Basque are obligatorily inflected for person, number, gender, tense, aspect, and mood (gender is only marked for second person informal forms with the ergative and dative morphemes). In a periphrastic verb, the participle is inflected for aspect, while the auxiliary is marked with the morphemes for person, number, and tense: (3) liburu asko ema-​ten d-​izki-​o-​t book many give-​IMP AUX-​ABS3PL-​DAT3SG-​ERG1SG ‘I (often) give him/​her many books’ The order of the morphemes on a finite Basque verbs (present tense) can be seen in (4): (4)

Absolutive-​Root-​Dative-​Modal-​Ergative-​Tense

(Laka 1988)

Basque verbal agreement distinguishes between three persons, as well as singular and plural number, as illustrated in Table 26.1 (adapted from Ezeizabarrena 1996: 45).

 

Acquisition of Basque ergative and dative morphology    649 Table 26.1 Verbal person markers in Basque (present tense)

Prefixes for absolutive agreement (unaccusative argument structure)

Suffixes for dative agreement (dative experiencer argument structure) and ergative agreement (monotransitive argument structure)

1st singular

n-​ Example: naiz ‘ I (ABS) am’

-​t Examples: zait ‘To me (DAT)’ dut ‘I (ERG) have’

2nd singular (familiar)

h-​ Example: haiz ‘you are’

-​k(masculine) -​n (feminine) Examples: zaik, zain ‘to you (DAT)’ duk, dun ‘you (ERG) have’

3rd singulara

d-​ Example: da ‘he/​she/​it is’

-​o Example: zaio ‘to him/​her (DAT)’ -​Æ Example: du ‘he/​she/​it (ERG) has’

1st plural

g-​ Example: gara ‘we are’

-​gu Examples: zaigu ‘to us (DAT)” dugu ‘we (ERG) have”

2nd plural

z-​ + -​te (plural marker) Example:

-​zue Examples: zaizue ‘to you pl. (DAT)’ duzue ‘you pl. have (ERG)’

zarete ‘you(pl) are’ 3rd plural

d-​ Example: dira ‘they are’

-​e Example: zaie ‘to them (DAT)’ Example: -​te dute ‘they have (ERG)’

a Some authors consider -​d the realization of tense or mood, rather than a third person agreement

morpheme. See Laka (1988) and discussion in Ezeizabarrena (1996).

Number agreement for plural direct objects takes the form of an infix that is inserted into the auxiliary root, as seen in (5), where the morpheme -​it is inserted between the

 

650   Jennifer Austin auxiliary root prefix d-​and before the root vowel u and the final first person ergative marker -​t. (5) singular: plural

du-​t AUX.Erg1SG ‘I have it’ d-​it-​u-​t AUX.ABS3pl.AUX.ERG1SG ‘I have them’

The two Basque auxiliaries: izan ‘to be’ and edun ‘to have’, are selected according to the argument structure of the main verb. If the verb takes an ergative argument, edun is projected. If it does not, izan is selected. The number of arguments that can be inflected is up to three (for edun). The possible combinations of arguments that can be inflected are shown in (6): (6) Auxiliary 1 argument 2 arguments 2 arguments 3 arguments

izan ‘to be’ ABS ABS/​DAT -​-​-​-​-​-​ -​-​-​-​-​-​

edun ‘to have’ ERG ERG/​ABS ERG/​DAT ERG/​DAT/​ABS

26.2.1  Basque Case Marking In Basque three nominal cases can agree with the verb: the ergative, dative, and absolutive cases. An example of a ditransitive verb inflected with all three of these agreement markers is provided in (7): (7) Irakaslea-​k ikasle-​ei lan asko The teacher-​ERG the students-​DAT work a lot eman  d-​i-​e. give    AUX-​ABS3SG-​DAT.3PL-​ERG.3SG ‘The teacher has given the students a lot of work.’ Examples with the NP neska (‘the girl’) inflected with the ergative, dative, and absolutive case are shown in Table 26.2: Three of these morphemes are homophonous, ending in -​ k: absolutive plural, ergative singular, and ergative plural. In Basque, absolutive singular case is zero-​marked, as is generally true in ergative languages (Dixon 1994; Bittner & Hale 1996a).

 

Acquisition of Basque ergative and dative morphology    651 Table 26.2 Basque case suffixes Case

Morpheme

Singular (a)

Plural (k)

Absolutive

-​Ø

DP: nesk-​a ‘the girl’

DP: nesk-​ak ‘the girls’

Ergative

-​k

DP: nesk-​a-​k ‘the girl’

DP: nesk-​ek ‘the girls’

Dative

-​ri, -​ei

DP: nesk-​ari ‘to the girl’

DP: nesk-​ei ‘to the girls’

26.2.2  Default Case and Agreement in Adult Basque Several recent analyses of Basque morphosyntax assume that absolutive, which is zero-​ marked, is the morphologically default case in Basque (Preminger 2011b; Arregi & Nevis 2012; Rezac et al. 2014). Although Basque absolutive morphology appears in some of Schütze (2001)’s default case contexts (Rezac et al. 2014), others of Schütze’s tests such as dislocation, ellipsis, and gapping cannot be applied Basque, since case matching permits both ergative and absolutive morphology to be in used in non-​finite contexts, such as (8) and (9). (8) Left dislocation: Nik,  (nik)    tipo  hau    ezagutzen dut. I-​ERG I-​ERG guy   this-​ABS know      AUX-​ 1sgERG3sgABS “Me, I know this guy” (9) Ellipsis:

Nork nahi du          izozki    bat? Who-​ERG want AUX-​3sgERG3sgABS  ice cream  one-​ABS “Who wants an ice cream?” Nik/​ *ni I-​ERG/​I-​ABS “Me.” Nik ez/​*ni ez. I-​ERG NEG/​I-​ABS NEG “Not me.”

Both absolutive and ergative case are even possible in contexts where there is no finite verb in the discourse context, such as in the song title “Guk euskaraz, zuk zergatik ez?” (by Urko and Gabriel Aresti): (10) Gu-​k euskara-​z, zu-​k zergatik We-​ERG Basque-​INS you-​ERG why “We (speak) in Basque, why not you?”

ez? NEG

 

652   Jennifer Austin According to Preminger (2011b), the availability of absolutive case in both finite and non-​finite contexts in Basque demonstrates that “absolutive case in freely available in Basque, and is not dependent on agreement with-​or even the presence of-​any particular functional node” (Preminger 2011b: 13). Arregi and Nevins’ (2012) analysis of the morphological structure of Basque auxiliaries is couched in the Distributed Morphology framework, and outlined in Figure 26.1. SYNTAX Merge & Move Agree-Link Cliticization Absolutive Promotion

POSTSYNTAX Exponence Conversion Agree-Copy Fission ... Feature Markedness Participant Dissimilation Plural Clinic Impoverishment ... Morphological Concord Have-Insertion Complementizer Agreement ... LINEARIZATION Linear Operations Clitic Metathesis and Doubling ... VOCABULARY INSERTION ...

Figure 26.1  Morphological structure of Basque auxiliaries Source: Reproduced from Arregi and Nevins (2012) with permission of Springer.

Rather than relying on the lexicon to supply the syntax with an assembled word, in Distributed Morphology, there are post-​syntactic vocabulary items which get selected; each vocabulary item has its own schema. This schema pairs a morphosyntactic feature bundle with the phonological string that represents the morpheme as in (11):

 

Acquisition of Basque ergative and dative morphology    653 (11) Vocabulary Item Schema signal context of insertion Example: /​-​k/​ [_​_​_​_​, +ergative]/​DP Under this model, morphemes are joined to form words at any syntactic level through head movement, adjunction, or head merger, rather than via separate word-​assembling mechanisms in an independent lexicon. In their analysis, Arregi and Nevins (2012) propose that absolutive arguments are caseless in the syntax and are assigned case features at the Exponence Conversion Stage (post-​syntactically). Thus, absolutive case is not related to a specific functional head, unlike ergative and dative case, which are assigned structurally in a particular position in Basque. As seen in (Figure 26.2), Arregi and Nevins claim that dative case is assigned in spec ApplP and ergative case is assigned in spec of vP rather than TP, a hypothesis which is compatible with the existence of both finite and non-​finite clauses containing ergative arguments in Basque. Ditransitive sentences

vP v’

AErg VP App1P IODat

v V

App1’

DOAbs

App1

Figure 26.2  Dative case assignment in Basque ditransitives Source: Reproduced from Arregi and Nevins (2012) with permission of Springer.

In addition, Arregi and Nevins argue that Basque auxiliaries are formed of clitics rather than agreement markers which are hosted by C and T, as shown in the auxiliary structure in (Figure 26.3). C T (C1Abs/Dat)

C TAgr

(C1Erg)

C Agr

C

Figure 26.3  The structure of Basque auxiliaries Source: Reproduced from Arregi and Nevins (2012) with permission of Springer.

 

654   Jennifer Austin The clitics are assembled as bundles of abstract features in the syntax, and then sent to post-​syntactic modules where morphological rules arrange them in the correct order, and provide them with a phonological exponent at the Vocabulary Insertion stage, where the exponent with the greatest number of matching features is selected. Based on evidence from Person-​Case Constraint effects in Basque, Arregi and Nevins claim the dative and absolutive clitics are in competition for cliticizing to T, and that absolutive agreement (which is default) surfaces when there is no competitor. In section 26.4 I will propose that this theoretical framework which can account for several patterns of inflectional errors seen in child Basque.

26.3  The Order of Acquisition of Case and Verbal Morphemes in Child Basque Research into the acquisition of Basque began with the pioneering work of Larrañaga (1994), Barreña (1995), and Ezeizabarena (1996). As part of a collaborative project between the University of the Basque Country and the University of Hamburg, these researchers collected longitudinal natural speech data from two bilingual children, Mikel and Jurgi. The data were collected every month in Basque and in Spanish between the ages of 1;06 to 4;00 years (Mikel) and 1:06 to 4;01 years (Jurgi). Larrañaga (1994) studied the development of Spanish and Basque nominal case in Mikel from age 1:06 up to the age of 2;06, and found that the use of absolutive case preceded the use of ergative case in child Basque. Ezeizabarrena and Larrañaga (1996) compared Mikel’s production of case to his production of verbal agreement in Basque. They found that Mikel began target-​like production of ergative agreement at 2;00, but did not begin producing ergative case regularly (in 87% of obligatory contexts) until 2;04. There was no corresponding discrepancy for absolutive singular case, but because this case is zero-​marked in the singular, it is difficult to know for certain when it is acquired. Absolutive singular case and verbal agreement emerged at the same time, in contrast to ergative case and agreement, although the absolutive plural was slow to be acquired both in verbal inflection and nominal marking. Ezeizabarrena and Larrañaga found that pre-​ consonantal contexts favored the children’s omission of both the absolutive plural determiner and the ergative case marker. In addition, these authors found the ergative case morpheme was omitted more often than the absolutive plural determiner before vowels. They concluded that the protracted development of ergative case in Basque was not due to an inherent difficulty of ergativity per se, since Mikel produced ergative verbal agreement from an early age. Rather, they suggested that there is a delay in discovering the relation between verbal agreement and case marking, leading the child to express the argument structure of the verb through verbal inflection only at first. In research based on these same data, Ezeizabarrena (1996) studied the development of verbal agreement morphology in Mikel and Jurgi. Barreña (1995) studied the development of Basque verbal agreement and case in three children: Mikel, Peru (another

 

Acquisition of Basque ergative and dative morphology    655 Spanish/​Basque bilingual child) and Oitz, a monolingual child learning Basque. The pattern that emerged from these studies was that absolutive verbal agreement was produced first, followed by ergative agreement, and dative agreement last of all (there was a discrepancy between the age at which the two authors considered Mikel to have first used absolutive agreement), as shown in Table 26.3. Table 26.3 The order of production of Basque verbal agreement morphemes in child speech Child

Age of first use of absolutive agreement

Age of first use of ergative agreement

Age of first use of dative agreement

Oitz

2;00

2;02

2;04

Mikel

1:10 (Barreña) 1;07 (Ezeizabarrena)

1:10

2;04

Peru

1;11

2;04

3;00

Jurgi

2;04

2;08

3;03

Sources: Barreña (1995); Ezeizabarrena (1996)

In previous research, I examined the production of finite verbs (Austin 2009, 2010, 2012, 2013) and case marking (Austin, 2007, 2013) in natural speech samples from 20 Basque/​Spanish bilingual children and 11 monolingual children learning Basque. These cross-​sectional data were collected in the Spanish Basque Country, and information about the children’s MLUs in each language, number of utterances, and length of sessions are provided in Tables 26.A1 and 26.A2 in the Appendix. Austin (2013) found that children acquiring Basque do not produce dative case and agreement in tandem, in addition to the discrepancy between the production of ergative case and agreement reported by Ezeizabarrena and Larrañaga (1996). In particular, many children omitted dative indirect object agreement while producing dative case. Interestingly, some children were able to produce dative experiencer agreement but not dative indirect object agreement. This can be seen in examples from RB, who produces an utterance with missing indirect object inflection in (12a), yet can produce target-​like dative experiencer agreement, as in (13). The auxiliary in (12a) is also missing the final morpheme (t) which indicates first person singular agreement. (12) a. *Bai esan-​go d-​u (t) aita-​ri eta amatxu-​ri    yes tell-​FUT AUX.ABS.3SG-​ERG.1SG dad-​DAT and mommy-​DAT ‘(I) will tell Dad and Mommy’ (RB 2;08) b. Target: Bai esan-​go di-​e-​t aita-​ri eta amatxu-​ri yes tell-​FUT AUX.ABS.3SG-​DAT3PL-​ERG.1SG dad-​DAT and mommy-​DAT ‘(I) will tell Dad and Mommy’

 

656   Jennifer Austin (13)

(Ez) z-​ai-​o (NEG) AUX.ABS.3SG-​DAT.3SG ‘(Doesn’t) like’

gusta-​tzen. like-​IMPERF

(RB 2;08)

26.3.1  Inflectional Errors in Child Basque Although children acquiring Basque begin to use finite verbs before two years of age, their ability to produce target-​like morphological forms continues to develop over the course of several additional years, during which time they continue to make inflectional errors. Austin (2009) found that children produced four types of errors with verbal agreement most often: (1) the omission of an entire auxiliary, which occurred in 58/​1548 or 4% of their utterances with verbs; (2) the omission of an ergative agreement person marker, which occurred in 44/​656 possible contexts, or 7% of the time; 3) the omission of dative agreement indirect object agreement, in 12/​23 or 52% of possible contexts; and 4), and the substitution of singular for plural agreement, which occurred in 8/​74 utterances (11% of the time). Children also frequently produced case errors in Basque. Bilingual and monolingual children omitted ergative case in 84/​188 obligatory contexts (45% of the time). In addition, bilingual children omitted dative case in 4/​23 or 17% of obligatory contexts. The most common case error of omission was the failure to produce ergative case with transitive verbs, as seen in example (14): (14) zein(ek) harrapatu which-​Ø catch ‘Which has caught it?’

du? AUX.ERG3SG.ABS3SG

[IA 3;02]

There were only two examples in which a child substituted ergative for absolutive case, as shown in (15): (15)

*ni-​k, nere  ohea-​n I-​ERG my    bed-​in ‘I gets in my bed’

sartu da get be.ABS3SG

[TC 2;05]

At younger ages, children substituted infinitives for finite verbs and 3rd person singular forms for other person and number combinations. These results are similar to the patterns of inflectional errors from child Spanish reported in Radford and Ploennig-Pacheco (1995); Clahsen, Aveledo and Roca (2002); Davidiak and Grinstead (2004) in which children substitute non-​finite verbs or 3rd person singular verbs for other parts of the verbal paradigm, but not vice-​versa.

 

Acquisition of Basque ergative and dative morphology    657

26.3.2 Factors Affecting the Production of Inflection in Child Basque Phonological difficulty and morphological complexity have been reported to influence the acquisition of case and agreement morphemes in children learning Basque. For phonological reasons, ergative case in Basque may be particularly difficult for children to produce. In Basque, ergative case is marked with the suffix -​k, and in adult Basque speech, consonants tend to undergo deletion when they are in pre-​consonantal contexts, a process known as Stop Deletion (Hualde 1991; Hualde & Ortiz de Urbina 2003). In early stages of acquiring Basque, monolingual and bilingual children omit the ergative case marker in all phonological contexts (Barreña 1995; Ezeizabarrena and Larrañaga 1996). Children continue to omit this morpheme well after they are producing ergative verbal agreement. In Austin (2007) I compared ergative case-​marking in child and adult Basque speech, and found that whereas adults only dropped the ergative case marker in pre-​consonantal position, children omitted the ergative case marker in pre-​vocalic and utterance-​final positions too. That study, which compared natural speech production in Basque in eight monolingual and 20 bilingual children age 2;00–​ 3:06, also found that bilingual children omitted ergative case significantly more than monolingual ones; bilingual children omitted ergative case in 52% of possible contexts, compared to 31% for monolingual children and 4% of possible contexts in the case of adults. By comparison, dative case, which is marked by a suffix that forms an open syllable (-​ri) was omitted to a far lesser extent than ergative case; bilingual children omitted dative case 17% of the time, and dative case was not omitted at all by monolingual children nor by adult speakers (Austin 2013).1 When children omitted ergative or dative, they produced a zero-​marked DP, which is homophonous with the absolutive singular form, as seen in (16): (16) inor(k) ez du ikus-​ten no one-​Ø NEG AUX.ERG3sg.ABS3sg see.IMP ‘No one can see’

(IA 3;02)

Phonological factors may also contribute to the late production of dative indirect object agreement in child Basque relative to other agreement morphemes. Dative inflection is a non-​final suffix, and crosslinguistic research indicates that middle affixes are

1 

The data examined for this study were collected from the same 20 bilingual and 8 monolingual children examined in Austin (2007) with the addition of three more monolinguals (for a total of 11 monolingual children).

 

658   Jennifer Austin slower to be acquired than prefixes or suffixes, perhaps in part because they are less phonologically salient (Slobin 1973; Pye 1983; Pye et al. 2007). The third type of agreement error is shown in example (17), the dative agreement morpheme is omitted and a transitive auxiliary is being used instead of a ditransitive one, even though there is a dative marked argument present: (17)

Dative agreement missing: *ba ematen d-​e-​t txorixu-​∅ well give-​IMPERF AUX-​ERG1SG.ABS.3SG chorizo-​ABS ‘Well, I give chorizo (sausage) to the cat’

katu-​ari cat-​DAT (NC 2;04)

In Austin (2010, 2013) I found that morphological complexity predicted the order in which auxiliaries were acquired in Basque (morphological complexity was defined as the number of morphemes encoded by the auxiliary). In child Basque, verbs that are more complex morphologically (such as ditransitive auxiliaries) were produced later in development than ones that encoded fewer arguments, such as unaccusative verbs. Children produced significantly more root infinitives with verbs that required ergative/​absolutive agreement than with verbs that required only absolutive agreement. When children repeated adult utterances (these instances were excluded from consideration in the analyses reported in this chapter), children made the same types of inflectional errors as in their natural speech, such as the omission of dative verbal agreement seen in example (18): (18)

Adult:

Aber zer let’s see what konta-​tzen d-​io-​gu-​n! tell-​IMPERF AUX-​ERG.1PL-​DAT.1SG-​ABS.3SG ‘Let’s see what we are telling him!’

Child:

Aber xxx kontatu-​ko d-​egu! let’s see tell-​FUT AUX-​ERG1PL.ABS.3SG ‘Let’s see what we will tell (him)’

[TC 2;05]

Austin (2013) examined frequency in the adult input as a potential additional factor which could influence children’s production of morphology. As seen in Table 26.4, I found that the order of acquisition of agreement and case morphemes in child Basque was not predicted by their frequency in adult speech:

 

Acquisition of Basque ergative and dative morphology    659 Table 26.4 Adults’ production of case and agreement in Basque Type of verbal agreement

Ergative

Absolutive subject agreement

Percentage produced by adults

684/​1488 45.9%

Order of production by children

Dative experiencer

Ditransitive agreement

573/​1488 38.5%

42/​1488

83/​1488

Second

First

Third

Fourth

Type of case

Ergative case

Absolutive case

Dative case

Percentage produced by adults

100/​1488 6.7%

596/​1488 40%

17/​1488 1%

Order of production by children

Second

First

Third

3%

6%

These results suggest that the order of acquisition of these morphemes in not driven primarily by frequency in the adult input, but by morphological complexity.

26.4  The Use of Default Morphology as a Repair Strategy in Child Basque The two most common inflectional errors in child Basque were the omission of ergative case, seen in (19a), and the omission of dative ditransitive agreement, seen in (19b). (19) a. Omission of ergative case: *Ni(k) baduakat hau etxi-​an I-​ÿ have-​ERG1sgABS3s that-​ABS house-​in ‘I have that at home’

(LA 3;0)

b. Omission of dative ditransitive agreement: *Bai esan-​go d-​u (t) aita-​ri eta amatxu-​ri yes tell-​FUT AUX.ABS.3SG-​ERG.1SG dad-​DAT and mommy-​DAT ‘(I) will tell Dad and Mommy’   (RB 2;08)

 

660   Jennifer Austin In errors such as these, children omit either verbal inflection or nominal case but not both. This suggests that the errors are not due to a syntactic deficit, in which case the child’s utterances would lack finiteness or certain functional categories altogether; instead, these errors seem to be attributable to a problem in producing target morphological forms. In each case, an unmarked absolutive form is substituted for a marked ergative or dative form. I propose that these inflectional errors in Basque find a natural explanation in the Distributed Morphology framework as elsewhere forms that surface when a child has difficulty selecting the appropriate form for whatever reason (phonological difficulty or complexity, for instance). Under such circumstances, the Subset Principle (Halle 1997: 428) allows the child to insert the next best fit, a vocabulary item which contains only a subset of the morpheme’s features: The phonological exponent of a vocabulary item is inserted into a morpheme … if the item matches all or a subset of the grammatical features specified in the terminal morpheme. Insertion does not take place if the vocabulary item contains features not present in the morpheme. Where several vocabulary items meet the conditions for insertion, the item matching the greatest number of features specified in the terminal morpheme must be chosen.

When no closer match can be found, a default vocabulary item is inserted via the Elsewhere Condition. In example (19a), the child has produced absolutive rather than ergative case. Under the analysis proposed here, the child has not selected the correct phonological exponent seen in (20a), and the absolutive elsewhere form is selected (20b):2 (20) a. ergative: b. absolutive:

/​k/​ [+motion, -​peripheral] /​Ø/​ [-​motion, -​peripheral]

The availability of the absolutive elsewhere form when a child runs into morphological trouble would explain why children’s case errors involve substituting absolutive for ergative or for dative, but not the opposite. This outcome (moving from marked to unmarked feature values) is similar to the result of ergative impoverishment rules in adult Basque discussed in Arregi and Nevins (2012). Another possible case error, substituting ergative for dative case (or vice-​versa), was never produced by any children. In examples such as (19b) where the child’s utterance is missing dative agreement, the child has selected a morphological form with a subset of the features of the ditransitive target auxiliary. The presence of dative case-​marked arguments (aita-​ri ‘Dad’and amatxu-​ri ‘Mommy’) in this utterance suggests that dative agreement has been generated in the syntax, allowing for dative case to be assigned, but morphologically something has gone astray in producing the target form. In Arregi and Nevins’ theory, this 2 

Case features proposed by Calabrese (2008).

 

Acquisition of Basque ergative and dative morphology    661 case-​assigning configuration (KP) would have the structure in (Figure 26.4) for ergative and dative case (nonparticipant arguments are third person), with a clitic generated in the spec of KP. Nonparticipant arguments with dative or ergative case KP Dcl

K’

DPArg –participant

K ±peripheral

–author

+motion

±singular

Figure 26.4  Case assignment in Basque Source: Reproduced from Arregi and Nevins (2012) with permission of Springer.

While there may several reasons why child has substituted default absolutive inflection for the target dative agreement, the presence of dative case suggests that the problem is one of morphological rather than syntactic competence. This analysis proposed here assumes that missing inflection in child language is the result of a morphological system which is still developing rather a syntactic system which has missing or underspecified functional projections, as was proposed by the earliest generativist accounts of inflectional development. These approaches interpreted missing verbal inflection to reflect a lack of syntactic knowledge of inflection on the part of the child, or a truncated clause in early stages of development (until age 2:06 or so) consisting of a VP with the upper functional projections missing (Rizzi 1994). While these earlier theories were based on data from English and other languages with little morphology, they made incorrect predictions for Basque and other highly inflected languages. The Truncation Hypothesis would predict the emergence of inflectional morphemes in children’s language to correspond to the hierarchy of functional morphemes in the Basque inflectional phrase (absolutive-​> dative-​> ergative). However, the pattern that we find in child Basque is different (absolutive–​ergative–​dative), and does not correspond to the order which we would expect if clauses were being built in increments from the bottom up. The Optional Infinitive Hypothesis (Wexler 1994, 1995) predicts that there should be no optional infinitives in pro-​drop languages, but children acquiring Basque produce them. Another advantage of the analysis proposed here is that it can account for a wide array of children’s inflectional errors, rather than just their production of nonfinite verbs in finite contexts. This account correctly predicts that children’s inflectional errors will be characterized by the substitution of less specified forms for more specific ones. It also predicts that the production of nonfinite forms will gradually decline in child grammars, rather than ending abruptly (Austin 2010).

 

662   Jennifer Austin

26.5 Conclusions The acquisition of verbal morphology and case in L1 Basque is a lengthy process that proceeds incrementally. Children’s inflectional errors in Basque consist of producing default inflection or morphological forms that contain a subset of the abstract features of the target forms. These errors suggest that in early stages of morphological acquisition children use default forms which encode a subset of the morphemes as a “best match” of the target forms to support their developing language when they are unable to produce or retrieve target forms. Children’s systematic substitution of default morphological forms for more complex ones suggests that they know enough about the properties of the abstract features encoded by the morphemes they are learning to avoid using them in the wrong context. This adherence to the Subset Principle also suggests that they are guided early on by knowledge of the argument structure of the verbs that they are acquiring; children never substituted ergative inflection for dative inflection, for example, despite the fact that ergative agreement is produced more frequently by adults in the input (see Table 26.4). The ability of children to avoid non-​subset errors supports the hypothesis that in L1 acquisition “syntax comes first, morphology later” Blom and Wijnen (2006). The inflectional errors that children make in acquiring Basque indicate that early in development, they are generating combinations of verbs and auxiliaries not found in the adult input, rather than using complex morphological forms in lexicalized chunks. Children’s production of these inflectional errors presents a challenge to usage-​based models of grammatical development (e.g. Tomasello 2003), since children are producing innovate combinations rather than replicating patterns found in the adult input. The complex possibilities afforded by Basque morphology make visible patterns in morphological development which cannot be observed in the acquisition of more commonly studied languages which have less rich verbal inflection. In the case of an highly inflected language like Basque, for instance, children produce auxiliaries with a simpler argument structure and fewer morphemes earlier than more morphologically complex ones; the acquisition of inflection is not necessarily an “all or nothing” process, as was implicitly assumed in earlier analyses of the acquisition of finiteness in English such as the Optional Infinitive Hypothesis (Harris & Wexler 1996). The ergative nature of Basque morphology also permits us to compare the development of subject and object inflection in child acquisition as well as the effect of verbal argument structure on the acquisition of different types of subjects. The finding that the acquisition of case marking and verbal agreement does not occur in parallel in Basque would not be possible to observe in a language with simpler morphology; for example, the production of ergative agreement precedes the production of ergative case, whereas the opposite pattern obtains for dative case and agreement. Insights such as these highlight the fact that while

 

Acquisition of Basque ergative and dative morphology    663 research on the development of morphology in English has yielded many important insights that are universally applicable, nevertheless it is important for comprehensive theories of L1 development that we also consider data from less commonly studied languages such as Basque.

Appendix Table 26.A1 Number of utterances and mean length of utterance in each language for bilingual children Child’s initials

Sex

GG

Recording length (in minutes)

Age

Basque utterances

Basque MLU

M

2;01

283

1.85

60

NI

F

2;01

78

1.18

60

NC

F

2;04

85

2.44

60

LH

M

2;05

71

1.95

60

TC

F

2;05

413

1.53

90

AI

F

2;06

205

2.14

60

ME

F

2;06

200

2.65

60

RM

M

2;07

64

2.20

90

IC

M

2;07

20

2.22

60

RB

F

2;08

364

4.55

100

AR

M

2;08

67

1.85

60

OH

M

2;08

362

3.16

90

IU

M

2;10

126

3.93

60

XO

M

3;00

315

3.16

90

LA

F

3;00

151

4.26

60

DG

M

3;01

67

3.15

60

MA

M

3;01

279

2.48

60

IA

M

3;02

237

2.95

100

AM

F

3;02

115

3.33

60

AB

F

3;04

86

3.24

30

 

664   Jennifer Austin Table 26.A2 Number of utterances and mean length of utterance for monolingual children in Basque

MLU

Length of recording (in minutes)

Child’s initials

Sex

Age

Total utterances in sample

AC

F

2;01

221

1.95

60

JH

M

2;01

126

1.88

60

MA

M

2;03

183

1.73

60

EC

F

2;05

319

1.65

90

AH

F

2;05

162

2.46

60

MC

M

2;08

278

1.80

60

AG

M

3;00

155

3.24

60

EG

M

3;00

117

2.42

45

ME

M

3;01

177

3.67

60

NS

F

3;02

282

5.07

60

AB

M

3;03

257

2.83

60

Acknowledgments Many thanks to the children and parents in the Basque Country who participated in this study, as well as the research assistants who worked with me for their invaluable help with data collection and transcription. I am also grateful to Edith L. Bavin, Susana Bejar, and Lisa Travis for their very helpful comments on this chapter. All remaining errors are my own.

Abbreviations 1, first person; 3, third person; ABS, absolutive; AUX, auxiliary; DAT, dative; DET, determiner; ERG, ergative; FUT, future; IMP, imperative; IMPERF, imperfective; INS, instrumental; NEG, negative; PL, plural; SG, singular.

 

Chapter 27

A c omparati v e st u dy of the ac qu i si t i on of nom inative and e rg at i v e al ignm ent in E u rope a n and M ayan l a ng uag e s Clifton Pye and Barbara Pfeiler

27.1 Introduction This chapter uses language data on the acquisition of nominative person markers in two European languages as an external test of the nature of person marking in four Mayan languages. Specifically, we compare the acquisition of nominative person markers in French and Spanish with the acquisition of person marking in the Mayan languages Wastek, Yukatek, Ch’ol, and K’iche’. This comparison clarifies the clitic status of the Mayan person markers and provides a better understanding of the acquisition of person marking in all six languages. Spanish uses verb suffixes to mark person, number, and tense (1). (1) Spanish person marking a. Bail-​o. dance-​1.sg.present ‘I dance.’ b. Bail-​as. dance-​2.sg.present ‘You dance.’ c. Bail-​a. dance-​3.sg.present ‘She/​he/​it dances.’

 

666    Clifton Pye and Barbara Pfeiler The person marking suffixes in Spanish are phonetically distinct and make the overt use of pronouns unnecessary in most contexts (cf. Grinstead 2004). In comparison with Spanish, the suffixes on French verbs are phonetically indistinct. French makes use of obligatory pronominal clitics to mark person (2). (2) French person marking a. Je danse. ‘I dance.’ b. Tu danses. ‘You dance.’ c. Il/​Elle danse. ‘He/​She dances.’ The Spanish person marking suffixes have some features in common with the French pronominal clitics. The person suffixes and the pronominal clitics both realize nominative person marking. Neither the suffixes nor pronominal clitics can be used in isolation (3). (3) Question responses in Spanish and French a. Spanish ¿Quién vino? *-​a. Who came? He/​She. b. French Qui est venu? Who has come?

*Il/​Elle. He/​She.

Despite these similarities, there are two crucial differences between the Spanish person marking suffixes and the French subject clitics. First, the French subject clitics occur in a preverbal position, whereas the Spanish person suffixes occur in a postverbal position. Second, the Spanish person suffixes may not be separated from the verb stem by a negation marker, whereas French places the negation marker ne between the subject clitic and the verb (4). Kayne (1975) discusses other properties of the French pronominal clitics. (4) Je ne danse I neg dance ‘I do not dance.’

pas neg

In this chapter, we use acquisition data on person marking in French and Spanish to make an operational distinction between agreement and clitic doubling. We compare

 

The acquisition of nominative and ergative alignment    667 the acquisition data in French and Spanish with acquisition data on person marking in four Mayan languages in order to investigate whether the acquisition of person marking on Mayan verbs resembles the acquisition of nominative clitics in French or nominative agreement markers in Spanish. Our study is partly motivated by generative accounts of language acquisition that attribute different developmental patterns to the affix/​clitic distinction in the adult languages (cf. Hamann et al. 1998: 329; Grinstead 2004). Our study is also motivated by proposals that identify ergative or absolutive person markers in Mayan and other languages with nominative person marking (cf. Johns 2006). Proposals that identify ergative marking as nominative include the Absolutive-​ S-​as-​Object approach of Trager (1946) and Bobaljik (1993b). Proposals that identify absolutive marking as nominative include the Passive approach of Schuchardt (1895) and Hale (1970) as well as the Inverse approach of Dixon (1972), Trechsel (1982), and Marantz (1984). More complex proposals include the Absolutive-​as-​Clitic approach of Woolford (2000), the Absolutive as Nominative and Accusative approach of Legate (2006) and the High and Low Absolutive approach of Coon et al. (2014). The precise nature of the distinction between agreement markers and pronominal clitics presents both descriptive and theoretical challenges (Spencer and Luís 2012). Clitics are notoriously difficult to describe precisely because they have properties of both bound and unbound morphemes. Descriptions of person markers use the term clitic or affix on the basis of one or two linguistic features instead of considering the full constellation of phonetic, morphological and syntactic diagnostics. The ambiguity between clitic doubling and agreement presents a challenge for linguistic description as well as for linguistic theory (cf. Kramer 2014a). This challenge is increased by grammaticalization processes that over time convert pronominal clitics to agreement markers. The literature on these topics is large and continues to expand. Generative theories analyze agreement as a realization of phi features on a phrasal head, whereas clitic doubling has been analyzed as the movement of a D head. Kramer (2014a: 593) notes that ‘In principle, these two phenomena are distinct, but in practice they can be difficult to distinguish.’ Spencer and Luís (2012) provide an overview of research on clitics, Corbett (2006) provides an overview of research on agreement and Siewierska (2004, 2011) provides overviews of research on person marking. Koopman and Sportiche (1991) provide a classic discussion of the relation between Case assignment and agreement. Our concern in this chapter is with the acquisition of person marking rather than these theoretical accounts. For this reason, we use the acquisition of nominative person marking in French and Spanish as evidence for the clitic status of the person markers in Mayan languages. The following section of the chapter presents the acquisition data for French and Spanish, and discusses its analysis in acquisition theories. The third section of the chapter presents acquisition data for the ergative person markers in the four Mayan languages. The fourth section of the chapter presents acquisition data for the absolutive person markers in the four Mayan languages. In the conclusion to the chapter

 

668    Clifton Pye and Barbara Pfeiler we discuss the implications of our findings for research on the acquisition of person marking.

27.2  The Acquisition of Person Marking in European Languages Research on the acquisition of grammatical inflections relies upon a measure of morpheme production relative to the morpheme’s obligatory contexts (Brown 1973). Following children’s percentage use of inflectional morphemes in their obligatory contexts allows researchers to compare morphological development between 2-​year-​old children, who do not produce many verbal utterances, and 3-​year-​old children, who produce many utterances during the course of a day. Children acquiring Italian, Spanish, and Catalan produce nominative person marking suffixes on verbs in over 90% of the obligatory contexts by the age of 2;0 (Hoekstra and Hyams 1998; Grinstead 2000). This claim was challenged by Gathercole et al. (1999), who observed that while children acquiring Spanish produce many verbs with subject markers, the vast majority of their verb forms are the third person present tense form (cf. Aguirre 2003). Acquisition researchers have appealed to the contrasting forms of the Spanish person marking suffixes to explain why children produce Spanish verb suffixes so early (Hyams 1986, 1989; Hamann 2002). Hoekstra and Hyams (1998: 87) ascribe Spanish children’s success to the contrasting person markers on verbs, which make visible a tense chain linking a tense operator in Comp to the tense inflection. The contrasting person markers on Spanish verbs allow children acquiring Spanish to rapidly converge on the adult grammar. Grinstead (2000: 132) simply states that ‘The Person Phrase … does appear to be active from the very beginning.’ Wexler (1998) proposed the Unique Checking Constraint (UCC) to account for the Spanish children’s relative success in producing inflected verbs. The UCC references a D-​feature on agreement and tense, and proposes that 2-​year-​olds can only check the D-​feature of a determiner phrase against a single functional category, i.e. tense or agreement, but not both. Wexler proposes that the D-​feature of nominative agreement in Spanish is [+interpretable] and therefore is not checked. The grammar of children acquiring Spanish only has to check the D-​feature of tense and therefore 2-​year-​old children can produce the agreement markers on verbs without violating the UCC. Children acquiring French exhibit a significant delay in the production of the pronominal subject clitics relative to the Spanish children’s production of person marking on verbs. Hamann et al. (1998: 329) provide information on the forms of the French child Augustin’s production of subject clitics. Table 27.1 shows the frequency of Augustin’s subject clitic production in their obligatory contexts.

 

The acquisition of nominative and ergative alignment    669 Table 27.1 Augustin’s use of French subject clitics in obligatory contextsa Age

Verbal Utterances

Subject clitic

Null subject

N

%

N

%

Percent use in obligatory context

2;0.2

57

17

29.8

16

38.1

51.5

2;0.23

30

4

13.3

6

42.9

40.0

2;1.15

22

4

18.2

2

25

66.7

2;2.13

55

16

29.1

11

28.9

59.3

2;3.10

45

12

26.6

9

31

50

2;4.1

62

10

16.1

25

52.1

26.5

2;4.22

54

11

20.4

14

36.8

44

2;6.16

116

25

21.6

30

32.3

45.5

2;9.2

175

80

45.7

28

21.1

74.1

2;9.30

115

99

63.4

12

9.6

89.1

a Based on Hamann et al. (1998)

The acquisition theories of Hoekstra and Hyams (1998) and Wexler (1998) also explain why children’s production of French pronominal clitics is delayed relative to children’s production of Spanish person suffixes. Hoekstra and Hyams would claim that the suffixes on French verbs do not make the tense chain visible to children. Wexler would claim that the D-​feature of nominative agreement in French is not [+interpretable] and therefore competes with the D-​feature of tense which in turn leads to UCC violations. We will use the results of the research on the acquisition of person marking in French and Spanish to analyze the acquisition of person markers in Mayan languages. We assume that if Mayan children acquire nominative person markers as agreement affixes they should display developmental changes that resemble the acquisition of the nominative agreement suffixes on Spanish verbs. In other words, the children should produce nominative agreement affixes in 90% of obligatory contexts by the age of 2;0. If Mayan children acquire nominative person markers as pronominal clitics they should display developmental changes that resemble the acquisition of nominative pronominal clitics in French. In this case, we expect Mayan children to produce nominative pronominal clitics with approximately 50% of verbs up to age 2;6 and in 90% of verbs at 3;0. We state this hypothesis in (5). (5) Child production of nominative person marking   1. Children produce nominative agreement on 90% or more of verbs at 2;0.   2. Children produce nominative pronominal clitics on approximately 50% of verbs up to 2;6 and in 80% of verbs at 2;9.

 

670    Clifton Pye and Barbara Pfeiler While there are many other factors that affect children’s inflectional development (cf. Pye 1983), the acquisition theories of Wexler and Hoekstra and Hyams tie children’s inflectional development directly to the structural realization of person marking. The hypothesis in (5) should hold to the extent that these acquisition theories are correct. In testing whether these theories account for the acquisition of person marking in Mayan languages we simultaneously test whether the theories provide a complete account of inflectional development. The acquisition data also provide evidence for identifying the Mayan person markers with the nominative person markers of French and Spanish. In the next section, we provide general information about the Mayan children and their language samples.

27.3  The Acquisition of Person Marking in Four Mayan Languages The Mayan language family has 28 languages that are still spoken in Mexico, Guatemala, Belize and Honduras. The family is divided into six major subgroups:  Wastekan, Yukatekan, Greater Mamean, Greater K’iche’an, Greater Tzeltalan, and Q’anjob’alan (Kaufman 1990a; Robertson 1992). We will compare children’s acquisition of ergative and absolutive subject markers in languages from four of these major subgroups: Wastek (Wastekan), Yukatek (Yukatekan), Ch’ol (Greater Tzeltalan), and K’iche’ (Greater K’iche’an). In this section, we provide general information about the children and their language samples.

27.3.1 The Language Samples The language samples for the four languages were recorded in and around the children’s homes and each recording session lasted approximately one hour. The participants included the children, various members of their families, the investigators, and visitors. The mothers and siblings were generally present during the recordings, but the fathers only participated occasionally. The families live in rural villages, and the children spend most of their day within the family compounds. The investigators were native speakers of the languages who interacted with the children to different degrees. The recording sessions included play with toys, natural objects, and picture books.

 

The acquisition of nominative and ergative alignment    671

27.3.2 Wastek Pfeiler and Pye analyzed 13 hours of recordings from a longitudinal database of three children acquiring Wastek living in the vicinity of Aquismón and Tancanhuits, in the state of San Luis Potosí, Mexico. The children were being raised in monolingual Wastek households although Spanish is widespread outside of the home. Table 27.2 provides basic statistics for these recordings. Table 27.2 Basic statistics for Wastek Children ELV

VLA

CAR

No. of utterances

No. of verbal utterances

Proportion

Duration (hrs)

2;3.11

490

199

.41

1

2;5.27

664

401

.60

1

2;11.26

952

742

.78

1

3;0.17

579

353

.61

1

2;2.26

285

220

.77

1

2;6.2

285

143

.50

1

2;6.14

235

139

.59

1

2;11.19

248

131

.53

1

3;0.19

426

211

.50

1

2;4.24

382

100

.26

1

2;4.27

143

55

.38

1

2;7.7

592

289

.49

1

3;0.8

213

137

.46

1

Age

27.3.3 Yukatek The Yukatek recordings were made in Yalcobá in the eastern part of the state of Yucatán, Mexico in the 1990s (Pfeiler and Martín Briceño 1997; Pfeiler 2003). For this study Pfeiler selected 17 hours of data from her longitudinal database of three children at two, two and a half and three years of age (Table 27.3).

 

672    Clifton Pye and Barbara Pfeiler Table 27.3 Basic statistics for Yukatek Children

Age

No. of utterances

No. of verbal utterances

Proportion

Duration (hrs)

SAN

2;0

380

109

.29

2

2;6

339

98

.29

1.5

3;0

741

125

.17

1.5

2;0

327

100

.31

2.5

2;6

325

64

.20

2.5

3;0

443

84

.19

3

2;0

429

99

.23

1.5

2;6

293

217

.74

1.5

3;0

249

158

.63

1

ARM

DAV

27.3.4 Ch’ol The Ch’ol analyses are based on a set of longitudinal recordings that were made in Tila in Chiapas, Mexico between 2005 and 2010 by native Ch’ol speakers. The investigators recorded three children living in a small neighborhood who were acquiring Ch’ol as their first language. Table 27.4 provides basic statistics for the Ch’ol language samples. Table 27.4 Basic statistics for Ch’ol language samples Children EMA

MAR

Age

No. of utterances

2;1.30

583

186

.32

1

2;6.16

519

147

.28

1

2;9.30

547

211

.39

1

3;0.17

584

228

.39

1

2;0.21

263

32

.12

1

65

17

.26

1

2;5.27

200

21

.10

1

2;11.23

507

187

.37

1

3;0.6

506

168

.33

1

1;11

230

63

.27

1

1;11.18

200

65

.33

1

2;7.3

83

7

.08

1

3;0.3

119

31

.26

1

2;2

MA

No. of verbal utterances

Proportion

Duration (hrs)

 

The acquisition of nominative and ergative alignment    673

27.3.5 K’iche’ The K’iche’ analyses are based on a set of longitudinal recordings that Pye recorded in the town of Zunil, Guatemala (Pye 1980; 1992). Table 27.5 provides basic statistics for the K’iche’ language samples. Table 27.5 Basic statistics for K’iche’ language samples Children

Age

No. of utterances

No. of verbal utterances

Proportion

Duration (hrs)

TIY

2;1

574

131

.23

2

2;7

594

214

.36

2

2;10

605

255

.42

2

LIN

2;0

501

159

.32

2

CHA

2;9

713

221

.31

2

3;0

825

432

.52

2

27.4  The Acquisition of Ergative Person Marking in the Mayan Languages In this section, we first describe the ergative person markers in Wastek, Yukatek, Ch’ol, and K’iche’ before we present an analysis of the children’s acquisition of ergative person marking. We follow this description with an analysis of how children acquire the ergative person markers in the four languages. We compare the developmental profiles for ergative person markers in these four Mayan languages in order to establish whether the children exhibit a common pattern of ergative person marking. We then compare the children’s ergative person marking to children’s nominative person marking in French and Spanish.

27.4.1 Ergative Person Marking in the Mayan Languages The ergative person markers in Mayan languages have properties of both affixes and clitics. The Mayan ergative markers are traditionally described as prefixes, e.g. Kaufman (1990a: 71); Lehmann (1993) analyzes the ergative markers of Yukatek as enclitics. Mayan languages use the ergative markers to cross-​reference the subjects of transitive verbs as well as the possessors of nouns. Woolford (2000: 172) identifies the ergative markers in the Mayan language Popti’ as nominative agreement prefixes. Coon et al. (2014) represent the ergative markers in Mayan languages as agreement prefixes that are marked for inherent case by the transitive v0.

 

674    Clifton Pye and Barbara Pfeiler The ergative markers in Yukatek, Ch’ol, and K’iche’, but not in Wastek, have allomorphs like affixes. Due to space limitations, we restrict our analysis in this chapter to the acquisition of the preconsonantal ergative allomorphs in Yukatek, Ch’ol, and K’iche’. Brown et al. (2013) present data on the acquisition of the prevocalic ergative allomorphs. The first person ergative marker in Ch’ol has the preconsonantal allomorphs k~j.1 Wastek uses the same set of ergative markers on verbs that begin with consonants and vowels, but requires a separate set of portmanteau markers for transitive verbs with first and second person objects. We excluded the Wastek portmanteau morphemes from our analysis. Table 27.6 lists the preconsonantal allomorphs of the ergative markers in the four languages. Table 27.6 Preconsonantal ergative person forms in Wastek, Yukatek, Ch’ol, and K’iche’ Number

Person

Wastek

Yukatek

Ch’ol

K’iche’

Singular

1

u

in

k~j

in

2

a

a

a

a

3

in

u

u

u

1 inclusive

i

in … o’on-​e’ex

la=k~j

qa

1 exclusive

i

in … o’on

k~j … -​lojoñ

qa

2

a … (chik)

a … e’ex

la=a

i

3

in … (chik)

u … o’ob

u … -​ob

ki

Plural

The ergative markers in Wastek, Yukatek and Ch’ol also have several attributes of clitics. Wastek, Yukatek and Ch’ol, but not K’iche’, allow a limited set of adverbial modifiers to occur between the ergative marker and the verb stem (6). (6) Preverb stem adverbs in Wastek, Yukatek and Ch’ol a. Wastek taam ti-ø k’al-​e ø-​in ejeet kinin-​iy-​al (Zavala 1994: 47) then ti-3sg.abs go-​it 3sg.abs-​3sg.erg slowly pull-​tt-​inc ‘Then he went pulling it slowly.’ b. Yukatek Ts’o’ok in    chéen hats’-​ik-​ech (based on Verhoeven 2007: 102) term 1sg.erg  just beat-​inc-​2sg.abs ‘I have just beaten you’ 1  All Mayan words are shown in the practical orthography developed by the Proyecto Lingüístico Francisco Marroquín (Kaufman 1976a) with a single exception: we use rather than for the glottal stop. The other orthographic symbols have their standard IPA values except:  = /​ϴ/​,  = /​c/​,  = /​tʃ/​,  = /​c’/​, =/​tʃ’/​,  = /​ʃ/​,  = /​x/​,  = /​ɨ/​.

 

The acquisition of nominative and ergative alignment    675 c.

Ch’ol tyi aw oraj-​tyaj-​a-​ø com 2sg.erg quickly-​find-​tt-​3sg.abs ‘You found it quickly.’

(Martínez Cruz 2007: 99)

The preconsonantal ergative markers in all four languages frequently syllabify with preceding clitics and form part of a complex phonological word. The Yukatek and Ch’ol verb complexes begin with an independent auxiliary aspectual verb that hosts the ergative subject markers (Lehmann 1993; Vázquez Álvarez 2002). K’iche’ verbs begin with an aspectual clitic that forms a syllable with the preconsonantal ergative subject markers (7). The transitive verb complex in Wastek begins with the ergative subject marker, which also syllabifies independently of the following verb stem. Wastek expresses aspect by means of thematic suffixes on the  verb. (7) K’iche’ aspect–​ergative phonological phrase k=ø in-​b’ii-​j inc=3sg.abs 1sg.erg-​say-​der.tv ‘I say it.’ We list the affix/​clitic properties of the ergative markers in Wastek, Yukatek, Ch’ol, and K’iche’ in Table 27.7. This table demonstrates that the ergative markers in these Mayan languages have characteristics of both affixes and clitics. Moreover, their properties differ across these four languages. The ergative person markers in Wastek do not have allomorphs and allow adverb insertion between the ergative markers and the verb stem. The ergative markers in Yukatek and Ch’ol have allomorphs, but also allow adverb insertion. Table 27.7 Affix/​clitic properties of ergative person markers in four Mayan languages Language

prevocalic/​preconsantal allomorphs

phonetic merger with aspectual verbs

adverb insertion

Wastek Yukatek

Clitic Affix

Affix Clitic

Clitic Clitic

Ch’ol

Affix

Clitic

Clitic

K’iche’

Affix

Clitic

Affix

The ergative markers in K’iche’ have allomorphs, but do not allow adverb insertion. The acquisition results from French and Spanish can potentially provide independent evidence of the agreement/​clitic status of the ergative person markers in these four Mayan languages. Assuming that children acquire the ergative person markers as nominative agreement markers, we expect the children to produce the ergative markers in over 90% of their obligatory contexts at 2;0. We predict a different acquisition outcome

 

676    Clifton Pye and Barbara Pfeiler for the children’s production of the ergative person-​marking clitics. Assuming that children acquire the ergative person markers in these languages as nominative pronominal clitics, we expect the children to produce the ergative clitics on approximately 50% of verbs up to 2;6 and in 80% of verbs at 2;9 based on the French data. We summarize these predictions in (8). (8) Predictions for the acquisition of ergative person markers in Mayan languages   a Mayan children acquire ergative person markers as nominative agreement markers and produce them in 90% of obligatory contexts at 2;0.   b Mayan children acquire ergative person markers as nominative clitics and produce them in approximately 50% of obligatory contexts between 2;0 and 2;6.

27.4.2 The Acquisition of Ergative Person Marking We analyzed the children’s use of the preconsonantal ergative person markers on transitive verbs. For the purposes of our study we grouped the children’s recordings into three groups of 2, 2-​and-​a-​half, and 3 years. The data for the children’s production of ergative markers are shown in Table 27.8. The cells in the table show the number of tokens produced, the number of obligatory contexts and the percent use of the marker in its obligatory contexts. Blank cells in Table 27.8 indicate that no data are available. Table 27.8 Ergative production on transitive verbs (preconsonantal) 2;0 Language

Child

Wastek

ELV CAR

Yukatek

Ch’ol

K’iche’

n/​contexts

2;6 %

n/​contexts

19/​159

12

28/​50

56

VLA

35/​67

3;0 %

n/​contexts

%

30/​213

14

210/​441

62/​89

69.7

33/54

61.1

52

39/​84

46

115/​165

70

48

ARM

4/​29

14

38/​61

62

65/​70

93

SAN

2/​8

25

33/​63

45

128/​136

94

DAV

4/​12

33

10/​35

29

29/​30

97

EMA

2/​31

6

3/11

27

21/50

42

MAR

3/​18

17

2/​4

50

9/​50

18

MA

4/​31

13

2/​3

67

7/​9

78

TIY

1/​23 4

4

13/88

15

19/76

25

LIN

6/​36

17 9/​177

5

23/​237

10

CHA

 

The acquisition of nominative and ergative alignment    677 The number of ergative clitics the children produced varied within each language as well as across the languages. We tested the similarity between the children’s ergative production within each language in order to determine the degree to which the linguistic structure determined the children’s production of the ergative person markers. We used the chi square statistic to analyze the children’s ergative clitic production within each language. We excluded any cells which did not meet the threshold of five tokens necessary to satisfy the requirements of the chi square test. The children acquiring Wastek and Yukatek had statistically different ergative productions (Wastek χ2 = 112, p < .01; Yukatek χ2 = 166, p < .01; K’iche’ χ2 = .62, p > .05). The children acquiring K’iche’ did not vary significantly in their ergative production. While we found significant differences in the children’s ergative production within Wastek and Yukatek, it is still possible that the average rate of ergative production follows a similar developmental pattern across the languages. Figure 27.1 displays the average production of ergative marking for each language. A chi square analysis of the ergative production summed across the three children in each of the four languages shows a statistically significant difference in ergative marking on transitive verbs (χ2 = 33.03, p < .01). This analysis demonstrates a significant difference in the children’s development of ergative marking across these four Mayan languages. Despite their formal similarity, the ergative markers present language-​specific learning problems of different magnitudes for Mayan children. 100.0%

Percent Use in Obligatory Contexts

90.0% 80.0% 70.0% 60.0% 2;0

50.0%

2;6

40.0%

3;0

30.0% 20.0% 10.0% 0.0%

Wastek

Yukatek

Ch'ol

K'iche'

Figure 27.1  Average production of preconsonantal ergatives

A pair-​wise test of the differences in ergative production using the chi-​square test shows that the Wastek results do not differ significantly from the results for Ch’ol and K’iche’ (Wastek and Ch’ol χ2 = 2.69, p > .05; Wastek and K’iche’ χ2 = 2.76, p > .05). The

 

678    Clifton Pye and Barbara Pfeiler comparison between Wastek and Yukatek was significant (χ2 = 27, p < .01). The Yukatek children exhibit a steady increase in ergative production from 2;0 to 3;0, whereas the children acquiring the other Mayan languages did not exhibit a marked increase in ergative production over this time period.

27.4.3 Testing the Ergative as Nominative Agreement Hypotheses In this section, we compare the ergative results with the production of person markers in French and Spanish to determine whether Mayan children acquire them as agreement markers or clitics. We examine whether any of the Mayan agreement profiles match children’s production of nominative agreement in Spanish or children’s production of nominative clitics in French. Table 27.9 shows the total number of ergative tokens the children produced as well as the average percentage of ergative production on transitive verbs in the four Mayan languages. Table 27.9 Ergative tokens and average percentage ergative production in obligatory contexts Wastek

Yukatek

Ch’ol

K’iche’

Age

Number

Percent

Number

Percent

Number

Percent

Number

Percent

2;0

82

29.7

17

20.4

9

11.3

7

11.9

2;6

131

33.9

69

50.9

7

38.9

22

8.3

3;0

358

54.2

118

94.1

37

33.9

42

13.4

The nominative hypothesis stated in (8) predicts that children acquiring Mayan languages would produce ergative person markers as nominative agreement markers in over 90% of their obligatory contexts at age 2;0. The nominative hypothesis stated in (8) predicts that children acquiring Mayan languages would produce ergative person markers as nominative clitics in approximately 50% of obligatory contexts between 2;0 and 2;6. Table 27.10 compares the percentage person marker production in obligatory contexts in the Mayan languages with the production in Spanish and French. Table 27.10 Average percentage person marker production in obligatory contexts Age

Spanish

French

Wastek

Yukatek

Ch’ol

K’iche’

2;0

90

51.5

29.7

20.4

11.3

11.9

2;6

45.5

33.9

50.9

38.9

8.3

3;0

89.1

54.2

94.1

33.9

13.4

 

The acquisition of nominative and ergative alignment    679 Table 27.10 shows a clear difference between children’s nominative person marker production in Spanish and the Mayan children’s production of ergative person markers. We interpret this difference as evidence that children do not acquire the ergative person markers on transitive verbs as nominative agreement markers. Table 27.10 also shows a clear difference between children’s nominative clitic production in French and the Mayan children’s production of ergative person markers. The Yukatek results at 2;6 and 3;0 come closest to the French clitic production, but differ at 2;0. We interpret these differences as evidence that children do not acquire the ergative person markers on transitive verbs as nominative clitics as in French. We discuss further implications of these results in the conclusion to our chapter. First, we present the results for the absolutive person markers in the following section.

27.5  The Acquisition of Absolutive Person Marking in the Mayan Languages We next turn our attention to the acquisition of the absolutive person markers in this section. First we describe the absolutive person markers and then analyze their acquisition. The final part of this section compares the children’s absolutive marker production with predictions from the nominative agreement hypothesis.

27.5.1 Absolutive Person Marking in the Mayan Languages The absolutive person markers in Mayan languages also have properties of both affixes and clitics. The Mayan absolutive markers are traditionally described as clitics in the literature, e.g. Kaufman (1990a: 71). Woolford (2000: 172) identifies the absolutive markers in the Mayan language Popti’ as nominative clitics. Following Legate (2008), Coon et al. (2014) claim that the absolutive markers on intransitive verbs in Mayan languages receive Nominative Case from Tense. They do not discuss the clitic status of the absolutive markers. The absolutive morphemes generally cross-​reference the subjects of intransitive verbs, objects of transitive verbs and the subjects of nonverbal predicates. The third person absolutive marker is a zero morpheme in each language, although Wastek also has non-​zero allomorphs of the third person absolutive marker. We list the absolutive morphemes for the four languages in Table 27.11.

 

680    Clifton Pye and Barbara Pfeiler Table 27.11 Absolutive person forms in Wastek, Yukatek, Ch’ol, and K’iche’ Number

Person

Wastek

Yukatek

Ch’ol

K’iche’

Singular

1

in

en

oñ

in

2

at

ech

ety

at

3

ø~u~i

ø

ø

ø

1 inclusive

u

o’on-​e’ex

la … oñ

uj

1 exclusive

u

o’on

lojoñ

uj

2

it … (chik)

e’ex

la … ety

ix

3

ø~u~i… (chik)

o’ob

ob

e

Plural

The absolutive markers precede the verb stem in Wastek and K’iche’, but follow the verb stem in Yukatek and Ch’ol. We provide examples of intransitive verbs with absolutive marking in (9). (9)

Examples of Mayan intransitive verbs a Wastek ʔin k’al-​e-​l 1sg.abs go-​it-​inc ‘I go.’

(Edmonson 1988: 117)

b Yukatek h ween-​en com sleep-​1sg.abs ‘I slept.’ c Ch’ol tyi ts’äm-​iy-​oñ com bathe-​it-​1sg.abs ‘I bathed.’ d K’iche’ k at inc 2sg.abs ‘You arrive.’

ul-​ik arrive-​ind.iv

The absolutive markers are suffixes in Ch’ol and Yukatek. The absolutive markers in Wastek can be separated from the verb stem by the same adverbs that separate the

 

The acquisition of nominative and ergative alignment    681 ergative markers in Wastek from the verb stem. Larsen (1988: 153–​158) discusses the clitic properties of the absolutive person markers in K’iche’. He concludes that the absolutive person markers in K’iche’ are clitics partly on the basis that movement clitics can be inserted between the absolutive person marker and the verb stem and introduce a dependent suffix on the verb stem (10). (10) Movement clitic in K’iche’ (based on Larsen 1988: 181) x=at=ul chakun-​oq com=2sg.abs=come work-​dep.iv ‘You came to work.’ We summarize the affix/​clitic properties of the absolutive markers in Wastek, Yukatek, Ch’ol, and K’iche’ in Table 27.12. This table demonstrates that the properties of the absolutive markers differ across these four Mayan languages. The absolutive markers in Wastek and K’iche’ have more clitic properties than the absolutive markers in Yukatek and Ch’ol. The absolutive markers in Wastek allow adverb insertion, while the absolutive markers in K’iche’ allow the insertion of movement verbs. At the same time the third person absolutive marker in Wastek is the only absolutive form that has allomorphs, whereas the absolutive markers in the other three languages do not. Table 27.12 Affix/​clitic properties of absolutive person markers in four Mayan languages Language

posthead position

allomorphs

phonetic merger with aspectual verbs

adverb/​movement verb insertion

Wastek Yukatek

Clitic Affix

Affix Clitic

Affix Affix

Clitic Affix

Ch’ol

Affix

Clitic

Affix

Affix

K’iche’

Clitic

Clitic

Clitic

Clitic

The zero marked third person absolutive marker in Mayan languages greatly reduces the number of intransitive verbs that have overt absolutive person marking. Yukatek and Ch’ol also extend ergative markers to mark subject agreement on intransitive verbs in incompletive aspects (Larsen and Norman 1979). Frequent extended ergative marking gives the appearance of an ordinary nominative-​accusative alignment. As Bricker (1981:  84)  observed for Yukatek, extended ergativity occurs so frequently that some linguists have analyzed the extended (accusative) paradigm as ‘basic.’ We analyzed the

 

682    Clifton Pye and Barbara Pfeiler children’s acquisition of extended ergative marking elsewhere (Pye et al. 2013), and so restrict our analysis in this section to the children’s acquisition of absolutive subject marking on intransitive verbs. Ch’ol differs from the other three languages in distinguishing between an agentive and non-​agentive class of intransitive verbs. Ch’ol restricts absolutive marking to the non-​ agentive class of intransitive verbs and uses ergative markers on the light transitive verb cha’l ‘do’ to indicate the subject of agentive intransitive verbs (11). The agentive intransitive verb class further reduces the use of the absolutive subject markers in Ch’ol relative to the other three languages. (11) Agentive intransitive verb in Ch’ol (Kaufman 1990a: 104) tyi=k cha’l-​e k’ay com=1sg.erg do-​tt song ‘I sang.’ The acquisition results from French and Spanish can potentially provide independent evidence of the clitic status of absolutive person marking in these Mayan languages. Assuming that Mayan children acquire the absolutive person markers as nominative agreement markers, we expect them to produce the absolutive markers in over 90% of their obligatory contexts at 2;0. Assuming that Mayan children acquire the absolutive person markers in these languages as nominative pronominal clitics, we expect them to produce the absolutive markers on approximately 50% of verbs up to 2;6 and in 80% of verbs at 2;9. This analysis predicts that these children will produce absolutive subject clitics at a rate that is similar to the production of subject clitics in French. We summarize these predictions in (12). (12)

Predictions for the acquisition of absolutive person markers in Mayan languages    a Mayan children acquire absolutive person markers as nominative agreement markers and produce them in 90% of obligatory contexts at 2;0.    b Mayan children acquire absolutive person markers as nominative clitics and produce them in approximately 50% of obligatory contexts between 2;0 and 2;6.

27.5.2  The Acquisition of the Absolutive Person Markers The children’s absolutive results are shown in Table 27.13.

 

The acquisition of nominative and ergative alignment    683 Table 27.13 Absolutive production on intransitive verbs 2;0 Language

Child

Wastek

Yukatek

Ch’ol

K’iche’

2;6

3;0

n/​contexts

%

n/​contexts

%

n/​contexts

%

ELV

3/​19

15.8

7/​23

30.4

32/​67

47.8

CAR

2/​4

50

4/​8

50

2/​3

67

VLA

3/​4

75

5/​8

63

7/​15

47

ARM

11/​13

91

7/​7

100

7/​7

100

SAN

7/​7

100

7/​7

100

DAV

4/​4

100

0/​1

0

1/​2

50

2/​2

100

EMA

1/​1

100

MAR

0/​1

0

MA

3/​3

100

TIY

4/​11

36

LIN

3/​11

27

CHA

8/​20

40

19/​52

36

0/​11

0

12/​43

28

The children did not produce enough absolutive tokens to meet the requirements of the chi square test. For this reason, we relied upon qualitative assessments of their accuracy in producing the absolutive markers as measured by the percent correct production in obligatory contexts. The Yukatek and Ch’ol children generally had a more accurate absolutive production than the Wastek and K’iche’ children. The children acquiring Ch’ol limited their production of the absolutive suffixes to a few contexts, notably the evidential verbs che’oñ ‘I say’ and che’ety ‘You say’ (13). These verbs are exceptional in that they are added to the end of an utterance, whereas Mayan speakers usually put verbs at the beginning of an utterance. The use of the absolutive suffixes on these verbs violates the extended ergative constraint in Ch’ol, which limits the use of absolutive clitics on intransitive verbs to the completive aspect. Ch’ol children probably acquire the evidential verb as an unanalyzed form outside of their developing absolutive grammar. The Ch’ol children did not produce absolutive markers on other intransitive verbs in incompletive contexts.

 

684    Clifton Pye and Barbara Pfeiler (13)

EMA (2;1.14) Sister mu’tyo ich’ämo’   ma achi’es    tyiya. mu’=tyo i-​ch’äm-​o’ ma a-​ chi’es tyiya inc=still  3sg.erg-​carry-​tt go 2sg.erg-​nance  aunt ‘They are still going to carry your nances aunt.’ EMA chety. = che’-​ety say-​2sg.abs ‘You say.’

The absolutive data in Table 27.13 show that extended ergativity in Yukatek and Ch’ol as well as the class of agentive intransitive verbs in Ch’ol had a minor impact on the number of absolutive tokens the children produced. We had expected the children acquiring Wastek and K’iche’ to produce many more absolutive tokens than the children acquiring Yukatek and Ch’ol. Up through the age of 2;6 the children acquiring all four languages produced similar numbers of absolutive tokens. There is a difference in the number of contexts of use for the absolutive markers between the languages. As expected, the children acquiring Wastek and K’iche’ produced more utterances that required absolutive marking on average than the children acquiring Yukatek and Ch’ol. Only the Yukatek child ARM approached the Wastek and K’iche’ children in the number of absolutive contexts of use although the Wastek child ELV and the K’iche’ child TIY far surpass ARM at 2;6. Although children acquiring Yukatek and Ch’ol had relatively few opportunities to produce absolutive markers, they still produced as many absolutive tokens as the children acquiring Wastek and K’iche’. Obviously, the relative frequency of absolutive marking in the four languages did not predict the children’s use of the absolutive person markers. Pye et al. (2013) provide further information on the lack of an effect of the input frequencies. One feature of the children’s absolutive production that is similar across the languages is their constant rate of production over the one-​and-​a-​half-​year period. The children acquiring Yukatek began producing a high rate of absolutive marking at 2;0, and do not change, whereas the children acquiring Wastek and K’iche’ produce absolutive marking at lower rates, but do not show much improvement over the period between 2;0 and 3;0.

27.5.3 Testing the Absolutive as Nominative Agreement Hypothesis In this section, we compare the Mayan children’s production of the absolutive person markers with the production of person markers in French and Spanish. Table 27.14 shows the total number of absolutive tokens the children produced as well as the average percentage of absolutive production on intransitive verbs in the three age periods in the four Mayan languages.

 

The acquisition of nominative and ergative alignment    685 Table 27.14 Absolutive tokens and average percentage absolutive production in obligatory contexts Wastek

Yukatek

Ch’ol

K’iche’

Number

Percent

Number

Percent

Number

Percent

Number

Percent

2;0

8

29.6

11

91

4

80

7

31.8

2;6

16

41

18

100

8

25.8

3;0

12

48.2

14

100

31

32.8

3

75

We predicted in (12) that Mayan children acquiring the absolutive person markers as nominative agreement markers would produce them in over 90% of their obligatory contexts at age 2;0. We also predicted that Mayan children acquiring the absolutive person markers as nominative clitics would produce them in approximately 50% of obligatory contexts between 2;0 and 2;6. Table 27.15 compares the percentage person marker production in obligatory contexts in the Mayan languages with the production in Spanish and French. The results in Table 27.15 support the hypothesis that Yukatek children acquire the absolutive person markers as nominative agreement markers similar to children acquiring nominative agreement suffixes in Spanish. The results are less clear for the children acquiring Wastek and K’iche’. The Wastek and K’iche’ children clearly differ from children acquiring Spanish and Yukatek in their production of absolutive person markers. This difference supports the hypothesis that Wastek and K’iche’ children do not acquire absolutive person markers as nominative agreement markers. Table 27.15 Average percentage person marker production in obligatory contexts Age

Spanish

French

Wastek

Yukatek

Ch’ol

K’iche’

90

51.5

29.6

91

80

31.8

2;6

45.5

41

100

3;0

89.1

48.2

100

2;0

25.8 75

32.8

The development of the Wastek and K’iche’ children’s production of the absolutive person markers does not match that for the French nominative clitic markers either. The Wastek and K’iche’ children’s absolutive production lags behind the production of the French clitics. Even 3-​year-​old Wastek and K’iche’ children continue to produce relatively few numbers of absolutive person markers in their obligatory contexts. We interpret these results as indicating that the acquisition of absolutive person markers

 

686    Clifton Pye and Barbara Pfeiler in Wastek and K’iche’ is distinct from the development of nominative agreement in Spanish and nominative clitics in French. We conclude that children acquiring absolutive person markers in Mayan languages follow at least two distinct developmental routes. Children acquire absolutive person markers in Yukatek in a manner that closely matches that of children acquiring nominative agreement suffixes in Spanish. Children acquire absolutive person markers in Wastek and K’iche’ in a way that differs from both the pattern for nominative agreement markers and nominative clitics. Unfortunately, there was not sufficient data to assess the acquisition of absolutive person marking in Ch’ol. We discuss the implications of these results in the conclusion to our chapter.

27.6 Conclusion We discussed the clitic properties of the ergative and absolutive markers in some detail in order to introduce some of the features of person marking in Mayan languages that fall outside the scope of acquisition theories that focus exclusively on the functional architecture of syntactic representations. Clitics have long posed a problem for linguistic theory precisely because of the ways that clitics disrupt the parallel between phonetic and morphosyntactic representations (Spencer and Luís 2012). We suggest that acquisition data provides independent evidence for the clitic status of person markers. We outlined the affix/​clitic properties of the ergative person markers in Table 27.7. The ergative person markers are described as affixes in the grammars of Mayan language, although Lehmann (1993) analyzes the markers as enclitics in Yukatek. With the exception of Wastek, the ergative markers have prevocalic and preconsonantal allomorphs. With the exception of Wastek, the preconsonantal ergative markers merge phonetically with preceding aspectual markers. With the exception of K’iche’, the languages permit adverb insertion between the ergative markers and the verb stem. The overall similarity of the affix/​clitic properties of the ergative markers would predict that Mayan children acquire them in a uniform fashion. As shown in Table 27.16, the acquisition data provide evidence that the ergative person markers present distinct challenges to children acquiring these four Mayan languages. The different acquisition patterns do not support a uniform analysis of the ergative person markers as either agreement affixes or clitics. The Mayan acquisition data differs substantially from the acquisition of agreement marking in Spanish as well as from the acquisition of pronominal clitics in French. The Mayan acquisition data demonstrate that an understanding of ergative person marking in these languages is more complex than a simple contrast between agreement and clitic predicts.

 

The acquisition of nominative and ergative alignment    687 Table 27.16 Average percentage ergative and absolutive production in four Mayan languages Wastek

Yukatek

Ch’ol

K’iche’

Age Ergative Absolutive Ergative Absolutive Ergative Absolutive Ergative Absolutive 2;0

29. 7

29.6

20.4

91

11.3

2;6

33. 9

41

50.9

100

38.9

3;0

54. 2

48.2

94.1

100

33.9

80

75

11.9

31.8

8.3

25.8

13.4

32.8

We outlined the affix/​clitic properties of the absolutive person markers in Table 27.12. The absolutive person markers are described as clitics in grammars of the Mayan languages (Kaufman 1990a), and have notable differences across the four Mayan languages. The absolutive person markers precede the verb stem in Wastek and K’iche’, but follow the verb stem in Yukatek and Ch’ol. The absolutive person markers in Wastek and K’iche’ can be separated from the verb stem by an intervening adverb or movement verb, whereas this is not the case in Yukatek and Ch’ol. The absolutive markers in K’iche’ merge phonetically with preceding aspectual markers, whereas this is not the case in Yukatek and Ch’ol. These properties predict a similar acquisition of the absolutive markers in Wastek and K’iche’ that is distinct from the acquisition pattern for Yukatek and Ch’ol. This prediction is supported by the results shown in Table 27.16. The Mayan data support distinct acquisition patterns for the absolutive person markers. While the acquisition data for Wastek and K’iche’ are not similar to the acquisition of person marking in either Spanish or French, the acquisition data for Yukatek suggests that the absolutive markers in that language are acquired as agreement affixes. Children acquiring the Mayan languages Tzeltal and Tzotzil also produce absolutive verb suffixes at high rates of accuracy (Brown et al. 2013), which indicates that the absolutive suffixes in those languages are also acquired as agreement markers. Theories that identify the ergative or absolutive markers with nominative person marking in European languages do not predict the full range of the Mayan results. The Mayan acquisition data indicate that ergative person marking in all four languages and absolutive person marking in Wastek and K’iche’ differs categorically from nominative person marking in French and Spanish. Only the acquisition of the absolutive markers in Yukatek, Tzeltal, and Tzotzil would support an analysis of absolutive suffixes in those languages as nominative agreement markers. Massam (2006) concluded that Niuean ergative and absolutive case markers could not be identified with nominative case markers. One explanation for our results may be that too many variables intervene between the children’s grammar and their actual realizations to make acquisition data useful as an independent test of grammatical theory. After all, we found significant variation

 

688    Clifton Pye and Barbara Pfeiler between children acquiring the ergative person markers in Wastek and Yukatek. The children’s abilities and activities during the recording sessions may have contributed to their frequency of morpheme production. This conclusion implies that all theoretical claims for acquisition data should be abandoned as acquisition data is too noisy to support structural theories. In particular, such a conclusion calls into question acquisition research on European languages that claims to support structural accounts of tense and agreement (e.g. Haegeman 1998; Wexler 1998; Grinstead 2004). We are not ready to abandon acquisition data as a source of linguistic insight into the structure of the adult languages. Inspection of children’s data from the European and Mayan languages suggests that the children’s person marking in specific languages varies within a restricted range, and that the differences between the languages are real. A gap between theory and data suggests that rather than questioning the acquisition data, questions should be raised about the structural orientation of some acquisition theories. The developmental profiles demonstrate that children have a more varied production than a simple contrast between nominative and ergative marker would predict. The focus on structural relations between morphemes neglects a myriad of other factors that affect the realization of person markers in children’s speech. Bybee (2010) suggests that grammaticalization processes provide a better understanding of the adult grammar as well as the forms that children produce. We began with the search for the cross-​linguistic identity of nominative agreement. The Mayan languages with their complex verb structure reveal features of grammaticalization that remain obscure in European languages with simple stem-​affix structures. Our results show that children acquire person marking differently in our sample of four Mayan and two European languages and does not support a categorical identification of nominative and absolutive person markers or a categorical distinction between absolutive and ergative markers. We suggest, instead, that the grammaticalization of person markers as determined by their specific combination of clitic and affix properties predicts children’s production of the person markers more accurately than their categorical status as absolutive or ergative, clitic or affix. Our results demonstrate how acquisition data can contribute to a better understanding of the adult languages. Future research is necessary to determine the precise factors involved in the grammaticalization of clitics and their effect on language acquisition.

Acknowledgments Data collection for Wastek and Yukatek was funded by the Consejo Nacional de Ciencia y Tecnología of Mexico (SEP-​CONACYT:  105596, 4639-​H, and 27893-​H) and from the Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica-​Universidad Nacional Autónoma de México (PAPIIT-​UNAM IN401207) to the second author. Data collection for Ch’ol was funded by grants from the National Science Foundation (BCS-​0613120 and BCS-​0515120) and the General Research Fund of the University of Kansas to the first author. Data collection for K’iche’ was supported by grants from the Organization of American States and the Wenner-​Gren Foundation to the first author. All of these projects would not have

 

The acquisition of nominative and ergative alignment    689 been possible without the aid and support of the children’s families as well as the efforts of the team of Wastek investigators: Leonarda Hernández Gutiérrez, Magdalena Martínez Enríquez, Alicia Hernández Martínez, and Andrés Cruz Cruz; the Yukatek investigators: Neifi Vermont Vermont, Andrés Dzib Dzib, and Adiel Mena Keb; the Ch’ol investigators: Pedro Gutiérrez Sánchez, Asunción López Pérez, and Melba del Carmen Martínez Pérez; and the K’iche’ investigators Augustin Huix Huix, Pedro Quixtan Poz, Emilio Quiej Huix, and Santos Quiej Huix. We would also like to thank the editors of this volume for suggestions that have substantially improved our initial drafts. They are not responsible for any remaining errors.

Abbreviations 1, First person; 2, Second person; 3, Third person; abs, Absolutive; com, completive aspect; comp, complementizer; dep, Dependent mood; sg, Singular; der, Derived suffix; erg, Ergative; imp, Imperative mood; inc, Incompletive aspect; ind, Indicative mood; it, Intransitive thematic suffix; iv, Intransitive verb; neg, Negation; present, Present tense; term, Terminative aspect; ti, Complementizer; tt, Transitive thematic suffix; tv, Transitive verb.

 

 

Experimental

 

 

Chapter 28

Pro cessing erg at i v i t y: Behaviora l a nd e l ectrophysi ol o g i c a l eviden c e Adam Zawiszewski

28.1 Introduction Theoretical aspects of ergativity have been widely studied in linguistics (Dixon 1994; Johns, Massam, & Ndayiragije 2006; Laka & Fernández 2012; Legate 2012a), but less is known about its representation and processing. Here I present an overview of works that explored ergativity from an experimental perspective. The chapter is organized as follows: I first describe the methodology the studies are based on; next, I analyze the literature on the experimental aspects of ergativity and discuss the findings reported therein; and finally I end by highlighting the implications of the results for further research on ergativity.

28.1.1 Language Processing: A Short Introduction to Research Methodology Experimental studies on ergativity mentioned in this section are based on three experimental methods: self-​paced reading, event-​related potentials (ERPs) and functional magnetic resonance imaging (fMRI).1 In the self-​paced reading technique 1 

This section focuses only on the methodology used to investigate ergativity. For other approaches, such as the eye-​tracking, magnetoencephalography (MEG), positron emission tomography (PET), or transcranial magnetic stimulation (TMS), see, for instance Ahlsén 2006; Traxler & Gernsbacher 2006; Bornkessel-Schlesewsky & Schlesewsky 2009.

 

694   Adam Zawiszewski participants are required to read sentences (= stimuli) on a computer screen, word-​ by-​word (or phrase-​by-​phrase) by pressing a button. At each button press only one piece of the sentence (word /​phrase) is shown while the rest remains hidden (words are replaced by asterisks, lines, or dashes). The task of the participant is to read sentences in a most natural (fast) way and the time he/​she spends reading each word (or phrase) is assumed to reflect its processing cost. The subsequent steps when reading the sentence Yesterday the director arrived late to work are illustrated by the following example: (1)

********* *** ******** ******* **** ** ****. Yesterday *** ******** ******* **** ** ****. ********* the ******** ******* **** ** ****. ********* *** director ******* **** ** ****. ********* *** ******** arrived **** ** ****. ********* *** ******** ******* late ** ****. ********* *** ******** ******* **** to ****. ********* *** ******** ******* **** ** work.

There are different modalities of this technique: the stimuli can be displayed in a cumulative or non-​cumulative way, that is, the items read by the participants may remain revealed while the following ones are being successively uncovered (cumulative), or they can be hidden again, meaning that only one segment is visible to the participant at time (non-​cumulative), as shown in the example (1). The usefulness of self-​paced reading technique has been consistently confirmed in many psycholinguistic experiments -​it has been shown, for instance, that temporarily ambiguous, complex, or ungrammatical sentences require significantly more time to read than non-​ambiguous, simple or grammatical sentences (Mitchell 1994; Trueswell & Kim 1998; Pickering & Van Gompel 2006). Longer reading times are interpreted in terms of larger processing cost of a given structure (word, phrase, sentence, etc.), while shorter reading times reflect smaller processing cost (for a more detailed description, the history of this technique and the discussion, see Marinis 2003; Jegerski 2014). Event-​ related brain potentials (ERPs) are another technique widely used in order to investigate language processing (for an overview, see Kutas, Van Petten, & Kluender 2006; Kutas & Federmeier 2007). Given their high temporal resolution (milliseconds), they are an appropriate tool to measure the time-​course of the processes underlying language comprehension and production. Technically speaking, ERPs are potential changes in the electroencephalogram (EEG), triggered by sensory or cognitive events—​in the case of language—​words, phrases, or sentences (Kaan 2007; Bornkessel-Schlesewsky & Schlesewsky 2009). In a standard setup experiment, the EEG is recorded from a set of electrodes (e.g. 32, 64, 128, or more) secured in an elastic cap while a participant is reading or listening to the stimuli. These potential changes are averaged over a large number of trials of the same type yielding the ERP. After the averaging procedure has been performed for each participant, a “grand average”

 

Processing ergativity   695 is computed over these individual averages. The final results are interpreted on the basis of these “grand averages.” An example of such a design is illustrated in the Figure 28.1. ONGOING EEG AMPLIFIER

S1

S2

S1

S2

S1

S2

S1

ONE SECOND EVENT RELATED POTENTIAL –8µV

STIMULI (S) SIGNAL AVERAGER

S1 S2

+8µV STIMULUS ONSET

200

400

600

800

1000

TIME (MILLISECONDS)

Figure  28.1  Typical setup of an ERP experiment. Brain activity is recorded and amplified (ongoing EEG) while the participant is presented with visual or auditory stimuli (S). The averaged stimuli (S1 and S2) are represented as voltage changes in microvolts (y-​axis) over time (x-​axis). Positive potential changes are plotted downward and negative potential changes are plotted upward.

The waveforms are distinguished depending on their polarity (positive or negative), latency (milliseconds) and topography (scalp distribution) (Kaan 2007; Bornkessel-Schlesewsky & Schlesewsky 2009). For instance, as shown in Figure 28.1, the stimulus of the second type (S2) elicited a larger negativity than the stimulus of the first type (S1) and this difference is largest around 400 milliseconds after the stimulus onset. This type of wave has been labeled a N400. There is also a difference between both waves starting about 600 milliseconds after the stimulus onset, that is, the S2 elicited larger positivity in comparison to the S1. Following the conventional labeling, this waveform is described as a P600. Several ERP components (waveforms corresponding to specific processes) have been identified in relation to language, but in the subsequent section I will briefly describe the most studied ones: Left Anterior Negativity (LAN), N400, and P600. LAN, starting about 300 ms after the stimulus onset and observed over the left anterior site of the scalp, has usually been reported in syntactically illicit or complex contexts (Neville et al. 1991; Münte, Heinze, & Mangun 1993; Friederici 1995; Fiebach,

 

696   Adam Zawiszewski Schlesewsky, & Friederici 2002), rule violations (Ullman 2004), as well as in long-​distance dependencies (Phillips, Kazanina, & Abada 2005). LAN has been also claimed to index working memory load (Martín-​Loeches et al. 2005) and has been found in semantically reversible sentences (Meltzer & Braun 2013). However, the nature of this component is still not well known, as many studies failed to identify it in syntactically ungrammatical structures (for more details, see Molinaro, Barber, & Carreiras 2011; Molinaro et al. 2014; Tanner & Van Hell 2014; Tanner 2014). The second component mentioned, the N400, is mostly distributed over centro-​ parietal electrodes and similarly to LAN, starts around 300 ms after the stimulus onset. This negativity has been generally interpreted as a response to semantic, pragmatic, or thematic hierarchy violations or in general, as a correlate of non-​rule-​based lexically stored information (Kutas & Hillyard 1980; Kutas & Federmeier 2000; Frisch & Schlesewsky 2001, 2005; Hagoort et al. 2004; Lau, Phillips, & Poeppel 2008; Choudhary et al. 2009). Finally, the P600 is a positive component, distributed over the parietal electrodes and starting about 500 ms after the stimulus onset. It has been assumed to reflect reanalysis or integration processes taking place when syntactically ungrammatical, ambiguous, or structurally complex information is being parsed (Osterhout & Holcomb 1992). Besides self-​paced reading and ERPs, the functional magnetic resonance imaging method (fMRI) has been also applied to investigate different aspects of language processing. In a nutshell, the fMRI measures changes of neuronal activity related to an increase and decrease in local blood flow and volume, that is, it captures the changes in brain areas according to the level of oxygenation in the blood (Traxler & Gernsbacher, 2006; Bornkessel-​Schlesewsky & Friederici, 2007; Hunt & Thomas, 2008). Given that oxygenated hemoglobin has little effect on the magnetic field while the deoxygenated hemoglobin leads to a higher distortion in the magnetic field, the ratio of oxygenated to deoxygenated blood in a given volume (voxel) is reflected by the so-​called “blood oxygen-​level dependent contrast” (BOLD) (Bornkessel-​Schlesewsky & Schlesewsky 2009). This BOLD contrast is obtained by subtracting the activation in a control condition from that in an experimental condition. As a result, the pictures of the brain are displayed where the colors reflect the probability associated with the difference between two conditions. In a standard fMRI experiment the participants are presented (either aurally or visually) with stimuli while lying inside the scanner. During the experimental session they are required to perform a task (e.g. reading /​listening /​speaking /​pressing a button /​resting, etc.) and at the same time the scanner takes pictures of the brain. Despite its poor temporal resolution, the fMRI method is very accurate to detect where the changes occur. So far researchers have been able to measure and localize some language-​related processes, such as the sensitivity of distinct brain regions to the syntactic and semantic processing of speech (Friederici, Meyer, & von Cramon, 2000; Meyer, Friederici, & von Cramon 2000), the representation of the native and second languages in the human cortex (Kim, Relkin, Lee, & Hirsch 1997) or the effects of age of acquisition and proficiency

 

Processing ergativity   697 level on neural correlates of grammatical and semantic judgments in bilingual populations (Wartenburger et al. 2003). When it comes to the specific areas of the brain related to language, the data acquired from many studies suggest that mostly two regions are engaged in both language comprehension and production: left posterior superior temporal gyrus (STG, corresponding to “Wernicke’s area”) and left inferior frontal gyrus (IFG, corresponding to “Broca’s area”). Although these are not the unique brain areas involved during language processes (see, for instance Binder et al. 1997; Ullman 2001; Booth et al. 2007), in general the findings support the idea that language processing engages mostly a network of (left) fronto-​temporal brain regions (Bornkessel-​Schlesewsky & Schlesewsky 2009) (for an overview of the brain basis of language processing, see Bornkessel-Schlesewsky & Friederici 2007; Indefrey 2007; Friederici 2011).

28.2  Ergativity and Language Processing The experimental techniques mentioned in the previous section have successfully been used to investigate processes underlying different phenomena in language comprehension, such as verb agreement, case morphology, filler–​gap dependencies, word order, and many others (Coulson, King, & Kutas 1998; Fiebach, Schlesewsky, & Friederici 2001; Matzke et al. 2002; Erdocia et al. 2009; Molinaro et al. 2011). However, until recently, ergativity has not received much attention from a psycholinguistic or experimental perspective. The aim of the present chapter is to highlight and bring together the findings of those (few) studies that tackled ergativity from an experimental approach. This topic has been examined from different viewpoints—some works, as Choudhary et al. (2009) or Díaz et al. (2011), for instance, focused on the electrophysiological correlates of ergativity and tested whether and to what extent the ERP pattern corresponds to that reported previously for nominative-​accusative languages. Other studies, such as Carreiras et al. (2010) and Polinsky et al. (2012) used ergative languages in order to examine the universality of language processing strategies and tested whether subject relative clauses are universally easier to process than object relative clauses, as suggested by previous experimental data on nominative-accusative languages. Ergative languages have also been employed to test how language processing is influenced by syntactic and semantic cues and which cortical networks are involved in syntactic and semantic computation—as in the case of Dillon et al. (2012) and Nieuwland, Martin, & Carreiras (2012). Finally, in Zawiszewski et al. (2011) ergativity has been used as a testing ground to investigate to what degree native vs. non-​native differences in language processing are due to parametric differences between the first (L1) and the second (L2) languages.

 

698   Adam Zawiszewski

28.2.1 Electrophysiological Correlates of Ergativity Choudhary et  al. (2009) aimed to examine whether the assumption that syntactic processes are reflected by LAN components while lexical-​semantic processes are indexed by the N400 holds also for ergative languages such as Hindi, where case marking impacts on semantic interpretation. More precisely, they tested the hypothesis that some N400 components reported previously in the ERP literature (Frisch & Schlesewsky 2001; Haupt et al. 2008) might be engendered by syntactic information that is interpretatively relevant. To this end, they ran an ERP experiment in which the native speakers of Hindi listened to grammatical and ungrammatical sentences and performed a grammaticality judgment task. The participants were presented with the following type of sentences: (2) (a) shikshak maalii-​ko teacher-​NOM gardener-​ACC ‘The teacher sees the gardener.’

dekh-​taa hai see-​IPFV-​3SG.M AUX

(b) shikshak-​ne maalii-​ko *dekh-​taa teacher-​ERG gardener-​ACC see-​IPFV-​3SG.M ‘The teacher sees the gardener.’

hai AUX

(c) shikshak maalii-​ko *dekh-​aa teacher-​NOM gardener-​ACC see-​PFV-​3SG.M ‘The teacher has seen the gardener.’

hai AUX

(d) shikshak-​ne maalii-​ko dekh-​aa hai teacher-​ERG gardener-​ACC see-​PFV-​3SG.M AUX ‘The teacher has seen the gardener.’ In Hindi, an ergative-​marked argument (2b and 2d) receives an agentive reading in a perfective aspect while a nominative-​marked argument can be interpreted as an actor of a two-​participant event in an imperfective aspect, or an actor or undergoer in a one-​ participant event (e.g. the teacher is ill). Choudhary et al. (2009) follow Butt & Holloway King (2005) in the assumption that subjects are assigned nominative case by default while the ergative is a “semantic” case that is assigned under more restricted circumstances. The experimental design shown in the example (2)  allowed the authors to investigate whether the response elicited by the violation would be modulated by the misapplication of a default (nominative) vs. a non-​default (ergative) rule. They hypothesized that if the violation of rule-​based linguistic knowledge lead to LAN effects, both ungrammatical conditions should yield LANs as compared to their grammatical counterparts (2b vs. 2a = 2d vs. 2c). However, if the ERP correlate of a violation depended on the type of linguistic rule (default vs. non-​default), the incorrect usage of the nominative (default) (2c) should elicit a LAN in comparison to (2d), while the incorrect usage of the ergative (non-​default) (2b) would yield a N400 in comparison to (2a). Finally, if

 

Processing ergativity   699 the rule-​based information were interpretatively important, the corresponding violations should yield a N400. According to the literature, the authors also expected to find a P600 for both violation conditions (Osterhout & Holcomb, 1992). Electrophysiological results of the experiment revealed a N400 for both types of violations, followed by a broad positivity (P600) in the ungrammatical ergative as compared to the grammatical nominative condition (2b vs. 2a), but no such a positivity was observed when comparing the ungrammatical nominative condition to the ergative grammatical condition (2c vs. 2d). Importantly, the N400 effect was larger for the ungrammatical ergative than for the ungrammatical nominative condition suggesting that the interpretative problem caused by the rule misapplication was greater in the former than in the latter case. These findings lend support to the third hypothesis put forward by the authors, according to which syntactic processes that impact on semantic interpretation can elicit N400 effects. Also, the P600 elicited by the ungrammatical ergative condition as compared to the grammatical nominative condition suggests that late positive ERP effects may be highly sensitive to rule exceptions, that is, to occur only in response to principled incompatibilities between grammatical features such as case and aspect. In sum, these findings indicate that the dichotomy between the rule-​based and lexically stored information is not necessarily reflected by a LAN vs. N400 components. Rather, the ERP signature is determined by the type of syntactic information where formal rule violations such as subject-​verb agreement lead to a LAN, while interpretatively relevant rule violations, as subject case-​marking in Hindi, yield a N400. Neural language architecture was also the main topic studied by Díaz et al. (2011) who investigated the cross-​linguistic validity of electrophysiological correlates of morphosyntactic processing in Basque. During the ERP experiment native speakers of Basque listened to grammatical and ungrammatical sentences (double case violations) and were required to respond whether the sentences were grammatical or not. The sample of the materials is shown in (3): (3) (a) Mikelen arreb-​ek egunkari-​a saski-​an ekarri dute kiosko-​tik. Mikel’s sister-​ERG.PL newspaper-​ABS.SG basket-​in brought have kiosk-​from. (b) Mikelen arreb-​ek *egunkari-​ek saski-​an ekarri dute kiosko-​tik. Mikel’s sister-​ERG.PL    newspaper-​ERG.PL basket-​in brought have kiosk-​from. ‘Mikel’s sisters have brought the newspaper in a basket from the kiosk.’

While (3a) is grammatical, (3b) is not because the object of the sentence egunkariek ‘newspapers’ bears the same case mark -​ek as the subject arrebek ‘sisters’, yielding ungrammaticality. Double case violations tested previously in other (nominative-​ accusative) languages elicited a centro-​parietal negativity (N400) followed by the P600 component (Frisch & Schlesewsky 2001, 2005; Mueller, Hirotani, & Friederici 2007). Based on this evidence the authors aimed to examine whether the ERP correlates of ergative case processing are similar to those found in nominative languages. Behavioral results revealed no differences in accuracy between the grammatical and

 

700   Adam Zawiszewski the ungrammatical conditions (95.1% vs. 95.2%). In comparison to the grammatical sentences, double violations elicited larger positivity (P600) between 400 and 1250 ms at the critical word position (egunkaria ‘newspaper-​ABS.SG’ vs. egunkariek ‘newspaper-​ERG.PL’). In this sense, the results are similar to those reported for nominative languages (Frisch & Schlesewsky 2001; Mueller, Hahne, Fuji, & Friederici 2005). The authors did not replicate the N400 reported in Hindi by Choudhary et al. (2009) and argue that the differences between the results might be accounted for either by the differences between Hindi and Basque (Hindi is a split-​ergative language while Basque is not) or by the type of materials used in both experiments. In sum, according to the authors, the presence of the P600 component supports the view that ERP signatures engaged in the detection of case violations are similar across languages and do not depend on the argument alignment type.

28.2.2 Universal Processing Strategies and Ergativity: Subject vs. Object Relative Clauses Among the works that have tested universal processing strategies (Demiral, Schlesewsky & Bornkessel-​Schlesewsky 2008; Kwon et al. 2013), Carreiras et al. (2010) were the first who took advantage of ergativity to investigate whether subject relative clauses (SR) are universally easier to process than object relative clauses (OR), as suggested by previous results on the topic (King & Kutas 1995; Weckerly & Kutas 1999; Friederici et al. 2001; Kwon et al. 2013). To this purpose Basque, an ergative, head-​final language with pre-​nominal relative clauses, spoken in the northeastern Spain and southwestern France was used as a testing ground and the authors used both self-​paced reading moving window (Experiments 1 and 2) and ERP techniques (Experiment 3). The following conditions were compared: (4) (a) Irakasleak aipatu dituen ikasleak lagunak ditu. [e1 irakasle-​ak aipatu ditu-​en] ikasle-​a-​k1 lagun-​ak ditu [e1 teacher-​pl mentioned has-​rel] student-​sg-​S1 friend-​pl has ‘The student that mentioned the teachers has friends’ (b) Irakasleak aipatu dituen ikasleak lagunak dira. [irakasle-​a-​k e1 aipatu ditu-​en] ikasle-​ak1 lagun-​ak dira [e1 teacher-​sg-​S e1 mentioned has-​rel] student-​pl1 friend-​pl are ‘The students that the teacher mentioned are friends’

(SR)

(OR)

Both sentences are disambiguated toward a SR or OR interpretation at the auxiliary verb position. The results of the first experiment showed that OR sentences were easier to read than SR sentences, that is, the participants needed less time (115 ms) to read the auxiliary verb dira ‘are’ than the verb ditu ‘has.’ In the second experiment the authors used

 

Processing ergativity   701 the modified version of the stimuli from the previous study in order to make sure that the effects were not due to sentence final wrap-​up effects (Just & Carpenter 1980) and obtained very similar results: ORs were easier to process than SRs (a 109 ms difference). Finally, the aim of the last experiment was to provide more detailed evidence on the time-​ course of SR vs. OR processing by using ERPs. The data revealed significant differences between the conditions at the critical word position (dira ‘are’ vs. ditu ‘has’), that is, the waves elicited by subject relative clauses were more positive than those elicited by the object relative clauses. Given its latency and distribution, this effect was labeled as a P600. According to the authors, the advantage in processing object relatives over subject relatives found in Basque may be accounted by the fact that if morphological unmarkedness provides a processing advantage in language (Badecker & Kuminiak 2007; Laka 2012), then differences would be expected for ergative languages as compared to languages with the nominative-​accusative case system. More precisely, in Basque objects are the unmarked form and the (transitive) subjects are the marked one while in other tested languages (English, Spanish, Japanese, Korean) objects are the marked class and the subjects the unmarked one. This typological difference might thus explain the processing advantage reported for object relatives in Basque, in contrast to previous findings who revealed subject relatives to be processed with a greater ease than object relatives (see also Erdocia et al. (2009) for a similar advantage of object over subject processing in canonical and non-​canonical sentences in Basque). Altogether, these findings indicate that rather than being universal, subject /​object processing strategies are impacted by language specific aspects of grammar, in this case ergative-​absolutive vs. nominative-​accusative case alignment. Polinsky and collaborators (2012) (for an extended description, materials, and discussion see Longenbaugh and Polinsky, Chapter  29, this volume) provided another piece of evidence in order to clarify whether or not SRs are universally easier to process than ORs by testing subject preference and ergativity in Avar, a language spoken in the northwest and central regions of the Republic of Dagestan. The authors used the self-​paced reading (moving window) method to investigate how native speakers of Avar process ergative subject, absolutive object and absolutive subject relative clauses. The main aim of this study was to investigate whether subject gaps are easier to process than object gaps (regardless of the case form) or, whether the processing preference is driven by surface case considerations. Both hypotheses make different predictions: according to the former ergative subject and absolutive subject relative clauses should be easier to process than object absolutive clauses, whereas the latter predicts ergative subject vs. absolutive subject and absolutive object differences. Self-​paced reading times revealed no significant effect of case marking at the head noun region. Only a marginally significant effect of grammatical function was found, that is, the intransitive subject was read faster than the ergative subject and the absolutive object, and these two were read at similar rate. The same effects were observed at the subsequent region. In sum, in light of these results Avar does not show a processing difference between the ergative subject and the absolutive object. The authors

 

702   Adam Zawiszewski interpret these data by arguing that processing strategies in Avar are driven by two preferences:  the one for subject relatives and the other for morphologically cued gaps. The former makes the ergative and absolutive subject gap to be processed easier than the absolutive object gap while in the latter the ergative case serves as a cue which allows the parser to predict the structure of the clause (the missing absolutive object), making thus the absolutive object gap processing easier. Since these two preferences cancel each other out, the reading times for the ergative subjects and absolutive object relative clauses are similar. In sum, in light of these data the question whether or not subject relative clauses are easier to process than object relative clauses in Avar and in other ergative languages remains open and needs to be further tested (but note that the interpretation of the results is based on a null effect, and should be taken with caution).

28.2.3 Syntactic vs. Semantic Processing, and Ergativity Similarly to Choudhary et al. (2009), Dillon et al. (2012) also investigated the electrophysiological correlates of ergativity in Hindi, but under a slightly different approach. The main objective of this study was to examine the effects of syntactic and semantic cues on tense /​aspect processing. During the ERP experiment, native speakers of Hindi read sentences word-​by-​word on the screen and had to decide whether they were grammatical or not. The shortened version of the materials used by the authors is presented in (5): (5) (a) Haalaanki us bunkar-​ne (…) bun-​aa, (…) although that weaver-​ERG weave-​PFV ‘Although that weaver wove (…) , (…)’ (b) Haalaanki us bunkar-​ne (…) *bun-​e-​gaa, (…) although that weaver-​ERG weave-​AGR-​FUT ‘Although that weaver will weave (…), (…)’ (c) Haalaanki pichle shaam (…) gir-​aa, (…) although last night fall-​PFV ‘Although last night (…) fell, (…)’ (d) Haalaanki pichle shaam (…) *gir-​e-​gaa, (…) although last night fall-​AGR-​FUT ‘Although last night (…) will fall, (…)’ In the first two examples (5ab), the cue for the tense /​aspect of the verb is provided by the ergative case marker while in (5cd) the tense /​aspect of the verb is cued by the semantic information of the adverb. ERPs to grammatical and ungrammatical verb forms were measured and the results showed that in the conditions

 

Processing ergativity   703 where syntactic cues predicted verbal morphology (5ab), the tense /​aspect violations elicited a right-​lateralized anterior negativity (RAN) within 300–​500 ms time window, followed by a P600 component. In contrast, in the context where verbal morphology was predicted by semantic cues (5cd), the ungrammaticality yielded an early posterior negativity (200–​400 ms) and a small P600 effect (600–​800 ms). The comparison between both syntactic and semantic conditions revealed that the P600 was significantly larger and more broadly distributed in the syntactic (ergative) cue conditions than in the semantic (adverbial) cue conditions. According to the authors, these qualitative and quantitative differences in the response to the two types of cues indicate that the processing system is able to rapidly recognize and distinguish between different potential error causes and also suggest that the predictions generated by ergative case marking are stronger than those induced by the semantic cue condition. Another study that focused on syntactic vs. semantic aspects of language processing in an ergative language is that of Nieuwland et al. (2012) who used event-​related fMRI to investigate the cortical networks involved in ergative case, number agreement, and semantic processing in Basque. In this study native speakers of Basque read sentences shown in (7abcd) word-​by-​word in the scanner and judged their grammaticality by pressing a corresponding button. (6)

(a) Gizon-​a-​k lehiatila-​n jaso dit-​u sarrer-​ak Man-​the-​ERG.SG box office-​loc received them-​root-​he ticket-​the-​ABS.PL

goizean morning

(b) Gizon-​a-​k lehiatila-​n jaso dit-​u *sarrer-​ek Man-​the-​ERG.SG box office-​loc received them-​root-​he ticket-​the-​ERG.PL

goizean morning

(c) Gizon-​a-​k lehiatila-​n jaso dit-​u *sarrer-​a Man-​the-​ERG.SG box office-​loc received them-​root-​he ticket-​the-​ABS.SG ‘The man at the box office has received the tickets in the morning.’

goizean morning

(d) Gizon-​a-​k lehiatila-​n jaso dit-​u *begi-​a-​k Man-​the-​ERG.SG box office-​loc received them-​root-​he eye-​the-​ABS.PL ‘The man at the box office has received the eyes in the morning.’

goizean morning

(6a) is grammatical, as the morphology of both the subject (gizonak ‘the man’) and the object (sarrerak ‘tickets’) is correctly reflected in the auxiliary verb. On the contrary, (6b) is ungrammatical because it contains an incorrectly (ergative) case-​marked object sarrerek ‘'tickets’ instead of the grammatical absolutive object sarrerak ‘tickets.’ In (6c) the verb ditu ‘them-​root-​he’ does not agree in number with the object sarrera ‘ticket’ yielding ungrammaticality and in (6d) the object begiak ‘eyes’ does not fit in the previous context, making the sentence semantically implausible. This design allowed the authors to compare the areas of the brain involved in processing case (6b vs. 6a) and number agreement morphology (6c vs 6a) on the one hand, and semantic processing (6d vs. 6a) on the other hand.

 

704   Adam Zawiszewski Results from the grammaticality judgment task showed that participants responded more accurately and faster to ergative case violations than to other conditions (correct control, number agreement violations, semantic violations). As for the neuroanatomical correlates of the phenomena tested in the experiment, besides small differences, the overlapping neural circuits (parietal regions and left /​right inferior parietal gyri) were involved in processing both the ergative case and the number agreement suggesting that similar cortical networks are recruited during these processes. In contrast, different brain regions (left /​right anterior prefrontal gyri) were involved in semantic processing, indicating that syntactic and semantic processing rely on qualitatively different brain circuits. All in all, the data suggest that the neural consequences of a thematic problem generated by an ungrammatical use of ergative case are different from those engendered by a thematic problem in which an argument cannot bear the thematic role it is assigned.

28.2.4 Ergativity and Native vs. Non-​Native Language Processing In addition to native populations, ergativity has also been investigated among non-​ native speakers. Zawiszewski et al. (2011), for instance, used ERPs to examine the extent to which parametric differences between the first (L1) and the second (L2) language of the bilinguals influence the way the L2 is processed. To this purpose Basque-​Spanish and Spanish-​Basque bilingual populations were tested. Given that Basque and Spanish diverge with respect to the case system—Basque is ergative-​absolutive while Spanish is nominative-​accusative—the authors tested how native speakers of Basque and highly proficient Spanish-​Basque bilinguals (L2 learned before the age of 3) deal with grammatical and ungrammatical sentences, as these shown in the examples (7ab): ni-​k I-​ERG

denda-​n. shop-​in

(b) Goiz-​ean ogia erosi dut *ni Morning-​in bread bought have I-​ERG ‘This morning I bought bread in the shop.’

denda-​n. shop-​in

(7) (a) Goiz-​ean Morning-​in

ogia bread

erosi dut bought have

While (7a) is grammatical, in (7b) the subject of the sentence ni ‘I’ lacks the ergative marker (-​k), yielding ungrammaticality. The participants were required to read the sentences and to judge (by pressing a corresponding button) whether the sentences were grammatical or not. Behavioral results showed that the non-​native speakers were significantly less accurate than the native speakers (85% vs. 97%). The electrophysiological pattern also differed between both populations: among the L1 speakers of Basque the ungrammaticality elicited a N400 followed by a P600 component (cf. Choudhary et al. 2009 for similar findings) whereas in the L2 speakers only a N400 was found (see Figure 28.2).

 

Processing ergativity   705

Natives

Nonnatives

Fz –5 +5

Fz

µV 500

Cz

ms Cz

FZ CZ Pz

PZ

Pz

Grammatical Ungramatical

Figure 28.2  The ERP pattern elicited by the ergative case violations (dashed line) as compared to the grammatical sentences (continuous line) among the native speakers (left side of the panel) and the non-​native speakers of Basque (right side of the panel). The waves correspond to the critical word position (nik ‘I’ vs. *ni ‘I’). Negativity is plotted upward. The N400 component (similar in both groups) can be observed at the Cz electrode between 300 and 500 ms after the stimulus onset. The P600 component (more positive wave for the ungrammatical than for the grammatical condition) is present in the native group (Pz electrode, left side), but absent in the non-​native group (500–​800 ms time window).

These results indicate that despite high proficiency and early L2 acquisition onset, the non-​native speakers of Basque process the ergative case differently from the native speakers, as shown by both the behavioral and electrophysiological data. This, in turn, suggests that native vs. non-​native differences obtain if a syntactic parameter of the non-​ native grammar diverges from the native grammar (see also Ojima, Nakata, & Kakigi 2005 and Chen et al. 2007 for verb agreement; Weber-​Fox & Neville 1996 for subjacency effects). In sum, these results reveal that the case parameter is an important factor to be taken into account when comparing native and non-​native language processing, particularly in those circumstances where it takes a different value in the bilinguals’ first and second languages.

28.3  Discussion and Conclusion The purpose of the present chapter was to present those studies that approached ergativity from a psycholinguistic perspective. The work by Choudhary et al. (2009) showed that the violations of rule-​based linguistic knowledge do not necessarily lead to LAN effects. Rather, as suggested by the evidence from Hindi, the ERP pattern triggered by

 

706   Adam Zawiszewski the violations depends on whether the violated grammatical rule is interpretatively relevant or not. In other words, the processing of rule-​based linguistic knowledge correlates with an N400 when the misapplication of the rule has interpretative consequences, otherwise a LAN is expected as response to the violations of rules which do not induce comprehension problems. The results of this study were fully corroborated by the ERP data from native speakers of Basque (Zawiszewski et al. 2011), where ergative case violations elicited a similar N400–​P600 pattern to that reported for Hindi. In that sense, it seems that the violations of ergative case yield larger interpretative problems than case violations in nominative-​accusative languages, where usually left anterior negativities followed by the P600 component have been found (e.g. Coulson et al. 1998). On the other hand, the late positivity reported by Díaz et al. (2011) for double ergative case violations in Basque, found also in Choudhary et al. (2009) and Zawiszewski et al. (2011) indicates that repair and reanalysis processes do not depend on case alignment and are crosslinguistically uniform. On the other hand, the experimental data from subject and object relatives reviewed here seem to cast doubts on the theories claiming subject relative clauses to be universally easier to process than object relatives. As shown by Carreiras et  al. (2010) and contrary to previous findings in nominative-​accusative languages, in Basque the object relatives are processed with greater ease than the subject relatives. As argued by the authors, a plausible explanation of this finding lies on the fact that if morphologically unmarked arguments require less effort than marked arguments to be processed, subject relatives are processed faster than object relatives in nominative-​accusative languages, as subjects are morphologically unmarked while objects are marked. In contrast, in ergative languages a different pattern is expected:  given that the objects are an unmarked class, object relatives are easier to process than transitive (marked) subject relatives. This, in turn, challenges the hypothesis on the universality of subject over object preference and suggests that subject /​object processing strategies may vary depending on the case alignment. These data are similar to those reported by Gutiérrez-​ Mangado (2011) who revealed that 4 and 6 year children respond to ORs with greater accuracy than to SRs when performing a comprehension task in Basque. Likewise, Munarriz, Ezeizabarrena, & Gutiérrez-​Mangado (2014) tested a Spanish-​Basque Broca’s aphasic bilingual patient and reported higher accuracy in a comprehension task for ORs than for SRs in Basque. However, these findings were not confirmed by Polinsky et al. (2012) who did not observe a similar tendency in Avar. Although the authors explain the lack of contrast between subject and object relatives in terms of two (opposite) processing preferences which cancel each other out, further evidence is still needed in order to confirm either theory (for the discussion on the topic, see also Laka 2012). Within the third group of studies mentioned here Dillon et al. (2012) measured how verb processing is modulated by syntactic (ergative-​marked arguments) or semantic (tense adverbials) information in Hindi. With respect to the early effects, the negativity elicited by semantic cues was similar to a classic N400 component, while the ungrammatical use of the ergative marking yielded a right anterior negativity (RAN). The latter result differs significantly from previous evidence on ergativity violations (Choudhary

 

Processing ergativity   707 et al. 2009; Díaz et al. 2011; Zawiszewski et al. 2011). Although the authors interpret this negativity as an index of processing demands during the resolution of the morphological dependencies between the verb and its arguments (see also Ueno & Kluender (2009) for a similar reasoning when dealing with anomalies during the processing of Japanese wh-​questions), it is not clear why similar manipulations elicited an N400 in the study by Choudhary et al. (2009). The differences could be accounted by the type of materials used in both studies (simple vs. complex sentences) or to the presentation modality (aural vs. visual), but more detailed research is needed in order to clarify this issue. Importantly, Dillon et al. (2012) found also a P600 component as response to the ungrammatical use of ergative morphology. This finding is consistent with all the ERP studies where ergativity was manipulated and indicate that late (repair /​reanalysis) processes are crosslinguistically similar and do not depend on a specific case alignment setting. Furthermore, the study by Nieuwland et al. (2012) provides additional evidence on the neuroanatomical correlates of ergativity. The authors contribute by investigating the cortical networks involved in case, agreement, and semantic processing, and thus, shed more light on the mechanisms involved during these operations. The fMRI data support, to a large extent, previous findings in fMRI literature (Kaan & Swaab 2002; Kuperberg et al. 2003): neural circuits associated with syntactic processing differ significantly from those engaged during semantic processing. Also, in the light of the neuroimaging evidence, case and agreement violations draw upon overlapping neural circuits suggesting that similar brain regions are involved during case and agreement comprehension. These findings are largely compatible with the ERP data on case and agreement violations: usually both manipulations yield similar biphasic ERP pattern—the P600 preceded by a negative component. In sum, the fMRI evidence together with the ERP findings reveal that when dealing with ungrammaticality, processing mechanisms are similar across languages. In addition, the study by Zawiszewski et al. (2011) demonstrates how ergativity is processed by the native speakers and the non-​native high-​proficiency L2 bilinguals. While both groups of speakers were tested when processing ergative case violations in Basque, only the native group displayed a N400—P600 pattern. The non-​natives showed a N400 but no P600 effect, despite their high proficiency and early age of acquisition (AoA) onset (3 years). These differences are attributed to both the delay in L2 acquisition and to the parametric variation in case alignment setting between the L1 and the L2 of the non-​native group. In other words, according to the authors, native vs. non-​native effects obtain even at high level of proficiency and relatively low AoA only for those aspects of grammar which substantially differ between the L1 and the L2, such as case setting, whereas other, superficially divergent morphological aspects of language such as verb agreement do not seem to be sensitive to age of exposure. Summarizing, the experimental evidence on ergativity reported here provide a more complete picture of language processing architecture. In general, the electrophysiological pattern found when processing ergative case violations corresponds to that revealed during similar case violations in accusative languages (Frisch & Schlesewsky 2001, 2005) and thus indicate that

 

708   Adam Zawiszewski the mechanisms underlying language comprehension are comparable across languages with a different case morphology. Also, the findings reported in Carreiras et al. (2010) and Polinsky et al. (2012) shed more light on the subject vs. object processing strategies in ergative languages and show that the theories claiming subjects to be universally easier to process than objects need to be reconsidered by taking into account languages with different typologies and characteristics in order to successfully explain general principles of language processing mechanisms.

Acknowledgments This research has been supported by the Spanish Ministerio de Economía y Competitividad and the Spanish Ministerio de Ciencia e Innovación (FFI2010-​20472; FFI2012-​31360); the Basque Government (IT665-​13), the European Union’s Seventh Framework Programme for research, technological development and demonstration under grant agreement no. 613465 (AThEME). I would also like to thank Kepa Erdocia, Itziar Laka and Mikel Santesteban for their helpful comments on the previous versions of this manuscript.

Abbreviations 3 3rd person; ABS, absolutive; ACC, accusative; AGR, agreement suffix; AUX, auxiliary; e gap; ERG, ergative; FUT, future; GER, gerund; IPFV, imperfective; LOC, locative; M, masculine; NOM, nominative; O, object; OBL, oblique; PFV, perfective; PL, plural; PRTCP, participle; S, subject; SG, singular.

 

Chapter 29

Ex pe rim ental a pproac h e s to ergative l a ng uag e s Nicholas Longenbaugh and Maria Polinsky

29.1 Introduction The objectives of this chapter are twofold. First, we present and analyze experimental work on ergative languages; our discussion combines a review of existing work and the presentation of new experiments on Niuean and Avar. Second, we identify possible future avenues of experimental investigation where ergative languages can make a significant difference, as compared to languages without morphological or syntactic ergativity. We focus primarily on psycholinguistic and computational research on the languages under discussion. Due to space limitations, we will not address work on language acquisition (see Chapters 25, 26, 27). Until recently, experimental linguistic work has by necessity been confined to a small number of easily accessible languages, with English being the most prominent. Such work has also been predominantly limited to the university setting, with researchers testing undergraduates who are already comfortable with relevant aspects of experimentation (familiar with computers, experienced with test-​taking, etc.). Data from these types of participants may include some individual differences, but experimental work on familiar topics such as passives, past tense, or relative clauses suggests that individual differences are negligible in relation to more general patterns. Experimental investigations of ergative languages, in contrast, often take researchers out of their comfort zone and, in many cases, call for a creative methodological approach. Some ergative languages are endangered or spoken by very small communities, presenting a challenge for statistical analyses. Moreover, speakers in such communities are often bilingual, speaking both the minority language and the dominant language of their society. Bilinguals are of course a reasonable population for testing, but comparisons between such speakers and the monolingual speakers whose data abound in experimental work may require extra provisions. Indeed, at least two studies

 

710    Nicholas Longenbaugh and Maria Polinsky have found that bilingual speakers respond differently than their monolingual counterparts when exposed to identical stimuli (Zawiszewski et al. 2011; Clemens et al. 2015; Zawiszewski, Chapter 28, this volume). As an additional confound, many ergative languages are used only for oral communication, making reading-​based studies difficult or impossible. The remoteness of some relevant communities likewise makes it impossible to use more complex technology, such as neuroimaging. Finally, although ergative languages are by no means unique in this regard, they are often spoken in communities whose speakers lack formal education. Lack of formal education has an effect on the pattern of responses and results, for example, causing lower accuracy and slower response times (as we show below for Niuean and Avar). These challenges need to be given special consideration when such languages are studied (see Christianson and Ferreira 2005, Gagliardi 2012, Clemens et al. 2015, Wagers et al. 2015 for discussion). Existing experimental work on ergative languages can be roughly divided into the following categories: “accidental tourist” studies, experimental work on the competition between accusative and ergative alignment, work on long-​distance dependencies, and work on agreement. In “accidental tourist” studies, the choice of a given language is due to properties other than ergativity. For example, Skopeteas et al. (2012) examine the processing of dative subjects and objects in Georgian and conclude that case-​marking plays a more important role in online parsing than word order does. This is an important result, but Georgian is chosen because of its parallels to Icelandic, not because of its ergative alignment. Similarly, Duñabeitia et al. (2007) compare morphological decomposition in Spanish and Basque, but the choice of Basque is motivated by its agglutinative properties, not by its ergativity; Erdocia et al. (2012) likewise offer an impressive study of verb-​final and verb-​medial orders in Basque, which are orthogonal to ergativity. We will not discuss “accidental tourist” studies further in this chapter (although we will briefly return to Skopeteas et al. 2012 in section 29.5). The remainder of the chapter devotes one section to each of the three remaining types of studies mentioned above: in section 29.2 we survey work that explores the competition between accusative and ergative alignments; in section 29.3 we discuss experimental work on long-​distance dependencies in ergative languages; and in section 29.4 we take up experimental approaches to agreement. We outline possible directions for future experimental work in section 29.5.

29.2  Competition between Ergative and Accusative Alignment Accusative languages can exist without ergativity, but few languages exhibit consistent ergative alignment, whether through case-​marking or agreement, without incorporating some other alignment (accusative or neutral) in one of their subsystems. The result is the well-​known phenomenon of splits, where ergative alignment

 

Experimental approaches to ergative languages    711 is observed only in a subset of aspectual forms, or with a subset of DPs (see Coon and Preminger, Chapter 10, this volume). Indeed several researchers have proposed that ergative alignment systems are a “recessive feature” cross-​linguistically, susceptible to degradation and total loss in diachronic development and in contact scenarios with neighboring accusative languages (see Nichols 1993; van de Visser 2006; Maslova and Nikitina 2007; Bickel 2008; Bickel and Witzlack-Makarevich 2008; amongst others); split systems, in contrast, have been contended to be stable over time (Jäger 2007). Experimental work on alignment has focused on deriving the typological preference for accusativity from general constraints on processing or learnability. The underlying assumption—​which is not unquestionable itself—​is that phenomena that are harder to process or learn should be less common. If ergative alignments are found to be more difficult to process or to learn, this would offer a compelling explanation for their cross-​linguistic infrequency. From the processing perspective, the co-​occurrence of accusative (or neutral) and ergative alignment in split-​alignment languages makes it possible to test the relative processing costs incurred by the two alignments. Bornkessel-​Schlesewsky et al. (2008) attempt to measure just this contrast in Hindi, which features an alignment split along aspectual lines (see also Nevins et  al. 2007, and Dillon et  al. 2012 discussed below). The agreement system of Hindi operates as follows. Verbs agree in number and gender with the structurally highest unmarked argument in the clause (Mohanan 1994b; Keine 2010, 2012). Thus, if a subject appears in the nominative (unmarked), then the verb agrees with that subject; if the subject appears in the ergative, marked with -​ne, the verb agrees with the unmarked absolutive object. If the subject and object both appear with case-​marking, the verb shows default third-​person masculine singular agreement (see Mahajan, Chapter 4, this volume). Such default agreement may be overridden by so-​called long-​distance agreement (LDA), where a verb agrees with the agreement trigger of its clausal complement. LDA is common in control structures, where a matrix control verb may agree with the object in the embedded clause, provided the embedded clause verb also agrees with that object. Bornkessel-​Schlesewsky et al. (2008) provide the following paradigm illustrating some possible agreement variations that arise under subject control.1 These data also serve as the basis of their experiment—​although, as we discuss below, the pattern of results does not fully reflect the Hindi paradigm. (1) a. Raam1 [∅1 saikal calaa-​naa] caah-​taa hai. Raam.m.nom nom cycle.f.nom ride-​inf.m want-​ipfv.m aux (embedded verb agrees with its null subject; matrix verb agrees with its subject) b. Raam1 [∅1 saikal calaa-​nii] caah-​taa hai. Raam.m.nom erg cycle.f.nom ride-​inf.f want-​ipfv.m aux (embedded verb agrees with its object; matrix verb agrees with its subject) 1 

In the examples that follow, we represent the presumed subject of the infinitival clause atheoretically as a null element and show its presumed case as it is done in Bornkessel-​Schlesewsky et al. (2008).

 

712    Nicholas Longenbaugh and Maria Polinsky c. *Raam1 [∅1 saikal calaa-​nii] caah-​tii hai.    Raam.m.nom erg cycle.f.nom ride-​inf.f want-​ipfv.f aux    (embedded verb agrees with its object; matrix verb shows LDA with embedded object) d. Raam-​ne1 [∅1 saikal calaa-​naa] caah-​aa hai. Raam.m-erg nom cycle.f.nom ride-​inf.m want-​pfv.m.dflt aux (embedded verb agrees with its null subject; matrix verb shows default agreement) e. *Raam-​ne1 [∅1 saikal calaa-​nii] caah-​aa hai.    Raam.m-erg nom cycle.f.nom ride-​inf.f want-​pfv.m.dflt aux    (embedded verb agrees with its object; matrix verb shows default agreement) f. Raam-​ne1 [∅1 saikal calaa-​nii] caah-​ii hai. Raam.m-erg nom cycle.f.nom ride-​inf.f want-​pfv.f aux (embedded verb agrees with its object; matrix verb shows LDA with embedded object) ‘Raam wants to ride a bicycle.’ Bornkessel-​Schlesewsky et al.’s (2008) study is based on an experimental paradigm in which sentences whose initial argument displays a subject-​object ambiguity incur a steep processing cost when that argument is resolved as an object. No such cost is associated with a subject resolution, indicating that “subject” is the default value for ambiguous initial arguments as perceived by the processor (Kuperberg 2007; Miyamoto 2008). There is likewise no processing cost associated with a sentence-​ initial unambiguous object. Bornkessel-​Schlesewsky et  al. (2008) claim that an accusative processing preference should be detectable in this same way: if a sentence-​ initial argument is a nominative subject and the verb turns out not to agree with it (so that the sentence is ungrammatical), there should be an increase in processing costs, given that the initial assumption of accusative alignment was revised. According to this hypothesis, the contrast between (1-​a) and (1-​c), where (1-​a) is processed more easily, should support the accusative preference. By contrast, they speculate that no increased cost will appear in the context of an ergative subject because the processor knows there will be no accusative alignment after it encounters the first word (compare (1-​d) and (1-​f ) above). The authors also investigate whether an accusative preference in the main clause can be modulated by the alignment patterns in the embedded control clause: an accusative alignment should “strengthen the [accusative alignment] preference for the matrix subject” while an ergative alignment should “counteract it” (Bornkessel-​Schlesewsky et al. 2008: 411). The authors hypothesize that there should be notable differences, measurable in reading time and neural activity, in the processing of sentences like (1-​a), where the control clause and the matrix clause have accusative agreement, and sentences like (1-​b), where the control clause has absolutive agreement. Their experiment employs sentences

 

Experimental approaches to ergative languages    713 like those in (1)  to test these hypotheses using self-​paced reading, electrophysiology (ERP), and acceptability judgments. While we share the authors’ interest in the potential differences in the processing of ergative and accusative alignments, we are concerned that their experimental paradigm does not directly address the relevant question. All else being equal, we agree that observing a processing-​time increase when the processor is forced to revise its initial assumption of accusative alignment would be a substantial result. However, the so-​ called revision in their experiment involves a sentence that is not grammatical. The increase in processing cost may thus result from the participants’ confusion regarding the acceptability of the sentence, or from their surprise at hearing an ungrammatical form. In order to meaningfully compare the processing costs of two sentences in the context of a proposed accusative alignment preference, these sentences must be minimally different in their alignment. The stimuli should thus include clauses that are initially ambiguous between an accusative and an ergative alignment, with that ambiguity resolved later in the clause. Including an ungrammatical sentence in the minimal pair introduces an unnecessary confound. With respect to the authors’ second hypothesis, the data do suggest that ergative alignment in a controlled clause affects the processing of matrix accusative alignment. In examples with accusative alignment in both matrix and embedded clauses, the appearance of a sentence-​final agreeing verb triggers a neural response (a high P300 value) consistent with anticipation by the processor (see Polich 2007 for an overview of P300). No such response is triggered when the embedded clause is ergative. That said, it is not immediately evident how this result bears on the original question motivating the experiment. That ergative alignment can interfere with the processors ability to predict the alignment of the matrix clause does not on its own entail the existence of a differential processing cost associated with accusative/​ergative alignments. The two phenomena seem to be unrelated. We conclude that Bornkessel-​ Schlesewsky et al.’s (2008) results, while interesting in their own right, do not allow us to directly answer the question of whether the processor favors accusative or ergative alignment. Computational linguists have also taken up the issue of competition between alignment systems, specifically from the perspective of learnability. Van Everbroeck (2003) explicitly sets out to test the relative learnability of accusative versus ergative alignment, taking in to account alignment expressed via agreement, case-​marking, or both. In the experiment, artificial neural networks were tasked with learning a number of toy languages, which varied according to three parameters meant to simulate the typological variety attested across the world’s languages: clausal word order, word-​order correlations, and head-​versus dependent-​marking. Each toy language consisted of a fixed set of SVO, SOV, VSO, VOS, OVS, or OSV sentences, and marked the grammatical roles of its arguments via agreement on the verb (head marking), case-​marking on the arguments (dependent-​marking), both, or neither. Both accusative and ergative alignments were represented. Finally, each toy language varied based on whether it had postpositions or prepositions and whether relative clauses were pre-​or post-​nominal (word-​order

 

714    Nicholas Longenbaugh and Maria Polinsky correlations). A separate toy language was constructed for every unique combination of the parameter settings. A neural n ​ etwork was exposed to a subset of 3000 sentences drawn randomly from the corpus associated with each language (the training set), then provided with the remaining sentences in the corpus (the test set) and tasked with parsing these sentences according to the grammar learned from the training set. Languages with ergative alignment were not significantly harder to learn in the subject-​before-​object cases, where the learned grammar was able to almost perfectly parse the data in both the accusative and ergative alignment cases. Object-​before-​ subject order made accusative languages slightly more difficult to learn, as compared to the subject-​before-​object languages. Object-​before-​subject languages with ergative alignment, however, were significantly harder to learn, as compared to both the corresponding accusative object-​before-​subject languages and the ergative subject-​before-​ object languages. The results were particularly striking in the VOS and OSV cases. As van Everbroeck (2003) notes in his discussion, however, there are a number of complications that prevent us from drawing definitive typological conclusions from this study. First, if there was a direct correlation between the ease of learning these toy languages and the typological distribution of the various characteristics controlled in the experiment (word order, head-​/​dependent-​marking, and alignment), we would expect, for example, for VSO and SVO languages to be more prevalent than SOV languages cross-​linguistically since the latter are significantly more difficult to learn. Likewise, all else being equal, VOS languages should be much more likely to surface with accusative alignment than ergative alignment. In reality, SOV word order is basic in approximately 50% of the world’s languages (per van Everbroeck’s 2003 discussion), and VOS languages can have all types of alignment, with ergative VOS observed across Mayan, Salish, and Austronesian languages (Coon 2013a; Polinsky 2016; Aldridge, Chapter  21, this volume; Aissen, Chapter 30, this volume; Pye and Pfeiler, Chapter 27, this volume). Beyond these difficulties in extrapolating from the learnability results to actual typological variation, there are also conceptual obstacles that sound a note of interpretive caution. Significantly, several categories of information available to a human learner were absent in the experiment, including semantic content, prosodic cues, and information regarding animacy and agency of arguments. These omissions limited the learning algorithm in a variety of ways; for instance, it could not capitalize on the universal tendency for subjects to be animate and agentive, nor could it group words into part-​of-​speech categories based on semantic information. Despite these difficulties, van Everbroeck’s (2003) data offers tantalizing, albeit tentative, evidence in favor of a learnability-​based explanation for the cross-​linguistic distribution of ergative and accusative languages. To summarize, split-​alignment languages provide an ideal testing ground for the hypothesis that ergative alignments are more difficult to process than their accusative counterparts. While we believe it is possible to devise an experimental paradigm to test the validity of this hypothesis, the work to date in this area has been inconclusive. Studies addressing whether ergative alignments are intrinsically more difficult to learn than accusative alignments can address the difference between alignments from

 

Experimental approaches to ergative languages    715 another angle, and here the experimental work of van Everbroeck (2003) on the learnability of toy languages by artificial neural-​networks is a first step.

29.3  Subject-​Object Asymmetries Long-​distance dependencies have long been a central concern for linguists, and have played an important role in the ongoing dialogue between theoreticians and experimentalists. As both groups of researchers share an interest in the motivation for empty categories and the nature of filler-​gap relationships, A’-​dependencies (topicalization, focusing, wh-​question formation, and relative clause formation) have been widely explored both experimentally and theoretically. One major result, supported by a large body of research, is the recurrent subject-​object asymmetry with respect to A’-​extractions. Owing to their near universality cross-​linguistically, relative clauses (RCs) have proved an especially fertile domain for the study of such asymmetries. In the processing literature, extensive experimentation has revealed two major generalizations concerning subject-​object asymmetries. The first generalization has two components: (i) unambiguous subject RCs are easier to process than unambiguous object RCs; (ii) when speakers are faced with an RC that is ambiguous between a subject-​ headed interpretation and an object-​headed interpretation, there is a strong preference for resolving the ambiguity in favor of subject-​headedness (see Kwon et al. 2010 for an overview). We will refer to these two asymmetries collectively as the Subject Processing Advantage (SPA). In accusative languages, the SPA favors the nominative argument, which is typically the subject. The second generalization revealed by studies of subject-​object asymmetries pertains to the strategies used by the parser to anticipate the upcoming material –​specifically the use of a particular form or category as a cue for another form or category that can be projected. In an accusative language, the presence of an accusative-​marked argument in a clause informs the parser that the clause is transitive and that a nominative-​ marked argument should be projected; the presence of a nominative does not have the same effect, because nominatives can occur with intransitive verbs. More specifically, within a RC, nominative gaps are predicted to be easier to parse because the presence of accusative case “cues” the parser to the presence of a nominative argument. Thus, if this argument is absent, it is likely to be the head of the RC. Several studies provide empirical evidence in favor of this cueing effect in RCs, most notably in Korean (Kwon et al. 2006) and Japanese (Ueno and Garnsey 2008).2 These cueing effects should result in a processing advantage for nominative gaps in a relative clause of an accusative language.3 2  See also Bornkessel-​Schlesewsky and Schlesewsky (2009: 159) and Skopeteas et al. (2012) for evidence that morphological case is a strong cue in processing. 3  Of course, not all subjects in accusative languages are nominative, but a large portion of experimental work has focused mainly on the contrast between nominative subjects and accusative objects.

 

716    Nicholas Longenbaugh and Maria Polinsky A limitation of the above experiments arises from the fact that nominative case and subject status align in accusative languages, independent of transitivity. As such, the effects of the SPA and of case-​cueing converge in all cases, rendering it difficult to test experimentally if they are truly independent phenomena. In ergative languages, transitive and intransitive clauses exhibit different mappings between grammatical function and morphological case. Indeed, the case-​cueing effect, if it exists in these languages, should be expected to be triggered by ergative, but not absolutive, case: the presence of an ergative argument informs the parser that there must be an absolutive argument projected in the relative clause.4 This means that the processing of an absolutive gap should be easier than the processing of an ergative gap, at least on the basis of case-​cueing alone. In contrast, the SPA predicts that ergative-​subject gaps and absolutive-​subject gaps should be easier to process than absolutive-​object gaps. The two principles are thus at odds, allowing a careful experimenter to distinguish their relative impact on RC processing. Here, then, ergativity has the potential to offer novel insights into the mechanisms of long-​distance dependency processing. (2) a. SPAerg: ergative-​subject gaps and absolutive-​subject gaps should be easier to process than absolutive-​object gaps b. Case-​cueingerg: an absolutive-​object gap should be easier to process than an ergative gap; the presence of an ergative argument cues the parser to the presence of a absolutive argument Studying the processing of RCs in ergative languages can also afford us insights into the nature of syntactic ergativity and extraction constraints in general. Syntactic ergativity is a phenomenon attested in a subset of morphologically ergative languages whereby transitive (ergative) subjects cannot undergo A’-​extraction by leaving a gap in the base position, while intransitive subjects (absolutive) and transitive objects (absolutive) may freely do so. It remains an open question why some but not all morphologically ergative languages exhibit this behavior (see e.g. Campana 1992; Manning 1994; Aldridge 2004; Coon et al. 2014; Polinsky 2016, in press). No equivalent pattern obtains in accusative languages, where the accessibility hierarchy (AH) for A’-​extraction seems to directly track grammatical function: if an intransitive subject can extract, so can a transitive subject (Keenan and Comrie 1977; Comrie and Keenan 1979).5 One line of reasoning, in the spirit of Hawkins (1994, 2004, 2014), suggests that certain grammatical constraints follow 4 

This logic applies to those ergative languages that do not have split ergativity; if the ergative can encode subjects of transitive and unergative predicates, then the predictions are different. 5  In their seminal paper on accessibility to relativization, Keenan and Comrie acknowledge that ergative languages pose a potential challenge to their hierarchy; their solution is to interpret ergative clauses as synchronically passive-​like (where the ergative “subject” is more like a by-​phrase, hence lower on the AH) or at least as related to passives diachronically. This solution, however, has some problems. First, the ergative is not like a by-​phrase with respect to other typical subject properties (binding, imperative formation, nominalizations). Second, not all ergative languages have developed from the passive, and even if all the syntactically ergative languages have done so, it is unclear how this diachronic development can still influence a language learner.

 

Experimental approaches to ergative languages    717 from processing constraints; when the latter become grammaticalized, the grammar is forced to follow a given principle. If we extend this logic to ergativity, we might hypothesize that syntactic ergativity stems from the more general processing load associated with ergative-​argument gaps. Under this account, in languages without syntactic ergativity, ergative-​argument gaps should impose a heavier processing load than absolutive-​ argument gaps, and should therefore be dispreferred, just as other types of constructions that tax the processor are dispreferred (compare Kluender 1998 for a processing account of island effects). Thus, whereas other languages merely constrain the use of constructions that are difficult to parse, syntactically ergative languages take this constraint to the logical extreme and forbid such constructions entirely. If this is correct, we should expect to see a general processing difficulty of ergative-​argument extraction in those languages that are not syntactically ergative. Finally, ergative languages also allow us to compare the processing of case and agreement. Some languages express their alignment not via case-​marking on nominals (dependent-​marking) but rather via agreement morphology on the verb (head-​marking). In principle, the same logic that justifies the existence of cueing effects of dependent-​marking extends to languages with head-​marking. Thus we might expect a cueing effect to arise in languages with head-​marking as well. This effect is difficult to test for in accusative languages, given that both the SPA and cueing via dependent-​marking serve to make subject RCs easier to parse. Just as with case cueing, in ergative head-​marking languages we can dissociate the effects of subjecthood and of head-​marking triggered cueing by comparing the relative ease of parsing RCs with ergative gaps versus absolutive subject gaps. While addressing these research questions in an experimental paradigm is straightforward, it requires some modifications of the standard research approach to the processing of RCs. First, RC-processing experiments in accusative languages have historically been limited to transitive clauses, which are sufficient for assessing subject-​object asymmetries. In an ergative language, however, intransitives must be included as well to establish comparison between absolutive subjects (intransitive) and absolutive objects (transitive). Of course, experimental stimuli must be balanced by length (measured by the number of words, syllables, noun phrases, etc.); this means that intransitive RCs included in processing experiments should include a PP in order to match transitive clauses as closely as possible. Second, in order to rule out possible frequency biases, one needs to ascertain the distribution of different types of RCs in a given language. If we limit our statistics to subject and object RCs in accusative languages, the following distinction typically emerges: intransitive subject RCs are most common, followed by transitive subject and object RCs (see Gordon and Hendrick 2005 for English). As far as we can tell, the distribution of these three types of RCs in Niuean and Avar, which we discuss below, is quite comparable; intransitive subject RCs constitute, respectively, 40% and 43% of all subject and object RCs; the frequency of transitive subject RCs is 31% and 26% respectively, and object RCs are at 31% and 28%.6 In Basque, Carreiras et al. (2010) estimate the 6 

The data are based on our text counts of 23,000 Niuean clauses from elicited dialogues, and 30,000 Avar clauses from a mixed corpus.

 

718    Nicholas Longenbaugh and Maria Polinsky occurrence of object RCs at 36%; they do not provide the breakdown of intransitive and transitive subject RCs that comprise the remaining 64%. Future studies of RC processing in ergative languages need to include RC frequencies as well. The final modification pertains to types of populations tested in the experiments, an issue we alluded to in the introduction. A number of ergative languages lack a robust reading tradition—​or any reading tradition whatsoever—​which makes it impossible to use self-​paced reading tasks, a paradigm that has been immensely successful in languages such as English, German, or Spanish. In such cases, one effective experimental method is sentence-​picture matching, which has been shown to yield largely compatible results to self-​paced reading (Clemens et al. 2015).

29.3.1 Basque Basque provides a good opportunity for dissociating the effects of case-​cueing and the SPA. Both subjects and objects in Basque may relativize with a gap. The verb agrees with both the absolutive and the ergative arguments, although some researchers suggest that at least the ergative markers may be instances of clitic-​doubling (Preminger 2009; Arregi and Nevins 2012). Basque also exhibits case homophony; in particular, the affixal exponent -​a-​k, an agglutinative marker formed from -​a-​, the definite determiner, and -​k, the ergative marker, is indistinguishable from -​ak, the marker of absolutive plural (for more discussion of Basque ergativity, see Laka, Chapter 7 this volume; Berro and Etxepare, Chapter 32, this volume). Carreiras et al. (2010) exploit this ambiguity in an experiment designed to test for effects of the SPA. Crucially, the ambiguity of -​ak permits the construction of transitive RCs in which the clausal argument is ambiguous between an ergative and an absolutive. This ambiguity is not resolved until the penultimate word of the matrix sentence (W6 in the examples below). In spoken language, the ambiguity can be resolved prosodically, but in reading that is impossible. (3) a. [_​_i​ irakasle-​ak aipatu ditu-​en] ikasle-​a-​ki lagunak ditu    teacher-​abs.pl mention aux-​adn student-​det-​erg friends has orain. now ‘The student who mentioned the teachers has friends now.’ b. [irakasle-​a-​k _​_​i aipatu ditu-​en] ikasle-​aki teacher-​det-​erg mention aux-​adn student-​abs.pl dira  orain. are    now ‘The students that the teacher mentioned are friends now.’

lagunak friends

Carreiras et al. (2010) tested this type of stimulus in a self-​paced reading experiment with 54 native speakers. The reading time data are presented in Figure 29.1.

 

Experimental approaches to ergative languages    719 1300 1200 1100 1000 900

SR OR

800 700 600 500 400

W1

W2

W3

W4

W5

W6

W7

Figure 29.1  Results from Carreiras et al. (2010: 85)

As expected, the reading times for each of the stimuli are identical up to W6, which is the first word where the stimuli diverge in content. At W6, there is a severe slowdown for resolution of the ambiguity in favor of an ergative-​subject gap compared to resolution in favor of an absolutive-​object gap. This suggests that object RCs are easier to process, possibly due to case-​marking (as Carreiras et al. 2010: 91 propose). That said, we feel that the case-​marking of the extracted argument in Basque is not the only factor that influenced these experimental results. At least two other factors may have played a role. The first is that the disambiguating words differed in transitivity: ditu “has” (3-​a) versus dira “are” (3-​b). While it is critical to study transitive and intransitive RCs in any experiment on RC processing in ergative languages, the difference in transitivity here is in the matrix clause, not in the RC. The object RC modifies the subject of an intransitive matrix clause, and the subject RC modifies the subject of a transitive clause. This variation potentially introduces a poorly understood transitivity confound to the data; note that in most experiments on extraction (both in accusative and ergative languages), the properties of the head noun are either kept constant or are balanced across conditions. Second, we question the authors’ implicit assumption that the parser delays resolution of the ambiguity in the RC until there is enough information to decide one way or the other (W6). There is significant evidence from other studies that the parser will commit to one interpretation immediately and revise if necessary, rather than delay commitment (see Clemens et al. 2015 for discussion). This initial commitment is usually determined on the basis of prior experience and statistical preferences, both of which favor the more common ergative singular over the less common absolutive plural (Austin 2007; Clemens et al. 2015). In fact, an independent experiment (Laka and Erdocia 2012) also confirms a preference among Basque speakers to interpret the potentially ambiguous N-​ak segments as ergative rather than absolutive DPs. For the experiment discussed here, if the parser assumes the ambiguously-marked argument to be

 

720    Nicholas Longenbaugh and Maria Polinsky an ergative, it must commit to the interpretation of the missing (gapped) argument in the RC as an absolutive object. The ergatively-interpreted argument serves as a cue for the parser to project the absolutive object. This cueing effect may be weakened, because Basque also has ergative intransitives (Laka 2006b), but it cannot be ignored; once the parser reaches the RC predicate, it can ascertain that an absolutive object is indeed projected. At the disambiguation points (the head noun and/​or the predicate of the main clause), the object RC interpretation thus requires no revision of the initial commitment. In contrast, the subject RC interpretation requires a revision down the line, which yields higher processing costs. Carreiras et al. (2010) also conducted a neuroimaging (ERP) study of 22 subjects using the same stimuli as in the reading experiment. They propose that the response to subject RCs in Basque evokes a P600 effect (Carreiras et al. 2010: 88–​89), a somewhat puzzling result which suggests that subject RCs are associated with a perception of ungrammaticality.7 However, closer examination reveals that the reported ERP data do not fit the distribution of a standard P600. We suggest that the distributional irregularity can be accounted for by a left anterior negativity (LAN) to object RCs. Indeed, the authors entertain this interpretation themselves, but ultimately reject it, arguing that a LAN-​based explanation is inconsistent with the behavioral results presented above; however, as we have illustrated, those results should be treated cautiously. Within the actual ERP data, the main effect is observed in the left anterior region, which is a signature LAN distribution. In contrast, P600 is normally found in the central posterior region, sometimes skewed to the right hemisphere (Hagoort et al. 1999). Figure 4 in Carreiras et al. (2010: 88) shows negativity between 300 and 500 ms, a typical LAN latency, well within anterior-​negativity bounds. Anterior negativity triggered by long-​ distance dependencies with an object gap has been observed in many languages, from English (King and Kutas 1995) and German (Fiebach et al. 2001) to Japanese (Ueno and Garnsey 2008) and Korean (Kwon et al. 2013). If the LAN interpretation is correct, the Basque results are pleasingly consistent with the electrophysiological responses to object RCs observed in other languages; as the object is structurally lower than the subject in clause structure, integrating it with the remainder of the RC and then with the head noun imposes a heavier processing load.

29.3.2 Avar and Niuean The Basque study reported above inspired a reading study of another head-​final language with morphological ergativity: Avar (Nakh-​Dagestanian). Avar case morphology distinguishes between ergative and absolutive; the language is consistently ergative and has agreement in gender with the absolutive argument (indicated in Roman numerals

7 

P600 is typically associated with syntactic and/​or morphological violations (Hagoort et al. 1993, 1999); see also Brouwer et al. (2012) for a more recent overview.

 

Experimental approaches to ergative languages    721 in the glosses). Unlike Basque, there is no ergative marking in intransitive clauses. Both subjects and objects relativize by leaving a gap at the extraction site. Polinsky et al. (2012) conducted a self-​paced reading study of Avar RCs where they tested participants on three types of extraction:  absolutive subject (4-​a), absolutive object (4-​b), and ergative subject (4-​c). In each condition, the head noun of the relevant RC appeared as the subject of an intransitive clause (not shown). (4)

a. [_​_i​ bercinay artistka-​yal-​da ask’o-​y č’:u-​n y-​ik’-​ara-​y]     beautiful actress-​obl-​loc near-​ii standing-​ger ii-​be-​ptcp-​ii ‘the girl that stood next to a beautiful actress’ b. [bercinay artistka-​yal _​_​i repetici-​yal-​de y-​ač:-​un beautiful actress-​erg     rehearsal-​obl-​loc ii-​bring-​ger yasi... girl.abs ‘the girl that a/​the beautiful actress brought to a rehearsal’

yasi... girl.abs

y-​ač’-​ara-​y] ii-​come-​ptcp-​ii

c. [_​_i​ bercinay yas repetici-​yal-​de y-​ač:-​un y-​ač’-​ara-​y]       beautiful girl.abs rehearsal-​obl-​loc ii-​bring-​ger ii-​come-​ptcp-​ii artistkai... actress.abs ‘the actress that brought a/​the beautiful girl to a rehearsal”

Figure 29.2 shows residual reading times from this experiment; each data point represents the average time spent on a given word. Participants processed RCs with absolutive-​subject gaps faster than the corresponding RCs with absolutive-​object gaps or ergative-​subject gaps. This difference manifests most clearly at the head noun of the RC, and in the spill-​ over region. There was a clear advantage for the absolutive-​subject gap, but the ergative and absolutive-​object gaps were processed at roughly the same reading speed, albeit with a different time course. With object RCs, there was a clear case-​cueing effect of the ergative DP; readers slowed down significantly at an ergative inside the RC, but not at an absolutive in same position in the RC. The slowdown at the ergative suggests that readers spend time projecting a transitive clause with an absolutive argument. The SPA and case-​cueing effects, clearly visible at the individual word points following and within the RC, cancel each other out, yielding a roughly comparable response time to ergative-​subject and object RCs. Across the board, reading times in the study were excessively long, due to the status of Avar as a primarily spoken language. Because reading was not entirely natural, the present authors also conducted a picture-​matching experiment involving three conditions: absolutive-​subject gap (with a PP in the RC in order to balance the length of the stimuli with the transitive condition), absolutive-​object gap, and ergative-​subject gap (see Longenbaugh and Polinsky 2016 for full discussion). There were 45 participants (avg. age 44;6) who took the experiment in Dagestan; they were asked to listen to 36 stimuli recorded by a native speaker of the standard dialect of Avar. In a within-​subjects design, subjects saw pictures on a computer screen depicting three participants (Figure 29.3). In the transitive condition, one kind of participant (X1) acts on another kind of participant (Z), who acts on another participant of

 

722    Nicholas Longenbaugh and Maria Polinsky 1100 1050 1000 Absolutive Subject Gap Absolutive Object Gap Ergative Gap

950 900 850 800

W1

W2

W3

W4

W5 (RC W6 (Head W7 (Spill Predicate) Noun) Over)

W8

Figure 29.2  Avar reading time measurements (Polinsky et al. 2012: 272) (a) Intransitive (X is in front of Z)

(b) Transitive (X is licking Z)

Figure 29.3 Experimental images

the first kind (X2). In the intransitive condition, two kinds of participants, X1 and Z, do one thing, and another one (X2) does something else. Subjects saw the pictures and heard a question such as “Where is the X that is VERB-​ ing the Z/​that Z is VERB-​ing?”; the RCs had the same structure as in the reading experiment (cf. (4)). (See Niuean stimuli in (5).) The questions about the middle participant, the one that is unambiguously identified because it has no pair, served as fillers. The subjects had to select the appropriate participant in the picture. The number of correct and incorrect answers was recorded as a measure of accuracy; response times (RT) were measured based on the time elapsed between the start of the sound recording and the mouse-​click on the relevant portion of the picture. In Avar, there was no significant difference between the response time for correctly selected RCs with an ergative-​subject gap (4812 ms, s.d. 4300) and correctly selected RCs with an absolutive-​objects gaps (5128 ms, s.d. 8950), p =.15. However, correctly selected RCs with an absolutive-​subject gap were processed much faster, at 2943 ms, s.d. 4100 (p =.009). This is consistent with the results from the self-​paced reading experiment. The error rate of responses to the three types of RCs was about the same (73.7% correct responses in the absolutive-​subject condition, 76.1% in the ergative-​subject condition, and 74.3% in the absolutive-​object condition). These numbers are lower than what we

 

Experimental approaches to ergative languages    723 normally find in experiments on familiar languages, but they are consistent with the lower pattern of responses observed in older, less literate populations that are not used to experimental testing. There was no difference in RTs between correct and incorrect answers, which supports the reliability of the data. All told, the picture-​matching experiment results replicate the results obtained by reading by Polinsky et al. (2012). No difference was found between the processing of ergative and absolutive-​object gaps. Basque and Avar are both head-​final languages with prenominal RCs; case-​cueing in such RCs occurs early, which may result in special processing benefits for the missing absolutive. To determine whether word order may influence the processing of filler-​ gap dependencies in ergative languages, we conducted an experimental study on the Polynesian language Niuean, which has unmarked VSO word order with postnominal RCs (see Longenbaugh and Polinsky 2016 for a full discussion). Ergative alignment, which is consistent throughout the language, is expressed via case-​marking on nominals: both absolutive and ergative are marked overtly. Niuean lacks agreement. Both subjects and objects may undergo relativization with a gap. In this regard, Niuean is different from its close Polynesian relatives, where the ergative cannot relativize with a gap. Given that Niuean is predominantly used as a spoken language, we decided to conduct an auditory picture-​matching study using the same paradigm as reported above for Avar. 24 stimuli were normed with three native speakers and were recorded by another native speaker. Forty-​one L1 speakers of Niuean (avg. age 42;8), recruited in Auckland, New Zealand, took part in the experiment. Example stimuli corresponding to Figure 29.3 are shown in the following.8 (5)

a. Ko fe e putii [ne mua _​_​i he where abs cat nfut be.ahead     obl ‘Where is the cat that is in front of the dog?’ b. Ko fe e putii [ne epoepo he where abs cat nfut lick erg ‘Where is the cat that the dog is licking?’

kuli]? dog

kuli _​_​i]? dog

c. Ko fe e putii [ne epoepo _​_​i e kuli]? where abs cat nfut lick   abs dog ‘Where is the cat that is licking the dog?’ The pattern of results is very similar to that found in Avar. The response accuracy in the three conditions was 72.1% for absolutive-​subject gaps, 71.8% for ergative gaps, and 79.5% for absolutive-​object gaps. Again, these numbers are much lower than what we normally find in experiments with undergraduates, but they are consistent with the lower pattern of responses observed in populations that are not used to experimental testing. There was no difference in RTs between correct and incorrect answers, which supports the reliability of the data. The difference between the response time for ergative 8 

Wh-​questions in Niuean necessarily involve pseudo-​clefts (see Potsdam and Polinsky 2011).

 

724    Nicholas Longenbaugh and Maria Polinsky gaps and absolutive-​object gaps was not significant (Wilcoxon test, p=0.96). We interpret this result as indicating that morphological cueing (whereby the ergative case serves as the cue in projecting the absolutive argument) and the SPA are roughly equivalent in terms of their effects on the processing of RCs. With ergative-​subject gaps, the SPA acts in favor of easier processing, while case-​cueing acts against easier processing. Meanwhile, the difference between the response time for absolutive-​subject gaps versus the two other types of gaps was highly significant (Wilcoxon test, p Dependent Case > Lexical/​Oblique Case

Nominative and absolutive are usually taken to be unmarked cases, while ergative and accusative are dependent cases. As such, if only unmarked case is accessible for agreement, then agreement will always be with the nominative (as in English) or the absolutive (as in Nakh-​Dagestanian languages). In languages with ergative case-​marking and nominative agreement, the claim is that both unmarked and dependent case are accessible for the determination of agreement morphology. As such, the verb will always agree with the structurally highest argument bearing ergative or absolutive case, which will be the subject, yielding an accusative agreement pattern. In contrast, this system predicts that accusative case-​marking cannot co-​occur with ergative agreement:  even if nominative and accusative are both accessible for agreement, the structurally superior subject will always end up being the trigger. If Bobaljik (2008) is on the right track, we once again predict that a language with accusative case-​ marking and ergative agreement should be not just rare, but impossible. In turn, this means that a computational model should treat this pattern differently from patterns that are attested but merely rare.

29.5.2 Structural vs. Inherent Case A majority view in syntactic theory is that ergative is an inherent case, licensed either by a transitive v or Voice head (Aldridge 2008a; Legate 2008), by a P head (Markman and Grashchenkov 2012), or by both types of head (Polinsky 2016). That said, a small number of researchers have suggested that ergative may be a structural case; this analysis has been advocated by some researchers for Basque (Rezac et al. 2014, but see Laka 2006b for the inherent-​case analysis) and Burushaski (Baker, Chapter 31, this volume). In general, the distinction between structural and inherent case has received relatively little attention in the experimental domain. Most work in this area has been conducted on German, where the structural cases (nominative and accusative) are compared with the inherent dative (e.g., Jacobsen 2000). Similarly, for Japanese, researchers have investigated the processing of nominative-​ marked vs. dative-​ marked wh-​phrases (Aoshima et al. 2004). A recent study on Georgian (Skopeteas et  al. 2012)  also compared the structural nominative and the inherent dative. The results of the existing studies have been largely inconclusive, although Jacobsen (2000: 148–149) indicates that structural but not inherent case evoked a late positivity in German, while violations of structural case evoked a higher negativity. Given that the distinction between structural and inherent case is still theoretically precarious (cf. the discussion in Bobaljik and Wurmbrand 2008), it would be desirable to

 

732    Nicholas Longenbaugh and Maria Polinsky determine whether this distinction has processing correlates. If such correlates are identified, they could then be used to assess the case status of ergative vis-​à-​vis absolutive in individual languages.

29.5.3 Ergative Languages as a Heterogeneous Class Just as nominative-​accusative languages are not all the same, ergative languages form a heterogeneous class as well. Approaches to such heterogeneity vary, with some researchers placing the burden on the variation in the ergative case assignment, and others, on the absolutive. Adopting the latter approach, ergative languages can be divided into two broad classes (Aldridge 2008a; Legate 2008, 2012a). In the first class, the absolutive is a morphological default (abs=def). abs=def languages simply lack nominative and accusative case morphology, and at the point of vocabulary insertion, the nominative and accusative structural cases are realized by the absolutive simply because it is the most appropriate morphological default. Warlpiri and Hindi seem to be examples of this type. In these languages, the absolutive form is used for DPs that arguably lack abstract case altogether (such as hanging topics); objects may appear as the sole absolutive argument in a non-​finite clauses; agreement tracks the subject or the highest unmarked structural case (illustrated earlier for Hindi); when conditions for a particular case assignment are not met, default absolutive can appear, resulting in multiple absolutives in a single clause; there is no evidence of syntactic ergativity. The second type of ergative languages has abs=nom; in such languages, the accusative is never assigned, and what is called the “absolutive” is simply a reflex of the nominative case. Georgian may be considered an abs=nom language. The distinction between abs=def and abs=nom leads to an important testable prediction. From the processing standpoint, a morphological default is treated as the absence of a certain feature or value, which means that a mismatch where the default is expected (and does not appear) is parsed differently from a mismatch where a particular, specified, case is expected (see Clahsen 1999 for a general discussion of psycholinguistic work on defaults, and Penke et al. 1997 for neuroimaging evidence of defaults). Conversely, if the default appears in unexpected contexts, the parser is predicted to respond differently than it does when one specified form replaces another. This suggests that in a abs=def language, there could be two patterns of violation response. The first should occur when case X is mismatched and case Y appears instead; the second should surface when case X is expected and the absolutive default appears instead. In abs=nom languages, the absolutive is a genuine case, not a morphological default, so its appearance in lieu of another case would be treated as a featural violation. An experimental study along the lines described here would not only confirm the division of ergative languages into subclasses, but would also serve as a valuable tool to test the specific problem of the morphological default status of the absolutive.

 

Experimental approaches to ergative languages    733

Acknowledgments The authors of this chapter are listed in alphabetical order. Some of the work reported in this chapter was supported by the Max-​Planck Institute for Evolutionary Anthropology and by NSF grants BCS-​114223, BCS-​137274, BCS-​1414318 to the second author. Any opinion, findings, or conclusions expressed here are those of the authors and do not necessarily reflect the views of the NSF or other agencies. We are grateful to Jason Brown, Kamil Ud Deen, Ömer Demirok, Itziar Laka, Ellen Lau, Jason Merchant, Lea Nash, Colin Phillips, Eric Potsdam, and Yakov Testelets for helpful discussion of the ideas presented in this chapter. We would also like to thank Zuzanna Fuchs for help with statistical analyses, Granby Siakimotu and Kara Tukuitonga for help with the Niuean study, and Magomed Abdulkhabirov, Karim Karimov and Magomed Magomedov for help with the Avar study. All errors are our responsibility.

Abbreviations abs, absolutive; adn, adnominal; aux, auxiliary; det, determiner; dflt, default; erg, ergative; f, feminine; ger, gerund; inf, infintive; ipfv, imperfective; loc, locative; m, masculine; nfut, non-future; nom, nominative; obl, oblique; pfv, perfective; pl, plural; ptcp, participle.

 

 

PA RT  I V

C A SE ST U DI E S

 

 

Chapter 30

C orrel at e s of e rgativit y i n Maya n Judith Aissen

30.1 Introduction The title of this chapter references an earlier paper, ‘Correlates of ergativity in Mayan grammar’ (Larsen and Norman 1979), which identified already many of the issues that have driven research on ergativity in Mayan since 1980. As they observed, there is no question that Mayan languages are morphologically ergative. The question is to what extent ergativity is an organizing principle at other levels and, if it is, how that is related to the morphology. Larsen and Norman (1979) provided a lucid discussion of the issue. They pointed out that not every process which treats S and O alike, and differently from A, is a syntactic correlate of ergative morphology—​for such processes are found in languages of various alignment types. An example from discourse structure is the preference to introduce new discourse referents as S and O, but not as A (Du Bois 1987b, Chapter 2, this volume). Although Du Bois argues that this preference can be grammaticalized through ergative morphology, the preference itself is not restricted to morphologically ergative languages.1 Only a syntactic process which aligns S and O against A and which is found discriminately in languages with ergative alignment is a good candidate for a syntactic correlate of ergativity. The best candidate for such a correlate in Mayan involves constraints on extraction. Larsen and Norman noted that while absolutives freely extract in Mayan, some

1  Recent work on preferred argument structure in Tsotsil and Ch’ol (Martínez 2012; Vázquez and Zavala 2013) has argued that the preference does not actually pick out S and O. Rather it picks out Os and inactive (or unaccusative) S (see also Durie 1988, 2003). Thus, the underlying contrast in Tsotsil and Ch’ol is active/​inactive, not ergative/​absolutive.

 

738   Judith Aissen languages restrict the extraction of ergatives. Their approach was typological and they proposed that only languages with ergative morphology would treat ergatives and absolutives asymmetrically in extraction. I  will refer to this general phenomenon as the ergative extraction constraint (eec). Larsen and Norman did not address the question how the eec is related to ergative morphology nor what exactly ‘causes’ the eec. Two main approaches to these questions have been explored in subsequent work. One derives the eec from the way nominal arguments are Case-​licensed in eec-​sensitive languages (Campana 1992; Ordóñez 1995; Coon et al. 2014). The key structural assumption is that in such a language, O comes to occupy a position above A in canonical transitive clauses. In this view, the eec reflects a distinctively ergative syntax, one which is also reflected in the morphology. The other approach links the eec to the availability of distinct, dedicated morphosyntax which is used when the external argument is extracted. In contexts where this morphosyntax is required, extraction from the corresponding transitive is blocked, yielding the eec (especially Stiebels 2006; also Aissen 2003a; Erlewine 2016). In this view, no special syntax is implied for languages which are eec-​sensitive and there is no bifurcation of languages (or constructions) into those which are ‘syntactically ergative’ and those which are not. Central to any study of the eec in Mayan are the constructions which are used instead of canonical transitives to extract A—​what are called agent focus (af) constructions in the Mayan literature. The aim of this chapter is to lay out some of the basic facts concerning the eec and its relation to af. There is a good deal of variation in the ‘reach’ of the eec in Mayan and, correspondingly, in the distribution of af. We describe some of that variation in 30.3 and then turn in 30.4 to the two approaches sketched above. A note on terminology: I use A, S, and O as follows: A for the external argument in a 2-​argument clause (including the subject of an antipassive clause), O for the internal argument in a 2-​argument clause (including the ‘demoted’ object in an antipassive) and S for the sole argument in a one-​argument clause. The term ergative refers to the subject of a transitive clause and absolutive to the object of a transitive clause or the subject of an intransitive. Hence, when I speak of ‘ergative extraction,’ I mean extraction from a transitive clause, i.e., one with canonical transitive morphosyntax.

30.2 Background 30.2.1 Historical Relations The Mayan family includes some 30 extant languages, 20 spoken primarily in Guatemala and 10 in Mexico. A  commonly accepted genetic grouping, involving four primary branches, is shown in Table 30.1, following Kaufman (1976a). There is disagreement on some points, in particular, the affiliation of Tojolab’al (whether it should be classified

 

Correlates of ergativity in Mayan    739 with Q’anjob’alan or with Tseltalan). In addition to the genetic groupings, there are two groups in which language contact has resulted in diffusion of features across genetic lines. One is the so-​called ‘Greater Lowland Mayan Subgroup’ which includes the Yucatecan, Ch’olan, and Tseltalan languages and, more peripherally, Tojolab’al, Q’eqchi,’ Poqom, and possibly Ixil (see Law 2014 for discussion). The other is the Huehuetenango diffusion area, where speakers of Mamean and Q’anjob’alan languages interacted. These groupings will be relevant for what follows.

Table 30.1 Mayan family Huastecan

Huastec Chicomuceltec

Yucatecan

Yucatec Lacandon Mopan Itzaj Q’anjob’alan (Q)

Popti’ (Jakaltek) Akatek Q’anjob’al Mocho’ Tojolab’al Chuj

Western Maya Ch’olan-​Tseltalan

Tseltalan (T)

Tsotsil Tseltal

Ch’olan

Ch’ol Chontal Ch’orti’ Q’eqchi’ Poqomchi’ Poqomam

K’ichean (K) Central K’ichean Eastern Maya

K’iche’ Sipakapense Sakapultek Tz’utujil Kaqchikel Uspantek

Mamean

Mam (Tekti)tek Awakatek Ixil

 

740   Judith Aissen

30.2.2 Sets A and B Typologically, Mayan languages are fairly agglutinative, with rich derivational morphology. They are head-​initial and exhibit a high degree of head-​marking; personal pronouns are usually not pronounced unless emphatic. As Mayan languages are head-​ marking, the locus of morphological ergativity lies in the agreement system. Agreement in all Mayan languages involves two sets of morphemes, so-​called Sets ‘A’ and ‘B.’ Finite intransitive verbs contain one morphological position for agreement, finite transitive verbs contain two. Intransitive verbs use Set B to index the subject (S). Transitive verbs use Set A to index the external argument (A) and Set B to index the internal one (O). This pattern is summarized in Table 30.2 and is illustrated by (1) from Tsotsil. Table 30.2 Ergative pattern Set A S A O

(1)

✓

Tsotsil a. Ch-​i-​‘abtej av-​u’un ok’ob. icp-​b1s-​work a2-​rn tomorrow ‘I will work for you tomorrow.’

Set B ✓ ✓

(Haviland 1981: 146)

b. L-​i-​s-​vula’an j-​me.’ cp-​b1-​a3-​visit a1-​mother ‘My mother visited me.’ c. Ta j-​maj. icp a1-​hit ‘I will strike him/​her/​it/​them.’ Note that the marker for 1st person S in (1a) and 1st person O in (1b) is the same and distinct from the marker for 1st person A in (1c). These examples illustrate two further properties of Mayan agreement. First, in most languages, 3rd person singular S/​O are not indexed by any overt marker (see (1c)). I assume there is no Set B3 marker and do not represent any such marker in examples. Second, Set A markers are also used to index nominal possessors, as in (1b): the 1st person prefix marking the possessor in (1b) is identical to the 1st person prefix marking A in (1c). Nominal possession also plays a role in licensing oblique arguments and adjuncts. These relations are often expressed in Mayan by so-​called ‘relational nouns’ (rn), as

 

Correlates of ergativity in Mayan    741 in (1a), where -​u’un denotes a benefactive relation (it has other meanings as well). The nominal that bears the relation, e.g. the 2nd person in (1a), functions syntactically as possessor of the rn and, like other nominal possessors, is indexed by Set A. The use of rn’s to express oblique relations will be relevant to what follows. The Set A markers are prefixing in all Mayan languages, but the position of Set B varies. In some languages, Set B markers precede the agreeing stem (2a), in others, they follow it (2b), and in yet a third set, their position is variable (see also Table 30.3). (2) a. Tz’utujil x-​at-​nuu-​ch’ey cp-​b2s-​a1s-​hit ‘I hit you’ b. Tseltal la     s-​koltay-​at cp  a3-​help-​b2s ‘S/​he helped you’

(Dayley 1985: 89)

(Polian 2013: 127)

Tsotsil and Q’eqchi’ (distantly related) are ‘variable’ languages. In the presence of a preverbal aspect prefix or particle, Set B suffixes to the aspect prefix, as in (1a,b); otherwise, it suffixes to the verb, as in (3a,b). (3) Tsotsil a. tal-​em-​on come-​pf.iv-​b1s ‘I have come’ b. j-​chuj-​oj-​ot a1-​tie-​pf.tv-​b2s ‘I have tied you up’ A plausible scenario is that the Set B markers were second-​position clitics in Proto-​ Mayan (T. Kaufman p.c. 1978; see also Robertson 1992). As aspect markers generally developed from independent words or clitics, Set B clitics would have attached to the right of the aspect marker, if one was present (4a), and to the predicate if one was not (4b). (4) a. asp=b (a)-​v b. (a-​)v=b These patterns were then grammaticized in different ways in different languages, with Tsotsil and Q’eqchi’ reflecting something close to the original distribution.

 

742   Judith Aissen Finally, note that it is a consequence of various features of Mayan (verb-​initiality, head-​marking, the way agreement is organized) that a single nominal expression in a transitive clause may be ambiguous as to grammatical function. In (5), for example, verb inflection indicates that both A and O are 3rd person, but there is nothing in the structure itself—neither case, agreement, nor word order—​which indicates whether the single overt argument is A or O. (5) Tsotsil (Haviland 1981: 254) I-​s-​mil li Antun-​e. cp-​a3-​kill det Antonio-​enc ‘He killed Antonio’ or ‘Antonio killed him/​it’ (5) contrasts with (1b) where the persons of A and O are different and the 3rd person argument is unambiguously identified as A by agreement morphology.

30.3  Correlates of Ergativity 30.3.1 Ergative Extraction Constraint There is one class of constructions which, in a subset of Mayan languages, distinguish absolutives and ergatives. These are constructions involving extraction of a core argument to a non-​argument position (what is known as ‘wh-​Movement,’ ‘A-​bar movement,’ etc.). Constructions formed with A-​bar movement in Mayan include interrogatives, focus constructions, relative clauses, and certain indefinite constructions. The basic observation is that while absolutives can be extracted in all Mayan languages without modification to the verb or to clausal organization, there is substantial variation when it comes to extraction of ergatives. Some permit ergative extraction (i.e. from a transitive clause) without modification, while others require the use of a dedicated construction to extract the external argument, at least in some contexts. The morphosyntactic properties of this construction and the conditions under which it is required vary from language to language. Regardless of its properties, I will refer to constructions dedicated to agent extraction as agent focus (af), i.e. af refers to the function of the construction, not to its form. Consider first Ch’ol, a Ch’olan language which lacks af morphosyntax and permits ergative extraction. (6) Ch’ol (Coon et al. 2014) Maxki tyi y-​il-​ä jiñi wiñik? who asp a3-​see-​tv det man ‘Who saw the man?’ or ‘Who did the man see?’

 

Correlates of ergativity in Mayan    743 Example (6) is ambiguous between extraction of A and extraction of O, as the gramma­ tical functions of the core arguments are not identified by case, word order, or agreement (both arguments being 3rd person). This ambiguity is resolved by agreement when the two arguments differ in person (Coon et al. 2014). (7) Maxki tyi y-​il-​ä-​yety? who asp a3-​see-​tv-​b2 ‘Who saw you?’ Languages which lack af morphology and permit ergative extraction include most of the Yucatecan languages (all except Yucatec), the Tseltalan-​Ch’olan languages (except, to a limited degree, Tsotsil), and Tojolab’al. At the other extreme are most of the K’ichean, Q’anjob’alan, and Mamean languages, which have dedicated af morphosyntax and require it to extract A. Formally, af constructions come in two main types. Both involve detransitivization of the verb, but they differ in how O is syntactically realized. In one type, O is realized as an oblique; in the other, it is realized as a direct argument. Consider first Q’eqchi’ (K’ichean). From the transitive clause in (8a), extraction of A is impossible, *(8b). (8) Q’eqchi’ (Berinstein 1985: 162-​165) a. X-​in-​x-​sak’ li wiinq. rcp-​b1s-​a3s-​hit det man ‘The man hit me.’ b. * Ha’ li wiinq k-​in-​ix-​sak.’ foc det man cp-​b1s-​a3s-h ​ it (‘That’s the man who hit me.’) To extract A in Q’eqchi’, the construction in (9a,b) is required. The verb is intransitive and O is oblique. (9) a. Ha’ li wiinq ki-​sak’-​o-​k foc det man cp-​hit-​af-​iv ‘That’s the man who hit me.’ b. Ani ta-​paab’aa-​n-​q who pot-​ask-​af-​iv ‘Who will ask him?’

w-​e. a1s-​dat

r-​e? a3s-​dat

Morphological evidence that the verbs in (9a,b) are intransitive includes the fact that they do not carry Set A markers (nor in these cases any overt Set B marker, since the sole direct argument, the subject, is 3rd singular). On the other hand, they do carry

 

744   Judith Aissen aspect-​related ‘status’ suffixes (-​k/​-​q) which are restricted to intransitive stems.2 There is also syntactic evidence that (9a,b) are intransitive, namely the fact that O is oblique. Formally, O is realized as possessor of the rn -​e which licenses indirect objects in Q’eqchi’ (recall the discussion of rn’s in connection with (1a)). Putting this all together, it is clear that af clauses in Q’eqchi’ have the form of a cano­ nical demotional antipassive (see Polinsky, Chapter 13, this volume). In Q’eqchi,’ antipassive structures with O realized as oblique are found only when A is extracted. Hence it is a structure which is specialized for agent extraction, possibly a ‘last resort’ to enable extraction of A.3 The second type of af construction also involves detransitivization of the verb but now with no visible demotion of O. Example (10a) is a canonical transitive clause in Akatek (a vso, Q’anjob’alan language). In contrast to Ch’ol, extraction of a core argument from a transitive clause can only be interpreted as O-​extraction (10b). To extract A, the verb is detransitivized by the suffix -​on (10c). (10) Akatek (Zavala 1992: 279; (10b) was kindly provided by R. Zavala (p.c.)) a. [X]‌-​s-​ma’ ix malin naj xhunik. cp-​a3-​hit cls:det Maria cls:det Juan ‘Maria hit Juan.’ b. Maj [x]‌-​s-​ma’ naj xhunik? who cp-​a3-​hit cls:det Juan ‘Who did Juan hit?’ (not ‘Who hit Juan?’) c. Maj [x]‌-​ma’-​on who cp-​hit-​af ‘Who hit Juan?’

naj cls:det

xhunik? Juan

The verb in (10c)—​like the ones in (9a,b)—​is morphologically intransitive: it lacks a Set A marker and, if it occurs sentence-​finally, carries an intransitive status suffix (see (11)). But in this case, the O argument is not oblique. An obvious possibility is that (10c) is an antipassive, but with a ‘masked’ oblique. However, the verb can agree with O, showing that O is, in fact, maintained as a direct argument. Agreement is via Set B (Zavala 1992: 279):4

2  Status suffixes, which index the distinction between transitive and intransitive, often conflated with aspect and mood categories, are common in Mayan. In some languages, their occurrence or form is prosodically determined (Henderson 2012). 3  The detransitivizing suffixes in (9a,b), -​o and -​n are used elsewhere in Q’eqchi’ for other antipassive constructions (without oblique O) and they have cognates elsewhere in Mayan which are associated with antipassive and af. The choice in Q’eqchi’ between -​o and -​n is determined by whether the verb stem is a ‘root’ transitive (-​o) or a ‘derived’ transitive (-​n). ‘Root’ transitive stems are cvc in form, ‘derived’ transitives are usually longer. 4  No Set B marker occurs in (10c) because O is 3rd person singular.

 

Correlates of ergativity in Mayan    745 (11) Maj x-​in-​ma’-​on-​i? who cp-​b1s-​hit-​foc-​iv ‘Who hit me?’ Structures like (11)—​ with a morphologically intransitive verb but two direct arguments—​are possible only when A is extracted. I refer to both the oblique af (e.g. 9a,b) and the ‘direct’ one (e.g. 10c, 11) as agent focus (af), distinguishing them as afobl and afdir when necessary. The availabi­ lity of the two af structures is summarized in the following paragraphs (see Stiebels 2006 for extended discussion). Semi-​colons separate languages belonging to different branches. (12) a. only afobl: Mam; Q’eqchi’ b. only afdir: Popti,’ Q’anjob’al, Akatek, Chuj: Tsotsil; Poqom; Yucatec c. both afobl and afdir: K’iche,’ Tz’utujil In languages with both, the same morphology is involved, whether the construction is direct or oblique. An example from Tz’utujil is shown in (13). Note that the internal argument is indexed on the oblique marker in the afobl structure (13a) and on the verb itself in the afdir structure (13b). We will come back to this. (13) Tz’utujil (Dayley 1985: 350-​351) a. Jaa’ n-​ili-​n w-​xiin. she icp-​serve-​af a1s-​obl ‘She’s the one who serves me.’

afobl

b. Jaa’ n-​in-​ili-​n-​i. she icp-​b1s-​serve-​af-​iv ‘She’s the one who serves me.’

afdir

The previous discussion may have suggested that Mayan languages can be bifurcated into two distinct classes: those which lack af morphosyntax and permit ergative extraction, and those which have af morphosyntax and disallow ergative extraction. However, the situation is more complex. For one thing, at least one language, Tsotsil (Zinacantec dialect), has af, but nonetheless permits (in fact, requires) ergative extraction in most contexts. For example, ergative extraction is possible in examples like (14); as in Ch’ol, the result is ambiguous (cf. 6). (14) Tsotsil (Aissen 1999a: 459) Buch’u i-​s-​kolta li tseb-​e? who cp-​a3-​help det girl-​enc ‘Who helped the girl?’ or ‘Who did the girl help?’

 

746   Judith Aissen But af is used when, roughly, only the wrong interpretation is likely to result without it. af is possible only when both arguments are 3rd person—​in any other case, agreement fully determines grammatical function. And then, af is only used if O has the properties statistically associated with A—​when O is higher in animacy than A, (15a), or when O is the sentence topic (15b) (Aissen 1999a, 2003a). (15)

Tsotsil (Aissen 1999a: 464, 469) a. K’usuk nox tij-​on-​uk li j-​malal-​e. whatever just awaken-​af-​irr det a1-​husband-​enc ‘Just anything wakes my husband.’ b. Pero buch’u i-​mil-​on? but who cp-​kill-​af ‘But who killed her (i.e. the woman, mentioned in previous sentence)?’

Beyond Tsotsil, all languages which are subject to the eec appear to have contexts in which ergative extraction is possible, sometimes, in fact, the only option. These ‘exceptional’ contexts need to be accounted for and they provide several dimensions along which alternative accounts can be compared. Three are discussed in the next section. Two involve binding and one involves certain A-​O person settings.5

30.3.2 Exceptions to the EEC 30.3.2.1 Reflexive Clauses Reflexive clauses look transitive across Mayan. The verb is transitive (it inflects with Set A markers and carries transitive status suffixes) and it takes the reflexive anaphor as object. Much like English my-​self, your-​self, etc., the reflexive anaphor is a possessed noun whose possessor agrees in person and number with its antecedent. The features of the possessor are indexed on the anaphor by Set A, while those of the antecedent are indexed on the verb, also by Set A, as expected for transitive subjects. The anaphor, being always 3rd person singular, is not indexed by any overt Set B marker. Example (16), from K’iche,’ is typical. (16) K’iche’ (López Ixcoy 1997: 236) X-​aw-​il    aw-​iib’  pa  ja.’ cp-​a2s-​see  a2-​rr in water ‘You saw yourself in the water.’ While reflexive clauses are transitive, many Mayan languages otherwise subject to the eec permit (or require) that the external argument be extracted as an ergative from a transitive construction. In K’iche,’ for example, A extracts as ergative from a reflexive clause. 5  A further exceptional context is noted in Aissen (2011): K’iche’ permits ergative extraction when O is a bare nominal. af is also possible. The choice appears to have interpretive consequences related to specificity, though this needs to be further investigated.

 

Correlates of ergativity in Mayan    747 (17) K’iche’ (Mondloch 1981: 233) a Aree xun kumatz u-​b’aq’ati-​m r-​iib.’ foc one snake a3s-​roll-​prf.tv a3s-​rr (‘It was a snake that coiled itself (around the tree).’) b. * Aree xun kumatz b’aq’ati-​n-​naq r-​iib.’ foc one snake roll-​af-​pf.iv a3s-​rr (‘It was a snake that coiled itself (around the tree).’

30.3.2.2 Extended Reflexive Clauses These are clauses in which A binds the possessor of O, e.g. hei saw hisi (own) sister, whoi saw hisi (own) sister? They are also transitive. Again, some languages which restrict ergative extraction from a transitive clause nonetheless permit (or require) it when the object is an ‘extended reflexive.’ K’iche’ is one that requires ergative extraction, precluding the use of af. (18) K’iche’ (Mondloch 1981: 235, 237) a. Aree lee Axwaan x-​u-​k’at foc det Juan cp-​a3s-​burn ‘Juan is the one whoi burned hisi foot.’ b. (*)Aree lee Axwaan x-​k’at-​ow foc det Juan cp-​burn-​af ‘Juan is the one whoi burned hisj/​∗i foot.’

r-​aqan. a3s-​foot r-​aqan. a3s-​foot

The paradigm in (18) clearly has to do with binding, as (18a) is possible only with a bound reading, and (18b) is possible only under a disjoint reading. Table 30.3 shows how A is extracted from reflexive and extended reflexive clauses in several eec-​sensitive languages where the facts have been documented. It appears that reflexive clauses permit ergative extraction more readily than extended reflexive ones. Note that this domain is characterized by a good deal of optionality, perhaps reflecting change in progress.

Table 30.3 Agent extraction from (extended) reflexive clauses Tsotsil (T) K’iche’ (K) Q’anjob’al (Q) Popti’ (Q) Q’eqchi’ (K) Tz’utujil (K) Chuj (Q)

reflex

ext reflex

references

tv tv tv tv tv tv/​af tv/​af

tv tv tv tv/​af afobl afdir,obl tv/​af

Aissen 1999a Mondloch 1981: 232ff. Pascual 2007: 75ff. Craig 1977: 217ff. Berinstein 1985: 102ff. Aissen 2011 Hou 2013

 

748   Judith Aissen

30.3.2.3 Person of A and/​or O Some languages which otherwise restrict ergative extraction nonetheless permit (or require) it under certain person combinations of A and O (Stiebels 2006). ◦ Local A or Local O: In Tsotsil, af is possible only when both arguments are 3rd person. In any other setting, ergative extraction is the only possibility (Aissen 1999a) ◦ Local A: In Popti’ and Q’anjob’al, af is impossible when the external argument is 1st or 2nd person. In this case, ergative extraction is fully grammatical (Craig 1979: 59; Pascual 2007). ◦ Local A & Local O: In dialects of K’iche’ with access only to afdir structures, af is impossible if both A and O are 1st or 2nd person. In this setting, A extracts as an ergative (Mondloch 1981: 223) (see section 30.4 for further discussion and a refinement). It appears that person-​based restrictions arise only for afdir, not for afobl, a point we come back to.

30.4  Approaches to the EEC 30.4.1 Case-​Based Approach Within the tradition of Government Binding Theory/​ Minimalism, Case-​ based approaches to the eec have been pursued since the early 1990s. In this approach, the eec is a consequence of the way Case is assigned to the object in a transitive clause. In a syntactically ergative language, absolutives –​both S and O –​enter into a Case-​licensing relation with the same head, one usually associated with finiteness. I refer to it here as ‘finite I.’ Crucially, while O starts below A, it moves to a structural position above A in order to enter into the proper structural relation with finite I. For locality-​related reasons (relativized minimality or phasehood/​barriers), the resulting structure traps the ergative, preventing it from moving to a higher A-​bar position (Campana 1992; Ordóñez 1995; Bittner and Hale 1996a,b; Aldridge 2004; Coon et al. 2014). (19) shows the problem schematically; the path represented by the arrow is blocked by the intervening absolutive. (19) Absi (O)

I fin

Erg (A)

V

ti

 

Correlates of ergativity in Mayan    749 af constructions function then as a ‘last resort’ which permit O to be Case-​licensed without raising over the ergative and thereby allow extraction of the ergative. It is straightforward to see how afobl (i.e. antipassive) remedies the situation: O is Case-​licensed in a low position by a preposition or relational noun and does not raise. Hence no intervening argument prevents A from moving to its target A-​bar position. It is less obvious how the afdir construction avoids the problem posed by (19), but Ordóñez (1995) made the influential suggestion for Popti’ that af morphology itself Case-​licenses the internal argument, permitting it to remain low.6 Since af morphology does not occur freely in Popti,’ he took it to be a ‘last resort,’ to permit extraction of the external argument while still case-​licensing the internal one. What about the ‘morphologically ergative’ languages which lack af constructions and where ergative extraction from a transitive clause is possible (e.g. Ch’ol)? Early approaches assumed that S and O were licensed by the same head (e.g. finite I) in ergative languages (accounting in that way for their morphological identity), but that the position in which O was licensed could vary. In syntactically ergative languages, it raised above A, while in morphologically ergative ones, it remained below A (Bittner and Hale 1996b). More recent work has suggested that there are deeper differences between the two types of languages, that in syntactically ergative languages, finite I licenses both S and O in a high position, while in morphologically ergative ones, finite I licenses S, but a lower head, e.g. v, licenses O (Aldridge 2004; Legate 2006, 2008). Legate calls the case associated with finite I ‘nominative’ and the one associated with v ‘accusative.’ In morphologically ergative languages, the syntactic distinction between nominative and accusative is neutralized in the morphology, yielding what we call ‘absolutive.’ Coon et al. (2014) (cmp) adopt this framework and apply it to Mayan. Their analysis of languages which observe the eec follows that of earlier work. These are the languages in which O is licensed by finite I, requiring raising of O to a high position, and trapping the ergative. This is the source of the eec. Further, as in earlier work, af constructions permit O to be licensed in a low position, either by a relational noun or by a v head associated with af morphology. Because af occurs only when ergative extraction is impossible, these constructions are a ‘last resort’ option. What is new is their proposal for languages which permit ergative extraction. cmp propose that in these, O is licensed by a lower head, v, and remains low. cmp call the two types of languages ‘high-​abs’ and ‘low-​abs’ respectively. The Case-​centered approach to the eec has some very attractive features. Since the eec is a consequence of the way O is Case-​licensed, it predicts Larsen and Norman’s implicational universal: if a language treats ergative and absolutive differently in extraction, it has ergative morphology. It provides a structural Case-​based explanation for why some languages are not subject to the eec. And finally, it can (and does) appeal to exactly the same explanation for areas of exceptionality within languages that are subject to the eec: in these contexts, for whatever reason, O does not raise past A for case-​ licensing and hence does not ‘trap’ A. 6 

In Ordóñez’s account, the af suffix is an incorporated preposition, so licenses Case as a preposition.

 

750   Judith Aissen At the same time, the cmp analysis and the larger framework in which it is embedded raise a number of issues, three of which I discuss in the next section. Two of these pertain to particular analyses which put Case at the center of further (possibly) eec-​related pheno­ mena, and the third more generally to the role played by ‘last-​resort.’

30.4.2 Issues 30.4.2.1 Case and the Position of Set B cmp’s proposal that licensing of O involves I in some Mayan languages, but v in others is motivated in part by a very interesting observation made in Tada (1993). The observation was that there is a correlation between the position of the Set B marker and sensitivity (or not) to the eec. The correlation is not exact, but roughly, it is that languages which are sensitive to the eec attach Set B high, while languages which are not attach it low. Table 30.4, taken from cmp, shows the correlation. Table 30.4 Position of Set B marker and sensitivity to eec + extraction asymmetry high attachment

Q’anjobal, Akaktek, Popti,’ Chuj, Q’eqchi,’ Uspantek, K’iche,’ Poqomam, Poqomchi,’ Kaqchikel, Tz’utujil, Sakapultek, Sipakapense Mam, Awakatek

low attachment

Yucatec, Ixil

– extraction asymmetry

Lacandon, Mopan, Itzaj, Ch’ol, Chontal, Tseltal, Tojolab’al

Source: Coon et al. (2014)

cmp derive the presence/​absence of the eec and the position of the Set B marker from the same underlying factor, namely, the head involved in licensing O. They propose that low attachment of Set B reflects in situ case licensing by v, high attachment reflects raising to finite I for licensing. Since for them it is this raising which underlies the eec, they seem to predict the pattern seen in Table 30.4: languages with low Set B markers will permit ergative extraction from transitive clauses, while languages with high Set B markers will not. Taken at face value, the position of the Set B marker should correlate with its licensor. But this is not always the case. For example, O in an af clause in Q’anjob’al is licensed by v (per cmp), but its Set B marker attaches high (see (11)). cmp assume an epp feature to force the internal argument to a high position. This results in high attachment of Set B, but now the connection between the licensing head and the position of Set B has become opaque. An alternative possibility is that the position of Set B is a matter of the morphology, and that Q’anjob’al is a language in which Set B attaches high, regardless of how the corresponding argument is licensed.

 

Correlates of ergativity in Mayan    751 The indexing of intransitive S in the low attachment languages presents the converse problem. cmp assume that S is always licensed by I in finite clauses, even when O is licensed by v. If the position of Set B identifies the licensing head, Set B should attach high when it indexes S and low when it indexes O. But with few exceptions, the position of Set B is the same, whether it indexes S or O, i.e. its position is fixed in a language, regardless of what the licensing head is. If S and O are in fact licensed by different heads, morphological rules are needed in any case to position Set B, as its placement is not in fact determined by the licen­ sing head. The syntactic account of cmp then does not really explain the correlation shown in Table 30.4. There is an alternative, historical explanation for Tada’s generalization: the languages in the upper left cell are from the Q’anjob’alan, K’ichean, and Mamean branches, while the languages which populate the lower right cell are Yucatecan, Tseltalan and Ch’olan, the branches which make up the Greater Lowland Mayan Subgroup.7 This latter group shares a range of properties as a result of areal diffusion, e.g. a phonemic distinction between /​p’/​and /​b’/​and a distinction between 1st plural inclusive and exclusive. Set B suffixation and loss of af might be other independent developments shared by these languages for areal reasons (Law 2014).

30.4.2.2 Case and Reflexives Here we consider two of the systematic ‘exceptional’ contexts that permit ergative extraction in some languages, namely reflexives and extended reflexives. The strategy for the Case-​based approach is clear: explore analyses in which the problematic configuration (19), does not arise. Ordóñez (1995) proposed for Popti’ that reflexive objects do not raise for Case but are licensed through incorporation (Baker 1988). Consequently, the ergative is not trapped, it extracts without a problem and last-​resort Case licensing is not triggered. cmp suggest this analysis for both reflexives and extended reflexives in Q’anjob’al. Reflexive objects in a number of Mayan languages do show a kind of quasi-​incorporation in that they occur adjacent to the verb (even if the corresponding non-​reflexive would not). This is true in Popti’ and Q’anjob’al and provides some evidence for an incorporation account of reflexives (though not for extended reflexives, which fail to show the same adjacency effects). Incorporation would also unify reflexives and extended reflexives with bare n (possibly np) objects which, at least in some languages, also constitute an exceptional context and permit ergative extraction (see n. 5). However, the incorporation solution is problematic for Q’anjob’al (likely also for Popti’) because reflexive and extended reflexive O’s do not otherwise share the distribution of incorporated nouns. Q’anjob’al has two canonical incorporation constructions which permit as object only an n (or possibly np) without a determiner. One involves a finite antipassive verb formed with the suffix -​w (20a), and the other involves infinitives

7 

af in Yucatec is probably an innovation, since it is not found in the other Yucatecan languages. Further the form it takes is different from that of the rest of the family. Ixil has become a suffixing language under influence from the languages which populate the lower right cell (see text).

 

752   Judith Aissen (20b). (Determiners in Q’anjob’al, as well as personal pronouns, are drawn from a set of classifiers and vary according to features of the head noun or the referent.) (20) Q’anjob’al (Pascual 2007: 69, 91) a. k’am=to ch-​in=’uk’-​wi neg=cl icp-​b1s-​drink-​ap ‘I still don’t drink (*the) liquor.’

(*an) cls:det

an liquor

b. k’am mak x=y-​i=toq y-​etoq [say-​oj (*xim) ixim]. neg who cp=a3-​take=dir a3-​with seek-​inf cls:det corn ‘He didn’t take anyone with him to look for (*the) corn.’ As neither context permits a full dp object, the obvious hypothesis is that these minimal objects do not require Case, perhaps because they are incorporated, or for some other reason. However, reflexives and extended reflexives cannot function as the object in either construction.8 (21)

Finite antipassive a. * Chi lotze-​wi s-​b’a icp feed-​ap a3-​rr (‘She feeds herself.’)

ix. pro:fem

b. * Ch uk’-​wi y-​an naq. icp drink-​ap a3-​liquor pro:m (‘Heidrinks hisi/​j liquor.’) (22) Infinitive a. * Max s-​cheq-​toq ix heb’ naq cp a3-​send-​dir pro:fem pl cls:det (‘She sent the men to help each other.’)

winaq man

[kol-​oj s-​b’a]. help-​inf a3-​rr

b. * Max toj ix ix [lotz-​oj s-​kalnel]. cp go cls:det woman feed-​inf a3-​sheep (‘The womani went to feed heri/​j sheep.’) Hence with respect to Case, reflexive and extended reflexive objects pattern with dp’s, not with incorporated n’s, suggesting that Case (or at least Case alone) is not responsible for the exceptional behavior of (extended) reflexive clauses with respect to the eec.9 Another question arises in those languages where Agent extraction from reflexive or extended reflexive clauses has two options and can come either from a transitive clause or an af clause (see Table 30.3). If extraction from a transitive clause is possible, how 8 

9 

I am indebted to B’alam Mateo Toledo for these examples. Analogous problems arise in K’iche’. Space limitations preclude discussion.

 

Correlates of ergativity in Mayan    753 does the last resort option get triggered? This seems to require two modes of generating clauses with reflexives, one of which precludes ergative extraction and one which does not, together with an appropriate conception of how ‘last resort’ is triggered.

30.4.2.3 Person Exceptionality and ‘Last Resort’ We saw earlier that in some languages, the eec is inactive under particular A-​O person settings—​permitting ergative extraction—​and that these settings vary according to the language. The question for the Case-​based account is how just these structures avoid the problematic configuration which blocks ergative extraction (19). cmp discuss Q’anjob’al, where the eec is inactive when A is 1st or 2nd person. They suggest that because of its indexical properties, a local person A might originate in a higher position than usual, in particular, higher than the position to which O moves. The local argument is not ‘trapped,’ and can move to a higher A-​bar position (if different from its initial position). Here I want to consider a different case of person exceptionality, that of K’iche’ and Tz’utujil. The facts are different from those of Q’anjob’al and there is no reason to think that the same solution would apply. However the Case-​based approach to the Q’anjob’al problem gives us a starting place. Extraction of A in K’iche’ and Tz’utujil requires af in 3–​3 settings, as well as in 3-​Local and Local-​3 settings. Recall that af verbs have only a single position for agreement and that agreement is by Set B. In K’iche’ and Tz’utujil, Set B can index either argument, but only one of them. Which argument is indexed in afdir clauses is determined not by grammatical function, but by the person-​number hierarchy in (23), with Set B indexing the higher-​ranked argument (see Dayley 1981 and Mondloch 1981 for fuller descriptions; Stiebels 2006 and Preminger 2011a for recent theoretical accounts; and Coon and Preminger, Chapter 10, this volume, for a wider perspective). (23) local ≫ pl ≫ 3 Hence, when one argument is local and the other is 3rd, the appropriate Set B marker indexes the local person, whether it is A or O. (24a,b) show this in a striking way: the verb forms are the same, with Set B1s indexing A in (24a) but O in (24b). (24) K’iche’ (Mondloch 1981: 114, 64) a. In x-​in-​il-​ow-​ik. pro.1s cp-​b1s-​see-​af-​iv ‘I was the one who saw it.’ b. Jachin x-​in-​il-​ow-​ik? who cp-​b1s-​see-​af-​iv ‘Who saw me?’ The examples are unambiguous, since the af verb signals that the extracted (preverbal) argument must be A.

 

754   Judith Aissen This is very interesting, but our concern is with another aspect of the interaction of af and person, namely the outcome when both arguments are local. Speakers who have access to an afobl structure use that construction (all Tz’utujil speakers, some K’iche’ speakers). A is indexed on the verb (by Set B), and O on the oblique relational noun (by Set A). (25) Tz’utujil (Dayley 1985: 350) Inin x-​in-​ch’ey-​o aw-​xiin. pro.1s cp-​b1s-​hit-​af a2s-​rn ‘I was the one who hit you.’ However for speakers (or dialects) who have access only to an afdir structure (e.g. those speakers described in Mondloch 1981), extraction of A proceeds from a transitive clause. (26) K’iche’ (Mondloch 1981: 223) a. In k-​at-​in-​to’-​oh. pro.1s icp-​b2s-​a1s-​help-​tv ‘I myself will help you.’ b. At x-​in-​aa-​yoq’-​oh. pro.2s cp-​b1s-​a2s-​mock-​tv ‘You yourself mocked me.’ The question is how to generate these sentences in the Case-​based approach. In the logic of that approach, ergative extraction is expected unless it is blocked. In K’iche’ it is blocked in every other A-​O person setting because, per CMP, O raises over A for licensing. What then is syntactically special about a clause with two local person arguments such that the associated structure does not involve raising of O over A for licensing? Rather than speculate on this, I turn to a different perspective on the possibility of ergative extraction in (26), one which relates it to the priority that most Mayan languages place on indexing 1st and 2nd person arguments through agreement (Stiebels 2006). Given what we have observed, it is clear that this requirement cannot be satisfied in afdir clauses when both arguments are local, as only one of them can be indexed by means of a Set B marker. In the dialect described by Mondloch (1981), the resolution is to extract the ergative from a transitive clause so that both arguments are indexed on the verb.10 Several observations suggest that this is the relevant perspective. First is the fact that person-​based exceptions to the eec arise only when both arguments are local. If either A or O is 3rd person, the eec is in full effect, ergative extraction is blocked, and af is required. 10  According to López Ixcoy (1997: 369), some speakers permit an extracted local person A argument to be unindexed on the af verb.

 

Correlates of ergativity in Mayan    755 (27) K’iche’ (Mondloch 1981: 226) a. * Laa at k-​aa-​b’an lee ch’ajo’n? q pro.2s icp-​a2s-​do det washing (‘Are you the one who does the washing?’) b. Laa at k-​at-​b’an-​ow lee ch’ajo’n? q pro.2s icp-​b2s-​do-​af det washing ‘Are you the one who does the washing?’

tv

af

On the morphological account, this is expected, as it is only in Local–​Local settings that two positions are required for agreement. Whether it is expected on a syntactic account would depend on the details of such an account and, as far as I know, none has been proposed. Second is the fact that person exceptionality only arises for afdir structures, not afobl structures. It does not arise in Mam and Q’eqchi,’ which have only the oblique version, nor in Tz’utujil, which has both. From the morphological perspective, the reason is clear: the oblique construction makes available two agreeing heads, the verb and the oblique relational noun. A is indexed on the first, O on the second, and the requirement that local persons be indexed through agreement is satisfied. And third, K’iche’ has a 2nd person formal category which is indexed by a clitic which does not belong to either Set A or Set B. Rather than occupying the positions associated with Sets A and B, these morphemes (lah in the singular and alaq in the plural) cliticize to the right of the head. Consequently when one of the two arguments in an af clause is 2nd person formal, there is still space available on the verb to index the other argument. Hence even when both arguments are local in K’iche,’ as long as one of them is 2nd person formal, the eec is in full force and extraction of A proceeds from the af clause, not the transitive. (28) K’iche’ (Mondloch 1981: 221) In x-​in-​ch’aab’e-​n alaq. pro.1s cp-​b1s-​talk.to-​af pro.2p.f ‘I am the one who talked to you (pl).’ In short, it is not the indexical character of 1st and 2nd person that makes Local-​Local clauses exceptional, but their morphological properties. It seems highly likely then that in K’iche,’ ergative extraction from a transitive clause with local A and O is the solution to a morphological problem, not the reflection of a special syntax which is available only when both arguments are local. This problem is worth dwelling on because it bears on the ‘last resort’ character of af in the Case-​based approach. For a morphological problem to arise with af, an af derivation must exist. This implies in turn that a derivation with ergative extraction (from a transitive clause) must have failed. But how then can the transitive derivation be

 

756   Judith Aissen resurrected to ‘repair’ the morphological problem? However ‘last resort’ is understood, it needs to be able to deal with cases like these.11

30.4.3 A Marking Approach to the EEC A different approach to the eec is suggested by the typological fact that various unrelated languages use special marking of some kind when the subject is focused. Hartmann and Zimmermann (2007) discuss several Chadic languages of this type. They explain the special status of subject in terms of information structure: the default status for subjects is topic, not focus. Obligatory ‘focus’ marking serves then to signal their non-​canonical status (see also Schultze-​Berndt, Chapter 44, this volume). It is not hard to see why Mayan languages in particular might have specialized constructions for extraction of A. As we observed earlier, in the absence of disambiguating agreement morphology, extraction from a transitive clause is structurally ambiguous. Given the canonical status of transitive subjects as topics, when all else is equal, focus, interrogative, and indefinite extraction from such a clause is more likely to be interpreted as O-​extraction than A-​extraction. Hence there are clear functional reasons why special marking would exist for A-​extraction. In a marking account of the eec in Mayan, there is no structural problem with ergative extraction from a canonical transitive clause. Rather, the expected output is blocked by the existence in the language of an alternative structure which is specialized for extraction of A. Languages which are sensitive to the eec would be ones which have af morphosyntax, the other languages would not. When such a construction exists, it will usually be preferred when A is extracted. If, however, af is blocked for some reason, either syntactic or morphological, then the language will fall back on a less specialized construction, e.g. a canonical transitive clause. Since there is no structural problem with extraction from a transitive clause, extraordinary means to permit it (e.g. incorporation of reflexives and extended reflexives, an unusually high position for a local person A) are not required. Further, since there is no structural problem with either ergative extraction or af, the possibility that the two might coexist in some contexts (i.e. optionality) is not so problematic. The marking approach has been most fully developed in Stiebels (2006). Her account is fundamentally morphological and has little to say about the syntax of af, which she assumes to be essentially the same as the syntax of a transitive clause. Raising of O for Case-​related reasons plays no role. In Stiebels’ implementation, the af morpheme competes with the Set A morpheme to index the external argument in a transitive clause as both are restricted to ergatives. The key difference between them is that the af morpheme has a foc feature, but no phi (ϕ) features, while the Set A marker has ϕ features but no foc feature. Embedded in an OT architecture, the choice then between the transitive verb and the af verb is determined by a language-​particular ranking of constraints which enforce faithfulness to, or penalize the markedness of, particular features. 11 

See also the discussion in Erlewine (2016) which makes the same point.

 

Correlates of ergativity in Mayan    757 Languages which show no eec effects are ones in which faithfulness to ϕ features outranks faithfulness to the foc feature. In those which show eec effects, faithfulness to the foc feature outranks faithfulness to at least some ϕ features. As Stiebels’ account is intended to deal with morphological constraints on af verbs, it has an elegant account, for example, of the fact that particular A-​O person settings block the af form in some languages. Further, in this logic, af is not a ‘last resort.’ The transitive verb and the af verb compete and depending on constraint ranking and input, which emerges as optimal will vary. Thus the problem of ‘resurrecting’ the transitive form to repair a problem with the af form does not arise (30.4.2.3). The OT architecture also makes it possible to see how Tsotsil, a language which has af morphosyntax, but uses it very sparingly, is related to the other languages with af. The transitive form competes with the af form, but the former is always preferred over the latter, as long as the intended interpretation is recoverable. When it is not, the af form emerges as optimal, a state of affairs which can be modelled through bidirectional optimization (see Aissen 2003a for such an analysis).12 What is lacking in Stiebels’ analysis is enough of a syntax to account for those properties which are probably determined by the syntax or semantics, e.g. exceptional contexts involving reflexives, extended reflexives, and bare nouns, or the differences between afdir and afobl. Stiebels does not propose a formal analysis of reflexive or extended reflexive clause exceptionality. She assumes that af morphology arose to disambiguate clauses in which A is extracted (i.e. in 3–​3 clauses) and that it became grammaticized and extended in K’ichean, Mamean, and Q’anjob’alan to clauses where no ambiguity could arise, e.g. to clauses with one or more local persons. Reflexive and extended clauses are other clause types where, for binding reasons, no ambiguity can result from extraction as only the external argument can extract. The idea then is that the extension of af morphology through grammaticization is expected to reach such clauses late and to varying degrees. The optionality of af in these contexts (see Table 30.3) might indicate that grammaticization is still in progress. This makes some sense of why we find exceptionality and optionality in precisely this domain, but does not shed light on how it might be formally represented. The ‘marking’ framework needs to be harnessed to a syntax which is capable of accounting for the non-​morphological factors that are relevant to the distribution of af.

30.5 Conclusions We started with the proposal that there is at least one correlate of morphological ergati­ vity in Mayan, namely asymmetric restrictions on the extraction of ergatives and absolutives. Our main task here has been to understand what implications that has for the 12 

Roughly, bidirectional optimization identifies the optimal form from which the intended meaning is recoverable. See, for example, the papers in Benz and Mattausch (2011).

 

758   Judith Aissen syntax of these languages. The situation is complex and interesting because there is a good deal of variation across the family regarding the contexts in which ergative extraction is blocked and the properties of the alternative af structures. We have discussed here two approaches, both interesting, both as yet incomplete. One, based on Case, has a substantial tradition within generative syntax, it is linked to formal work on other ergative languages, and it has been elaborated in far-​reaching directions by recent work. The other, based on ‘marking,’ is more closely linked to typological work and to traditional ideas about agent focus in Mayan. It too has been developed in very promising ways. With respect to the question what the eec and af tell us about syntactic organization, the two approaches diverge. Under the Case-​based account, transitive clauses in eec-​sensitive languages have a distinctive ergative syntax, one in which O ends up in a position higher than A. The eec is thus a bona fide syntactic correlate of morphological ergativity. Under the marking approach, the eec is not a syntactic correlate of morphological ergativity. First, it implies nothing distinctive about the syntax of the eec-​sensitive languages—​ergatives appear to be treated differently from absolutives only because af structures exist as alternatives to transitive clauses, not intransitive ones. And second, since special marking of subject extraction is not found discriminately in morphologically ergative languages, it is not clear, under the marking approach, that the eec would be a correlate of ergativity at all. Much has been learned about the variation in this domain since the publication of Larsen and Norman (1979), but much remains to be documented. How is agent extraction realized in the exceptional contexts that have already been identified? are there further contexts in which ergative extraction is (unexpectedly) possible? how does agent extraction proceed from complement clauses of different types? does ‘salience’ play a role in the distribution of af in languages other than Tsotsil? We can hope that with further empirical and analytical work, the relations among the various phenomena discussed here will become clearer, bringing us closer to a better understanding of the larger questions.

Acknowledgments I would like to thank Jessica Coon, Amy Rose Deal, and Yuko Otsuka for their helpful comments and suggestions. I am solely responsible for the content.

Abbreviations a, Set A; act, active; af, agent focus; ap, antipassive; asp, aspect; b, Set B; cls, classifier; cp, completive; dat, dative; det, determiner; dir, directional; enc, enclitic; f, formal; fem, female; foc, focus; icp, incompletive; inf, infinitive; irr, irrealis; iv, intransitive status suffix; m, male; neg, negation; obl, oblique; p, plural; prf, perfect; pot, potential; pro, pronoun; q, polar question particle; rcp, recent past; rn, relational noun; rr, reflexive/​reciprocal; s, singular; TV, transitive verb or transitive status suffix.

 

Chapter 31

Ergative c ase i n Burushaski: A de pe nde nt case ana lysi s Mark C. Baker

31.1 Introduction Burushaski, spoken in a mountainous region of Northern Pakistan, is traditionally categorized as an isolate (although recently there have been claims that it is of Indo-​ European stock). As such, it is has received a fair amount of attention from descriptive and typological linguists: see Lorimer (1935), Berger (1974), Tiffou and Pesot (1989), Willson (1996), Grune (1998), Munshi (2006, n.d.), Yoshioka (2012).1 This descriptive work makes it clear that Burushaski is a canonical ergative language, in which the subject of a transitive clause bears an overt case marker (-​e) that distinguishes it from both the subject of an intransitive clause and the object of a transitive clause, which are morphologically unmarked (absolutive). (1) a. In gucar-​imi. he.ABS walk-​PST.3mS ‘He walked.’

1 

(Munshi n.d.: 15)

When I say for convenience that “all sources agree” about a certain grammatical point, I have in mind these sources, except for Berger (1974), which I do not have the language skills to read for myself. I take illustrative examples primarily from Willson (1996) and Munshi (n.d.), as the two modern sources that best cover the range of phenomena under consideration here. I also cite pages from Lorimer (1935) as the first comprehensive source on Burushaski, but do not present examples from that work because I would have to construct my own fallible glosses, and later authors express some concern about Lorimer’s phonetic transcriptions.

 

760   Mark C. Baker b. Hilés-​e dasin mu-​yeéts-​imi. boy-​ERG girl.ABS 3fO-​see-​PST.3mS ‘The boy saw the girl.’

(Willson 1996: 3)

First, Burushaski qualifies as a canonical ergative language in that it does not have monadic intransitive verbs that take ergative subjects (Yoshioka 2012: 246–​247); it is not an active language. Second, the marking of the subject usually does not vary with the tense–​aspect of the clause, as happens in so-​called split ergative languages of the Indo-​Aryan and Iranian families (but see section 31.5 for a partial exception). Third, there is very little differential subject marking in the language: nominals of all types bear the same ergative case affix -​e, including pronouns, the only exception being the second person singular pronoun un (Tiffou and Pesot 1989: sec. 4.3.3; Grune 1998: 6). Part of Burushaski’s interest, then, is that it is a prototypical ergative language that is not obviously related to other ergative languages. It thus provides an opportunity to test and broaden the empirical base of generative theories of ergativity, which have been developed with other languages in mind. With this goal in mind, this chapter presents the facts about morphological ergativity in Burushaski that are most relevant to such theories. Indeed, I  have one particular theory that I  am interested in testing and replicating: a version of the so-​called dependent case theory of Marantz (1991), as developed by Baker (2014a, 2015) (see also in this volume Baker and Bobaljik, Chapter 5; Coon and Preminger, Chapter 10; Nash, Chapter 8). According to this theory, ergative case is not assigned by a functional head: it is neither a standard structural case assigned by a functional head like T under Agree, nor an inherent case assigned by a theta-​marking head like v (e.g. as proposed in this volume by Laka, Chapter 7; Legate, Chapter 6; Sheehan, Chapter 3; Woolford, Chapter 9). Rather, ergative and absolutive case are assigned by the simple configurational rules in (2). (2) a. If NP1 c-​commands NP2 and both are contained in the same domain (say, clause), then value the case feature of NP1 as ergative. b. Otherwise NP is absolutive. This dependent case theory is attractive in that it can account for both ergative and accusative languages with pleasing simplicity and symmetry; an accusative language is simply one that chooses to value the case feature of NP2 rather than NP1 in its version of (2a). Example (2) easily accounts for the basic data in (1): in (1b), the subject c-​commands another NP (the object) in the same clause (see (8)), so it is marked ergative, whereas the object in (1b) and the subject in (1a) c-​command no other NP, and are thus left to be absolutive. One can, however, worry that (2) is too simple a theory of ergative case assignment, given that real live ergative languages present complications and apparent exceptions when one takes a closer look. This is certainly true of Burushaski. In particular, case

 

Ergative case in Burushaski    761 marking in Burushaski seems to vary somewhat depending on what verb heads the clause. This is most evident in Willson (1996), which I make heavy use of here, because his framework (Relational Grammar) leads him to attend to these differences, and he presents them in an accessible way. However, the major variations are also attested in Lorimer (1935) and Munshi (n.d.). For example, although most two-​argument verbs have ergative subjects, as in (1b), there are a few for which both arguments are absolutive, as in (3) (Willson 1996: 43). (3)

Jé káman peesá d-​á-​can-​abaa. I.ABS some money.x.ABS D-​1sO-​need-​1sS.PRES ‘I need some money.’

There is also lexical variation among verbs that take a complement in dative case. The majority of such verbs have absolutive subjects, as expected (Willson, 1996: 36, see also Munshi 2006: 142): (4) Zamindáar [tshíl yál-​as-​e gáne] farmer.ABS water apply-​INF-​OBL for ‘The farmer went to the field to water (it).’

mál-​e-​r field-​OBL-​DAT

ni-​imi. go-​3mS.PST

However, a few such verbs have ergative subjects, resulting in an ergative–​dative pattern in which absolutive case is apparently not assigned.2 (5) Hilés-​e dasín-​mo-​r barén-​imi. boy-​ERG girl-​OBL-​DAT look.at-​3mS.PST ‘The boy observed the girl.’

(Willson 1996: 4)

Because of data like this, it is often said in descriptive terms that different case frames need to be stipulated in the lexical entries of different verbs. In contrast, a dependent case theory built around (2) leads to the strong expectation that ergative case assignment is purely structural. I argue that the dependent case theory is correct in this, despite appearances. The verbs in (3) and (5) do not call for special case assignment directly; rather, their thematic properties cause special clause structures to be projected, and those clause structures determine how case assignment happens in purely structural terms. In this, I  pursue a common analytical strategy within Chomskyan structure-​ based theories. I also extend the account to cover one additional seeming anomaly in Burushaski’s ergativity: the fact that clauses in future tense may have absolutive case subjects even if they are transitive. 2 

This particular verb crucially always takes a dative object, even when the observed thing is inanimate, as seen in (19b). In contrast, verbs like ‘hit’ never take a dative object, even when the object is animate and definite, as seen in (6). Therefore, Burushaski is not a so-​called differential object marking language, the way Urdu and many other languages of the region are.

 

762   Mark C. Baker

31.2  Agreement in Burushaski Before facing these puzzles about ergative case marking, it is worthwhile to review a major resource that Burushaski offers to linguists investigating its grammatical structure: its rich agreement system, which has attracted significant attention from all general descriptive sources, as well as more specialized studies like Anderson and Eggert (2001) and Baker (2008). Although nothing is known about c-​command tests like variable binding, anaphora, and quantifier scope in Burushaski, agreement can be used to test structural hypotheses. Burushaski agreement is rich in that it has both subject agreement and object agreement, as all sources agree. Both can be seen in (1b), and the similar example in (6) (Munshi n.d.: 16; see also Lorimer 1935: 63–​67, 219). (6)

In-​e hiles i-​del-​umo. she-​ERG boy.ABS 3mO-​hit-​3fS ‘She hit the boy.’

Subject agreement and object agreement show up in quite different places within the verbal morphology: subject agreement is a suffix, as is the tense inflection with which it varies, whereas object agreement is a prefix and is independent of tense inflection. It is thus reasonable to say that these two forms of agreement are the result of two different functional heads, each agreeing with the closest NP. T agrees with the subject, whereas v agrees with the object, as in Chomsky (2000, 2001) and related work. Confirmation for this comes from the fact that in infinitival clauses subject agreement is absent, whereas object agreement is unperturbed, as shown in (7)  (Willson 1996:  30; see also Grune 1998: 13 and Munshi n.d.: ex. (64a)).3

3  Miriam Butt and Maria Polinsky both ask if what I call “object agreement” could really be a type of incorporated (cliticized) pronoun, and if so, how the analysis would be affected. The issue is interesting, and I cannot fully resolve it here. The position and in some cases the form of the object marker might indeed suggest a clitic analysis (see Grune (1998: 5–​6) for a comparison of the forms). But there are some good reasons to adopt an agreement analysis, as in fact all the descriptive sources do. First, object marking on the verb is required (depending on the lexical properties of the verb) even if the object is a nonspecific indefinite (Munshi n.d.: ex. (118)) or a nominal without independent referential power, like a reflexive anaphor (Willson 1996: 18 ex. (36)). Second, assuming that the putative incorporated pronouns have absolutive case, then they disagree in case with the overt nominal that doubles them in examples like (9). Such case mismatches are not common in the clitic-​doubling literature. Third, it is not clear how a clitic-​doubling approach would explain the presence of clitics doubling the subjects of unaccusative verbs but not unergative verbs (see (10) and (11)). Even if object markers do turn out to be cliticized pronouns, they can still provide a test for syntactic structure, as long as Chomsky’s Agree is a precondition for cliticization, as in the theory of Kramer (2014a).

 

Ergative case in Burushaski    763 (7)

Ú-​e [dasin-​ants mi-​e-​r mi-​ú-​as] rái a-​é-​t-​c-​aan. 3p-​ERG girl-​PL.ABS 1p-​OBL-​DAT 1pO-​give-​INF want NEG-​3yS-​do-​NPST-​3pS ‘They do not want to give girls to us (in marriage).’

This is what we expect if subject agreement is a property of finite T, whereas object agreement is a property of a distinct syntactic head. The structure of a clause like (6) is thus roughly (8). TP

(8) vP NP

v’

T 3ms

v boy VP 3fs ERG NP V girl see (ABS) We may then tentatively assume that the inflected verb is built by head movement, which right-​adjoins the verb root to v, and then left-​adjoins the v+V complex to T. The result is that the object agreement in v ends up as a prefix,4 and tense morphology together with subject agreement ends up as a suffix to the verb stem (although alternatives exist; the details are not crucial). We need to bear in mind, however, is that there is some lexical variability concerning object agreement in Burushaski. Some verbs simply prohibit it, some permit it with countable nouns (gender x) but not mass nouns (gender y) (Willson 1996: 13, 26–​27, 33), and some use it only with human nouns (Yoshioka 2012: ch. 9, who refers also to Berger). This variation complicates somewhat one point made below. Another useful property of agreement in Burushaski is that both subject and object agreement are insensitive to the case of the agreed-​with NP, as recognized by all sources, and emphasized in Baker (2008). Thus T agrees with the subject in the same way regardless of whether it is ergative (in transitive sentences) or absolutive (in intransitives), as shown by underlining in (1a,b). Likewise, v agrees with the highest NP inside VP in the same way regardless of whether it is or absolutive, as in monotransitives like (6), or 4 

One might wonder whether v ever has an overt morphological realization in Burushaski beyond object agreement, as Tense does. Possible candidates are the d-​ and u-​prefixes found on intransitives in some anticausative pairs (Willson 1996: 12, 27–​29), and s-​, a transitivizer/​causativizer found only with unaccusative roots (Willson 1996: 57–​58; Munshi n.d.: (97)). Indeed, all three show up as prefixes near the verb root, as object agreement does.

 

764   Mark C. Baker dative, as in ditransitives like (9) (and (7)) (Willson 1996: 35; see also Lorimer 1935: 219; Grune 1998: 13; Anderson and Eggert 2001: 239; Munshi 2006: 138–​139; Munshi n.d.: 30; Yoshioka 2012:  242). Example (9)  has clear object agreement (mu-​) with the feminine singular dative object, but no sign of object agreement with its x gender theme (i-​); indeed verbs never bear agreement with more than one object in Burushaski. (9) a. Ja-​a in-​mo-​r hán tofá-​an mu-​ú-​abayam. I-​ERG her-​OBL-​DAT one gift.x-​INDEF.ABS 3fO-​give-​1sS ‘I have given her a gift. b.

TP vP NP

v’

T 3ms

v I VP ERG 3fs NP V' her NP V DAT one gift give (ABS) Therefore, agreement provides a window into the structure of the clause that is independent of morphological case. Finally, Burushaski is quite unusual in that the sole argument of an unaccusative verb—​a verb that selects a theme argument but no agent—​triggers both object agreement and subject agreement on the verb. This can be seen in (10); contrast the unergative examples in (1a) and (11), where the sole argument triggers only subject agreement (see also Lorimer 1935: 218–​219). (10) a. Acaanák hilés i-​ír-​imi. suddenly boy 3mO-​die-​3mS.PST ‘Suddenly the boy died.’

(Willson 1996: 19)

b. Uwe hilešo u-​waal-​uman. DEM boy.PL 3pO-​lost-​3pS ‘Those boys [got] lost.’ (11)

a. Síruf hir-​i girát-​c-​aan only man-​PL.ABS dance-​NPST-​3pS ‘Only men dance here.’

(Munshi n.d.: 15)

akhóle. here

(Willson 1996: 19)

 

Ergative case in Burushaski    765 Willson (1996:  19–​21) describes this explicitly as a difference between unaccusative verbs and unergative verbs; others present what I take to be the same generalization in more semantically oriented terms. For example, Munshi (2006: 130–​132) says that verbs with double agreement have subjects that are experiencers or undergoers of the action, whereas verbs with single agreement involve some kind of volition on part of the subject. See also Grune (1998:  13), Anderson and Eggert (2001:  241–​242), and Yoshioka (2012: 250–​252) for similar observations. The standard Chomskyan treatment of unaccusative verbs is that their theme arguments are generated in the same position as direct objects are, but then may move to the subject position to fulfill the need for a clause to have a subject (Burzio, 1986). It is significant, then, that both these grammatical functions are recorded in Burushaski by the two kinds of agreement. A possible syntactic structure is shown in (12b), compared with the unergative structure in (12a); note that v must agree strictly downward in this language to account for the contrast (see Baker 2008: 216). (12) a.

TP vP NP

v’

b. T 3ms

v he VP ABS (no V Agr) walk

TP vP ----

v’ VP

NP

V

T 3ms v 3ms

boy die ABS

This crosslinguistically rare property depends on at least two more basic properties of Burushaski. First, the fact that agreement is not related to case assignment in Burushaski is a precondition for having this double agreement, since if the two agreeing heads put incompatible case requirements on the NP that they agreed with, then they could not both agree with the same NP. However, double agreement on unaccusative verbs is even more restricted than this; it is not found in Bantu languages, for example. We may describe this by saying that many languages impose a requirement such that a single morphological form cannot manifest more than one agreement with the same NP at PF (see Carstens (2005), who follows Kinyalolo (1991)), whereas Burushaski clearly does not. However, the precise details of why most languages do not show this theoretically very useful form of double agreement are not crucial here. What is important is that object agreement in Burushaski gives us a good independent test for whether a given NP is internal to VP (inside the domain of v) or not: if it is, it should trigger object agreement, regardless of its case properties or whether it subsequently moves to the subject position. This is important below.

 

766   Mark C. Baker

31.3  Absolutive–​absolutive verbs I turn now to the first major puzzle for a structural account of ergativity in Burushaski: the fact that a small number of verbs take two absolutive arguments and no ergative argument, as in (3). Another example from Willson (1996: 44) is (13a); (13b) is from Munshi (n.d.: 41) (note that the second argument is a nominalized clause); more are found in Lorimer (1935: 220). (13) a. Jé ué tsum búT peesá a-​yá-​am.5 I.ABS DEM.PL from much money.x.ABS 1sO-​obtain-​PST.1sS ‘I obtained much money from them.’ b. Je [un d-​uko-​Sqalt-​a nu-​se] d-​a-​yal-​a bayam.6 I.ABS you.ABS D-​2sO-​arrive-​1sS PTCP-​say D-​1sO-​listen-​1sS be.PST.1sS ‘I heard that you arrived.’ Other verbs that behave like this include ‘find,’ and ‘perceive’ (Willson 1996: 44, 54 n. 81). Why are these verbs special in this way? In Baker (2014a, 2015), I studied an analogous phenomenon in Shipibo, an ergative language of Eastern Peru. I related the fact that some dyadic verbs in Shipibo have two absolutive arguments to the robust fact that all triadic verbs in Shipibo have two absolutive arguments: the theme and the goal. Indeed, Burushaski also has triadic verbs with two absolutive arguments, as shown in (14) (Willson 1996:  33, 61)—​although the way that three-​argument verbs are realized is more variable in Burushaski than in Shipibo, because of Burushaski’s lexical dative case (compare (9); also Munshi 2006: 197). (14) a. Hilés-​e dasín taswíir móo-​ltir-​imi. boy-​ERG girl.ABS picture.y.ABS 3fO-​show-​3mS.PST ‘The boy showed the girl the picture.’ b. BúT sis-​e hamíid sawáal-​iN d-​é-​Garus-​uman. many people-​ERG Hamid.ABS question.y-​PL.ABS D-​3mO-​ask-​3pS.PST ‘Many people asked Hamid questions.’

5 

Note that ‘money’ has gender x, not y, the gender that most often fails to trigger object agreement in Burushaski. Hence the absence of object agreement with ‘money’ in (13a) is not due to its lexical properties. 6  The prefix d-​here is a kind of preverb, which appears on a large number of verbs in Burushaski, as a semi-​idiosyncratic lexical property. It can appear on verbs of any argument structure (Grune 1998: 13), although with some roots it is has a detransitivizing effect (Willson 1996: 12, citing Berger). It appears on some ABS–​ABS verbs, but not all. I do not know why ‘arrive’ in this example is glossed as having a 1sS suffix rather than 2sS; this could be a typo in the source.

 

Ergative case in Burushaski    767 Even unremarkable-​seeming examples like (14) present a challenge to the simplest version of dependent case assignment, since the goal argument c-​commands the theme argument, as revealed by the fact that v agrees with the goal, not the theme (the structure is identical to (9b), except that the higher object is not marked dative). Why then doesn’t the goal argument receive ergative case, in accordance with (2a)? The hope is once the puzzle posed by (14) has been addressed, the same solution will work for (3) and (13) in Burushaski, as in Shipibo. My account is as follows. The theme does not trigger ergative on the goal in (14) because both NPs are contained in VP, and ergative case does not apply when VP domains are spelled out, but only when TP domains are. The theoretical infrastructure for this is Chomsky’s (2000, 2001) notion of derivation by phase. According to phase theory, v is a phase head that triggers the spell out of its VP complement, and C is a distinct phase head that triggers the spell out of its TP complement. Suppose further that dependent case is assigned to c-​command pairs (NPx, NPy) at spell out, just as linear order is fixed at spell out by such pairs in Chomsky’s (1995) version of Kayne’s (1994) Linear Correspondence Axiom. Now what kind of spell out domain do NPx and NPy need to be in in order for NPx to receive ergative case? The answer, apparently, is that they must be in a TP domain. In Baker (2014a), I observe that the kind of dependent case that is assigned to the higher of two NPs in a single domain can vary across the domains: it can be ergative in TPs, genitive in NPs, and structural dative in VPs. In addition, these distinctions can be neutralized, completely or in part,7 and some of these domains may simply lack a rule of dependent case assignment. What (14) shows us, then, is that VP domains in Burushaski (like Shipibo) do not have high dependent case assigned: they do not have their own distinctive case (the goal is not consistently dative), nor do they use the same case that is assigned in another domain (the goal is not ergative or genitive). We can also infer that when there is dative case on the goal argument, as in (9), that must be a lexical case, assigned idiosyncratically by some verbs but not others, given that the gross structures of (9) and (14) are the same, as evidenced by the similar patterns of verb agreement. Dative in Burushaski is different in this respect from structural dative in Sakha, as analyzed by Baker and Vinokurova (2010).8 In addition to these assumptions about domains, we also need to appeal to a kind of strict cycle condition. Burushaski is a uniformly ergative language, not a differential case marking language in which case marking depends on whether the object undergoes object shift out of VP or not (contrast Baker (2015: sec. 4.2.1) on Ika and 7 

Indeed, Burushaski itself partially neutralizes the distinction between ergative and genitive: the two cases are homophonous except on nouns with feminine gender, where ergative is -​e but genitive/​oblique is -​mo. See Baker (2015: sec. 4.4) for an analysis of partial syncretisms like this within a theory in which ergative is high dependent case assigned within TPs and genitive can be high dependent case assigned within DPs. 8  Similarly, lexical dative case is assigned to the internal argument of a dyadic verb in (5) but not in (1b), and to the experiencer subject in (24) but not in (13). Additional evidence that dative is not structural dependent case assigned to the higher NP in VP in Burushaski is that it is never assigned to the causee of a productive causative construction, as happens in Sakha (e.g. see Willson 1996: 59–​60).

 

768   Mark C. Baker Ostyak). Therefore, the object is always part of the representation of the clause when TP is spelled out, even though it was already spelled out with VP. In other words, v in Burushaski is what Baker (2015) calls a soft phase head: it triggers the spell out of its VP complement, including the fixing of word order and structural case properties, but the contents of VP are still carried forward into the representation of TP. When there are two arguments inside VP, as in (14), both are carried forward. We thus need to ask why the c-​command pair (girl, picture) in (14a) does not result in ‘girl’ receiving ergative when TP is spelled out, even though it does not receive ergative (or dative) when VP is spelled out. The reason for this, I claim, is because that pair was already considered for case at the spell out of VP, and only new c-​command pairs are considered at the spell out of TP—​only pairs that include the agent ‘boy,’ which was not present when VP was spelled out. This is reminiscent of the Strict Cycle Condition of Mascaro (1976) and Kiparsky (1982b), where phonological rules apply only in environments which are created for the first time on the relevant cycle. If we assume the same thing for how structural case is derived by phases, then the case properties of (14) are accounted for. The derivation goes as in (15). (15)

Structure spelled out a. [vP [vp girl [picture see ] v]

New c-​c pairs Case assigned (girl, picture) girl: -​-​, picture: -​-​

b. [CP [TP [vP boy [vp girl [picture see] v] T] C]

(boy, girl) (boy, picture)

boy: ERG girl: -​-​, picture: -​-​

c. After default case assignment: boy: ERG, girl: ABS, picture: ABS

This analysis also applies to dyadic clauses with two absolutive arguments, like those in (3) and (13), as long as we assume that their verbs are dyadic unaccusative verbs—​verbs with two internal arguments and no external argument. In Baker (2014a), I  claimed that this assumption was thematically plausible for Shipibo, because the relevant verbs do not have agentive subjects, but rather subjects that are experiencers or possessors, thematic roles that can be closely aligned with the goal role. For example, experiencer verbs often have dative subjects in languages like Icelandic, the same case that marks goal arguments in ditransitives. It is reasonable to assume that the experiencers are projected in the same syntactic position as goal arguments (cf. the UTAH of Baker 1988)—​in SpecVP rather than SpecvP. Indeed, ABS–​ABS verbs have nonagentive experiencer/​possessor subjects in Burushaski as well. Given this, the derivation is in (16). The pair (I, money) is considered for case when VP is spelled out, but there is no high dependent case for this domain in Burushaski. Both NPs are forwarded to TP, and TP is spelled out. Now ergative case is available in principle, but the only c-​command pair is (I, money) and that is not considered on this cycle, because it was already considered at the spell out of VP. Therefore, ergative case does not apply, and both arguments get default absolutive case.

 

Ergative case in Burushaski    769 (16) Structure spelled out a. [vP [vp I [money need ] v] b. [CP [tp [vP [vp I [money need] v] T ] C]

New c-​c pairs (I, money)

Case assigned I : -​-​, money : -​-​

none

I : -​-​, money : -​-​

c. After default case assignment : I : ABS, money : ABS Agreement in Burushaski provides strong independent confirmation for this analysis. Recall that monadic unaccusative verbs have special agreement behavior, in that their arguments trigger both prefixal (v) agreement and suffixal (T) agreement on the same verb. It so happens that ABS–​ABS verbs in Burushaski show the same kind of double agreement with their experiencer arguments as ordinary unaccusative verbs do with their theme arguments, whereas the second arguments of these verbs are not agreed with at all. Willson (1996: 43–​44) is explicit about this generalization; it is also visible in (13b) from Munshi (n.d.), and the double agreement is what leads Lorimer (1935: 220) to notice these verbs as a special class. ABS–​ABS verbs differ from ERG–​ABS verbs in this respect, the latter showing prefixal agreement with the absolutive argument and suffixal agreement with the ergative argument. Now this special agreement behavior is just what we should expect when we combine the analysis of agreement in section 31.2 with the analysis of ABS–​ABS verbs based on Baker (2014a). My account of why the higher argument lacks ergative case depends crucially on it already being present inside VP when VP is initially spelled out. If it is inside VP, then v should agree with it. Moreover, it blocks v from agreeing with the other argument (the theme) by the intervention condition on Agree, given that the experiencer is the closer target to v. T also agrees with its closest target, which is the experiencer again, resulting in double agreement with the experiencer. The structure of (3) is (17), with the agreements marked. TP

(17) vP --

v’ VP NP

V'

T 1sg v 1sg

I NP V ABS money need ABS In contrast, ordinary agentive transitive verbs have one argument inside VP where v agrees with it and the other argument in SpecvP where T agrees with it ((8)).

 

770   Mark C. Baker Burushaski’s agreement patterns thus give us independent evidence that verbs like ‘need’ and ‘hear’ appear in clauses with a different syntactic structure than verbs like ‘see’ and ‘hit.’ This difference in structure induces a difference in how dependent case assignment applies within a cyclic derivation. A descriptive theory that is content to stipulate that verbs can have idiosyncratic case frames or case patterns determined by lexical semantics without appealing to structure (cf. Butt, Chapter 33, this volume, and Butt and Deo, Chapter 22, this volume, on other South Asian languages) does not explain this correlation between special case marking and special agreement. This sort of independent evidence was not available in Shipibo, which has a very limited agreement system. Therefore, Burushaski makes a distinctive contribution to the theory of ergativity in this respect.9 It also is worth noting that there is no sign of Burushaski having dyadic verbs that take two ergative arguments and no absolutive argument in any of the sources. If lexical items could stipulate special case frames, it is not clear why there could not be such verbs: why should the frame [ERG ERG V] be banned but [ABS ABS V] allowed? But this gap follows from a dependent case view based on (2), where lexical items project certain syntactic structures, and those structures are the proximal causes of structural case marking. On this view, a single clause could only have two ergative arguments if it had two NPs such that NPx c-​commands NPy and NPy c-​commands NPx. But there can be no such configuration under current theories of clause structure, which bar mutual c-​command between two arguments of the same predicate. Not all ergative languages allow dyadic clauses with an ABS–​ABS case pattern, the way Shipibo and Burushaski do. For example, Kalaallisut, Chukchi, and Ingush (Nichols, 2011:  sec. 21.2) do not. In Baker (2015), I  tacitly assumed that this was because languages could lack the relevant structures; they simply might not have verbs with two VP-​internal arguments and no external argument. However, new data from Baker and Bobaljik (Chapter 5, this volume) (B&B) suggests that this assumption might not be general enough. For example, the Yup’ik example in (18a) has ergative case on the malefactee argument, even though that seems to be an internal argument, based on comparison with the similar applicative with an agent argument in (18b). (18)

a. Ing-​um maklagaq kic-​i-​lq-​aa. that.one-​erg bearded.seal.abs sink-​appl-​pst-​3sO.3sS ‘The bearded seal sank on that guy.’ b. Qimugte-​m ner-​i-​a angun akuta-​mek. dog-​erg eat-​appl-​3sO.3sS man.abs mixture-​abl ‘The dog ate some “akutaq” on the man.’

9  These agreement facts also tell against an analysis of examples like (13) in which they have the structure of ordinary transitive clauses but the verb assigns lexical absolutive case to its subject (a possibility raised by Miriam Butt).

 

Ergative case in Burushaski    771 Where then does this crosslinguistic variation come from? Similar questions may be posed by B&B’s example (10) from Chukchi. It would be very strange to say that derivations in some languages obey a strict cycle condition, whereas those in another language do not (and my hard phase/​soft phase distinction is no help here). Rather, the best thing to say might be that in Yup’ik ergative case applies when VP is spelled out as well as when TP is—​an option Baker (2015: sec. 4.2) uses for Ika and Ubykh. In other words, ergative case assignment happens already at the stage equivalent to (16a) in Yup’ik. The potential problem with this proposal is that it predicts that ergative case assignment should also happen at the (15a) stage in the derivation of a ditransitive construction in Yup’ik, with the result that the goal argument of the ditransitive verbs should also have ergative case. This is true for Ika and Ubykh, but not for Yup’ik or Chukchi. However, this is not an actual problem if Yup’ik happens to lack ditransitive constructions of the relevant kind—​if no transitive verbs have the subcategorization frame [_​_​NP NP], but only [_​_​NP PP]. (This seems to be true in Finnish, for example; see Kiparsky (2001: 341).) Then the problematic-​looking prediction turns out to be vacuous. And indeed that seems to be true for Yup’ik and Chukchi: they have no verbs that appear with an ERG–​ABS–​ABS case frame; rather, all triadic verbs appear with an ERG–​ABS–​OBL pattern, including some kind of oblique case (e.g. allative, instrumental) that can be analyzed as a PP (Nedjalkov 1976, Jonathan Bobaljik, p.c.). My tentative prediction, then, is that any language that has ergative case on the higher of two NPs in a dyadic unaccusative construction will also lack a ditransitive construction with two absolutive arguments. Niuean illustrates the other side of the predicted correlation: like Burushaski, it has ABS–​ABS case patterns with (derived) dyadic unaccusatives (see B&B’s example (13)), and like Burushaski it has ERG–​ABS–​ABS case patterns in ditransitive constructions (Massam, 2006: 33).10 Preliminary results are thus consistent with my prediction, although it should be investigated more fully in future research.

31.4  Ergative–​Dative Verbs The second major wrinkle to ergativity in Burushaski is the fact that some dyadic verbs that take a dative complement have an absolutive subject, whereas others have an ergative subject, as shown in (5). The ABS–​DAT pattern seen in (4) is the normal one, whereas the ERG–​DAT pattern is restricted to a handful of verbs. Another example

10  Another factor within my theory that influences whether a dyadic unaccusative verb heads an ERG–​ABS clause or an ABS–​ABS clause is which of the two arguments moves to SpecTP to fill the subject role. If for some reason the lower theme argument does this rather than the experiencer argument, then a new c-​command pair (theme, experiencer) is present at the spell out of TP, triggering ergative on the theme. This is why in Shipibo applicatives of unaccusatives have an ergative argument, but reciprocals of ditransitives do not (Baker 2014a, 2015). There is, however, no apparent reversal of the arguments in (18a) from Yup’ik, so ergative on the malefactee probably cannot be attributed to this.

 

772   Mark C. Baker from Willson (1996: 37) is (19a); (19b) is one from Munshi (n.d.: 37); see also Lorimer (1935: 78).11 (19) a. Ín-​e hilés-​e-​r she-​ERG boy-​OBL-​DAT ‘She scolds the boy.’

Garí-​c-​ubo. talk.badly-​NPST-​3fS.PRES

b. …daa hurmat-​e mal basiy-​ar barey  and Hurmat-​ERG field garden-​DAT look.after ‘…and Hurmat looks after the garden and field.’

bay. be.PRES.3mS

What then accounts for this seemingly lexical difference? This issue does not arise in Shipibo, but it does in Warlpiri, and Baker (2015) discusses it from the perspective of dependent case theory. In particular, I claim that the dative phrases that go with absolutive subjects are really PPs, whereas the dative phrases that go with ergative subjects are NPs with oblique (lexical) case.12 This distinction seems semantically plausible, in that the dative expression in (4) is more of a canonical directional argument than the datives in (19), which are on the borderline between themes and goals. It is also morphologically plausible, in that the overt expression of what I am calling dative case is somewhat ambiguous between what one would expect of a true PP and what one would expect of a case suffix. For example, Willson (1996) writes r as a separate postposition, glossed ‘to,’ which assigns oblique case to its NP complement, but he remarks (1996: 10) that r is not a phonologically independent word. In contrast, Munshi (n.d.) writes noun+r as one word and uses the gloss ‘DAT.’ The distinction between PP and NP is highly relevant because PPs do not trigger ergative case on their co-​arguments: the PP itself does not bear the right kind of nominal features, and its NP complement is rendered invisible by P being a (hard) phase head that causes the NP to be spelled out and thus removed from the representation. The inertness of PPs for dependent case can be seen for uncontroversial PPs in Burushaski in (20), where the subjects are absolutive. (20) a. Dasín há-​e le hurút-​umo. girl.ABS house-​OBL in sit-​PST.3fS ‘The girl sat in the house.’ b. Huk bel-​e yoondal hal del-​imi dog.x.ABS wall-​OBL above jump hit-​3xS ‘The dog jumped above the wall.’

(Willson 1996: 3)

(Munshi n.d.: (18))

Example (4) can be analyzed in the same way. 11  Lorimer mentions explicitly that ‘look at’ in Burushaski is special in taking a dative complement, but his examples have null subjects, so it is not clear that the subject is ergative. 12  Miriam Butt points out that essentially the same distinction, made in less structural terms, exists in Lexical Functional Grammar as the difference between the grammatical functions OBJ(ect) and OBL(ique).

 

Ergative case in Burushaski    773 In contrast, NPs with oblique case are potential triggers of ergative case, as far as category features and structure are concerned. Marantz (1991) said that NPs with oblique case do not trigger dependent case on another nearby NP, thinking of Icelandic, but Baker (2015: sec. 5.1) claims that this is parameterized. Burushaski then would have the same parameter setting as Warlpiri rather than Icelandic in this respect. Indeed, the sense that it is the dative NP that triggers ergative case on the subject is strengthened by Willson’s (1996: 37) observation that the verb ‘curse’ in (19a) can be used without a dative complement, and then the subject is absolutive, not ergative.13 (21) In búT Gar’-​c-​ubo. she.ABS much talk.badly-​NPST-​3fS.PRES ‘She curses a lot.’ This strongly suggests that it is the dative NP that triggers ergative on the subject, rather than the verb stipulating in its lexical entry that its subject must be ergative, or ergative being triggered by some kind of covert theme argument. So (19) is support for a dependent case view, but it requires that some dative expressions count as NPs rather than PPs. The contrasting structures are in (22).

TP

TP

(22)

vP NP

v’

T 3ms

boy VP v (3fs) ERG NP V girl observe DAT

vP NP

v’

T 3ms

v farmer VP (*Agr) (ABS) PP V field to go DAT

Once again, we can look to Burushaski’s agreement system for independent support for the crucial assumption. We saw in section 31.2 that agreement in Burushaski is insensitive to case features. Therefore, it is possible that functional heads in this 13 

Note that we must say that the internal argument that is present in (19a) is completely absent in (21), not present as an empty category, since the indefinite null objects of verbs like ‘eat’ still trigger ergative case on the subject, judging by (i) from Munshi (n.d.: (59b)) (see Baker (2015: sec. 5.3) on this as a point of crosslinguistic variation in ergative languages). This seems reasonable semantically, since the gloss of (21) is not equivalent to ‘She scolded someone.’ (i) yuul i-​xoliʂ xaa hurmat-​e 3-​stomach 3-​pain until Hurmat-​ERG ‘Hurmat ate until it hurt his stomach.’

ʂi-​mi. eat-​3mS

 

774   Mark C. Baker language could agree with an NP with oblique dative case. In contrast, we expect agreement to be ruled out with PPs in Burushaski, as in most other languages: the PP itself does not have the relevant phi-​features, and it is a phase head that renders its NP complement invisible to the outside world. Indeed, v does not agree with (the objects of) uncontroversial PPs like those in (20) (nor does T). We predict, then, that dative expressions should vary in their agreement properties in Burushaski: some should participate in agreement and some should not, depending on their underlying category. This is true. Examples (7) and (9) already gave examples in which v agrees with the dative argument of a three-​argument verb. But Willson (1996: 17, 36) shows that there are other ‘give’-​type verbs where v does not agree with the dative, but must agree with the theme. Example (23) gives two examples (see also Lorimer (1935:  77–​78) and Anderson and Eggert (2001: 239)).14 (23) a. Hilés-​e dasín-​mo-​r toofá-​muts píish o-​t-​imi. boy-​ERG girl-​OBL-​DAT gift.x-​PL.ABS present 3pO-​do-​3mS.PST ‘The boy presented gifts to the girl.’ b. Jé-​e gó-​e-​r cái tayáar é-​t-​am.15 I-​ERG you-​OBL-​DAT tea.y.ABS prepare 3yO-​do-​1sS.PST ‘I prepared tea for you.’ Agreement on T can also reveal the NP/​PP distinction. ‘Find’ in (24) is a “dative subject” verb in which T agrees with the dative NP; ‘memorize’ in (25) is a comparable two-​ argument verb in which T agrees not with the dative argument, but with the absolutive NP, which acts like the subject in (at least) this sense. (24) ó-​ltalik dish-​míN ulo ín-​e-​r sawáabkuSh d-​ée-​Gurk-​c-​ai. 3p-​both place.y-​PL.ABS in he-​DAT reward.y.ABS D-​3mO-​find-​NPST-​3mS ‘In both places he will find reward.’ (Willson 1996: 42) 14 

Another reason why some dative expressions might trigger agreement in Burushaski and others not could be that some are generated lower than the theme argument, whereas others are generated above the theme. The higher sort of dative would be the closest target for v to agree with, as in (9), but the lower sort would not be, since the theme intervenes between it and v. This may well also be a factor in some examples—​ (24) versus (25), for example. However, this possibility would not render the NP/​PP distinction superfluous. For example, sentences like (4) have a dative expression but no theme argument at all, so it cannot be the position of the dative relative to the theme that blocks agreement, but rather the presence of PP structure. 15  The examples in (23) (and (25)) are instances of a light verb construction (LVC), where a complex predicate is formed from a combination of noun and semantically bleached verb, as is common in South Asian languages and beyond. Sometimes nouns in such constructions count as direct objects, and sometimes they do not but are simply adjoined to the verb and inert for morphosyntax. Examples (23a) and (23b) look like instances of the later type, because ‘present’ and ‘prepare’ do not trigger object agreement on ‘do,’ nor do they prevent ‘gift’ and ‘tea’ from doing so. If so, the internal structure of the predicate can be ignored for current purposes. However, a general study of LVCs in Burushaski needs to be done, and it is conceivable that some clause types of interest either require or forbid the predicate to have an LVC structure, for reasons that would need to be discovered.

 

Ergative case in Burushaski    775 (25) Tók gitáap jé-​e-​r záp b-​ilúm. entire book.y.ABS I-​OBL-​DAT memorize be-​3yS.PST ‘I memorized the entire book.’ (Willson 1996: 40; see also Lorimer 1935: 78, Munshi 2006: 133–​134, Yoshioka 2012: 58) Note that in (24), v also agrees with the argument that T agrees with, as expected given that both are VP-​internal arguments. (Similar double agreement might be expected in (25), but here the agreeing verb is the copula ‘be,’ and this verb idiosyncratically fails to show object agreement in Burushaski.) We see, then, that dative expressions in Burushaski do vary in their agreement behavior, just as we expect if some of them are NPs and others are PPs. This confirms an important assumption of the dependent case view. Again, there is no reason to say that verbs directly stipulate case frames; rather they project certain structures in accordance with their lexical semantic properties, and those structures determine how dependent case applies. Other syntactic processes are sensitive to the same structural differences, including agreement. This argument is, however, somewhat indirect, because the predicates used to investigate the agreement properties of dative expressions are different from the ones used to see their dependent case properties. It would be nice to close the gap by showing that the very verbs that take an ABS–​DAT case pattern block object agreement with the dative, whereas the ones that take an ERG–​DAT case pattern permit object agreement with the dative. The first side of this prediction is true: there is no prefixal agreement with the goal in (4), even though the agent subject triggers only subject agreement, not object agreement (‘go’ counts as unergative in Burushaski). But the second side of this prediction is unfortunately not true: Willson (1996: 38) states that barénas ‘look.at’ also does not show object agreement with its dative object (see (5), (19b)). Here we need to recall that object agreement is not entirely regular and uniform in Burushaski, as mentioned in section 31.2; even some ordinary transitive verbs fail to show agreement with their objects for no obvious syntactic reason. I must say, then, that barénas happens to be one of these morphologically defective verbs that does not manifest object agreement, even though it would be permitted syntactically (and so is Garis ‘scold,’ apparently). One might hope that some other verb in this class would show the predicted object agreement, but that may not be true. Since only a small number of dyadic verbs take dative objects in the first place (Willson mentions only two), it is possible that all of them have this same morphological defect—​especially if historical factors play a role.16 Despite this failure to connect all the dots, I consider it significant that, looking at Burushaski as

16  For example, it is very possible that dative–​object verbs evolved historically from dative-​PP verbs, the PP being reanalyzed as an NP. Then the older version of the verb would not have had object agreement for grammatical reasons, and acquirers of the new version could have interpreted it as being one of the defective verbs, lacking object agreement for morphological reasons.

 

776   Mark C. Baker a whole, dative expressions vary in whether they behave like PPs or like NPs for both agreement and dependent case assignment.17

31.5  Absolutive Subjects in Future Tense There is one further wrinkle to straightforward ergativity in Burushaski that is worth discussing. This is the fact that clauses in the future tense sometimes have absolutive subjects even when the verb is transitive. Willson (1996: 17) gives the following contrasting pair: (26) a. Hilés-​e dasin mu-​yeéts-​imi. boy-​ERG girl.ABS 3fO-​see-​PST.3mS ‘The boy saw the girl.’ b. Hilés dasín mu-​yeéts-​c-​i(mi). boy.ABS girl.ABS 3fO-​see-​NPST-​3mS.FUT ‘The boy will see the girl.’ There has been some confusion in the descriptive literature about the scope of this phenomenon. Lorimer (1935: 64–​67) claimed that past/​perfective clauses had ergative subjects and nonpast/​imperfective clauses did not in Burushaski, as in Hindi and some other South Asian languages (see in this volume Coon and Preminger, Chapter 10, Nash, Chapter 8, and Woolford, Chapter 9 for various views on this relatively common kind of split ergativity). However, Tiffou and Morin (1982) argue that Lorimer was misled in this because of difficulties in transcribing stress and vowel length. They claim that the phenomenon is much more narrow: only first and second person pronouns in future tense (and perhaps imperatives) are absolutive where one would expect ergative on the basis of transitivity; see also Munshi (2006: 143–​144) and Yoshioka (2012: 247). Willson (1996) agrees with Tiffou and Morin that the transitive subject is absolutive only in the future, but he does not limit this to local pronouns; (26b) has a third person subject in absolutive case. So, there seems to be some variation regarding this phenomenon, which I cannot get to the bottom of here. Nevertheless, there is enough data to make it worth asking what is different about future clauses, such that ergative case is used less reliably in them. This is another instance of one particular lexical item (a T?) apparently putting an idiosyntactic condition on the case of a nearby NP. Can it also be seen as being at bottom a difference in syntactic structure, as a pure dependent case theory suggests? 17  Similar issues may arise for some source/​malefactive arguments. Willson mentions that source/​ malefactive nominals marked with oblique case trigger an ergative subject with two verbs (‘bite’ and ‘arrest’; 1996: 39), and also that source/​malefactive arguments sometimes trigger object agreement on the verb (e.g. with ‘snatch,’ 1996: 4). I thus tentatively say that the source/​malefactive arguments selected by certain verbs can also count as oblique NPs rather than PPs for case and agreement.

 

Ergative case in Burushaski    777 In Baker (2015), I discussed two instances of tense–​aspect governed split ergativity of this general sort, in Coast Tsimshian and Semelai. Both languages provide good evidence that something structural at work, since clauses with an ABS–​ABS case pattern differ from clauses with an ERG–​ACC case pattern not only in their tense–​ aspect, but also in the positions of the NPs inside the clause. Therefore, I argued that certain tense–​aspect heads can be specified as being phase heads, in addition to the universal phase heads C and v. As such, they induce an additional phase boundary in the clause (see also Coon and Preminger (Chapter 10, this volume) for a closely related idea). If the subject is very high (as subjects are in SVO order in Semelai) and T or Asp is a phase head, then the subject is in a different domain from the object at spell out. The result is that the subject does not see the object for purposes of (2), and ergative case marking does not happen. In contrast, if the subject is lower (as it is in VSO order in Semelai) and/​or if T or Asp is not a phase head, then the subject and the object are in the same spell out domain, and ergative case assignment happens as usual. How might this apply to Burushaski? Since Burushaski is a verb-​final language, we cannot expect to tell how high the subject is from simple facts about surface word order; we expect SOV order regardless of whether the subject is relatively high or relatively low (cf. Baker 2001: 136–​137). Nevertheless, on the dependent case view, we might expect to observe some structural way in which future tense is different from other tense–​aspect combinations that underlies the difference in case. Indeed, inspection of Burushaski’s verb paradigms reveals that there is something structurally special about the future tense:  it is the only tense–​aspect that is based on the present form of the stem and does not use the auxiliary ‘be.’ The future tense is thus special in having a sequence of tense–​aspect markers affixed to the same stem: V+c+m+AGR. Willson (1996) does not give a full paradigm, but other descriptions agree on this point. Example (27) is based on Tiffou and Pesot (1989); see also Grune (1998: 10–​11), Yoshioka (2012: 248), and Munshi (n.d.: 84). (27) a. Future: b. Present:

root+c+m+Agr Root+c+a/​u(m) be+Agr

(Present, Present progressive)

c. Imperfect: Root+c+a/​u(m) be+m+Agr (-​-​, past progressive) d. Aorist 1:

Root+Agr

(Past, -​-​)

e. Aorist 2:

Root+m+Agr

(Past II, simple past)

f.

Root+a/​u(m) be+Agr

(-​-​, present perfect)

Perfect:

g. Pluperfect Root+a/​u(m) be+m+Agr

(-​-​, past perfect)

h. Past

Root+aasc+m+Agr

(past III, -​-​)

i.

Past

Root Root be+m+Agr

(-​-​, -​-​)

j.

-​-​

Root-​c-​ume be+Agr

(-​-​, present perfect continuous)

 

778   Mark C. Baker Munshi’s (n.d.) forms are a bit different from Tiffou and Pesot’s, in that they lack the final /​m/​on the past and present participles and (therefore?) the participle contracts with the auxiliary ‘be’ in some parts of the paradigm (see also Yoshioka 2012). However, there is no substantive disagreement about the overall structure of the tense–​aspect system. It is notable that the future is the only nonperiphrastic tense built from the present/​continuous stem, the only one in which the main verb bears both the present/​continuous affix -​c and the agreement morpheme. In contrast, past tense verbs are simplex and bear agreement but have no analog of -​c, and progressive forms (both present and past) contain the morpheme -​c but are periphrastic, with primary agreement showing up on the auxiliary rather than the main verb. Therefore, there is room to say that having both -​c and agreement on the same stem creates a domain boundary that hides the object from the subject. This situation is interestingly different from the one I discussed for Coast Tsimshian and Semelai because there is no one morpheme that can be blamed for the lack of ergative marking; neither -​c by itself (in the present) nor Agr by itself (in the past) disrupts ergativity, but only the combination of the two. However, the effect can be captured rather nicely if we adopt the Phase Impenetrability Condition of Chomsky (2001: 14) rather than that of Chomsky (2000). This can be stated as in (28). (28)

If H is a phase head, and Z is the next higher phase head, then no element higher than Z can access anything inside the complement of H.

This version holds that material in the complement of one phase head remains visible until the next phase head is reached, and only then is it removed from the representation. Suppose, then, that -​c is an Aspect head (for concreteness; the exact label does not matter), the principal agreement-​bearing head is T, and both count as phase heads. Suppose further that there is an intermediate position (SpecPartP) for the subject to raise to that is above AspP but below the auxiliary in a complex tense,18 but there is no intermediate landing site for the subject in a simplex tense; in those clauses the subject raises straight from its thematic position inside vP to SpecTP. Then we have representations like the following, assuming that morpheme order is a good guide to syntactic structure here (Baker 1985). (I ignore v as a phase head, assuming that it is a soft phase head in the sense of Baker (2015).) In the simple past structure in (29a), T is the first relevant phase head, but its complement remains visible until the next phase head (C) is reached, so the subject in SpecTP sees the object in VP and is marked ergative. In contrast, in the future structure in (29c), Aspect is a phase head and Tense is a second phase head. Therefore, the object inside the 18  Some corroborating evidence for this assumption is that there is a modicum of person agreement on participles in Burushaski: the participle uses /​a/​in (27b,c,f,g) if the subject is first person, but /u/​ otherwise. According to the principles in Baker (2008), this implies that the subject merges with a projection of the participle head. In contrast, no lower person agreement is seen in the simplex tenses, only the outermost agreement associated with T.

 

Ergative case in Burushaski    779

a. Simple past

(29)

b. Periphrastic present or past progressive

TP

TP

NP

T'

boy vP ERG t

(NP) T

v'

(boy)

-m [ph]

VP

v

NP

V

girl

see

T' AuxP PartP

NP boy

AspP

future tense TP NP

T'

boy AspP T (ABS) vP Asp -m[ph] t

v' VP

-c [ph] v

NP

V

girl

see

Part

Asp v'

VP

c.

Aux (-m)[ph] Part' be

vP t

T

-c [ph] v

NP

V

girl

see

-u

 

780   Mark C. Baker vP complement of Aspect is not visible to the subject, which is higher than the second phase head T. Therefore, the object does not trigger ergative on the subject in (29b). This explains Willson’s contrast between (26a) and (26b).19 Finally, Aspect is a phase head in (29c) also, but here the subject lands in SpecPartP, below the next phase head T. The complement of Aspect is thus still visible to the subject in SpecPartP, and the subject is marked ergative (perhaps later raising to SpecTP). In this way, we can use a known grammatical principle to relate the fact that the future can allow absolutive subjects to its unique morphosyntactic structure within the Burushaski paradigm.20 I do not want to oversell this analysis. It is fragile in that it depends both on the idiosyncratic phase status of certain heads (T, Asp) and on a crucial assumption about clause structure (subjects can land in SpecPartP but not in, say, SpecAspP). The empirical facts are also fragile, in that other sources report more ergative subjects in the future tense than Willson does, and in that the distinction between periphrastic tenses and simplex tenses may be breaking down as auxiliaries undergo contraction with participles into a single word (Munshi n.d.: (85)). Nevertheless, I have shown that it is possible in principle to explain why a complex tense–​aspect combination might fail to have ergative subjects whereas simpler tense–​aspect combinations do have ergative subjects. Whether this tendency is found in other ergative languages remains to be seen.

31.6 Conclusion In this chapter, I have considered three apparent deviations from canonical ergativity in Burushaski: clauses with experiencer verbs that have two absolutive arguments, clauses that have an ergative subject and a dative complement, and future tense clauses that have an absolutive subject as well as an absolutive object. In each situation, I found evidence from the morphological characteristics of the verb—​either its agreement properties or its sequence of tense–​aspect morphemes—​that the structure of the clause is different from the structure of ordinary transitive sentences in a simple tense. This confirms that we can have a straightforward, structure-​based theory of ergative case assignment in which ergative is assigned to the higher of two NPs in the same local domain (dependent

19  Willson (1996: 17 n. 32) gives (i) as an exceptional example in which the subject of a future tense clause has ergative case, unlike (26b). I could potentially analyze this as an example in which the object, a likely bearer of focus, scrambles out of VP and thus escapes being spelled out with the complement of Asp, with the result that it does trigger ergative on the subject in SpecTP. Possible evidence in favor of the object being scrambled in (i) (but not (26b)) is that it appears before the adverb ‘also.’

(i)

20 

Ún-​e maasuum qaum ké you-​ERG innocent nation also ‘Will you also kill the innocent nation?’

o-​sqan-​c-​uma. 3pS-​kill-​NPST-​2sS.FUT

However, Baker (2015) uses Chomsky’s (2000) version of the PIC, so a unified version needs to be worked out—​a task that goes beyond what I can attempt here.

 

Ergative case in Burushaski    781 case). There is no need to say that lexical items can stipulate directly which NPs are ergative and which are absolutive. This replicates the results of Baker (2014a, 2015) for Shipibo and other ergative languages.

Acknowledgments For helpful comments on this chapter, I thank Jonathan Bobaljik, Miriam Butt, and Jessica Coon. For contributions to the larger project of which this is a part, see the acknowledgments in Baker (2015).

Abbreviations I have harmonized the glosses of examples taken from different sources. Agreement morphemes are glossed with a triple consisting of a roman numeral indicating person (1, 2, 3), a lowercase letter indicating gender or number (m, f, x, y, p), and an uppercase letter indicating the grammatical function (S, O) of the agreed with NP. x and y are special genders in Burushaski, distinguishing roughly concrete count nouns (x) and abstract or mass nouns (y). I have made no attempt to harmonize phonological transcription practices. ABL, ablative; ABS, absolutive; APPL, applicative; D, idiosyncratic verbal prefix (see n.  6); DAT, dative; DEM, demonstrative; FUT, future; INF, infinitive; NEG, negative; NPST, nonpast; OBL, oblique; PL, plural; PRES, present; PST, past; PTCP, participle.

 

Chapter 32

E rgativit y i n Bas qu e Ane Berro and Ricardo Etxepare

32.1 Introduction Basque is a language isolate spoken in the Basque Country by about 700,000 people at both sides of the Pyrenees across the Spanish–​French border. Most of the Basque speakers are concentrated on the Spanish side, in the Autonomous Community of the Basque Country, where the language has full official status, and in the region of Navarre, where it has a more limited status. Basque is spoken across the French border by about 55,000 people in the historical regions of Labourd, Low Navarre and Pays de Soule, which together comprise the western half of the Département des Pyrénées Atlantiques. The language has no official status in France, where Basque is rapidly declining. Virtually all Basque speakers (leaving aside a small percentage among the very youngest) are bilingual, with either Spanish or French as their other language. Despite the strong pressure exerted by its surrounding romance languages, the Basque language maintains its traditional typological profile (Trask 1998): it is overwhelmingly head-​final, strongly agglutinative, discourse configurational (in the sense of Kiss 1995), and ergative in its case and agreement morphology. Ergative case-​marking is at the same time a prominent feature of Basque and a foreign one in the European linguistic domain, a fact that has contributed to the attention paid to the language during the nineteenth and early twentieth centuries. Ergativity is also a foreign notion to the classical grammatical tradition based on Latin and Greek morphosyntactic categories. Early Basque grammarians had noted that the case-​alignment of Basque was different from both Latin and Romance languages, but had difficulties in moving away from the case categories inherited from that tradition. Some early studies of ergativity in Basque nevertheless were able to approach the ergative phenomenon with a considerable amount of ingenuity. Thus, the souletin jurist Oihenart provides in his Notitia (1638) a very modern characterization of Basque ergativity (see Oyharçabal 1993), that we will use to present the basic case-​marking pattern of the language here. He

 

Ergativity in Basque   783 distinguished an ‘active’ case and a ‘direct’ case. The ‘active’ case (realized as the suffix-​k) marks the subject of transitive clauses: (1) Gizona-​k egi-​ten man.det-​erg do-​ipfv ‘The man does it.’

d-​u. TNS-​root

Although Oihenart did not fully develop the concept of ‘direct case,’ this category was explicitly proposed in the absence of accusative case in the language. The direct case, which is morphologically unmarked, covers all those cases not covered by the ergative and by the other argument-​related cases dative and genitive. The ‘direct case’ thus covers the object of transitives and the subject of intransitive predicates: (2)

a. Gizona dator. man.det comes ‘The man comes.’ b. Gizona-​k hori egi-​ten d-​u. man.det-​erg that.abs do-​ipfv tns-​root ‘The man does that.’

Translating Oihenart’s schema into modern terms, Case-​marking in Basque presents an ergative alignment, whereby subjects of intransitives and objects of transitive predicates constitute a class for case-​marking and agreement, in contrast to subjects of transitive clauses. At the syntactic level, there is widespread agreement that ergative arguments are hierarchically higher than absolutive arguments in Basque. Absolutive arguments have no special syntactic status as opposed to ergative ones (besides case and agreement). Restrictions affecting controlled arguments, implicit arguments in coordination, reciprocal anaphors, unmarked word order and scope relations, as well as A-​bar extraction, either do not make a difference among the arguments of the verb, or they classify together subjects (both ergative and absolutive) in opposition to objects (e.g. in control, a fact noted by Anderson 1976 for ergative languages generally). We will not linger on the evidence gathered in this sense during the 1980s and 1990s. We refer the reader to Levin (1983), Ortiz de Urbina (1989), and Oyharçabal (1992) for a general discussion of (the lack of) syntactic ergativity in Basque. The chapter is organized as follows: the next two sections concentrate on the paradigmatic properties of ergative case and agreement. We will review the morphological configuration of nominal and verbal inflection in detail, by pointing at those phenomena, like patterns of syncretism and changes in case-​agreement correlations, which are relevant to ergativity. The fourth section examines the distribution of ergative case-​marking in unergative predicates across dialects, and the lexico-​semantic concepts which have been claimed to correlate with the observed dialectal variation. The fifth section critically examines the idea that ergative is a lexical or inherent case in Basque, related to

 

784    Ane Berro and Ricardo Etxepare specific aspects of the meaning of the predicate, by extending the discussion to predicates headed by adjectives and nouns, and by analyzing the distribution of ergative case in contexts of raising. Section 32.6 introduces the notion of ‘marked case’ in relation to ergative case-​marking in Basque, as developed in recent work by Rezac et al. (2014). A cautionary note on our use of terms making reference to dialectal domains: we assume the classification proposed by Zuazo (1998), which distinguishes six basic dialectal areas, called Western (Biscayan), Central (Guipuscoan), High Navarrese, Eastern Navarrese, Navarro-​Labourdin, and Souletin. In this chapter, we will focus on Western, Central, Navarro-​Labourdin, and Souletin varieties, which still possess a relatively healthy profile in sociolinguistic terms. We will furthermore lump together the two main dialects spoken in the French side of the border (Navarro-​Labourdin and Souletin) into a single category ‘Eastern.’ This idealization is justified by the fact that the two dialects behave in identical or very similar fashion with regard to ergative marking. This way we keep reference to dialectal variation in manageable levels. For a monograph specifically devoted to ergativity in Souletin, the reader may want to check Coyos (1999).

32.2  The Nominal Paradigm and Ergative–​Absolutive Syncretisms The Basque noun phrase presents a maximal structure that can be linearly represented as follows (adapted from Santazilia 2013): (3) a. Qcardinal

Noun Adjective

Qnon-​cardinal

Det Number Case

b. Etxe   handi guzi-​e-​n gaitza. house big all-​det.pl-​gen problem.det ‘The problem of all the big houses.’ c. Bi    zamaketari handi-​e-​z two porter big-​det.pl-​ins ‘Helped by the two big porters.’

lagunduta. helped

The case endings -​n (genitive) and -​z (instrumental) in (3b–​c), attach to the last element in the template, be it a noun, an adjective or a quantifier. Definiteness and number are lexicalized together by means of a portmanteau form (-​e-​) in overtly case-​marked plural definite DPs. In Basque, case-​inflection is phrasal in scope, but it is morphologically expressed by means of bound forms attached to the last word in the noun phrase (Hualde 2003: 171). The two other morphologically overt cases in Basque besides the genitive and the instrumental are the ergative -​k (1) and the dative -​(r)i (which attaches to indirect objects and experiencer subjects of some psychological verbs).

 

Ergativity in Basque   785 Basque nominal inflection presents some cases of ergative/​absolutive syncretism. Those cases comprise plural DPs with proximate determiners,1 plural demonstratives and 2nd person plural pronouns. Taking into account that the Basque determiner system has its origin in the demonstrative system (see Santazilia 2013; Manterola 2015), and that the latter is also used in place of the inexistent third person pronoun paradigm, we can conclude that the case-​syncretism configurations obey the general rule of being confined to the pronominal system (Baerman, Brown, and Corbett 2005: 40–​52). The standard paradigm for DPs, demonstratives, and pronouns in the absolutive and ergative cases is given in Tables 32.1–​32.3 (see Hualde 2003: 171–​186, for a more detailed presentation). Table 32.1 Determiner paradigm for the noun leku ‘place’ Definite

Indefinite

Singular  Plural    Proximate Absolutive

leku-​a     leku-​ak  leku-​ok

Ergative

leku-​a-​k  leku-​ek 

leku

leku-​ok

leku-​k

Table 32.2 Demonstrative paradigm for hau ‘this’ Singular

Plural

Absolutive

hau

hau-​ek

Ergative

hon-​ek

hau-​ek

Locative

hon-​etan

hau-​e-​tan

Table 32.3 Personal pronouns 1p.sing

2p.sing

2p.sing.coll

1p.pl

2p.pl

Erg

ni-​k

zu-​k

hi-​k

gu-​k

zu-​ek

Abs

ni

zu

hi

gu

zu-​ek

As shown in each of the tables, the ergative–​absolutive morphological case distinction tends to disappear in the plural, with the ergative and the absolutive sharing the unmarked form (absolutive). This is the case for the proximate plural determiners in Table  31.1, the absolutive and the ergative plural endings in the demonstrative paradigm (Table 32.2), and the second person plural pronoun in the pronominal paradigm 1 

Proximate determiners only occur in the plural productively in Basque.

 

786    Ane Berro and Ricardo Etxepare (Table  32.3).2 When Basque dialects present ergative–​absolutive syncretisms in the determiner phrase, those occur also in the plural. An illustrative example is that of central varieties spoken in Gipuzkoa, which extend the form of the absolutive plural to the ergative plural in both proximate and non-​proximate determiners (Table 32.4). Table 32.4 DPs, for the vowel ending noun leku ‘place’ in Gipuzkoan varieties Definite

Indefinite

Singular Plural  Proximate Absolutive

leku-​a   leku-​ak   leku-​ok

leku

Ergative

leku-​a-​k   leku-​ak   leku-​ok

leku-​k

Syncretisms between core cases are common in both ergative–​ absolutive languages (Central Yupik, Jacobsen 1995, in the work of Baerman et al. 2005; or Domaaki, Baerman et al. 2005: 48) and nominative–​accusative languages. Nominative/​accusative syncretisms, for instance, are quite general in Indo-​European in the dual and the plural (Baerman et al. 2005: 48). Most of the work on the ergative/​absolutive syncretism in Basque has been concerned with the diachronic process that may have led to such a state of affairs, a problem that is closely related to the emergence of the plural morpheme itself (Michelena 1987), and particular accounts differ depending on the source proposed for the underlying plural marker. For some (de Rijk 1981; Michelena 1987), the plural marker is at its origin a derivational suffix expressing abundance, and the addition of the ergative morpheme triggers a number of phonotactic processes resulting in the two basic dialectal forms for the ergative plural. For others (Trask 1997; Manterola 2015), the syncretism is motivated by dialect-​specific phonological processes targeting different underlying demonstrative forms for the singular and the plural. From the point of view of its wider typological scope, which has well-​attested correlates in nominative–​accusative languages, the solutions found within the exclusive realm of Basque morphophonological processes may be a bit too limited. Demonstrative paradigms in Basque are furthermore characterized by something we can call stem syncretism in the ergative and absolutive plural: thus, as shown in Table 32.2, whereas the singular ergative demonstrative, like the inessive one, presents the locative stem hon-​, the plural ergative is based on the absolutive singular demonstrative hau ‘this,’ to which a plural ending is attached. As far as we can tell, this particular stem alternation has been ignored by most of the accounts devoted to 2 

The plural attached to the second person pronoun zu ‘you’ is a recent innovation. Zu was the plural 2nd person pronoun in Basque, contrasting with singular hi. Subsequently zu was recruited as a plural of respect (cf. Spanish vos), and the singular hi was confined to the colloquial register. In order to recreate a non-​colloquial second person plural pronoun, the singular of respect zu was then combined with a plural suffix -​ek.

 

Ergativity in Basque   787 the case-​syncretism issue. One exception is Etxepare (2013), which tries to relate both problems. Etxepare follows Lakarra’s hypothesis (2005) that the ergative marker -​k is historically related to the locative suffix -​ga, used presently with animate NPs when a local case is added: (4) Lagun-​a-​ga-​n. Friend-​det-​suff-​iness ‘In the friend.’ Locative phrases typically involve the presence of a (sometimes silent) spatial bare noun (Aurnague 2001; Svenonius 2006) in Basque. Spatial bare nouns, a rich closed class set in the Basque locative paradigm (see Aurnague 2001), have well-​known restrictions regarding modifiability and counting. Among other things, such bare nouns are incompatible with number. Etxepare (2013) suggests that the Basque syncretism could be accounted for if ergative case (originally a locative) implied the presence of an underlying spatial noun, and this was incompatible with the presence of plural number. Singular number, unlike plural number, is not morphologically marked in Basque. This approach must nevertheless address the issue that western dialects (Biscayan) show stem syncretism in favor of the locative form in the plural demonstrative paradigm (see Jacobsen 1972, for a detailed discussion of ergative–​absolutive syncretisms in Basque).

32.3  Agreement and Ergative Displacement In this section, we will focus on the general properties of the inflected verbal forms and their morphological make-​up, including agreement morphemes and their relation to the case-​marked arguments of the clause. The shape and distribution of agreement morphemes in Basque correspond closely to case-​alignment, and the overall pattern of verbal inflection is thus an ergative one. There is a well-​known exception to this pattern, known as Ergative Displacement (after Laka 1988; see also Ortiz de Urbina 1989; Laka 1993a), that we examine in section 32.3.2.

32.3.1 The Basque Agreement System Inflected verbal forms in Basque can be of two types: synthetic or analytic (called periphrastic in the Basque linguistic tradition). Synthetic verbs incorporate all inflection along with the lexical verbal root, while in analytic verbs, the lexical verb is only marked for aspect, and agreement, tense and modality morphemes are conveyed in a separate auxiliary. Analytic forms are the only productive verbal pattern in Basque these days, so we will concentrate on such forms. There are five common auxiliaries in standard Basque

 

788    Ane Berro and Ricardo Etxepare (Euskaltzaindia 1997):  intransitives izan, *edin, *edun, *ezan, and ditransitive -​i-​. Most of the variation found in inflected forms across dialects has to do with the root effectively used in ditransitive constructions, which shows considerable variation (Laka 1988; see also Fernández 2013). Izan ‘be’ and *edun ‘have’ are used in the indicative and their distribution resembles the be/​have auxiliary alternation found in some nominative languages. On the other hand, *edin and *ezan are used in the subjunctive, potential and imperative. In Basque, the agreement complex may include up to four instances of person agreement: ergative, absolutive, dative, and allocutive.3 Grammatical encoding of the person and number features of the arguments is obligatory in Basque, apart from some dialectal exceptions involving dative agreement (see e.g. Fernández and Ortiz de Urbina 2010; Etxepare and Oyharçabal 2013; Etxepare 2014). The allocutive morpheme does not correspond to an argument, but to the addressee of the speech situation, and it is optional. Example (5a–​b) represent a ditransitive and a ditransitive-​plus allocutive verbal form, respectively: (5) a. Ni-​k zu-​ri liburua-​k ekarr-​i d-​i-​z-​ki-​zu-​t. I-​erg you-​dat book.det-​pl.abs bring-​prf tns-​root-​pl-​df-​2dat-​1erg ‘I brought you the books.’ b. Ni-​k   Ane-​ri liburua ekarr-​i z-​i-​o-​a-​t. I-​erg Ane-​ dat book.det.abs bring-​prf tns-​root-​3dat-​allo-​3erg ‘I brought the book to Ane (and I am speaking to you).’ In Table 32.5 (from Albizu 2002: 3), we show the morphemes corresponding to each person in standard Basque. The category coll stands for the colloquial register, available for the second person and obligatory for the allocutive. Table 32.5 Person agreement markers in Standard Basque abs 1 SG 2 SG MASC/​FEM COLL

n-​ h-​

erg

dat

allo

-​t /​ -​da-​ -​k /​ -​n -​a-​ /​ -​na-​ o /​ -​a

3 SG 1 PL

g-​

gu

2 SG NON-​COLL

z-​

zu

2 PL

z-​

zu

3 PL

3 

See Austin (2012) for the relevant stages in the progressive acquisition of absolutive, ergative, and dative agreement.

 

Ergativity in Basque   789 As pointed out in Albizu (2002), ergative, dative, and allocutive agreement morphemes are homophonous in the colloquial register of the 2nd person singular. The colloquial forms also present a gender distinction. Ergative and dative morphemes are also homophonous in the 1st person and in the 2nd non-​colloquial singular and plural person. 3rd person agreement morphemes are inexistent, except for the dative (where -​o-​is plausibly related to the proximate determiner -​o, see Gómez and Sainz 1995). Singular number is not morphologically expressed. Person affixes in the plural are either identical to pronominal forms, as in the ergative and dative rows, or reduced forms of the pronominal set, as in the absolutive row. Person affixes in the singular are distinct from their pronominal correlates. The plurality of arguments is indicated in the agreement complex by means of a separate set of plural affixes, shown in Table 32.6 (adapted from Albizu 2002: 4). Table 32.6 Plural markers abs

erg

dat

1 PL

-​z-​, -​it-​, -​zki-​, -​tza-​, -​de-​, -​ra-​​

2 SG

-​z-​, -​it-​, -​zki-​, -​tza-​, -​de-​, -​ra-​

2 PL

-​z-​, -​it-​, -​zki-​, -​tza-​, -​de-​, -​ra-​ … -​te-​

-​e-​

-​e-​

3 PL

-​z-​, -​it-​, -​zki-​, -​tza-​, -​de-​, -​ra-​

-​te-​

-​e-​ /​ -​te-​

As can be seen in Table 32.6, the distribution and shape of the plural morphemes is another feature that distinguishes ergative and dative agreement from the absolutive. As noted in Laka (1993a), the presence of these pluralizing morphemes represents evidence for the separation of person and number features. Contrary to person morphemes, (i) pluralizing affixes have several shapes depending on the tense and the auxiliary root; (ii) their position varies—​ they can precede the root, as in the case of the pluralizing -​it-​(6), or follow it, as in the case of -​zki-​(7); and (iii) they may occur even if absolutive person marking does not (7). (6)

(Zu-​k) (gu) ekarr-​i g-​a-​it-​u-​zu. you-​erg we.abs bring-​prf 1plabs-​root-​2erg ‘You has brought us.’

(7)

(Hark) (ni-​ri) liburua-​k ekarr-​i d-​i-​zki-​t. (S)he.erg I-​dat book.det-​pl.abs bring.prf tns-​root-​pl-​1sgerg ‘(S)he has brought me the books.’

In the absence of 1st or 2nd person absolutive, the absolutive slot is occupied by morphemes expressing Tense or Mood (Trask 1977; Laka 1993a; Gómez and Sainz 1995; Rebuschi 1999; Albizu and Eguren 2000; Arregi and Nevins 2013), as d-​ for present Tense in (7).

 

790    Ane Berro and Ricardo Etxepare The Basque verbal complex also includes tense and mood suffixes. Past tense is overtly realized by the marker -​(e)n (8b). Furthermore, present or past tense is also conveyed indirectly by other markers, like the shape of certain morphemes around the root, called class markers by Albizu (2002). According to Albizu (2002), class markers are morphemes which vary depending mainly on tense and verbs, but can also be influenced by the person of the absolutive argument. Compare the class marker -​a-​of the present tense (8a) and -​ind-​of the past tense (8b). (8) a.  Ikus-​i n-​a-​u-​zu.   see-​prf 1sgabs-​cl-​root-​2erg   ‘you have seen me.’ b. Ikus-​i n-​ind-​u-​zu-​n. see-​prf 1sgabs-​cl-​root-​2erg-​pst ‘you saw me.’ The sequential order of the morphemes just described is the following (adapted from Laka 1993a): (9) Tense/​Mood Absolutive Class Root Dative Modal Ergative Tense (If ABS 3rd) 1/​2 person Marker As shown in (9), absolutive affixes precede the root, while dative and ergative affixes follow it. Assuming the Mirror Principle (Baker 1985), Laka (1988, 1993a) suggests that the sequence of agreement morphemes in the verb is a surface manifestation of structure-​ dependent hierarchical relations. This way, the morphological sequence abs-​root-​dat-​erg mirrors the hierarchical relations of the arguments in the clause: Subjerg—​IOdat—​DOabs—​V.

32.3.2 Ergative Displacement Agreement closely aligns with case-​marking in Basque. An exception to the general pattern is Ergative Displacement (Heath 1976; Ortiz de Urbina 1989; Laka 1993a) (henceforth ED), a phenomenon whereby certain ergative DPs are encoded in the auxiliary by means of an absolutive prefix. ED takes place when: (i) the absolutive argument is 3rd person; (ii) the ergative argument is 1st or 2nd person; and (iii) the clause is in the past tense or in the irrealis mood. As a consequence of ED, the ergative morpheme disappears from the verbal complex. Examples (10a–​b) illustrate the phenomenon in the past tense: (10) a. (Ni-​k) liburua-​k ekarr-​i n-​it-​u-​en. I-​erg book.det-​pl.abs bring-​prf 1sgAbs-​pl-​root-​pst ‘I brought the books.’

 

Ergativity in Basque   791 b. (Ni-​k) liburua ekarr-​i d-​u-​t. I-​erg book.det.abs bring-​prf tns-​root-​1sgerg ‘I have brought the book.’ In (10a), the ergative argument is 1st person, the absolutive is 3rd person and the clause is in the past tense. ED has taken place, and the ergative argument is cross-​referenced by the n-​prefix, a morpheme which canonically agrees with absolutive 1st person arguments (see Table 32.5). Compare this to the present form d-​u-​t [tns-​root-​1erg] in (10b), where the suffix -​t cross-​references the ergative argument. Note that the pluralizer -​it-​, agreeing with the plural absolutive argument, is still present in the verbal form in (10a), suggesting that the absolutive argument is visible, at least partially, for syntactic agreement (Laka 1993a). If the person conditions are not met, ED does not occur, for example, when the absolutive argument is 2nd person (11a) or the ergative argument is 3rd person (11b). In (11a) the ergative argument is cross-​referenced by the affix -​da-​. In (11b), since both the ergative and the absolutive arguments are 3rd person, there is no personal morphology, but the absence of ED is manifested by the plural affix -​te-​encoding ergative agreement (Ortiz de Urbina 1989; Fernández and Albizu 2000). (11) a. (Ni-​k) (zu) kotxe-​z I-​erg you.abs car-​ins ‘I brought you by car.’

ekarr-​i z-​intu-​da-​n. bring-​prf 2abs-​root-​1sgerg-​pst

b. (Haiek) Mikel kotxe-​z ekarr-​i z-​u-​te-​n. They.erg Mikel.abs car-​ins bring-​prf tns-​root-​pl-​pst ‘They brought Mikel by car.’ As noted in the literature about ED, this agreement split has no syntactic consequences: the argument in the clause keeps the ergative case, even though it is cross-​ referenced by regular absolutive morphemes, and the auxiliary root is not altered. Laka (1993a) observes further that ED has no effect in configurational relations relevant to binding or scope. ED is generally viewed as a purely morphological operation (see Rebuschi 1999; Hualde 2002). Many recent analyses account for ED within a Distributed Morphology approach (Albizu and Eguren 2000; Albizu 2002; and Arregi and Nevins 2013). A few analyses link ED to aspects of the syntactic derivation, more concretely to properties of the Agree operation. For Fernández (1997) and Fernández and Albizu (2000), ED shows that case and phi-​feature checking occur separately—​against different functional heads—​and this explains the case/​agreement split. For Rezac (2003), ED makes manifest some of the formal properties of the Probe–​Goal relation: (absolutive) person Probes can Agree downward (preferentially), giving rise to the ordinary agreement pattern, or upwards, in the absence of an appropriate [+person] Goal in their complement domain, giving rise to ED. None of the syntactically oriented approaches addresses the Tense–​Mood restriction.

 

792    Ane Berro and Ricardo Etxepare

32.4  Unergative Predicates 32.4.1 Simple Unergative Predicates Unlike those ergative languages where the sole argument of intransitive predicates always surfaces as absolutive, ergative marking in Basque is sensitive to the unaccusative–​ unergative distinction (see Perlmutter 1979, Perlmutter and Postal 1982, Burzio 1986 for the distinction; Levin 1983, Etxepare 2003, Berro 2010 2012 2015, for Basque). Thus, in central and western varieties of Basque, subjects of intransitive predicates expressing volitional acts, verbs of emission and involuntary bodily processes typically take the ergative case (12), whereas predicates expressing change of state, predicates of existence or happening, and (some) aspectual predicates license absolutive case (13): (12)

a. Jon-​ek     dantza-​tu d-​u. Jon-​erg  dance-​prf   tns-​root ‘Jon has danced.’

(volitional act)

b. Urrezko hortza-​k dirdira-​tzen d-​u. golden tooth.det-​erg glitter-​ipfv tns-​root ‘The golden tooth glitters.’ c. Zurrunga-​tu d-​u-​zu. snore-​prf tns-​root-​2erg ‘You have snored.’ (13)

a. Jon eror-​i Jon.abs fall-​prf ‘Jon has fallen.’

(verb of emission)

(involuntary bodily action)

d-​a. tns-​root

b. Ezbehar bat gerta-​tu d-​a. accident one.abs occur-​prf tns-​root ‘An accident has occurred.’ c. Jon liburua lei-​tzen has-​i d-​a. Jon.abs book.abs read-​ger start-​prf tns-​root ‘Jon has started reading the book.’

(change of state)

(verb of occurrence)

(aspectual verb)

Such systems have been variously called Split S (Dixon 1979, 1994), Split Intransitive (Merlan 1989), Extended Ergative (Ortiz de Urbina 1989), Active (Bittner and Hale 1996b; see also Sapir 1917), Agentive (Mithun 1991b), or Semantically Aligned (Donohue and Wichmann 2008). The case alignment of western and central dialects is unlike the one we find in eastern dialects (see Lafitte 1962; Oyharçabal 1992; Aldai 2008, 2009; Berro 2010, 2012). Thus,

 

Ergativity in Basque   793 in eastern varieties, some of the very same intransitive verbal classes take an absolutive subject. This is the case for intransitive volitional manner of motion predicates: (14) a. Xabier promena-​tu d-​a. Xabier.abs stroll-​prf tns-​root ‘Xabier went for a stroll.’ b. Gizona asto-​gain-​era jauzi d-​a. man.det.abs donky-​top.all jump.prf tns-​root ‘The man has jumped on the donkey.’ (From the Basque General Dictionary) The verb classes involved in this dialectal split include, besides volitional manner of motion verbs, speech verbs (mintzatu ‘speak,’ solastatu ‘chat’), volitional activity verbs (borrokatu ‘fight,’ dantzatu ‘dance,’ jokatu ‘behave,’ jostatu ‘play’), and meal related verbs (bazkaldu ‘have lunch,’ afaldu ‘have dinner’). In this sense, Eastern dialects seem to be closer to those ergative systems which generalize the unmarked case (absolutive) to all intransitive subjects. Aldai (2008 2009) suggests that the dialectal divide within Basque corresponds to the divide between two case-​marking systems: one based on the ergative alignment of case, represented by eastern dialects, and a semantically aligned one, represented by western and central dialects. Levin (1983: 290ff.) was the first to point out that there is a semantic split conditioning the case-​marking of intransitive predicates in western and central Basque. She proposed that the ERG/​ABS split in Basque intransitives was based on the agentive/​non-​agentive status of the subject (Levin 1983: 298). Agentivity has been considered a determining factor for the division of unergative and unaccusative verbs (Perlmutter 1978). There are well-​known problems with this simple-​minded division, however: one is that it does not straightforwardly account for the category of emission verbs (see 18b), as the subject of those verbs is an inanimate entity; the other problem, pointed out by Aldai (2009), is that the agentive dimension, understood as the ability of the subject to willingly carry out the event named by the verb, is also present in many intransitive verbs of the unaccusative type, like verbs of directed motion (cf. etorri ‘come,’ joan ‘go,’ or abiatu ‘depart’). Aldai suggests that rather than agentivity, it is the patientive status of the single argument of intransitives which determines whether it will be mapped as an absolutive noun phrase in central and western dialects. Aldai defines patientivity along the lines of the Patient Proto-​Role proposed by Dowty (1991). The semantic properties associated to the argument bearing this role involve at least: (i) undergoing a change of state or position; (ii) being causally affected by another participant; (iii) being incremental in the sense of determining the temporal boundaries of the event. This provides an appropriate basis to distinguish ergatively marked activity verbs such as jostatu ‘play’ (which are atelic and do not involve a change of state) from absolutive-​marked intransitives such as etorri ‘to come’ or abiatu ‘to depart’ (which involve a change of position and are telic) in those dialects. The special status of the patientive property as opposed to the agentive one is

 

794    Ane Berro and Ricardo Etxepare crucial in Aldai’s characterization of western/​central dialects as semantically aligned dialects, as it allows him to extend this account to emission verbs such as dirdiratu ‘to shine.’ The single argument of those verbs does not show the typical properties of patientive arguments (it does not undergo a change of state or position, and it is not causally affected), but it does not naturally belong in the opposite agentive proto-​role either, as it does not present any volitional involvement in the event or state and does not cause any change of state in another participant. This is important, as the dialectal facts show that verbs of emission must constitute a natural semantic grouping: in eastern dialects, those verbs fail to show up as absolutive, unlike, say, agentive or volitional manner of motion verbs: (15)

a. Eguzkia-​k argi-​tzen sun.det-​erg light-​ipfv ‘The sun shines.’

d-​u. tns-​root

b. Urrezko hortza-​k dirdira-​tzen golden tooth.det-​erg glitter-​ipfv ‘His/​her golden tooth glittered.’

z-​u-​en. tns-​root-​pst

This is against the idea that eastern dialects are purely ergative systems. Aldai’s characterization of the eastern case-​alignment is also questioned by the behavior of inanimate manner of motion verbs such as kurritu ‘flow’ or funtzionatu ‘work, function,’ which take an ergative DP as their sole argument in eastern dialects (16), as well as by verbs directly expressing duration, like iraun ‘persist’ or intransitive berandu ‘to delay’: (16) a. Ura-​k kurri-​tzen d-​u. water.det-​erg flow-​ipfv tns-​root ‘The water flows/​is flowing.’ b. Hon-​ek    ez    d-​u     funtziona-​tzen. this-​erg  neg  tns-​root  work-​imp ‘This does not work/​is not working.’ c. Beran-​tzen z-​u-​en. delay-​ipfv tns-​root-​pst ‘He/​she was taking long.’ It seems therefore that the purported distinction between eastern and western/​central dialects in terms of a semantically based alignment versus an ergative alignment is too coarse, as not all unergative subjects in eastern dialects show up as absolutive. Table 32.7 reproduces the basic distribution of the verb–​case correspondence across dialects (adapted from Berro 2010: 43).4 4  Two notes concerning Table 32.7: verbs of communication in central and western varieties are complex predicates, so a one-​to-​one comparison with their eastern counterparts is not possible; then,

 

Ergativity in Basque   795 Table 32.7 Verb classes and subject case across dialects Type A

Type B

Type C

Subject case

Ergative

Absolutive

Ergative

Classes of verbs

Verbs of manner of motion Meal related verbs Animate activity verbs

Verbs of communication Verbs of manner of motion Meal related verbs Animate activity verbs

Inanimate emission verbs Inanimate manner of motion verbs Verbs of duration

Example

Jon-​ek dantzatu du (18a)

Jon dantzatu da

Eguzkiak argitzen du (21a)

Variety

Western/​Central

Eastern

All

Ultimately, central/​western varieties seem to invoke the ergative for both external and internal causation, in the sense of Levin and Rappaport-​Hovav (1995). L&RH argue that causation is one of the relevant factors involved in determining the mapping of the single argument of an intransitive predicate. Causation can be external—​in the case of causative verbs—​or internal. Externally caused verbs—​such as break—​denote eventualities which can be under the control of some external force. Internally caused verbs—​ like laugh, play, or speak—​on the other hand, are actions that can only be controlled by the individual engaged in them. For them, the concept of internal causation subsumes agency. Agentive verbs are a subclass of the broader class of verbs of internal causation. The inherent property responsible of carrying out the eventuality in agentive verbs is the will or volition associated to the sole argument of the predicate. This allows us to distinguish two natural subclasses in the category of internal causation: the class characterized by the presence of a volitional semantic feature, associated to verbs such as dantzatu ‘dance,’ jokatu ‘play, bet,’ or borrokatu ‘fight’ and the class not possessing that feature (verbs of emission or duration). Central/​western varieties tend to assign ergative case to all subjects of intransitive predicates expressing internal causation. Eastern dialects seem to distinguish volitional and non-​volitional subjects. The historical record shows that the central/​western system is an innovation, slowly taking ground over the eastern one (see Mounole 2008; Aldai 2009; Berro 2010). Borrowing patterns are a clear indication of this shift: as shown by Alberdi (2003), recently borrowed intransitive verbs in western/​central dialects and also in eastern ones adopt the unergative case pattern when the borrowed verb has an agentive subject or one which encodes the source of internal causation. Thus verbs such as abdikatu ‘abdicate,’ abortatu ‘abort,’ dudatu ‘doubt,’ emigratu ‘emigrate,’ protestatu ‘protest’ are borrowed in Basque with an ergative subject, as are verbs of internal causation such as zirkulatu ‘circulate (e.g. money),’ and funtzionatu ‘function, work.’ there are some important exceptions in the set of activity verbs; a verb like kantatu ‘sing’ does not take an absolutive subject in eastern varieties, as one would expect from Aldai (2009) or the classification proposed here. One possibility is that this verb belongs in the class of emission verbs.

 

796    Ane Berro and Ricardo Etxepare

32.4.2 Light Verb Constructions and the Implicit Object Hypothesis Morphologically simple unergative predicates such as those examined in section 32.4.1 are far less frequent in Basque outside the domain of borrowed verbs. As observed by Levin (1983), Uribe-​Etxebarria (1989), Laka (1993b), Hale and Keyser (1993), Etxepare (2003), Oyharçabal (2007), Aldai (2009), most unergative predicates in Basque are not simple verbal predicates, but complex ones built on the basis of a light verb and a bare noun. Leaving aside borrowing from the surrounding Romance languages, this is the only productive way to create new unergative predicates in Basque. Thus, verbs such as dantzatu ‘to dance’ or distiratu ‘to shine’ have the analytic counterparts dantza egin (do+dance) and distira egin (do+shine). All the complex predicates are formally transitive, and have an ergative subject: (17)

a. Jon-​ek dantza egin d-​u. Jon-​erg dance do.prf tns-​root ‘Jon has danced.’ b. Jon-​ek dantza-​tu d-​u. Jon-​erg dance.prf tns-​root ‘Jon has danced.’

(18)

a. Urrezko hortza-​k dirdira egi-​ten d-​u. golden tooth.det-​erg glitter do-​ipfv tns-​root ‘His/​her golden tooth glittered.’ b. Urrezko hortza-​k dirdira-​tzen golden tooth.det-​erg glitter-​ipfv ‘His/​her golden tooth glittered.’

d-​u. tns-​root

An influential approach to the central/​western alignment pattern described in the previous section tries to assimilate unergative case patterns to the more general transitive configuration. On the basis of alternative forms such as those in (17) and (18), Hale and Keyser (1993), following Laka (1993b), propose that unergative predicates may be treated generally as underlyingly transitive, with incorporation of the object noun into the verb in the lexical component. The presence/​absence of incorporation would underlie the distinction between (17a–​b) and (18a–​b). The transitivity hypothesis implies the idea that ergative case in ergative systems such as Basque is second-​to-​absolutive, a marked case which is assigned/​licensed once absolutive case has been (Laka 1993b; Bittner and Hale 1996b; see also Ortiz de Urbina 1989; San Martin and Uriagereka 2002). The implicit object hypothesis finds further support in the transitive form of the auxiliary (du ‘has’) in simple unergative predicates, which would seem to require an underlying object. Technically, the incorporation account poses the following

 

Ergativity in Basque   797 problem: incorporation typically results in the absorption of one of the case features of the verb (see Baker 1988), and under that premise, the subject of simple unergative predicates should be marked absolutive. An incorporation account is in this sense more natural for eastern dialects than for central/​western ones (Fernández 1997). Another problem raised by the underlying object hypothesis is that many unergative verbs do not seem to be amenable to a simple lexical decomposition of the sort that underlies verbs such as dantzatu or distiratu. As shown by Alberdi (2003) and Laka (2006), Basque borrowings of Spanish unergative verbs, such as funtzionatu ‘work’ or eskiatu ‘ski,’ which are simple unergatives, are not transparently built on nouns. Their borrowed part includes both the romance verbal root and its thematic vowel: (19) a. Hon-​ek ez d-​u funtziona-​tzen. this-​erg neg tns-​root work-​imp ‘This does not work/​is not working.’

(Sp. funcion-​a-​r ‘work’)

b. Miren-​ek ez d-​u eskia-​tzen. Miren-​erg neg tns-​root ski-​imp ‘Miren does not go skiing.’

(Sp. esqui-​a-​r ‘to sky’)

The relation between transitive auxiliary and underlying object is also weakened by the fact that the transitive auxiliary occurs even if the complement of the light verb is not a noun and therefore does not count for agreement or case. This complement can be an adpositional phrase (Etxepare 2003; Preminger 2012): (20) a. Dantza-​n egin d-​u. dance-​iness do.prf tns-​root ‘(S)he has danced.’ b. Pilota-​n egin d-​u-​te. Ball-​iness do.prf tns-​root-​pl ‘They have played handball.’ In both (20a–​b), the presence of the inessive conveys the idea that the event denoted by the predicate takes place more than once. Adpositional phrases in Basque do not agree with the auxiliary, which suggests that the morphologically transitive status of the auxiliary does not imply the existence of an underlying agreeing object. If this is the case, then ergative case in Basque cannot be considered second to absolutive. This conclusion is valid too for the several aspectual verbs which take adpositional aspectual complements, and show both a transitive auxiliary and ergative case in their subject: (21) a. Liburua-​k irakur-​tzen buka-​tu d-​u. book.det-​pl.abs read-​imp finish-​prf tns-​root ‘(S)he has finished reading the books.’

 

798    Ane Berro and Ricardo Etxepare b. Poema-​k errezita-​tzen jardun d-​u. poem.det-​pl.abs recite-​imp be.engaged.prf tns-​root ‘(S)he has been reading poems.’ Predicates which present ergative–​dative pairings also pose a problem to the idea that the presence of an ergative implies a previous absolutive argument, and more generally, to the idea that absolutive case is obligatorily discharged. Ergative–​dative pairings in Basque are limited to verbs of aiming,5 in which one of the participants exerts a volitional activity directed toward the other participant, without however triggering a change of state affecting the second participant: help, follow, beg, attack are an illustrative set of such verbs (see Fernández and Ortiz de Urbina 2010 and Mounole 2012 for a discussion including diachronic data). Those predicates present Erg-​Dat pairings in central and western Basque, and a transitive pattern Erg–​Abs in eastern ones: (22) a. Jon-​ek Miren-i lagun-​du d-​i-​o. Jon-​erg Miren-​dat help-​prf tns-​root-​3dat ‘Jon helped Miren.’ b. Jon-​ek Miren lagun-​du Jon-​erg Miren.abs help-​prf ‘Jon helped Miren.’

d-​u. tns-​root

(Western/​Central)

(Eastern)

In sum, the distribution of the ergative, at least in central and western dialects, cannot be straightforwardly accounted for by invoking an implicit absolutive object.

32.5  The Inherent Case Hypothesis and Its Problems An alternative view of ergative case in Basque is that it is determined by the lexical semantics of the verb, as an instance of lexical (Oyharçabal 1992) or inherent case (Holmer 1999; Laka 2006, Chapter 7, this volume). This view is in line with the hypothesis that the distribution of ergative case is related to the expression of semantic causation, with dialectal differences related to sub-​features of the causer argument, such as animacy. There is evidence however that ergative case is not strictly determined by the causative role held by a DP in the verbal predicate. This evidence concerns on the one hand, the existence of thematic roles other than those related to causation that may be 5  We leave aside the contact-​induced phenomenon of Differential Object Marking studied by Odria (2014) and Fernández and Rezac (2016) in central and western varieties.

 

Ergativity in Basque   799 realized with ergative case (section 32.5.1), and on the other, the distribution of ergative case in raising predicates and complements of perception verbs (sections 32.5.2 and 32.5.3).

32.5.1 Holders, Experiencers, Beneficiaries A number of analytic predicates in Basque are not based on the usual verbal forms, but are headed by bare nominal and adjectival predicates. Among these non-​verbal predicates, we find psychological predicates like maite izan ‘love’ and gorroto izan ‘hate,’ modal verbs like behar izan ‘must’ or nahi izan ‘want,’ or predicates expressing worth or value, like axola izan ‘matter,’ or balio izan ‘be worth.’ Here is an illustrative sample (from Berro 2016): (23) a. Noun headed axola izan balio izan behar izan falta izan merezi izan uste izan zor izan b. Adjective headed ageri izan aski izan gogoko izan gustoko izan maite izan atsegin izan

‘matter, lit. have importance’ ‘have worth/​value’ ‘must, lit. have need’ ‘lack, lit. have lack’ ‘merit, deserve, lit. have worthiness’ ‘think, lit. have (an) opinion’ ‘owe, lit. be debt’ ‘show, manifest, lit. be obvious/​have as obvious’ ‘be sufficient/​have as sufficient’ ‘like, lit. have (as) a pleasant thing’ ‘like, lit. have (as) a pleasurable thing’ ‘love, lit. have (as) dear’ ‘like, lit. have (as) a pleasurable thing’

These predicates cannot be headed by -​tu or -​tzen, but can compose directly with the auxiliary. These predicates are all stative, non-​dynamic, and non-​eventive, and the semantic roles attributed to their subjects cannot be subsumed under the notion of causation. Those subjects are, nevertheless, ergative. (24) a. Jon-​ek Mikel maite d-​u. Jon-​erg Mikel.abs love tns-​root ‘John loves Michel’ b. Jon-​ek Mikel gorroto d-​u. Jon-​erg Mikel.abs hatred tns-​root ‘John hates Michel’

 

800    Ane Berro and Ricardo Etxepare Bare non-​verbal predicates give rise to two types of thematic interpretation for the subject of the predication. In the case of predicates expressing worth or some related measurable condition (value, merit, quantity), we can define the relevant thematic role as holder of the situation denoted by the predicate: (25) a. Horr-​ek ez d-​u axola. that-​erg neg tns-​root matter ‘That does not matter.’ b. Zopa-​k gatza falta d-​u. soup-​erg salt.det.abs lack tns-​root ‘The soup lacks salt.’ In the case of psychological predicates, the animate subject has an experiencer role. We can include in this set the subject of modal predicates behar ‘need’ and nahi ‘want’ (Etxepare and Uribe-​Etxebarria 2012): (26) a. Jon-​ek txokolatea atsegin d-​u. Jon-​erg chocolate.abs love tns-​root ‘John likes chocolate.’ b. Jon-​ek txokolatea nahi/​behar d-​u. Jon-​erg chocolate.abs want/​need tns-​root ‘John wants/​needs chocolate.’ Together with the bare predicate constructions illustrated in (23), Basque can also construct complex predicates based on an agreeing nominal or adjectival predicate. Those complex predicates have a transitive auxiliary and an ergative subject, and they convey a situation or relation which is of interest for, or to the advantage of, the subject (Etxepare and Uribe-​Etxebarria 2012): (27) a. Ni-​k Jon eta Miren laguna-​k I-​erg Jon and Miren.abs friend.det-​pl ‘Jon and Miren are friends to me.’

d-​it-​u-​t. tns-​pl-​root-​1sgerg

b. Zu mutil azkarra z-​aitu-​gu. you.abs chap intelligent.det 2abs-​root-​1plerg ‘You are an intelligent chap to us.’ In (27), the transitive auxiliary does not directly select the object (despite the fact that it agrees with it), as (27a) does not imply (28): (28) Ni-​k Jon eta Miren I-​erg Jon and Miren.abs ‘I have Jon and Miren.’

d-​it-​u-​t. tns-​pl-​root-​1sgerg

 

Ergativity in Basque   801 The absence of such an implication suggests that the auxiliary selects an independent clause, containing the object and the agreeing nominal in a predication relation. Etxepare and Uribe-​Etxebarria (2012) suggest that the subject of the clause is generated outside of this predicative relation, by an independent adpositional head that they assimilate to an applicative function word: (29) …T [AuxP BE [ApplicativeP I P [Small Clause Jon and Miren friends]]] Adopting Kayne’s analysis of the be/​have alternation (1993), they take the adposition to incorporate into the copula, yielding the transitive auxiliary edun (‘have’). This transitive auxiliary licenses the absolutive case of the subject of the predicative clause, whereas the higher T licenses the ergative case of the subject. Etxepare and Uribe-​Etxebarria extend this analysis to modal verbs based on nouns behar ‘need’ and nahi ‘want,’ and Berro (2016) extends it to cover the whole set of de-​adjectival and de-​nominal complex predicates. Non-​verbal predicate constructions expand the set of roles that can be realized by an ergatively marked subject, and from this point of view, they undermine the inherent case hypothesis. The significance of this fact for the inherent case hypothesis of ergatives is not completely clear, however, as this expanded set of thematic roles not related to causation is claimed to be contributed by an abstract adpositional head. To the extent that this line of analysis is correct, an inherent approach could still be upheld. Note however that in the analysis proposed by E&UE, the ergative is not directly assigned/​licensed by this abstract head, but by T.

32.5.2 Complements of Perception Verbs Transitive subjects in Basque may bear ergative case in some nonfinite dependents, like clausal nominalizations, which behave in many respects as finite clauses. In nonfinite perception verb complements, however, transitive subjects are obligatorily absolutive (see Artiagoitia 2003). Thus, the nominalized clause in (30a) has ergative katuek in subject position corresponding to the absolutive katuak in the nonfinite tzen-​gerund in (30b). When we compare the two clauses, it is case, not the semantic role of the arguments, that is affected: (30) a. Katu-​ek sagua-​k harrapa-​tze-​a normala d-​a. cat-​det.erg mouse.det-​pl.abs catch-​nmlz-​det normal tns-​root ‘It is a normal thing that cats should catch mice.’ b. Katua-​k sagua-​k harrapa-​tzen ikus-​i d-​it-​u-​t. cat.det-​pl.abs mouse.det-​pl.abs catch-​ger see-​prf tns-​pl-​root-​1sgerg ‘I saw the cats catch the mice.’ Perception complements in Basque can correspond in principle to two very different syntactic configurations, as observed by Arteatx (2007 2012). The string

 

802    Ane Berro and Ricardo Etxepare Perception verb + DP + nonfinite VP can correspond to (31a) or (31b), structures that we will call adjunct and complement: (31)

a. [Vperception b. [Vperception

DPi] [PROi Tcomplete gerund] [DP Tdefective gerund/​infinitive]]

adjunct complement

In the adjunct structure (31a), the DP is an argument of the perception verb and the gerund is a clausal adjunct with an implicit argument (PRO) controlled by the DP. This structure entails perception of the referent of the DP. In the complement structure (31a), on the other hand, the object of perception is the event denoting [DP gerund] clause. There is no thematic relationship between the perception verb and the DP, so that perception of the referent of the DP is not entailed. It is this second structure that is of interest to the ergative–​absolutive alternation, because the DP belongs thematically to the nonfinite clause as its external argument. Consequently, if the DP behaves differently for case and agreement in it than it does in a finite clause or in clausal nominalizations, it cannot be due to any differences in the domain where thematic interpretation occurs, but to syntactic differences in higher structure. That perception verb complements in Basque may have the structure in (31b) in Basque can be shown by means of different semantic and syntactic tests. First, [DP gerund] can be isolated as a constituent (Arteatx 2007: 32f.), as in the question–​answer pair in (32). We assume that response fragments are constituents stranded by deletion (Merchant 2004). The absolutive form of Miren shows that it has been stranded by deletion of the matrix perception clause, because transitive subjects are not absolutive in any other context. (32) Constituency of DP + tzen A: Zer ikus-​i d-​u-​zu what.abs see-​prf tns-​root-​2erg

(*Miren)? Miren.abs

B : [Miren      pianoa     jo-​ tzen]. Miren.abs  piano.det.abs  play-​ger ‘What did you see (*Miren)? Miren play the piano.’ Then, the DP of perception can be shown to be an argument of the gerund only. This is so in (33a–​b). Here perception is of [DP gerund] and not of the referent of the DP, which is impossible or odd, so only [DP gerund] and not the DP is an argument of the perception verb. Transitive and unergative subject DPs in this situation instantiate external arguments that are absolutive rather than ergative, divorcing ergativity from external argumenthood. (33) Object of perception a. Miren-​ek tenperatura #(gora   egi-​ten) ikus-​i z-​u-​en. Miren-​erg temperature.det.abs     up  do-​ger see-​prf tns-​root-​pst ‘Miren saw the temperature #(go up).’

 

Ergativity in Basque   803 b. Jon-​ek beroa # (horma zart-​araz-​ten) entzun d-​u. Jon-​erg heat.det.abs ice.det.abs crack-​cause-​ger heard.prf tns-​root ‘Jon heard the heat #(breaking the ice).’ More strikingly still, the imperceptible DP may be an idiom chunk (Rezac et al. 2014). In (34a) ur-​, independently meaning ‘water,’ is part of an idiom meaning ‘time passes’, and then lacks the meaning ‘water.’ As a transitive subject, it is ordinarily ergative, but in a tzen gerund under a perception verb (34b) it is absolutive. The idiomatic meaning requires that ur-​be an argument of the gerund only, not of the perception verb which would entail perception of ‘water.’ In both (34a–​b) then, ur-​ has the same argument–​ predicate relations, yet its case switches from ergative to absolutive. (34) a. Kontu horretan, ura-​k bide egin d-​u azken urteotan. in this matter water.det-​erg way make.prf tns-​root in these last years ‘In this matter, things have advanced these last years.’ b. Kontu horretan, ura bide egi-​ten ikus-​i d-​u-​gu azken urteotan. water.det.abs way make-​ger see-​prf tns-​root-​1plerg ‘In this matter, we have seen things advance these last years.’

Perception verb complements show that ergativity is independent of argumenthood (thematic) relations. Assuming that complements of perception verbs are Tense-​ defective, Rezac at al. (2014) conclude that ergativity like nominativity requires a nondefective T-​system, present in finite clauses and in clausal nominalizations but not in non-​finite complements of perception verbs.

32.5.3 Raising Predicates A similar argument can be constructed out of raising structures. As noted by Artiagoitia (2001), raising verbs of the seem-​type in Basque take ergative subjects: (35) Jon-​ek dotore ema-​ten d-​u. Jon-​erg elegant look-​ipfv tns-​root ‘Jon looks elegant.’ The complement of the light verb eman ‘look’ is an adjective, so the transitive auxiliary does not signal the presence of a true bivalent verb. Even if the verb is followed by a noun, the noun behaves as a predicate, and cannot agree with the auxiliary (Artiagoitia 2001): (36) Jon-​ek eta Miren-​ek artista-​k ema-​ten d-​u-​te /​ *d-​it-​uz-​te. Jon-​erg and Miren-​erg artist.det-​pl look-​ipfv tns-​root-​ple /​ tns-​pla-​root-​ple ‘Jon and Miren look like artists.’

 

804    Ane Berro and Ricardo Etxepare Artiagoitia (2001) shows that for several argumenthood tests, eman does not assign an external theta-​role. Impersonal constructions, as other valency reducing operations like reflexivization or reciprocalization in Basque, are signaled by the presence of the intransitive auxiliary izan ‘be.’ No such construction can be built out of eman ‘look’ (Artiagoitia 2001): (37) *Gorbata jantzita, dotore tie.det.abs wearing, elegant ‘One looks elegant wearing a tie.’

ema-​ten look-​ipfv

d-​a. tns-​root

That the verb eman does not project an external argument is also shown by clausal complement constructions such as (38): (38) Ema-​ten d-​u [Xabier gaur dotore d-​ago-​ela]. look-​ipfv tns-​root Xabier.abs today elegant tns-​root-​comp ‘It seems that Xabier is elegant today.’ Example (38) does not admit an overt subject:6 (39) *Kontua-​k/​Miren-​ek ema-​ten d-​u [Xabier gaur dotore thing.det-​erg/​Miren-​erg look-​ipfv tns-​root Xabier.abs today Elegant d-​ago-​ela]. tns-​root-​comp ‘*The thing/​Miren looks like/​seems that Xabier is elegant today.’ Structures such as (39), as well as the semantic status of the verb eman (meaning ‘look like, seem’) suggest that the verb is a raising verb, and therefore that the ergative subject in cases like (35) is a derived subject, not related to an external argument. Artiagoitia (2001) extends the same analysis to the eastern raising verb irudi izan ‘seem.’ Rezac et al. (2014) extend Artiagoitia’s conclusion to other raising predicates in Basque, like necessity modals.7

32.6  Ergative Case as a Marked Case Rezac et al. (2014) observe that ergative agreement can cross-​reference an absolutive argument in necessity modal constructions, which they show to be raising 6  See Oyharçabal (1992) and Artiagoitia (2001) for discussion about the acceptance of structures like (39) by some speakers. 7  The conclusions on the derived status of ergative case, as uncovered by raising constructions and complements of perception verbs, converge nicely with recent ERP-​based neurolinguistic evidence suggesting that ergative-​case violation conditions in Basque trigger the same electrophysiological responses as violation conditions affecting the nominative in nominate–​accusative systems (see Diaz et al. 2011).

 

Ergativity in Basque   805 constructions (see also Goenaga 2006). They point out the natural occurrence of examples like (40): (40) Pintxo ona-​k egon behar d-​u-​te taberna horretan. tapa good.det-​pl.abs be must tns-​root-​pl pub that.iness ‘There must be good tapas in that pub.’ The auxiliary agrees in number with the only argument of the clause pintxo onak ‘good tapas.’ The agreement affix corresponds to ergative but the argument is marked absolutive. Example (40) contrasts with the also possible (41), in which the argument shows up as ergative: (41) Pintxo on-​ek taberna horretan egon behar d-​u-​te. Tapa good-​erg.pl pub that.iness be must tns-​root-​pl ‘The good tapas must be in that pub.’ The case alternation between (40) and (41) correlates with a difference in interpretation: the ergatively marked argument is interpreted as specific and definite; the absolutive one is interpreted as indefinite and non-​specific. The unmarked word order in each case is also different, with the ergatively marked argument outside the domain of the preverbal focus position, occupied by the locative taberna horretan. The immediate preverbal position in (40) on the other hand, is occupied by the absolutive argument. The difference in interpretation is reminiscent of the one we find in English there raising constructions between a raised subject and a there-​associate. The immediate preverbal position of the absolutive argument can then be compared to the embedded position of the nominal associate in there-​constructions. Rezac et al. (2014) take this to mean the following: whereas ergative agreement can be realized at a distance (cf. local conditions on Agree, Chomsky 2001); ergative case-​marking can only be licensed in closer association to T. By closer association, they mean internal merge into the TP projection. If this is correct, then there is a sense in which ergative case, as opposed to ergative agreement, is a more marked option. For a DP to surface with Ergative case in Basque it must (i) Agree with the inflected verb, and (ii) be merged to TP. We refer the reader to the more developed argument in Rezac at al. (2014). The marked status of ergative case in comparison to absolutive may account for its relative delayed acquisition: all early acquisition studies of ergativity in Basque for L1 and child L2 learners point out that ergative case-​marking is acquired later than ergative agreement (see Huarte 2007; Ezeizabarrena 2012).

Acknowledgments The authors are grateful to the generous review which helped us to considerably improve this chapter. All errors remain ours. The research leading to these results has received funding from the European Union’s Seventh Framework Programme for research, technological

 

806    Ane Berro and Ricardo Etxepare development, and demonstration under grant agreement no. 613465. We are glad to acknowledge also financial support from the Spanish Ministry of Economía y Competitividad (projects VALAL FFI2014-​53675-​P, FFI2011-​26906 and FFI2014-​51878-​P) and the Basque government (the pre-​doctoral grant BF109.203 and the project IT665-​13).

Abbreviations abs, absolutive; all, allative; allo, allocutive; cl, class marker; comp, complementizer; dat, dative; det, determiner; df, dative flag; erg, ergative; gen, genitive; ger, gerund; iness, inessive; ins, instrumental; ipfv, imperfective; nmlz, nominalizer; pl, plural; prf, perfect; pst, past tense; suff, suffix; tns, tense.

 

Chapter 33

Hindi/​U rdu an d re l at e d l anguag e s Miriam Butt

33.1 Introduction This chapter provides an overview of ergativity in Hindi/​Urdu and several related South Asian languages. Hindi/​Urdu is the South Asian language that has been studied most in terms of case marking and ergativity. It thus serves as a base for a chapter in a handbook. However, as South Asia forms a linguistic area, the situation in Hindi/​ Urdu is only one part of an overall pattern. This pattern can be studied more perspicuously if an awareness of the typological context is developed. As such, this chapter also takes a brief look at some of the other representative patterns found in South Asian languages. For further discussions and comparative data, see also in the volume Mahajan , Chapter 4, Legate, Chapter 6, Baker, Chapter 31, and Chelliah, Chapter 38. Consider, for example, the interaction between agreement and case marking. These two phenomena have been taken to be tightly intertwined and entire theories have been built around the assumption that case and agreement are connected up with one another very closely (e.g. see Butt 2006b, chs 4 and 5 for an overview discussion). However, the typological variation found in South Asian languages presents a challenge for such views. For example, Deo and Sharma (2006) go through some of the typological variation found within Indo-​Aryan, including Hindi/​Urdu and show that there is variation at the macro level that is mirrored at the micro level.

 

808   Miriam Butt Table 33.1 Subject marking and agreement in perfective clauses (Deo and Sharma 2006: 375) language

Erg marking

Agreement

Agr. features

Hindi

1sg, 2sg, 3sg, 1pl, 2pl, 3pl

nom S, nom O (when S is non-​nom)

gender, number

Nepali

1sg, 2sg, 3sg, 1pl, 2pl, 3pl

nom and erg S

person, number

Gujarati

1sg, 2sg, 3sg, 3pl

nom S, nom and acc O (when S is non-​nom)

gender, number

Marathi

3sg, 3pl

nom S, nom O (when S is non-​nom)

gender, number, person

Bengali

none

nom S

person, number

As can be seen in Table 33.1, not all Indo-​Aryan languages use the same agreement features (ϕ-​features). The gender/​number patterns are the result of a diachronic change that saw the old verb inflections erode and be replaced by old participial forms which had agreed in number and gender with their heads. The person/​number agreement tends to be brought into the system via the diachronic reanalysis of auxiliaries as verbal morphology. Beyond that, the languages differ as to which grammatical relation can be agreed with. This, in turn, is decoupled from the particular number and person of the (pro)noun that is to be agreed with. Deo and Sharma (2006) propose an Optimality-​ Theoretic analysis of the patterns, the interested reader is further referred to Klaiman (1987), Subbārāo (1999, 2012), and Verbeke (2013a) for typological surveys of agreement patterns in interaction with case in South Asian Languages. This chapter will not go further into the overall agreement patterns found in South Asian languages (beyond detailing what they are for whatever language is under discussion). However, the patterns depicted in Table 33.1 should serve as a caveat: in studying how ergativity works in one South Asian language, e.g. Hindi/​Urdu, it is instructive to keep the larger picture in mind, particularly as South Asian languages are known to have entered into a Sprachbund (Masica 1976) and form a linguistic area with shared features. In South Asia, the Indo-​Aryan, Dravidian, Tibeto-​Burman, and Munda languages have been in constant and close contact over several thousands of years. Although Masica (1976: 12) explicitly mentions ergativity as a feature that is not characteristic of South Asian languages as a whole (Dravidian languages do not employ an ergative), ergatives do alternate with other cases, including datives or genitives on experiencer subjects, which Masica does establish as a South Asian characteristic (see also Verma and Mohanan 1990). As all the South Asian languages I have looked at make use of differential case marking (dcm) to express (sometimes subtle) differences in meaning, I have argued in previous work that patterns of ergativity in South Asia cannot be understood without

 

Hindi/Urdu and related languages    809 keeping the larger dcm picture in mind (Butt and King 1991; Butt 2001; Butt 2006a; Butt and Ahmed 2011) and will therefore include a discussion of dcm patterns in this chapter as well. These dcm patterns work similarly across the language area (see also Malchukov, Chapter 11, this volume); however, agreement patterns tend to differ. Before delving into the language data, this section closes with a brief a note on geographical distribution. South Asian languages with ergatives are found mainly in the Northern and Central parts of South Asia, with ergativity present in Nepali and Dzongkha (Bhutan) to the East and in Sindhi and Balochi to the West and Pashto in the North (see Klaiman (1987) for a survey). As already mentioned, Hindi/​Urdu is the South Asian language that has been studied in greatest depth. The chapter will thus begin with describing the ergative patterns found in Hindi/​Urdu in detail and then by way of comparison, contrast these data with basic patterns found in Punjabi, Nepali, Gujarati, Haryani, and Bengali.

33.2  Hindi/​Urdu Officially, Hindi and Urdu are two separate languages. However, they are structurally so similar that linguistic results for one generally pertain to the other.1 As a result, researchers tend to refer to Urdu and Hindi under one name, for example: Urdu/​Hindi, Hindi/​ Urdu, Hindi-​Urdu. Having said that, however, there are some differences between Urdu and Hindi. Urdu is heavily influenced by Persian and Arabic, Hindi has seen a stronger (re)-​Sanskritization. This has mainly had effects on the lexical items used in the languages (Perseo-​Arabic origin vs. Sanskrit, though both also borrow heavily and similarly from English), but also on the phoneme inventory in that Urdu, for example, uses an [f]‌and an [x], which are not native to Hindi. There are also differences in morphology, mainly with respect to derivational morphology borrowed into Urdu from Persian and Arabic. In terms of case marking, both Urdu and Hindi function very similarly, though small differences do exist. With ditransitive verbs like ‘say to,’ for example, Hindi speakers prefer the instrumental/​comitative for the person being addressed, while Urdu can alternate between the instrumental/​comitative and the accusative. Given the small, circumscribed areas in which Urdu and Hindi differ, this chapter also refers to Urdu and Hindi under one term: Hindi/​Urdu.

33.2.1 The Basic Pattern Hindi/​Urdu is a morphologically split ergative language along the dimension of tense/​ aspect. The ergative appears on subjects of agentive (di)transitives and unergatives when the verb is marked with “perfective” morphology, as in (1b) and (2b).2 The ergative is 1  The situation is similar to that for Serbian, Bosnian, and Croatian, which used to be referred to together as Serbo-​Croatian before the breakup of Yugoslavia. 2  For an overview of the Hindi/​Urdu tense/​aspect system, see Butt and Rizvi (2010).

 

810   Miriam Butt generally obligatory on subjects of (di)transitives when the verb has perfective morphology (for exceptions, see (1) and (2)). It is optional on unergatives, where it alternates with nominative subjects (see section 33.2.4). (1)

a. ram gari cɑla-​ta hɛ Ram.M.Sg.Nom car.F.Sg.Nom drive-​Impf.M.Sg be.Pres.3.Sg ‘Ram drives a car.’ b. ram=ne gari cɑla-​yi hɛ Ram.M.Sg=Erg car.F.Sg.Nom drive-​Perf.F.Sg be.Pres.3.Sg ‘Ram has driven a/​the car.’

(2) a. nadya bılli=ko dud de-​ti hɛ Nadya.F.Sg=Erg cat.F.Sg=Dat milk.M.Nom give-​Impf.F.Sg be.Pres.3.Sg ‘Nadya gives milk to the cat.’ b. nadya=ne bılli=ko dud di-​ya hɛ Nadya.F.Sg=Erg cat.F.Sg=Dat milk.M.Nom give-​Perf.M.Sg be.Pres.3.Sg ‘Nadya has given milk to the cat.’ Example (1) illustrates a typical transitive, (2) a ditransitive. The basic Hindi/​Urdu case markers along with their distribution and further morphological effects are given in Table 33.2. Table 33.2 The Hindi/​Urdu case markers Clitic

Case name

Grammatical relation

Morphological effect

∅

nominative

subj/​obj

none

ne

Ergative

Subj

oblique np

ko

accusative dative

obj subj/​indirect obj

oblique np oblique np

se

instrumental

subj/​obl/​adjunct

oblique np

ka/​ki/​ke

Genitive

subj (infinitives) specifier

agrees with head noun oblique np

mẽ ‘in’ pɑr ‘on’ tɑk ‘towards’

Locative

obl/​adjunct

oblique np

The ergative marker ne is not homophonous with any other case marker in Hindi/​Urdu. In other languages, it is variably homophonous with the instrumental, the genitive, or the dative/​accusative (see section 33.3.3). The nominative is indicated by the lack of a case marker, while the accusative and dative share the marker ko, a polysemous use that is typical for South Asia. The instrumental se spans a range of functions, including

 

Hindi/Urdu and related languages    811 comitative, source, and ablative (cf. Khan 2009). The only case marker that inflects is the genitive—​historically it arose out of an inflecting participial construction. The genitive inflects to agree with the head noun (see Payne 1995 for a thorough discussion). Like the other overt case markers, it is also constrained to appear with the oblique form of a noun. In modern Hindi/​Urdu, this oblique morphology is only overt in masculine nouns ending in -​a, as shown in (3) for the noun lɑṛka ‘boy.’ (3) a. lɑṛk-​a gari cɑla-​ta hɛ boy-​M.Sg.Nom car.F.Sg.Nom drive-​Impf.M.Sg be.Pres.3.Sg ‘The/​a boy drives a car.’ b. lɑṛk-​e=ne gari cɑla-​yi hɛ boy-​M.Sg.Obl=Erg car.F.Sg.Nom drive-​Perf.F.Sg be.Pres.3.Sg ‘The/​a boy has driven a/​the car.’ The case markers themselves have been variously analyzed as postpositions or morphological affixes. Butt and King (2004) adduce evidence that the case markers are clitics and they are consistently glossed as such in this chapter. Finally, the attentive reader will have noted that the verb agrees with the subject when it is nominative, but not when it is ergative. When the subject is ergative, the verb agrees with the object but only if that object is also not case marked. Before moving on to a more detailed discussion of the interaction of agreement and case in section 33.2.3, a note on case terminology and notions of case alignment is in order.

33.2.2 Terminology and Case Alignment 33.2.2.1 Absolutive vs. Nominative The unmarked case is referred to as the nominative (following the line of argumentation in Mohanan 1994a). It is also referred to as the “absolutive” in some literature in keeping with ideas of ergative languages involving ergative/​absolutive systems while accusative languages involve nominative/​accusative systems (see Dixon (1994) for a recent overview discussion). Plank (1979: 4) concisely summarizes the basic idea as follows:3 1. A grammatical pattern or process shows ergative alignment if it identifies intransitive subjects (Si) and transitive direct objects (dO) as opposed to transitive subjects (St). 2. It shows accusative alignment if it identifies Si and St as opposed to dO.

3 

The more commonly used terms in the literature now tend to be A for St, S for Si and O for dO.

 

812   Miriam Butt Languages are thus expected to show a basic type of overall alignment, whereby this alignment could be expressed in terms of case marking or agreement. However, Hindi/​Urdu does not conform to the established ideas of ergative vs. accusative alignment and hence the absolutive nomenclature is also not appropriate for Hindi/​Urdu. This has been pointed out in various papers over the years (e.g. Pandharipande and Kachru 1977; Klaiman 1987; Butt and King 2003). Consider, for example, the fact that direct objects do not just appear in the unmarked form, but show dcm with accusatives. Accusatives in turn alternate with instrumentals (comitatives) with certain verbs, as shown below. The alternation depends on dialect preference (Hindi speakers disprefer the accusative) and on whether the meeting is more (instrumental) or less reciprocal (accusative) (Ahmed 2006). (4) a. nadya yasin=ko/​se Nadya.F.Sg.Nom Yassin.M.Sg=Acc/​Inst ‘Nadya will meet (with) Yassin.’ b. nadya yasin=ko/​se Nadya.F.Sg.Nom Yassin.M.Sg=Acc/​Inst ‘Nadya met (with) Yassin.’

mıl-​e-​g-​i meet-​3.Sg-​Fut-​F.Sg mıl-​a meet-​Perf.M.Sg

Note that although mıl ‘meet’ is transitive, it does not take an ergative in the perfective. Generally, verbs belonging to classes of reduced agentivity such as psych verbs or verbs of meeting as in (4) do not take the ergative. Davison (1999) lists the following transitives as exceptional in that they only allow for nominative subjects: bol ‘speak,’ la ‘bring,’ jit ‘win,’ lɑg ‘strike/​begin,’ ḍɑr ‘fear,’ lɑr ‘fight’ and mıl ‘meet.’ Some of these verbs can be explained via the “reduced agentivity” property, others cannot. No good synchronic explanation exists for the lack of an ergative on agentive verbs like la ‘bring’ and Butt and King (2004) treat them as instances of quirky case. Another small class of transitive verbs allows a nominative/​ergative alternation on subjects: samɑjh ‘understand, suppose,’ bhul ‘forget,’ jɑn ‘give birth (to),’ phãd ‘leap over,’ bɑk ‘to talk nonsense,’ har ‘lose, be defeated’ and pɑhcan ‘recognize’ (Davison 1999). These can be understood as expressing a semantic difference in that when the ergative is used, (more) responsibility for the action is ascribed to the subject. This is similar to the situation with unergatives (see section 33.2.4). However, the difference is not systematic for all speakers and the alternation may also be indicative of a system in flux (cf. Butt and Deo, Chapter 22, this volume). Patterns such as in (4), the existence of transitive agentive verbs that do not take an ergative and a further verb class that alternates between ergative and nominative subjects are not typical of ergative/​ absolutive alignment. Works such as Klaiman (1987) and Pandharipande and Kachru (1977) have thus argued that Hindi/​Urdu does not show “deep” ergativity, i.e. ergative alignment, but must be understood as basically accusative languages in which the ergative case marker is associated with semantic information, namely, animacy/​agency/​volitionality of the subject and whether the verb denotes a dynamic activity or not (Kachru 1981; Pandharipande 1981; Kachru 1987).

 

Hindi/Urdu and related languages    813

33.2.2.2 Dative vs. Accusative The homophony between the accusative and dative ko is very common across South Asia (and crosslinguistically). Some of the literature takes this as a reason to just gloss the ko as a dative. This is somewhat misleading, as the ko that appears on direct objects alternates with the nominative, whereas the ko that appears on indirect objects does not. That is, there are two distinct functionalities that are fulfilled by the same overt marker and these different syntactic functionalities can be expressed via a difference in glossing: dative vs. accusative.

33.2.3 Agreement 33.2.3.1 Basic Pattern This section describes the interaction between case and agreement. As illustrated in (5), the overall generalization is the following. When the subject is nominative (unmarked), the verb agrees with the subject (5a). If the subject is non-​nominative and the object is nominative, then the verb agrees with the object (5b). If both the subject and the object are non-​nominative, then the verb shows “default” masculine singular agreement (5c). (5) a. lɑṛk-​a gari boy-​M.Sg.Nom car.F.Sg.Nom ‘The/​a boy drives a car.’

cɑla-​ta hɛ drive-​Impf.M.Sg be.Pres.3.Sg

b. lɑṛk-​e=ne gari boy-​M.Sg.Obl=Erg car.F.Sg.Nom ‘The/​a boy has driven a/​the car.’

cɑla-​yi drive-​Perf.F.Sg

c. lɑṛk-​e=ne gari=ko boy-​M.Sg.Obl=Erg car.F.Sg=Acc ‘The/​a boy has driven the car.’

cɑla-​ya drive-​Perf.M.Sg

hɛ be.Pres.3.Sg hɛ be.Pres.3.Sg

Verbs generally agree with a noun in terms of gender and number (the relevant morphology descends from old participial morphology). An exception is the non-​past form of be (both as a copula and as an auxiliary) and the subjunctive, imperative, and future morphology on verbs. These are relics of the Old Indo-​Aryan tense system, which showed person and number agreement. The future is particularly interesting, as it combines person and number agreement with number and gender agreement. This is due to diachronic change in which a participial form of the verb ga ‘go’ gave rise to the modern future morphology -​g-​ (Butt and Lahiri 2013). In the examples below, the past and non-​past forms of be are exemplified in an auxiliary (6) and a copula use (7). (8) illustrates a subjunctive and (9) a future form. Full paradigms can be found in Butt and Rizvi (2010).

 

814   Miriam Butt    (6) a. ɑnjum adnan=ko mar-​t-​i hɛ Anjum.F.Sg.Nom Adnan.M.Sg=Acc hit-​Perf.f.Sg be.Pres.3.Sg ‘Anjum hits Adnan.’ b. ɑnjum adnan=ko mar-​t-​i th-​i Anjum.F.Sg.Nom Adnan.M.Sg=Acc hit-​Perf.F.Sg be.Past-​F.Sg ‘Anjum used to hit Adnan.’    (7) a. tʊm Pron.2.Nom ‘You are tall.’

lɑmbi ho tall.F.Sg be.Pres.2.Sg

b. tʊm lɑmbi th-​i Pron.2.Nom tall.F.Sg be.Past-​F.Sg ‘You were tall/​You used to be tall.’    (8)

ɑbhi a-​ũ? Pron.1.Sg.Nom now come-​1.Sg ‘Should I come (over) now?’ mε

   (9) tʊm xɑt lıkh-​o-​g-​i Pron.2.Nom letter.M.Sg.Nom write-​2.Sg-​Fut-​Fem.Sg ‘You will write a letter.’ The verb does not agree with the subject whenever there is non-​nominative marking. That is, the verb does not agree with dative, genitive or locative subjects. As illustrated in (10) for dative subjects, the agreement pattern in Urdu/​Hindi is thus not one that is specifically sensitive to ergativity, but to non-​nominative subjects in general. (10) a. mʊjhe Pron.1.Sg.Dat ‘I have a fever.’

bʊxar fever.M.Nom

hɛ be.Pres.3.Sg

b. nadya=ko dɑr lɑg-​a Nayda.F.Sg=Dat fear.M.Nom stick-​Perf.M.Sg ‘Nadya got scared.’

33.2.3.2 Agreement and Bare Nominals The nominative/​accusative alternation on objects illustrated in (5b) and (5c) is a general part of the language. The alternation is conditioned at least by specificity (Butt 1993b), animacy, and several other complex semantic and discourse-​related factors.4 4 

Differential Object Marking (DOM) based on specificity/​definiteness, animacy, and information structure is well-​established crosslinguistically (Bossong 1985, 1991; Aissen 2003b).

 

Hindi/Urdu and related languages    815 As seen in (5b) vs. (5c), when the object is marked with ko, it must receive a specific interpretation.5 Generally, when the direct object is animate, it must be marked with the accusative. However, there are exceptions and the best discussion to date on the semantics of nominative vs. accusative ko objects can be found in Dayal (2003), who focuses on understanding the semantics of bare nominals in Hindi. Understanding the interpretation and distribution of nominative (bare nominals) vs. accusative objects is further complicated by the fact that nominative objects can also have incorporated (pseudo-​incorporation) readings in Hindi/​Urdu (Mohanan 1995; Dayal 2011). Bare nominals can also act as parts of N-​V complex predicates, as shown in (11). Mohanan (1994a) argues that N-​V complex predicates show two patterns of agreement. When the subject is non-​nominative, the verb can either agree with the bare N (11a) or not (11b) (examples taken from Mohanan (1994a: 231)). (11) a. nadya=ne mohan=ki Nadya.F.Sg=Erg Mohan.M.Sg=Gen.F.Sg ‘Nadya remembered Mohan.’ b. nadya=ne mohan=ko Nadya.F.Sg=Erg Mohan.M.Sg=Acc ‘Nadya remembered Mohan.’

yad memory.F.Nom

k-​i do-​Perf.F.Sg

yad ki-​ya memory.F do-​Perf.M.Sg

Mohanan shows via several tests (e.g. passivization), that yad ‘memory’ in (11a) is the primary object of the sentence, which is why the verb also agrees with it. In (11b), the yad ‘memory’ is only part of the overall N-​V predicate, the primary object is the accusative Mohan. The verb ‘do’ in (12b) shows default masculine singular agreement since both the subject and the object are non-​nominative.

33.2.3.3 Long-​Distance Agreement Finally, Hindi/​Urdu allows for so-​called long-​distance agreement (lda; Davison 1985, 1988; Mahajan 1989; Butt 1993a, 1995; Bhatt 2005). lda occurs with embedded infinitives which are “nominative,” i.e. they do not bear case, they are not morphologically oblique and they are not followed by a postposition. Some verbs require lda while in others it is optional. The verb cah ‘want,’ for example, allows for optional lda, as illustrated in (12). (12) a. amad=ne [kıtab xɑrid-​ni] cah-​i Amaad.M=Erg book.F.Sg.Nom buy-​Inf.F.Sg.Nom want-​Perf.F.Sg ‘Amad wanted to buy a book.’ b. amad=ne [kıtab xɑrid-​na] cah-​a Amaad.M=Erg book.F.Sg.Nom buy-​Inf.M.Sg.Nom want-​Perf.M.Sg ‘Amad wanted to buy a book.’ 5 

The semantics work much like the system described for Turkish by Enç (1991), who makes a clear distinction between definiteness and specificity.

 

816   Miriam Butt Butt (1993a, 1995) analyzes lda as a series of local agreement relations. So, in (12a) the infinitive verb ‘buy’ agrees with its object ‘book’ and the finite verb in turn agrees with its object, the infinitive phrase ‘buy a book.’6 The optionality of agreement is attributed to a type of noun incorporation in (12b). Bhatt (2005) instead proposes an analysis by which the lda is licensed via a restructuring configuration. This analysis sees the infinitive as being essentially transparent for agreement purposes, thus allowing the matrix verb to agree directly with the embedded object. The agreement on the infinitive is seen as parasitic on that of the finite verb. An essential feature of Bhatt’s proposal is the decoupling of tense (T)  from case licensing, an assumption that breaks with previous tradition within this framework and potentially has major implications for a treatment of ergativity in Hindi/​Urdu. For example, Davison (1999) sees T as one of the licensing conditions for the Hindi/​Urdu ergative case, while Mahajan posits a tight connection between various SpecAgr positions and case marking (Mahajan 1989) in Hindi.

33.2.4 Unergatives and Unaccusatives Unergatives in Hindi/​Urdu generally take an ergative subject.7 The unergatives display dcm in that the ergative subject alternates with a nominative.8 Davison (1999) lists the following unergatives as showing this alternation: bhõk ‘bark,’ jhãk ‘peep, look into/​ through,’ khãs ‘cough,’ chĩk ‘sneeze,’ muskɑra ‘smile,’ thuk ‘spit,’ mut ‘urinate,’ hɑg ‘defecate,’ nɑha ‘bathe,’ ro ‘cry,’ hãs ‘laugh,’ ga ‘sing’ and so ‘sleep’ (see also Kachru 1978). This alternation is correlated with a type of agentivity (Tuite et al. 1985; Butt and King 1991), namely an ascription of responsibility for the initiation of the action to the subject. This is illustrated in (13), where the crying in (13a) could have been induced by an external agency so that Ram could not help but cry, but where in (13b) Ram is interpreted to be crying on purpose, i.e. to have initiated/​caused the crying event himself.

6 

See Butt (1993b) for argumentation that Hindi/​Urdu infinitives are verbal nouns. The nomenclature of unergative vs. unaccusative, originally due to Relational Grammar internal reasoning (see Butt 2006b: 38–​40), is unfortunate for languages like Hindi/​Urdu. 8  A reviewer objects to the use of dcm for case alternations found with subjects. One major concern is that the reviewer sees case alternations on objects functioning differently as compared to that on subjects. In particular, it is not clear that case alternations on subjects, in particular on unergatives, can be associated with different structural positions. This is a concern articulated from within a particular theoretical perspective. This chapter follows the state-​of-​the-​art in research on dcm, which at least since Aissen (1999b) sees case alternations on subjects and objects on a par. A recent overview is provided in Malchukov and de Swart (2009), a case study on dcm that includes alternations on both subjects and objects in Amharic can also be found in the same volume (Amberber 2009). With respect to structural correlations of different semantic interpretations of subjects, see as an example Ramchand’s (1997) proposal for two different subject positions for stage-​vs. individual-​level predication as well as her newer ideas which allows for a different structural representation of different flavors of unergatives and unaccusatives (Ramchand 2011). 7 

 

Hindi/Urdu and related languages    817 (13) a. ram ro-​ya Ram.M.Sg.Nom cry-​Perf.M.Sg ‘Ram cried.’ b. ram=ne ro-​ya Ram.M.Sg=Erg cry-​Perf.M.Sg ‘Ram cried (on purpose).’ In contrast, typical unaccusatives such as gır ‘fall,’ ja ‘go’ and a ‘come’ never take an ergative subject. This is illustrated in (14) for ja ‘go.’ (14) a. ram gɑ-​ya Ram.M.Sg.Nom go-​Perf.M.Sg ‘Ram went.’ b. *ram=ne gɑ-​ya Ram.M.Sg=Erg go-​Perf.M.Sg ‘Ram went.’ The initial paper on the Unaccusative Hypothesis (Perlmutter 1978)  and subsequent work on unaccusativity has established various syntactic tests for determining whether a verb is unaccusative or not, e.g. whether a verb allows for participle formation, impersonal passives, or reduced relatives. Ahmed (2010) carefully goes through these tests for Urdu and shows that they do not apply as such in that they are sensitive to two semantic dimensions: whether the subject is animate or not and whether the event described is telic or not. That is, verbs behave more as unaccusative when the subject is inanimate and the event is telic (cf. Kachru’s (1981) observation that the appearance of the ergative is conditioned by animacy, agency, volitionality, and whether the verb indicates a dynamic activity). Ahmed argues that verbs cannot be classified into unergatives vs. unaccusatives per se, but that semantic factors such as animacy and telicity play a large role in determining which syntactic structures a given verb may enter into, e.g. participle formation and nominative vs. ergative case marking of the subject. Ahmed’s approach represents a primarily lexical semantic approach to case, one that is also followed in Butt and King’s work (Butt and King 1991, 2003, 2004). Butt and King generally associate the ergative with some type of agency, or event initiation to use Ramchand’s (2011) term. Structural factors are acknowledged, but the primary focus is on lexical semantic factors. In contrast, researchers such as Davison (1999, 2010) and Mahajan (2000, 2004) invoke primarily structural factors (though lexical semantic factors are acknowledged by Davison) by working out a version of a dependent case analysis as proposed by Marantz (1991) (cf. also Baker, Chapter 31, this volume on Burushaski). Under this analysis, all unergatives are actually transitives with an unrealized object. However, no good explanation is provided for the ergative/​nominative DCM, which is quite systematic in the language.

 

818   Miriam Butt

33.2.5 Interaction with Light Verbs This section provides data on the interaction of case marking and light verbs. Light verbs combine with verbs, nouns, or adjectives to form complex predicates that act as a single predicational domain (Mohanan 1994a; Butt 1995). In V-​V complex predicates the light verb signals completion of an action (there is a result) and provides additional shades of meaning along the lines of suddenness, volitionality, forcefulness, etc. There is an extensive literature on V-​V complex predicates, with Hook (1974) being one of the central references. While the main verb contributes the main predicational content, the light verb acts as the head in terms of tense and agreement features. It also determines the overall “transitivity” of the complex predicate (Amritavalli 1979) in that it determines the case of the subject. For example, in (15) the transitive main verb ‘write’ carries perfective morphology and has an ergative subject, as expected. However, in V-​V complex predicates ‘write’ can be paired up with either one of two types of light verbs. When the light verb is of an unaccusative type, the subject of the overall predication must be nominative, as shown in (16a). In contrast, when the light verb is based on an agentive transitive verb, the subject must be ergative, see (16b).9 (15)

ʊs=ne/​*vo xɑt Pron.Obl=Erg/​Pron.Nom letter.M.Nom ‘He wrote a letter.’

lıkh-​a write-​Perf.M.Sg

(16) a. *ʊs=ne/​vo xɑt lıkh baiṭh-​a Pron.3.Obl=Erg/​Pron.3.Nom letter.M.Nom write sit-​Perf.M.Sg ‘He wrote a letter (completely).’ b. ʊs=ne/​*vo xɑt lıkh Pron.3.Obl=Erg/​Pron.3.Nom letter.M.Nom write ‘He/​She wrote a letter (completely).’

li-​ya take-​Perf.M.Sg

Light verbs in Hindi/​Urdu are always form-​identical with a full form version (Butt and Lahiri 2013) and the subject case marking conforms to the subject marking of the light verb when it is used as a main verb, as shown in (17). (17)

a. *ʊs=ne/​vo baiṭh-​a Pron.3.Nom/​Pron.3.Obl=Erg sit-​Perf.M.Sg ‘He got up.’ b. ʊs=ne/​*vo khana li-​ya Pron.3.Obl=Erg/​Pron.3.Nom food.M.Nom take-​Perf.M.Sg ‘He/​She took food.’

9 

Udaar et al. (2014) argue that this generalization does not hold. However, their counterexamples involve modal usages and aspectual auxiliary verbs, not the light verbs discussed here.

 

Hindi/Urdu and related languages    819 Butt (1995) understands this pattern as being due to the lexical semantics of the light verbs, which either add a dimension of less (16b) or more (16c) control over an action. In contrast, Davison (1999) posits that verbs lexically stipulate whether they select for an ergative subject or not via an [erg] feature. In complex predicates, the head (the light verb) determines the case marking of the head. The approaches have much in common, the main difference is that Davison eschews building in explicit lexical semantics into her analysis, while Butt argues that such an approach is necessary. Butt and Ramchand (2005) combine structural and lexical semantic considerations that represents (sub) sub-evental semantic information in the syntax. Another instance in which the light verb determines the case marking on the subject is found with respect to N-​V complex predicates (Mohanan 1994a). As illustrated in (18a), when the light verb is an agentive verb, the subject is ergative. When the light verb is non-​agentive, the subject is dative. (18) a. nadya=ne kɑhani yad k-​i Nadya.F.Sg=Erg story.F.Sg.Nom memory.F.Nom do-​Perf.F.Sg ‘Nadya remembered the story (actively).’ b. nadya=ko kɑhani yad a-​yi Nadya.F.Sg=Dat story.F.Sg.Nom memory.F.Nom come-​Perf.F.Sg ‘Nadya remembered the story.’ (lit. the memory came to Nadya) Mohanan (1994a) analyzes the difference in subject case marking as being due to the lexical semantics of the light verbs: agentive (kɑr ‘do’) vs. goal semantics (a ‘come’). In both cases, the light verb agrees with kahani ‘story.’ Patterns like this are wholly systematic in the language.

33.2.6 Dative/​Ergative Alternations This section introduces a further instance of the ergative alternating with the dative. As with the N-​V complex predicates, it is a semantically based alternation. The alternation was first pointed out by Masica (1990) and is illustrated in (19). It occurs in certain dialects of Hindi (mainly based in Delhi) and in the Urdu spoken in Pakistan. If the ergative is used, the subject is interpreted as wanting to perform the action, if the dative is used, the subject is generally interpreted as being under an obligation to perform the action. (19) a. nadya=ne zu ja-​na hɛ Nadya.F.Sg=Erg zoo.M.Sg.Obl go-​Inf.M.Sg be.Pres.3.Sg ‘Nadya wants to go to the zoo.’ b. nadya=ko zu ja-​na Nadya.F.Sg=Dat zoo.M.Sg.Obl go-​Inf.M.Sg ‘Nadya has/​wants to go to the zoo.’

hɛ be.Pres.3.Sg

 

820   Miriam Butt Based on a careful examination of current usage of the ergative in modern day Urdu TV dramas, Bashir (1999) concludes that the ergative in examples such as (19a) represents a marked case. When it is used, the subject is unambiguously interpreted as desiring the action. However, the dative is actually ambiguous between a ‘want’ and a ‘must’ reading. Note that the ergative and dative subjects (the fact that they are subjects can be established via the set of subject tests established by Mohanan (1994a) here appear in conjunction with an infinitive. The perfective morphology that otherwise is a necessary condition for the ergative is absent. The verb ‘be’ cannot be taken to license the ergative as it does not otherwise take an ergative (e.g., as a copula, auxiliary, or light verb in an N-​ V complex predicate). It similarly cannot be licensed by the verb ‘go’ since that also does not independently license an ergative. Nevertheless, Wunderlich and Lakämper (2001) analyze the ergative in (19a) as being licensed lexically by the copula in its modal reading. Mahajan (2004) in contrast posits an empty verb that takes a nominal complement headed by the verbal noun jana ‘to go.’ This empty verb cannot assign structural accusative case and thereby, by a version of the dependent case approach, a non-​nominative case must be assigned to the subject. That is, Mahajan (2004: 288) posits the general rule in (20). (20) In the configuration [XP YP [X′ X ZP]], assign YP a non-​nominative case marker if ZP, the complement of X, does not receive a structural Case from X. The sentence in (19a) is taken to have the configuration in (21). This exactly matches the structure specified in (20) and therefore the YP (=Nadya) is assigned a non-​nominative case. (21)

[VPe nadya=ne

Ve [DP PRO jana]]

hɛ

However, Mahajan’s analysis does not address the dative/​ergative alternation, i.e. what determines whether the non-​nominative case that is assigned is in fact the ergative and how could the dative in (20b) be licensed instead? Mahajan does assume that further lexical semantic factors govern the distribution of the case markers, however, he does not specify how these would integrate with the system he proposes. Butt and King (1991, 2004) take a completely different approach in that they follow a “constructive” perspective on case marking (Nordlinger 1998b). They point out that the only difference in the DCM sentences is the case marker itself. This holds for examples as in (19) and the specificity/​animacy alternation discussed previously, here illustrated by (22) as part of the desire infinitive. They therefore propose that the case markers themselves carry the relevant information, rather than just functioning as “spell outs” of features contributed by the verb or other parts of the syntactic structure. For example, the accusative ko in (22) is associated with a lexical entry which specifies that it can only appear on objects and that it is associated with a particular semantics (e.g. specificity).

 

Hindi/Urdu and related languages    821 (22) a. nadya=ne jiraf dekh-​na Nadya.F.Sg=Erg giraffe.M.Sg.Nom see-​Inf.M.Sg ‘Nadya wants to see a giraffe/​giraffes.’

hɛ be.Pres.3.Sg

b. nadya=ne jiraf=ko dekh-​na hɛ Nadya.F.Sg=Erg giraffe.M.Sg=Acc see-​Inf.M.Sg be.Pres.3.Sg ‘Nadya wants to see the giraffe.’ Similarly, the ergative ne specifies that it can only appear on subjects and that in situations where it is not structurally required, it contributes a particular semantics to the overall clause. In examples as in (22) and in the unergatives it plays out as denoting that the subject is responsible for the initiation of the action. For the particular case of the infinitive construction in (22), this produces a “desire” semantics. The lexical specifications contributed by the case markers interact with the structure and semantics of the remainder of the clause. The interaction is complex and mirrors what the bulk of the work on Hindi/​Urdu has concluded: case marking is due to an interaction of structural with lexical and/​or lexical semantic factors.

33.2.7 Summarizing Discussion The preceding sections have presented data pertaining to the distribution of the ergative. Along the way, some generalizations were brought out and several different approaches to understanding ergativity in Hindi/​Urdu were discussed. In Hindi/​Urdu, the ergative generally appears on subjects of (di)transitive verbs when these carry perfective morphology. Ergatives are generally optional on unergatives and a small number of transitives. Lexical semantics appear to play a role in licensing/​constraining the appearance of the ergative but these are not understood well enough to date. The ergative may also appear on desire infinitives, where the usual conditioning factors do not hold. Agreement in Hindi/​Urdu is only with unmarked subjects or objects. Interactions with long distance agreement, verbal nouns and complex predicate formation provide evidence for a complex system in which structural and semantic factors interact. On the whole, the literature on Hindi/​Urdu ergativity seems to have reached a consensus that the situation is complex and that structural as well as (lexical) semantic factors play a role. However, no consensus has been reached on just how these factors interact and which of the factors should be given more weight. There is also a consensus that attempting to understand the Hindi/​Urdu patterns in terms of case alignment systems is not useful. This was argued for early on by Kachru (1981), Pandharipande (1981), and Pandharipande and Kachru (1977), and was reiterated in Butt and King (2003). Hindi/​Urdu shows no signs of “deep” or syntactic ergativity, rather it must be understood as a basically accusative language with a complex case marking system. This point can be made even more forcefully by looking at the wider South Asian typological space. In what follows, I focus less on dcm and lexical

 

822   Miriam Butt semantic factors as these have not been explored to in sufficient depth for other South Asian languages with respect to the ergative (though it is clear that they all do exhibit dcm patterns). I focus more on the variety of agreement and case marking patterns that are found. They make the point that the overall morphosyntactic picture is complex and hint at further factors that need to be integrated into the overall picture of ergativity and its conditioning factors.

33.3  Patterns in Related Languages 33.3.1 Punjabi Punjabi is a language spoken by millions in the Punjab (which is divided across Pakistan and India) and in the Punjabi diaspora world-​wide. Despite its comparative size, there is very little work on Punjabi and closely related languages like Siraiki (Shackle 1976) or Potwari. The standard Punjabi grammar is Bhatia (1993), an older useful grammar is Cummings and Bailey (1912) and a recent dissertation is Akhtar (1999). An inventory of Punjabi case markers is given in Table 33.3. The shape of the case markers differs from the Hindi/​Urdu inventory. The only shared case marker is the ergative ne. However, the genitive da/​di/​de is presumably related to the Hindi/​Urdu ka/​ki/​ke. Table 33.3 Punjabi case markers Case clitic

Name

Grammatical function

∅

nominative

subject object

ne

ergative

subject

nũ

accusative dative dative

object subject indirect object

kolõ

instrumental ablative

argument/​adjunct subject

da/​di/​de

genitive

subject (infinitives), agrees with head noun specifier, agrees with head noun

te/​tõ

locative (from)

argument/​adjunct

Despite the difference in case markers, case and agreement in Punjabi appear to work much as in Hindi/​Urdu. However, there are a few interesting differences. As in Hindi/​ Urdu, agreement is with unmarked (nominative) direct arguments, both subjects and objects. If the subject is not available for agreement, the verb agrees with the object. If

 

Hindi/Urdu and related languages    823 that is not available, default masculine singular agreement is used. This is illustrated in (23) and (24). In (23) the subject is nominative and the verb agrees with it. In (24) the subject is ergative, the object is marked as well and the verb defaults to masculine singular. (23) a. ram muṇḍıã=nũ mar-​da Ram.M.Sg.Nom boy.M.Pl=Acc hit-​Pres.M.Sg ‘Ram is hitting the boys.’

ɛ be.Pres.3.Sg

b. laṛki muṇḍıã=nũ mar-​di ɛ girl.F.Sg.Nom boy.M.Pl=Acc hit-​Pres.F.Sg be.Pres.3.Sg ‘The girl is hitting the boys.’ (24) a. ram=ne muṇḍıã=nũ Ram.M.Sg=Erg boy.M.Pl=Acc ‘Ram has hit the boys.’ b. laṛki=ne muṇḍıã=nũ girl.F.Sg=Erg boy.M.Pl=Acc ‘The girl has hit the boys.’

mar-​ıa hit-​Past.M.Sg

mar-​ıa hit-​Past.M.Sg

si be.Past.3.Sg

si be.Past.3.Sg

Punjabi

Punjabi

Punjabi

Punjabi

As in Hindi/​Urdu, Punjabi is a split ergative language in which the ergative is sensitive to perfective marking. This is also illustrated above: the ıa in (24) is glossed as past, but it is historically the same piece of morphology that was glossed as perf in the Hindi/​Urdu examples. However, there is a difference with respect to pronouns. Although first and second person pronouns appear to be unmarked, the verb does not agree with them. Instead, the verb agrees with the nominative object. This is illustrated in (25)–​(27). (25) a. mε lakṛi vaḍ-​i Pron.1.Sg wood.F.Sg.Nom cut-​Past.F.Sg ‘I (male or female) cut the wood.’ b. tũ lakṛi vaḍ-​i Pron.2.Sg wood.F.Sg.Nom cut-​Past.F.Sg ‘You (male or female) cut the wood.’

Punjabi

Punjabi

(26) a. mε kampuṭar bech-​ia Pron.1.Sg computer.M.Sg.Nom sell-​Past.M.Sg ‘I (male or female) sold the computer.’

Punjabi

b. tũ kampuṭar bech-​ia Pron.2.Sg computer.M.Sg.Nom sell-​Past.M.Sg ‘You (male or female) sold the computer.’

Punjabi

 

824   Miriam Butt (27) a. o=ne lakṛi vaḍ-​i Pron.3.Sg.F/​M=Erg wood.F.Sg.Nom cut-​Past.F.Sg ‘He/​She cut the wood.’ b. o=ne kampuṭar Pron.3.Sg.F/​M=Erg computer.M.Sg.Nom ‘He/​She sold the computer.’

bech-​ia sell-​Past.M.Sg

Punjabi

Punjabi

In (25) and (26) the subject is a first and second person pronoun, respectively. No ergative marker appears on the subject, yet the verb does not agree with the subject. It instead agrees with the nominative object, just as in the ergative situation, cf. the same pattern found with the third person pronoun in (27). Since Silversteins’ (1976) seminal work, it is well known that ergative marking can also obey a person split. Punjabi conforms to the person split hierarchy described by Silverstein. However, the agreement pattern does not align with the person split. Despite there being no overt ergative on the first and second pronouns, the verb does not agree with them. An investigation into the history of the pronouns shows that the first and second person pronouns are old and are actually oblique forms. This is in contrast to the third person pronouns which are more recent and which have descended from demonstratives. They are not inherently oblique. Taking this historical information into account, it thus seems that Punjabi ergative split actually functions just as in Hindi/​Urdu. The difference is that the Hindi/​Urdu pronouns have been reanalyzed as nominative —​originally they stem from the same oblique pronominal ancestors as the Punjabi pronouns do. This process of reanalysis has not taken place in Punjabi —​the pronouns have kept their original oblique status, thus do not allow for verb agreement and also do not require additional marking by an overt ergative case. The Punjabi situation shows that although the particular inventory of cases may differ, the overall ergative distribution may be similar and that morphosyntactic patterns need to be investigated while keeping a diachronic perspective in mind. This is true for all of the Indo-​Aryan languages as the variation in case and agreement most likely reflect different historical stages or processes (cf. Butt and Deo, Chapter 22, this volume).

33.3.2 Nepali Nepali is another major South Asian language (Table 33.4). It is spoken mainly in Nepal, but also in Bhutan and India. It numbers millions of speakers as well, but very little work has engaged with Nepali. Some central references are Wallace (1982, 1985), the discussion here is based on Bickel (2004), Khan (2009), Butt and Poudel (2007), and Verbeke (2013a,b).

 

Hindi/Urdu and related languages    825 Table 33.4 Nepali case markers Case Clitic

Name

Grammatical Function

∅

nominative

subject object

le

ergative instrumental

subject argument/​adjunct

lai

accusative dative dative

object subject indirect object

saṅga

comitative

argument/​adjunct

ma

locative (in/​on)

argument/​adjunct

dekhi

ablative

argument/​adjunct

baṭɑ

ablative

argument/​adjunct

ko/​ka/​ki

genitive

specifier, agrees with head noun

The instrumental le doubles as the ergative in Nepali and the dative/​accusative form lai is very close in form to the instrumental/​ergative. Unlike in Hindi/​Urdu and Punjabi, the verb agrees with both the ergative and nominative subject. Verbeke (2013a: 153) characterizes it as a language with S/​A agreement, meaning that agreement is with subjects of intransitives and agentive transitives. This is illustrated in (28), taken from Deo and Sharma (2006: 377). (28) a. ma bas-​en Pron.1.Sg.Nom sit-​Perf.1.Sg ‘I sat.’ b. mai=le mero Pron.1.Sg=Erg my ‘I washed my clothes.’

Nepali

lugā clothes.Nom

dho-​en wash-​Perf.1.Sg

Nepali

Nepali does not, however, show agreement with dative subjects, see (29) from Bickel (2004: 91). When there is a nominative object, the verb agrees with that.10 (29) mɑla-​i tımi Pron.1.Sg-​Dat Pron.2.M.Hon.Nom ‘I like you.’

mɑn liking

pɑr-​ch-​au fall-​NonPast-​2.M.Hon

Nepali

As in Hindi/​Urdu, when there is perfective morphology on transitive (agentive) verbs, then the subject must be ergative. This is illustrated in (28b) and (30), taken from Khan 10 

Example (29a) involves an N-​V complex predicate, which is a common way of expressing psych predications in South Asian languages.

 

826   Miriam Butt (2009: 51). The same holds for unergatives, e.g. khok cough, mut-​urinate, hag-​defecate, nuhau-​bathe, and thuk-​spit, as illustrated in (31) for ‘urinate.’ Unergatives also allow for dcm in that the ergative alternates with a nominative subject, as shown in (32). Example (30) also illustrates the use of le as an ergative and an instrumental in the same clause. (30) ʊs=le cɑmsa=le bhat Pron.3.Sg=Erg spoon.M.Sg=Inst meal ‘He ate the meal with spoon.’ (31)

kha-​yo eat-​Perf.3.Sg

goru=le mut-​yo bull.M.3.Sg=Erg urinate-​Past.3.Sg. ‘The bull urinated.’

(32) a. hɑsɑn=le Hassan.M.Sg=Erg ‘Hassan danced.’

Nepali

Nepali

nac-​yo dance-​Past.3.Sg

Nepali

b. hɑsɑn nac-​yo Hassan.M.Sg.Nom dance-​Past.3.Sg ‘Hassan danced.’

Nepali

However, the ergative is not restricted to the perfective, but can in principle appear with every other tense (Verbeke 2013b). Earlier work reported that the ergative was generally optional in these non-​past/​non-​perfective situations; however, closer investigation has revealed systematic differences. Hutt and Subedi (1999: 116–​117) identify focus and question formation as well as the expression of a situation in which it is “part of the natural order of things for the subject to perform the verb …” Butt and Poudel (2007) analyze the latter observation in terms of stage-​vs. individual-​level predication (Kratzer 1995) and claim that Nepali has a ergative/​nominative dcm alternation in that the ergative expresses individual-​level predication while the nominative expresses stage-​level predication (Kratzer 1995). Consider the examples in (33). If one wants to express a temporary property of the subject, then the nominative is used. If a characteristic of that individual is to be expressed, then the ergative is used, as shown in (33a). Another example is provided in (34). The contrast here is that the ergative is used when referring to a driver whose job it is to drive (individual-​level predication) vs. a situation in which Hassan is driving a car, but where this is a property that is not characteristic of Hassan. (33) a. ram=le (#aja) ɑngreji Ram.M.Sg=Erg today English ‘Ram knows English (#today).’

jan-​da-​cha know-​Impf-​NonPast.3.Sg

b. ram (aja) ɑngreji bol-​da-​cha Ram.M.Sg today English speak-​Impf-​NonPast.3.Sg ‘Ram will speak English (today).’

Nepali

Nepali

 

Hindi/Urdu and related languages    827 (34) a. hɑsɑn=le gaṛi chalaun-​cha Hassan.M.Sg=Erg car.Nom drive-​NonPast.3.Sg ‘Hassan drives cars (that’s what he does).’ b. hɑsɑn gaṛi chalaun-​cha Hassan.M.Sg.Nom car.Nom drive-​NonPast.3.Sg ‘Hassan is driving a car/​cars.’

Nepali

Nepali

Verbeke (2013a,b) maintains that the stage-​vs. individual level analysis does not capture all of the ergative uses found with non-​past tenses in Nepali and proposes that a unifying conditioning factor can be understood in terms of aspectual considerations. However, it seems that Hutt and Subedi (1999) identification of two major factors, namely predication about the nature of the event in relation to the subject and focus provides a more fruitful avenue for potential explanations. Looking at examples like (35), taken from Verbeke (2013b: 594), it is not clear whether identifying the ergative with focus is the right analysis; however, information structural concerns such as topicalization do seem to be implicated. Indeed, the use of the ergative to express information structure has been documented in the region (Hyslop 2010). (35) a. bahırɑ ke-​ko khabal? outside what-​Gen noise ‘What’s the noise about outside?

Nepali

b. kɑrmı-​hɑru=le chana halı-​rɑh-​e-​ch-​an worker-​Pl=Erg roof lay-​Prog-​Pres-​3.Pl ‘The workmen are laying the roof.’

Nepali

As in Hindi/​Urdu, Nepali also appears to contain an alternation between the dative and ergative in modal expressions. Verbeke (2013a: 152) provides the examples in (36). While they do not constitute a minimal pair, they do provide evidence that the use of ergative/​dative alternations with expressions of modality are not a quirk of Hindi/​Urdu, but appear to reflect a more general pattern that needs to be accounted for. (36) a. mai=le mantri humu Pron.1.Sg=Erg minister be ‘I should be a minister.’

pɑr-​ch-​a fall-​NonPast-​3.Sg

b. hɑmı=lai tɑ jhɑn rıttai hınnu pɑr-​ch-​a ke Pron.1.Pl=Dat more empty walk fall-​NonPast-​3.Sg what ‘At least we can walk empty [without a load].’

Nepali

Nepali

Butt and Poudel (2007) provide this minimal pair, where the ergative alternates with the nominative. The use of the ergative expresses an obligation, or, as Abadie (1974) puts it, an “internalized necessity.”

 

828   Miriam Butt (37) a. ma aṛṛaa jaa-​nu par-​ch-​a Pron.1.Sg office go-​Inf fall-​NonPast-​3.Sg ‘I will have to go to the office.’

Nepali

b. mai=le aṛṛaa jaa-​nu par-​ch-​a Pron.1.Sg=Erg office go-​Inf fall-​NonPast-​3.Sg ‘I must absolutely go to the office.’

Nepali

In sum, the ergative in Nepali is obligatory with transitive agentive verbs when they carry past/​perfective morphology. There is optionality with unergatives and in non-​past tenses and this optionality appears to be conditioned by a number of semantic factors. In terms of agreement, Nepali illustrates yet another pattern of the interaction between case and agreement as the verb always agrees with ergative and nominative subjects, but not with other non-​nominative subjects.

33.3.3 Some Further Patterns The Indo-​Aryan languages and the greater South Asian language area offer up a rich area for research on the distribution of ergativity. The discussion so far has only presented a small segment of the overall picture. Existing overviews such as Masica (1976), Klaiman (1987), Masica (1990), Deo and Sharma (2006), and Verbeke (2013a) can provide the interested reader with readily accessible further data. However, the bulk of the South Asian patterns remain unexplored or only partially explored and documented. Much more field work is needed, in conjunction with formal analyses of the wide variety of patterns that are found. Table 33.5 provides an overview of the case forms of ergatives, datives, and genitives in a subset of the Indo-​Aryan languages.11 There are several noteworthy patterns that I would like to point out with respect to some of the languages. Table 33.5 Case markers across Indo-​Aryan Dative

Ergative

Instrumental

Genitive

Hindi/​Urdu Gujarati

ko ne

ne -​e

se -​e

Haryani Marathi Nepali Punjabi Sindhi Bengali Oṛiya Assamese

nae/​kae la lai nũ khe ke ku -​k

nae ne/​ni le ne -​e/​-​an/​-​in —​ —​ -​e

kae ne/​ni le kolõ -​e te -​e -​re

ka/​ki/​ke no/​ni/​nu/​ na/​nã ki (?) ca/​ci/​ce ko/​ka/​ki da/​di/​de jo/​ji/​je/​ja/​jyu -​(e)r -​rᴐ -​r

11  The forms are mainly taken from Masica (1991). The Assamese forms are from Devi (1986), the Haryani from Khan (2009).

 

Hindi/Urdu and related languages    829 Gujarati has an interesting agreement pattern in that it never agrees with a case marked subject, but always agrees with the object, regardless of whether this is case marked or not. This is illustrated in (38), taken from Deo and Sharma (2006: 379). (38) a. sita-​e kāgal Sita.F-​Erg letter.M.Nom ‘Sita read the letter.’

vāc-​yo read-​Perf.M.Sg

Gujarati

b. sita-​e raj-​ne pajav-​yo Sita.F-​Erg Raj.M-​Acc harass-​Perf.M.Sg ‘Sita harassed Raj.’

Gujarati

c. raj-​e sita-​ne Raj.M-​Erg Sita.F-​Acc ‘Rāj harassed Sita.’

Gujarati

pajav-​i harass-​Perf.F.Sg

The examples above also show that Gujarati uses the ne as a dative/​accusative marker. This is very likely the same ne that is used as an ergative in Hindi/​Urdu, Punjabi, Marathi and Haryani. Haryani moreover has been reported to use the same form, variously given as nae or ne for the ergative (39) and the dative/​accusative (40). (39) saḍh nae buḍhiaa ki jhũpṛii kii kun mae laat maaryi Sadhu Erg old-​lady Gen cottage Gen corner in leg hit ‘The Sadhu kicked the corner of the old lady’s cottage.’ (Singh 1970: 180) Haryani (40) yaah bi raam pyaarii this.Pl too Ram Piyari ‘Give these to Ram Piyari too.’

nae Dat

e Particle

de diye give give.Imp (Singh 1970: 180) Haryani

Shirani (1987) furthermore reports data in which both ergative and dative/​accusative appear in the same clause. This is illustrated in (41), (41) mɑn=ne sɑhɑb=ne Pron.1.Sg=Acc/​Dat Sahib.M.Sg=Erg ‘The Sahib hit me.’

mar-​a hit-​Perf.M.Sg

(Shirani 1987) Haryani

On the basis of the diachronic and synchronic patterns available in Indo-​Aryan, Montaut (2009) and Butt and Ahmed (2011) argue for an explicit connection between datives and ergatives. However, there is more to be said than just establishing a relationship between datives and ergatives. An inspection of Table 33.5 (and the larger patterns found in Indo-​Aryan) show that the same handful of forms keep cropping up in the languages (e.g. kX, lX, nX), but that they do not necessarily fulfill the same case function. These forms also do not occur in all the languages. For example, the Eastern languages Bengali and Oṛiya do not employ an ergative at all. The closely related language

 

830   Miriam Butt Assamese, on the other hand, still has vestiges of an ergative system. All of the kX, lX, nX forms have been innovated from about 1100 ad onwards. The -​e and other inflections found in Gujarati, Sindhi, and Assamese, for example, are vestiges of the Old Indo-​ Aryan case marking system. These vestiges can be found in older stages of Bengali and Oṛiya, but they have disappeared from the modern languages. Bengali also shows only person and number agreement and always agrees with the subject, see (42), taken from Deo and Sharma (2006: 381). (42) a. āmī sita-​ke dekh-​chī I.Nom Sita.F-​Acc see-​Pres.1.Sg ‘I see Sita.’

Bengali

b. āmī sita-​ke dekh-​lām I.Nom Sita.F-​Acc see-​Perf.1.Sg ‘I saw Sita.’

Bengali

c. anu sita-​ke dekh-​lo Anu.F.Nom Sita.F-​Acc see-​Perf.3.Sg ‘Anu saw Sita.’

Bengali

The Bengali pattern of agreement is of exactly the type found in Nepali. But Nepali has an ergative while Bengali does not. How can these patterns be reconciled? As already mentioned, Deo and Sharma (2006) propose an Optimality Theoretic account for the variation found in Indo-​Aryan with respect to case and agreement. However, there is no deeper explanation for why this variation should be found. My own intuition is that the deeper explanation lies in understanding and factoring in the semantics expressed by the case marking systems across the languages. We are at a point where this can be done, given that distracting arguments about case alignment and deep ergativity are being laid to rest and more fruitful perspectives on ergativity crosslinguistically are being developed (Legate 2008, 2012a; Coon 2013a).

33.4 Summary The goal of this chapter has been to provide a brief guide to ergativity as found in modern Indo-​Aryan languages. The discussion revolved mainly around Hindi/​Urdu, which is the language that has been worked on most extensively, but the discussion included some pointers to further patterns and languages. It should have become clear that a mixture of structural and semantic factors govern the appearance of the ergative in Hindi/​Urdu and related languages and that the interaction between these structural and semantic considerations is complex and in need of intensive further research.

 

Hindi/Urdu and related languages    831

Abbreviations ‘-​’, stands for a morpheme boundary; ‘=’, for a clitic boundary. 1, first person; 2, second person; 3, third person; Acc, accusative; Dat, dative; Erg, ergative; F, feminine; Fut, future; Gen, Genitive; Hon, Honorfic;ss Imp, Imperative; Impf, imperfective; Inf,infinitive; Inst, instrumental; M, masculine; Nom, nominative; NonPast, Non Past; Obl, oblique; Perf, perfective; Pl, plural; Pres, present; Prog, Progressive; Pron, pronoun; Sg, singular.

 

Chapter 34

E rg ativit y in I nu k t i t u t Richard Compton

34.1 Introduction This chapter examines the phenomenon of ergativity in Inuktitut, with particular focus on the manifestations of ergativity in the case system and the agreement morphology of the language. While the ergative-​absolutive case alignment in Inuktitut is often presented in the literature in the classic frame of an ergative-​marked agent and absolutive-​marked patient, data from ditransitives and causatives are used to highlight the structural nature of absolutive case assignment. Further indicative of the syntactic nature of ergativity, the agreement morphology in the language is shown to follow an ergative pattern insofar as the exponents of patient j-​features in transitives also index the subjects of intransitives. The use of the antipassive construction as a type of differential object marking for indefiniteness is also discussed. The rest of the chapter is organized as follows. Section 34.2 provides a brief overview of Inuktitut syntax. In section 34.3 I present the ways in which ergativity in Inuktitut is manifested in the language in terms of case marking and verbal agreement. Next, section 34.4 presents properties of the antipassive construction, including work by Johns (2001a, 2006) on dialectal variation in the use of the antipassive and Spreng’s (2012) proposal for a unified analysis of the antipassive and inceptive markers. I conclude in section 34.5.

34.2  Background on Inuktitut Inuktitut is part of the Eskimo-​Aleut language family, whose two main branches include Aleut, spoken on the Aleutian islands in Alaska, and the Eskimoan branch, which includes the Yupik and Inuit sub-branches. While Yupik languages are spoken in Alaska

 

Ergativity in Inuktitut    833 and Russia, Inuit forms a dialect continuum from Alaska, across the Canadian Arctic, to Greenland. Inuktitut typically refers to the Eastern Canadian portion of this Inuit dialect continuum:1 Eskimo-Aleut

(1)

Aleut

Eskimo

Yupik Languages

Inuit

Alaskan Iñupiaq

Western Canadian

Eastern Canadian

Greenlandic

Inuktitut, like all of Inuit, is highly polysynthetic. A  number of verbs trigger obligatory noun incorporation of their objects, a closed class of modifiers appear only inside nominal and verbal complexes, as do restructuring verbs, there is rich agreement, and virtually all grammatical markers (e.g. negation, modals, tense, aspect, passive, causative) are found only word-​internally. The polysynthetic character of the language is exemplified by the following single-​word utterance from the South Baffin dialect:2 (2) Iqalliariaqtuqattalauqsimagaluarivungattauq. iqalliaq-​riaq-​tu-​qattaq-​lauq-​sima-​galuaq-​gi-​vu-​nga=ttauq to.fish-​go.to.V-​very-​regularly-​dist.past-​perf-​although-​also-​indic.intr-​1sg=too ‘Although I also used to go fishing a lot too.’

1  English and French speakers in Canada often refer to any Canadian dialect in the continuum as Inuktitut, while the speakers of Western dialects typically use the names of their dialects or dialect groups (e.g. Inuinnaqtun). Similarly, the term Inuit is most frequently used to refer to the people. However, a number of researchers (Dorais 2003; Fortescue, Jacobson, & Kaplan 2010; MacLean 2014) as well as the government of Nunavut also use Inuit to refer to the dialect continuum as a language. 2  The literature of Eskimoan generally considers only the last element in this word to be an enclitic, as evidenced by the fact that it, and other enclitics, are always at the right edge of words and are flexible with respect to the category of the constituent to which they attach.

 

834   Richard Compton The language is both head-​marking and dependent-​marking. For instance, possession is marked by ergative case on the possessor as well as obligatory possessor agreement on the possessee indicating the number and person of the possessor:3 (3) Jaani-​up ataata-​nga John-​erg.sg father-​poss.3sg(abs.sg) ‘John’s father’ Similarly, as we will see in the next section, the language indicates grammatical relations simultaneous on heads and dependents by marking agreement with both ergative and absolutive arguments on verbs as well as employing case marking on nouns. The language exhibits what has been ‘radical pro-​ drop’ in the literature (e.g., Neeleman and Szendrői 2007). Pronouns are often avoided when their referents are recoverable from agreement or context, as in (4) where both subject and object pronouns are omitted. (4) Taku-​va-​git see-​indic.tr-​1sg.2sg ‘I see you.’ Furthermore, there are no dedicated third person pronouns, with demonstratives being used instead. Word order in the language is highly variable, to the point where one might consider Inuktitut to be a discourse configurational language. For instance, Gillon (1999: 20) notes that all of the six logically possible orders of the following sentence are possible.4 (5) Word Order a. Suusan Taivit-​mik Susan(abs.sg) David-​instr.sg ‘Susan loves David.’ b.

Taivit-​mik Suusan nagligusuk-​pu-q

c.

Taivit-​mik nagligusuk-​pu-q Suusan

d.

nagligusuk-​pu-q Taivit-​mik Suusan

e.

nagligusuk-​pu-q Suusan Taivit-​mik

f.

Suusan nagligusuk-​pu-q Taivit-​mik

nagligusuk-​pu-​q. love-​indic.intr-​3sg

(North Baffin)

3  Unless otherwise indicated, examples are the South Baffin subdialect of Inuktitut and were elicited by the author. 4  Glossing has been modified to match the conventions herein.

 

Ergativity in Inuktitut    835 However, as demonstrated in the examples throughout, there is a significant tendency towards SOV and SVO word orders (see also Sherkina-​Lieber 2004 on word order in wh-​questions in Inuktitut). In the next section I present the ways in which ergativity is manifested in the language, first in the case-​marking patterns found in the language, and then in terms of verbal agreement morphology.

34.3  The Exponence of Ergativity 34.3.1 Case marking Inuktitut exhibits an ergative-​absolutive case alignment, with agents (A)  of transitive verbs bearing ergative case and both patients of transitives (P) and also subjects of intransitives (S)  taking absolutive case, which is null for singular unpossessed arguments.5 (6)

Transitive a. Arna-​up niri-​ja-​nga aapu. woman-​erg.sg eat-​decl.tr-​3sg.3sg apple(abs.sg) ‘The woman is eating the/​an apple.’ b. Kina Jaani-​up taku-​lauq-​pau? who(abs.sg) John-​erg.sg see-​dist.past-​interr.tr.3sg.3sg ‘Who did John see?’

(7)

Intransitive a. Arnaq pisuk-​tu-​q. woman(abs.sg) walk-​decl.intr-​3sg ‘The woman is walking.’ b. Qimmi-​kka ani-​qqau-​ju-​it. dog-​poss.1sg.abs.pl go.out-​rec.past-​decl.intr-​3pl ‘My dogs went out.’

As illustrated in (6), the agent in a transitive clause (excluding the antipassive, which is discussed in section 34.3) always bears ergative case. The subject of an intransitive, as in (7), always bears absolutive case, regardless of the status of the verb as unergative or

5  While singular unpossessed ergative is marked with -​up and the absolutive singular is null, the unpossessed dual and plural forms do not distinguish between the ergative and absolutive. However, almost all of possessed forms distinguish between ergative and absolutive across all numbers.

 

836   Richard Compton unaccusative.6 For instance, in (8) the subject of the intransitive verb innginguaq ‘sing’ must be in the absolutive, with ergative case on the subject being ungrammatical, unless the verbal morphology is changed to make it transitive. In this case a specific referential patient is interpreted, as shown by the transitive form of the mood marker and the object agreement in (8-​c). (8) Unergative a. Arnaq innginguaq-​tu-​q woman(abs.sg) sing-​decl.intr-​3sg ‘The woman is singing.’

(Nunavik)

b. *Arna-​up innginguaq-​tu-​q woman-​erg.sg sing-​decl.intr-​3sg Intended: ‘The woman is singing.’ c.

Arna-​up innginguaq-​ta-​nga woman-​erg.sg sing-​decl.tr-​3sg.3sg ‘The woman is singing it.’7

As expected given the non-​agentive nature of the subject of unaccusative verbs, their subject must also be in the absolutive, as illustrated with the verb tikit ‘arrive’ in (10).8 (10) Unaccusative a. Qangattajuuk tiki-​kainna-​tu-​q airplane(abs.sg) arrive-​rec.past-​decl.intr-​3sg ‘The plane arrived.’

(Nunavik)

b. *Qangattajuu-​p tiki-​kainna-​tu-​q airplane-​erg.sg arrive-​rec.past-​decl.intr-​3sg Intended: ‘The airplane arrived.’ In simple monotransitives, the patient always bears absolutive case. However, if we expand our view to include ditransitives and causatives, we find that patients do not 6 

As noted by Allen (1996: 15), there is no language-​internal test for the unergative/​unaccusative distinction. However, Spreng (2012: 23) proposes that verbs whose sole argument in the intransitive is semantically an agent and which do not require the antipassive marker in the antipassive construction are unergatives. 7  A speaker offered this sentence as a repair to the ungrammatical ­example 8b and said that this could be used as an answer to whether the woman was singing a particular song. 8  In response to the ungrammatical example in (10) where the subject bears ergative case, a speaker offered the following repair in which the unacceptable ergative-​marked argument was transformed into an oblique and the subject was interpreted as pro-​dropped: (9) Qangattajuu-​kkut tiki-​kainna-​tu-​q airplane-​vialis.sg arrive-​rec.past-​decl.intr-​3sg ‘It arrived by plane.’

 

Ergativity in Inuktitut    837 always receive absolutive, with either goal or causee arguments appearing in it instead. For instance, and as observed by Creider (1978) for the Kivallirmiut dialect, while the patient in the ditransitive in (11-​a) bears absolutive case, absolutive can instead be assigned to the goal argument, as in (11-​b), with the patient instead receiving an oblique case.9 However, both may not bear absolutive simultaneously, as shown in (11-​c). (11) Ditransitive a. Jaani-​up aapu tuni-​ja-​nga Miali-​mut. John-​erg.sg apple(abs.sg) give-​decl.tr-​3sg.3sg Mary-​allat.sg ‘John gave the apple to Mary.’ b.

Jaani-​up tuni-​ja-​nga Miali John-​erg.sg give-​decl.tr-​3sg.3sg Mary(abs.sg) ‘John gave Mary the apple.’

aapu-​mit. apple-​obl.sg

c. *Jaani-​up tuni-​ja-​nga Miali aapu. John-​erg.sg give-​decl.tr-​3sg.3sg Mary(abs.sg) apple(abs.sg) Intended: ‘John gave Mary the apple.’ Similarly, in the causativized version of an intransitive or antipassive, it is the causee argument that takes absolutive case, even when both patient and goal arguments are also present, as in (12-​b): (12) Causative of intransitive or antipassive a. Ani-​ti-​qqau-​ja-​ra angunasukti. go.out-​cause-​rec.past-​decl.tr-​1sg.3sg hunter(abs.sg) ‘I made the hunter leave.’ b. Angunasukti tuni-​si-​ti-​qqau-​ja-​ra hunter(abs.sg) give-​ap-​cause-​rec.past-​decl.tr-​1sg.3sg arnar-​nut. woman-​allat.pl ‘I made the hunter give books to the women.’

uqalimaagar-​nit book-​obl.pl

While some dialects only permit intransitive and antipassive forms to be causativized (see Smith 1982b on Labrador Inuttut), other dialects also permit causatives of transitive stems (Johns 2001b). In these varieties the patient argument may take absolutive case and participate in object agreement, with the causee taking allative case, as shown in the following examples from Johns (1987:  13)  and Jenson and Johns (1989:  211), respectively. 9  While many dialects preserve the contrast between ablative case and instrumental cases, e.g. abl. sg -​mit and instr.sg -​mik, the dialect of the speaker from South Baffin collapses the two. I use the label obl(ique) for this neutralized case.

 

838   Richard Compton (13)

Causative of transitive a. Jaaani-​up Miuri-​mut ipuittuq angmaq-​ti-​ta-​a John-​erg.sg Mary-​allat.sg can(abs.sg) open-​cause-​decl.tr-​3sg.3sg ‘John made Mary open the can.’ b. Nutara-​up arna-​mut angut aktuq-​ti-​taa child-​erg.sg woman-​allat.sg man(abs.sg) touch-​cause-​decl.tr-​3sg.3sg ‘The child made a woman touch the man.’

To summarize the distribution of absolutive case, absolutive is assigned once and only once in every clause. It is only when another clause is added that an additional absolutive argument is possible, either overtly as in (14), or as a pro-​dropped argument as in (15).10 (14) Jaani uqa-​qqau-​ju-​q Miali John(abs.sg) say-​rec.past-​decl.intr-​3sg Mary(abs.sg) tuktu-​viniq-​tu-​qqau-​mmat. caribou-​former-​consume-​rec.past-​because.intr.ds.3sg ‘John said (earlier today) that Mary was eating caribou meat (earlier today).’ (15)

Pisu-​raja-​qati-​gi-​ɬuniuk walk-​go.for-​partner-​have.as-​contemp.ss.3sg.3sg tuki-​qqau-​ja-​nga. kick-​rec.past-​decl.tr-​3sg.3sg ‘While she was walking it, she kicked it (earlier today).’

It is worth noting that while it has been proposed that absolutive is assigned by T (e.g. Bok-​Bennema 1991, Campana 1992, Murasugi 1992, and Ura 2001), polysynthetic verbal complexes which appear to contain an embedded TP cannot license additional absolutive arguments. For instance, we can contrast (14) with (16), in which the TP-​ incorporating restructuring verb niraq ‘say’ is marked as recent past, while the embedded and incorporated verb niri ‘eat’ is marked as near future.11 And yet, among the three arguments of these two verbs, only one may be absolutive.

10 

That absolutive has been assigned to the unpronounced patient arguments of both verbs in (15) is supported by the concomitant object agreement, as will be shown in the next section. 11  While West Greenlandic has been argued to lack tense, and while this also appears to hold of Western Canadian Inuit dialects such as Kangiryuarmiut (at least for the past/​nonpast distinction), Hayashi (2011) demonstrates that South Baffin Inuktitut has an obligatory (and rather complex) system of tense distinctions. My experience with the North Baffin, Nunavik, and Aivilik dialects leads to be believe that this claim extends to Eastern Canadian Inuit generally, although to my knowledge no one has tested this in each dialect subgroup.

 

Ergativity in Inuktitut    839 (16) Saila-​up niri-​niar-​niraq-​qqau-​ja-​nga Alana Saila-​erg.sg eat-​near.fut-​say-​rec.past-​decl.tr-​3sg.3sg Alana(abs.sg) tuktu-​viniq-​mit caribou-​former-​obl.sg ‘Saila said (earlier today) that Alana will be eating caribou meat (later today).’ The contrast between (14) in which there are two sets of agreement morphemes—​one on each verb—​and (16) in which there is a single agreement morpheme, while nevertheless maintaining two instances of tense, is particularly revealing. Instead of there being a one-​to-​one correlation between tense-​marking (which is assumed to be a manifestation of the T head) and absolutive case, the correlation in Inuktitut appears to be with verbal agreement. Each verbal complex (or each CP as argued by Compton & Pittman 2010) licenses a single absolutive argument. This is further exemplified by the fact that clauses bearing certain dependent moods, such as the contemporative mood of the first cause in (15), lack the ability to mark tense (and are thus similar to non-​finites), but can nevertheless license an absolutive argument through agreement. The obligatory presence of absolutive case in every (full CP) clause in Inuktitut coincides with analyses in the literature that absolutive is an obligatory structural case (Bobaljik 1993a; Bittner and Hale 1996a,b; Legate 2008). An analysis of ergativity specifically addressing case-​assignment in the closely related dialect of West Greenlandic is proposed by Bittner and Hale (1996a,b).12 Working in the Government and Binding framework, they propose that case is assigned in a structural configuration cross-​linguistically, with both ergative and accusative cases being assigned in the presence of a “case-​competitor” within the same government domain. Nominative (which for them subsumes absolutive), on the other hand, is treated as a caseless DP or NP, lacking the KP layer of overtly-​marked cases. The difference, then, between ergative-​absolutive languages such as Inuit and nominative-​accusative ones, is that the latter have an adjoined D in the VP which acts as a case-​competitor, licensing accusative on the patient, while preventing the patient from serving as a case-​competitor for the agent, thus blocking the assignment of ergative. In their system, ergative is assigned in a structural configuration relative to INFL and nominative (caselessness) is licensed by C, although they are careful to separate the assignment of case from agreement on functional heads, proposing that the two operate independently. To support their analysis, Bittner and Hale point to the relative order of agreement affixes in subordinate moods in West Greenlandic (in cases where these are not fused), showing that the exponence of subject agreement is internal to that of object agreement, as expected given the Mirror Principle assumption that INFL is closer to the verb than C (see Baker 1985). For instance, they give the following pair of examples showing the relative order of agreement markers in the becausative mood (which they gloss as dependent past). 12 

The label “Inuit” used by Bittner and Hale is somewhat unfortunate since they are specifically dealing with data from West Greenlandic and the details of case-​marking—​as well as those of agreement and scope which they use as evidence in support of their analysis—​actually vary across the dialect continuum. In many respects, West Greenlandic is an outlier within Inuit and should not be taken to be representative of the continuum.

 

840   Richard Compton (17)

West Greenlandic (Bittner and Hale 1996a: 18, original glossing) a. Juuna nuannaar-​a-​mi _​_​ miiqqat [Juuna(nom)i happy-​dpst-​3sg.proxi] pro(erg)i children(nom)j kunip-​p-​a-​i. kiss-​ind-​[+tr]-​3sgi.3plj ‘Because Juunai was happy, hei kissed the children.’ b. Juuna-​p miiqqat taku-​ga-​mi-​git _​_​ [Juuna-​ergi children(nom)j see-​dpst-​3sg.proxi-​3plj] pro(nom)i nuannaar-​p-​u-​q. happy-​ind-​[–​tr]-​3sgi ‘Because Juunai has seen the children, hei’s happy.’

As well, they point to the presence of object agreement and the lack of subject agreement in nonfinite clauses, which nevertheless project a full clause, since they bear mood, as further suggestive that object agreement is on C.13 Bittner and Hale’s (1996a) analysis also extends to the nominal domain, proposing that D, the parallel of INFL in the extended projection of nouns, is responsible for assigning ergative case to possessors. Their later (1996b) paper goes further, tying their proposal to data from scope, agreement, and switch-​reference marking in West Greenlandic. While a full treatment of the analyses proposed by Bittner and Hale (1996a,b) is beyond the scope of this work, several points in their analyses as applied to Inuktitut are problematic. While West Greenlandic has been argued to lack tense, Hayashi and Spreng (2005) and Hayashi (2011) show that Inuktitut, in contrast, has genuine tense. Crucially, tense intervenes between the verb and its agreement morphology in Inuktitut (along with negation and aspect), suggesting that agreement morphology may not be on INFL, as they propose, unless separate INFL and T heads are assumed. A related issue involves the fact that the exponence of both subject and object agreement frequently co-​vary with mood, but not tense, again suggesting that agreement and case-​assignment higher in the clausal structure than they predict. The scope facts used by Bittner and Hale (1996b) as support for their analysis are also distinct in Inuktitut. In their analysis the ergative-​marked argument remains low inside the VP and is assigned case by INFL, and this low position is supported by the availability a narrow scope reading with respect to negation (although they propose it moves at PF to derive the default SOV word order). However, unlike in West Greenlandic, Wharram (2003) shows that in Inuktitut an ergative-​marked agent always takes wide scope with respect to negation, again suggesting it occupies a higher position. In the next subsection I present how ergativity is manifested in the verbal agreement morphology of the language. 13  However, Dorais (1988: 95–​96) presents alternative forms in the Nunavik dialect of Inuktitut that do in fact agree with both the subject and the object, thus calling into question the analysis of such forms by Bittner and Hale as being nonfinite.

 

Ergativity in Inuktitut    841

34.3.2 Agreement In addition to an ergative pattern of case marking, the language also displays an ergative alignment in its verbal agreement morphology. In a transitive sentence such as (18), the ergative-​absolutive case alignment must co-​occur with both the transitive form of the mood marker and agreement morphology that is conditioned by the j-​features of both the agent and the patient. Neither the intransitive form of the mood marker, nor an intransitive form of agreement indexing only the subject are possible. (18) Transitive agreement a. Arna-​up niri-​ja-​nga aapu. woman-​erg.sg eat-​decl.tr-​3sg.3sg apple(abs.sg) ‘The woman is eating the/​an apple.’ b. *Arna-​up niri-​ja-​q aapu. woman-​erg.sg    eat-​decl.tr-​3sg apple(abs.sg) c. *Arna-​up niri-​ju-​nga aapu. woman-​erg.sg eat-​ decl.intr-​3sg.3sg apple(abs.sg) d. *Arna-​up niri-​ju-​q aapu. woman-​erg.sg eat-​decl.intr-​3sg apple(abs.sg) Conversely, the form of agreement on an intransitive verb depends on the j-​features of the subject, as illustrated in (19) where the agreement covaries with the number of the subject. (19) Intransitive agreement a. Angutik palla-​tu-​q. man(abs.sg) trip-​decl.intr-​3sg ‘The man tripped.’ b. Qimmi-​kka ani-​qqau-​ju-​it. dog-​1sg.poss.abs.pl go.out-​rec.past-​decl.intr-​3pl ‘My dogs went out.’ The ergative nature of this conditioning is illustrated in Table 34.1, which summarizes the forms of agreement in the declarative mood in the Nunavik dialect of Inuktitut (adapted from Dorais 1988).14 Forms are presented in a phonemic IPA transcription.15 14 

In the literature of Eskimoan languages this mood is often called participial, since it can serve a nominalizing function. In West Greenlandic verbs in this mood cannot stand as the main predicate in a clause, but in Canadian dialects it functions as the default declarative mood. 15  While the normal orthography is quite close to a phonemic transcription, I use the IPA here to avoid any possible confusion that may result from the [ŋ] digraph.

 

842   Richard Compton Table 34.1 Agreement in declarative mood, Nunavik dialect Transitive

Patient

Agent

1sg

1du

1pl

2sg

2du

2pl

3sg

3du

3pl

1sg 1du 1pl 2sg 2du 2pl 3sg 3du 3pl

—​ —​ —​ ʁma ttiŋa tsiŋa aŋa aŋa aŋa

—​ —​ —​ ttiɣuk ttiɣuk ttiɣuk atiɣuk atiɣuk atiɣuk

—​ —​ —​ ttiɣut ttiɣut ttiɣut atiɣut atiɣut atiɣut

ɣit ttiɣit ttiɣit —​ —​ —​ atit atit atit

ttik ttik ttik —​ —​ —​ atik atik atik

tsi tsi tsi —​ —​ —​ asi asi asi

ʁa vuk vut it tik si ŋa ŋak ŋat

akka avuk avut akkik atik asi aŋik aŋik aŋik

kka vuk vut tit tik si ŋit ŋit ŋit

ɣuk

ɣut

tit

tik

si

q

uk

t

Intransitive ŋa subject

While a number of the transitive agreement forms appear to contain within them the intransitive agreement forms, it is notable that the similarities tend to correlate with the patient j-​features of the transitive forms. For instance, the transitive 2 {sg/​du/​pl}.1du form [ttiɣuk] and the 3 {sg/​du/​pl}.1du form [atiɣuk] both contain a string identical to the intransitive subject agreement 1du form [ɣuk]. This pattern, which holds for most of the transitive forms with first and second person patients, is crucially an ergative patterning: the same exponent that indexes the patient argument of transitives indexes the subjects of intransitives. Furthermore, insofar as the portions of the agreement forms that are not shared between the transitive and intransitive appear only in the presence of an ergative agent, these also constitute an ergative agreement pattern. For instance, transitive forms indexing a third person agent (and first or second person patient) all begin with [a]‌. The transitive forms indexing a third person patient are distinct in that they do not seem to bear the same resemblance to the intransitive forms that was observed for the first and second person patient forms discussed above. This is due to the fact that third person patient forms in the declarative mood were originally (absolutive) possessor agreement forms occurring obligatorily on possessed nouns. For instance, in his grammar of the phonologically conservative Kangiryuarmiut dialect of Inuinnaqtun (Western Canadian Inuit), Lowe (1985: 111) presents the one-​ to-​one match of the possessor agreement that obligatorily appear on possessees and the transitive agreement forms with third person patients, as summarized in Table 34.2.

 

Ergativity in Inuktitut    843 Table 34.2 Sample comparison of possessor marking and third person patient forms in the Kangiryuarmiut Inuinnaqtun dialect of Western Canadian Inuit Possessive suffixes

Transitive agreement

qaja-​ʁa qaja-​kka qaja-​tka

my canoe my two canoes my canoes (pl)

taku-​ja-​ʁa taku-​ja-​kka taku-​ja-​tka

I saw him/​her I saw them (du) I saw them (pl)

qaja-​n qaja-​kkin qaja-​tin

your (sg) canoe your (sg) two canoes your (sg) canoes (pl)

taku-​ja-​n taku-​ja-​kkin taku-​ja-​tin

you (sg) saw him/​her you (sg) saw them (du) you (sg) saw them (pl)

qaja-​a qaja-​ik qaja-​it

his/​her canoe his/​her two canoes his/​her canoes (pl)

taku-​ja-​a taku-​ja-​ik taku-​ja-​it

he/​she saw him/​her he/​she saw them (du) he/​she saw them (pl)

This pattern continues for plural possessors, showing that the third person patient agreement forms in the declarative mood have a distinct diachronic source from the rest of the paradigm. However, the distribution of these forms also constitutes in an ergative pattern insofar as they are reserved (in their use as verbal agreement markers) for clauses with an ergative-​absolutive case configuration. Thus, except for the forms originally repurposed from the possessor agreement paradigm, we observe an ergative patterning in the exponence of verbal agreement in Inuktitut. This ergativity is manifested in three ways. First, the agreement for intransitives matches the portion of the transitive forms that are (mostly) consistent across patients with corresponding j-​features. Second, the beginning portions of the transitive forms (as well as fused strings such as 2sg.1sg [ʁma]) occur only in the presence of an agent. Third, the system as a whole, including the forms derived from possessor agreement, only employs the transitive agreement paradigm in the ergative-​absolutive case alignment, and not with intransitives (or with antipassives as discussed in 34.4). In addition to ergativity being reflected in the agreement morphology of the language, the form of mood markers is also conditioned by case-​alignment insofar as only transitive clauses trigger the transitive forms of mood markers. For instance, in the above examples the transitive form of the declarative mood marker is ta or ja, after a consonant or vowel respectively, but the intransitive form is tu or ju. While at first this may seem to be purely a function of the transitivity of the verb, as we will see in the next section, semantically transitive verbs such as ‘eat’ take the intransitive forms of mood markers in the antipassive construction. For further background on the exponence and historical evolution of ergativity in Eskimo-​Aleut, see Fortescue (1995). In the next section I examine the properties of the antipassive construction in Inuktitut.

 

844   Richard Compton

34.4  The Antipassive In addition to transitive clauses exhibiting an ergative-​absolutive alignment, Inuktitut possesses a second bivalent construction: the antipassive. In an antipassive clause the case on the subject and the form of both mood-​marking and agreement morphology pattern with the intransitives presented above, and yet a (semantic) patient bearing oblique case is licensed. Some verbs, such as niri ‘eat’ in (20-​a) may be used in the antipassive without any special marking other than the intransitive mood and agreement forms, while others, such as kii ‘bite’ in (20-​b), require the overt antipassive marker si after the verbal root. (20) Antipassive a. Arnaq niri-​ju-​q aapu-​mit. woman(abs.sg) eat-​decl.intr-​3sg apple-​obl.sg ‘The woman is eating an apple.’ b. Qimmiq uvannit dog(abs.sg) 1.obl.sg ‘The dog bit me.’

kii-​si-​ju-​q. bite-​ap-​decl.intr-​3sg

Spreng (2012) argues that the appearance of overt antipassive marking on some verbs and not others is due to their argument structure. She shows that verbs that take require overt marking in the antipassive are either obligatorily transitive, such as quqiq ‘shoot’ in (22), or can alternate between being unaccusative and monotransitive, like surak ‘break’ in (23)—​in other words verbs whose (semantic) patient must always be expressed.16 (22) Obligatorily Transitive (South Baffin, Spreng 2012: 20, modified) a. Piita nanur-​mit quqir-​*(si)-​ju-​q Peter(abs.sg) polar.​bear-​obl.sg shoot-​*(ap)-​decl.intr-​3sg ‘Peter is shooting a polar bear.’ 16 

A possible exception to Spreng’s generalization that “verbs that have an obligatory internal argument require the AP marker” (p. 204) are the closed class of obligatory noun-​incorporation verbs in the language. Given that these verbs must incorporate their semantic patient, we might predict them to take the AP marker, and yet they do not: (21) Maqaitti illu-​liu-​lauq-​tuq. hunter(abs.sg) house-​make-​dist.past-​decl.intr-​3sg ‘The hunter made a house.’ For a comprehensive account of noun incorporation in Inuktitut see Johns (2007b) and references therein.

 

Ergativity in Inuktitut    845 (23) Unaccusative/​Transitive (North Baffin, Spreng 2012: 22, modified) a. Anautaq surak-​tu-​q stick(abs.sg) break-​decl.intr-​3sg ‘The stick broke.’ b. arnaq surak-​*(si)-​ju-​q woman(abs.sg) break-​*(ap)-​decl.intr-​3sg ‘The woman is breaking a stick.’

anautar-​mik. stick-​instr.sg

Conversely, she also shows that the verbs that do not require antipassive marking are unergatives which can optionally become transitive, like miqsuq ‘sew’ in (24)—​thus, verbs whose agents must always be expressed. (24) Unergative/​Transitive (North Baffin, Spreng 2012: 23, modified) a. Arnaq miqsuq-​tu-​q. woman(abs.sg) sew-​decl.intr-​3sg ‘the woman is sewing’ b. Arnaq miqsuq-​(*si)-​tu-​q woman(abs.sg) sew-​(*ap)-​decl.intr-​3sg ‘The woman is sewing a pair of pants.’

qarling-​nit. pant-​obl.du

As has been observed by Johns (2001a, 2006), the semantic effects of the antipassive vary across the Inuit dialect continuum. For instance, in the following antipassive and corresponding transitive examples from the Nunavik dialect, the translations provided by the speaker appear to correlate the antipassive with an indefinite interpretation of the object: (25) Definiteness/​Specificity of the object a. Annuraa-​nik uvva-​tu-​nga. clothing-​instr.sg wash-​decl.intr-​1sg ‘I am washing clothes.’

(Nunavik)

b. Annuraa-​t uvva-​ta-​kka. clothing-​abs.pl wash-​decl.tr-​1sg.3pl ‘I am washing the clothes.’ Such data suggest that the antipassive serves a differential-​object marking function, indicating that a patient is indefinite or perhaps non-​specific (see Aissen 2003b, de Swart 2007, references therein, and Johns and Kučerová, Chapter 17, this volume). However, as demonstrated by Johns, this phenomenon appears to vary across the dialect continuum. She observes that in Western dialects, names cannot appear as

 

846   Richard Compton the oblique-​marked patient in an antipassive construction, or yield an unexpected reading, presumably due to a clash between the inherent definiteness of proper names and a strong indefinite character of the either the oblique case in the antipassive construction.17 (27) Western dialects: AP incompatible with proper names a. *John tautuk-​tu-​q Mary-​mik. John(abs.sg) see-​decl.intr-​3sg Mary-​instr.sg Intended: ‘John sees Mary.’ (Iñupiaq, Manning 1996: 95) b. ?Alana-​mik Alana-​instr.sg ‘someone dressed up as Alana’ 

(Inuvialuktun, Johns 2001a: 133)

Notably, examples from Johns suggest that this definiteness is not an inherent property of the case marker itself (i.e. it does not appear that this case morpheme is simultaneously encoding an indefinite feature) since proper names with their expected definite interpretations can take the instrumental case in these dialects when serving as genuine instruments, as illustrated with the following example from Lowe (1985). (28) Ilruq niuvvaavi-​lia-​rami Uvvayua-​mik Ilruq(abs.sg) store-​travel-​because.ss.3sg Uvvayua-​instr.sg uqaqsiq-​tua-​q interpret-​decl.intr-​3sg ‘when Ilruq went to the store, she used Uvvayuaq as an interpreter’ (lit. she interpreted with Uvvayuaq) Conversely, despite the indefinite interpretation for common nouns illustrated in (25), Eastern dialects allow oblique-​marked names in the antipassive, as in (29).18

17  While Johns’s claim appears true for a number of Western dialects, at least one dialect, Kangiryuarmiut Inuinnaqtun, appears to be an exception. A speaker judged both the transitive and antipassive versions of the following sentence to be acceptable with a proper name as the internal argument:

(26) An exception to AP patient indefiniteness in a Western dialect (Kangiryuarmiut) a. Emili-​up Richard Emily-​erg.sg Richard(abs.sg) ‘Emily saw Richard.’

taku-​ya-​a. see-​decl.tr-​3sg.3sg

b. Emily Richard-​mik taku-​yu-​q. Emily(abs.sg) Richard-​instr.sg see-​decl.intr-​3sg ‘Emily saw Richard.’ 18 

Note that in the Labrador orthography uppercase is used to represent [q]‌and velars and uvulars are neutralized in coda position.

 

Ergativity in Inuktitut    847 (29) Eastern dialects: AP compatible with proper names a. Margarita Kuinatsa-​i-​ju-​k Ritsati-​mik. Margarita(abs.sg) tickle-​ap-​decl.intr-​3sg Richard-​instr.sg ‘Margarita is tickling Richard.’ (Labrador Inuttut, Johns 2006: 295) b. Jaani taku-​gunnaq-​paa Miali-​mit John(abs.sg) see-​can-​interr.intr.3sg Mary-​obl.sg ‘Can John see Mary?’

(South Baffin)

Johns (2006) argues that this variation across dialects is due to Western dialects maintaining a more prototypical ergative-​absolutive alignment, while Eastern dialects are undergoing a typological shift towards a more nominative-​accusative alignment. While the use of the antipassive in other languages is often connected to an aspectual split (see Coon 2013a and references therein), aspect is somewhat orthogonal to the antipassive in Inuktitut in that it is not the use of the transitive or antipassive construction that yields a perfective or progressive interpretation, but rather the aspectual properties of the verb itself, as observed by Hayashi & Spreng (2005). They observe and Hayashi (2011) further elaborates that durative verbs without tense marking receive a progressive interpretation, while punctual verbs receive a perfective interpretation. For instance, in the following examples, it is the verbal root that determines the aspectual interpretation, cutting across both constructions. In (30) the durative verb niri ‘eat’ yields a default progressive interpretation, while in (31) kii ‘bite’ yields a default recent past interpretation, regardless of whether the antipassive is used. (30) Durative = Progressive a. arna-​up niri-​ja-​nga woman-​erg.sg eat-​decl.tr-​3sg.3sg ‘The woman is eating the apple.’

aapu apple(abs.sg)

b. arnaq niri-​ju-​q aapu-​mit woman(abs.sg) eat-​decl.intr-​3sg apple-​obl.sg ‘The woman is eating an apple.’ (31) Punctual = Perfective a. qimmi-​up kii-​ja-​anga dog-​erg.sg bite-​decl.tr.3sg.1sg ‘The dog bit me.’ b. qimmiq uvannit kii-​si-​ju-​q dog(abs.sg) 1.obl.sg bite-​AP-​decl.intr.3sg ‘The dog bit me.’ However, Hayashi & Spreng (2005) and later Spreng (2012) also attribute an imperfective aspectual contribution to the antipassive marker, arguing that both it and a

 

848   Richard Compton homophonous morpheme that has typically been glossed as inceptive in the literature should receive a unified analysis. For instance, in the following example the contribution of si does not add the possibility of an oblique-​marked patient, but rather yields a present/​progressive interpretation to a punctual verb that would otherwise receive a perfective or immediate past interpretation in the absence of an overt tense-​marker. (32) Ani-​si-​ju-​q taku-​ga-​viuk. go.out-​incept-​decl.intr-​3sg see-​because-​2sg.3sg ‘He is leaving because/​since you saw him.’ While Spreng (2012) argues extensively for the antipassive marker being an aspectual morpheme, she also notes that only the inceptive version deletes a preceding consonant in North Baffin (p. 22). Furthermore, Spreng notes that both antipassive and inceptive si can occur simultaneously in a single verbal complex, as in (33) (p. 23). (33) Anguti kunik-​si-​si-​vuq man(abs.sg) kiss-​ap-​incept-​indic.intr-​3sg ‘The man starts to kiss a woman.’

arna-​mik. woman-​instr.sg

While her account treats such cases as the result of vocabulary insertion of the same morpheme into two distinct positions, a lower position for the antipassive and a higher aspectual head, using a Distributed Morphology approach, such data also seem compatible with an analysis in which the forms are merely homophonous, and perhaps share the same diachronic source. Given the extremely small phoneme inventory for Inuit of only three vowels and between thirteen and fourteen consonants—​roughly half the number of phonemes in English—​and given the frequent homophony of other grammatical morphemes in the language, the chances of a two-​phoneme-​long string being homophonous seem quite high. In addition, examples from Allen (1996) show that the Tarramiut Nunavik dialect possesses distinct forms for the two uses: tsi for the antipassive and si for what she glosses as present. For a more complete picture of her proposal and her arguments for treating the AP marker as an aspectual morpheme, I refer the reader to Spreng (2012). To summarize, the use of the antipassive construction Inuktitut serves as a form of differential object marking, with common noun oblique-​marked patients in the antipassive receiving an indefinite interpretation (cf. Spreng 2012).

34.5 Conclusion This chapter has presented the various ways in which ergativity is manifested in the nominal case marking and verbal agreement morphology of Inuktitut. In particular,

 

Ergativity in Inuktitut    849 the one-​to-​one correspondence between agreement on a verbal complex and the presence of a (sometimes pro-​dropped) absolutive argument, along with the impossibility of additional absolutive arguments when additional tense heads are present in a verbal complex, was taken to show that it is agreement, not tense, that licenses absolutive case in Inuktitut. This is further highlighted by the fact that clauses in subordinate moods that lack the ability to mark tense, and may thus be thought to a type of non-​finite, nevertheless license absolutive arguments because they bear agreement. The ergative nature of agreement morphology in Inuktitut was also highlighted. For a large portion of the declarative (and also indicative) paradigm, exponents of agreement indexing intransitive subjects also index patients of transitive clauses. Furthermore, a number of the exponents of verbal agreement are only present with the ergative-​ absolutive alignment, disappearing with intransitive verbs. Taken together, the ban against multiple absolutive arguments in a single clause, the possibility of non-​patient arguments receiving absolutive in ditransitive clauses, and the one-​to-​one relationship between verbal agreement and the presence of an absolutive argument point to an analysis of the absolutive as a structural case (as opposed to the morphological default of two structural case, as proposed by Legate 2008 for a subset of ergative languages). An analysis of absolutive as a single structural case and not a morphological default of distinct structural cases, in the sense of Legate’s (2008) ABS = DEF languages, further coincides with the observations by Creider (1978) and Johns (1987) showing that relativization also treats absolutive arguments as a syntactic class. The gap in such constructions is always in an absolutive case position, as illustrated in the following contrasting examples: (34) Angum-​mik [arna-​mik taku-​ju-​mik] taku-​vu-​nga man-​instr.sg [woman-​instr.sg see-​decl.intr-​instr.sg] see-​indic.intr-​1sg ‘I saw the man who saw the woman.’ (Kivallirmiut, Creider 1978: 100) (35) *Angut(-​up) [nanuq kapi-​ja-​a] ani-​ju-​q. man(-​erg.sg) polar.bear(abs.sg) stab-​decl.tr-​3sg.3sg go.out-​decl.intr-​3sg Intended: ‘The man who stabbed the polar bear went out.’ (S. Baffin, Johns 2007a) However, a lingering problem in analyzing the restriction on relativization to absolutive arguments as the result of syntactic ergativity is that the declarative/​participial mood used in these constructions is also used to nominalize verbs. For instance, the verb anijuq ‘he/​she went out’ in (35) can also mean ‘the one who went,’ as illustrated in (36). (36) Ani-​ju-​up niri-​ja-​nga go.out-​decl.intr-​erg.sg eat-​decl.tr-​3sg.3sg ‘The one who went out is eating the apple.’

aapu. apple(abs.sg)

 

850   Richard Compton Given the ability of verbal complexes in declarative/​participial mood to act as nominals, this calls into question whether they are in fact relative clauses, or whether they could be nominalizations in apposition with the nouns they modify, as suggested by the speaker’s spontaneous translation for the following example: (37) Ani-​ju-​up arna-​up niri-​ja-​nga aapu go.out-​decl.intr-​erg.sg woman-​erg.sg eat-​decl.tr-​3sg.3sg apple(abs.sg) ‘The woman who went out is eating the apple.’ (Speaker: “The one who went out, the girl, is eating the apple.”) As such, it is not clear whether the restrictions on relative clauses are the result of ergativity at the level of the clause or an ergative pattern of nominalization. Finally, this chapter briefly outlined the functioning of the antipassive in Inuktitut. While typically described in the literature as marking an object as indefinite or non-​ specific, Spreng’s (2012) proposal to treat the AP marker as an aspectual morpheme is particularly interesting given the growing body of work connecting case alignment splits in ergative languages to aspect (e.g. Polinsky 2013, Coon 2013a).

Acknowledgments My deepest thanks to the late Saila Michael, a speaker of the South Baffin dialect, who generously and patiently shared her language with me for over a decade and provided much of the original data presented here. I am greatly indebted to her. I would also like to thank Billy Meeko, a speaker of the Nunavik dialect, and Emily Kudlak, a speaker of the Kangiryuarmiut dialect of Western Inuit. Thank you to Alana Johns and Jessica Coon for their feedback on an earlier draft. This research was supported by a SSHRC Insight Development Grant.

Abbreviations 1, first person; 2, second person; 3, third person; a, agent; abs, absolutive case; allat, allative; ap, antipassive; because, becausative mood; c, complementizer; cause, causative; contemp, contemporative mood; d, determiner; cp, complementizer phrase; dist.past, distant past; dp, determiner phrase; decl, declarative mood; dpast, dependent past (becausative mood); du, dual; erg, ergative case; incept, inceptive; ind(ic), indicative mood; infl, inflection; instr, instrumental case; interr, interrogative mood; intr, intransitive; kp, case phrase; near.fut, near future; nom, nominative; np, noun phrase; o, object; obl, oblique case; p, patient; perf, perfect; pl, plural; pf, phonetic form; poss, possessive; prox, proximate; rec.past, recent past; s, subject; sg, singular; ss, same subject; t, tense; tp, tense phrase; tr, transitive; v, verb; vialis, vialis case.

 

Chapter 35

Ergativi t y i n Na kh–​D aghesta nia n Diana Forker

35.1 Introduction The Nakh–​Daghestanian (or East Caucasian) language family is the largest and the linguistically most diverse of the three autochthonous language families in the Caucasus. The more than 30 Nakh–​Daghestanian languages are primarily spoken in the southern parts of Russia and in northern Azerbaijan. A few speech communities are found in Georgia. The family can be divided into several subbranches: Nakh (Chechen, Ingush, Tsova-​Tush [Batsbi]), Avar-​Andic (Avar, Andi, Godoberi, Bagvalal, and some more), Tsezic (Tsez, Hinuq, Khwarshi, Hunzib, and Bezhta), Lezgic (Lezgian, Agul, Tsakhur, Kryz, Archi, Udi, and a few more), Khinalugh (sometimes grouped together with Lezgic), Dargi (traditionally considered to be one language, but consisting of several varieties that are mutually incomprehensible), and Lak (sometimes grouped together with Dargi). The languages have rather large consonant inventories that include ejectives. The morphology is agglutinating/​fusional. They are predominantly dependent-​marking, have ergative case and usually a vast array of spatial cases. In most Nakh–​Daghestanian languages gender is an important grammatical category. The languages have numerous finite and non-​finite verb forms (converbs, participles, infinitive, and masdar—​ a deverbal noun). The most common constituent orders are head-​final at clause as well as at phrase level. For recent overviews see van den Berg (2005) and Daniel and Lander (2011). Linguists working on languages of the Caucasus were among the earliest researchers to use the term ergative in its modern sense (cf. Dirr 1912, 1928). Languages that have been explored with respect to grammatical alignment and ergativity are Lezgian

 

852   Diana Forker (Job 1985; Mel’čuk 1988; Haspelmath 1991; Manning 1995), Agul (Ganenkov et  al. 2008), Archi (A. E. Kibrik 1979), Chechen (Nichols 1980; Molochieva and Witzlack-​ Makarevich 2008), Ingush (Nichols 2008), Tsezic languages (Comrie 2004; Comrie et al. 2011; Forker 2011), Avar (Crisp 1983), and Sanzhi Dargi (Forker in press). Furthermore, impressive comparative work has been done by Catford (1974), Kibrik (1979, 1985, 1997, 2001b, 2003), and Schulze (1988, 2000). The majority of scholars come to the conclusion that ergativity is largely restricted to morphology and that the syntax of Nakh–​Daghestanian languages shows, if any, more traces of accusativity than ergativity. Nichols’ (2008) paper on Ingush is an exception. She identifies a relatively large number of syntactically ergative traits and only very little accusative patterns. The surveyed properties and constructions differ from study to study so it is not completely surprising that the conclusions differ. Furthermore, since the investigated languages belong to different branches of Nakh–​Daghestanian we can expect some variation.

35.2  Valency Classes Before starting with the investigation of ergativity a few words on valency classes and some terminological clarifications will be helpful. Common valency classes in Nakh–​ Daghestanian are: intransitive (one single argument in the absolutive, abbreviated as S (1a) extended intransitive (one absolutive argument S and a further argument in a spatial case), (canonical) transitive (one ergative and one absolutive argument, abbreviated as A and P (1b) extended transitive (one ergative A, one absolutive P and one further argument in the dative or a spatial case) (8a, b) affective (one experiencer argument in the dative, affective or a spatial case and one stimulus argument in the absolutive, abbreviated as EXP and STIM (1c). (1)

a. Karata (Magomedova and Xalidova 2001: 78) ho-​w wu-​šãq-​a hač’e dem-​m m-​work-​cvb cop.neg ‘He does not work.’ b. Karata (Magomedova and Xalidova 2001: 55) qoči-​bi ra-​k’ũ-​da ho-​ššu-​l book.obl-​pl pl-​swallow-​prs dem-​obl.m-​erg ‘He swallows books.’ (i.e. He reads a lot) c. Karata (Magomedova and Xalidova 2001: 449) išši-​ja dãde wo-​kku hudu-​w 1pl.excl-​dat against m-​directed dem-​m ‘We met him.’

 

Ergativity in Nakh–Daghestanian    853 I will call argument of the types S, A, EXP subject-​like arguments, and P, STIM object-​ like arguments. All classes have in common that their members require at least one argument in the absolutive. However, many languages have exceptional valency classes with only a few members that deviate from the ones above in that they lack absolutive arguments and/​or mark arguments with cases other than ergative, absolutive, dative, or spatial cases or have more than one absolutive argument.

35.3  Head Marking: Agreement The vast majority of Nakh–​Daghestanian languages show gender/​number agreement. Exceptions are Lezgian, Agul, Udi, and southern Tabasaran varieties that have lost this feature. In contrast, person agreement is only attested in a few languages. The agreement rules for gender/​number agreement are basically identical for all languages whereas person agreement varies significantly not only with respect to its origins and morphological exponents but also with respect to its rules. The two agreement systems act completely independently of each other and are treated separately, beginning with gender/​ number agreement.

35.3.1 Gender and Number Agreement Gender is a chief category in the grammars of most Nakh–​Daghestanian languages. The number of gender varies from two genders in Tabasaran to five genders in all Nakh and some Tsezic and Andic languages (e.g. Hinuq, Khwarshi, and Chamalal). Apart from a few exceptions gender is not marked on the nouns, but becomes obvious through combined gender/​number agreement. Typical agreement targets are verbs, adjectives, and pronouns, but many languages also have agreeing numerals, adverbials, postpositions, and even case markers. However, all languages have the same agreement rules: within the phrase it is the head noun that triggers the agreement, and within the clause it is the absolutive argument, independently of the semantic role of the argument (1a–​c). If the verbal complex has more than one agreement slot, all exponents are governed by the same agreement trigger. Every language has a default gender that shows up if the clause does not contain an argument in the absolutive. For instance, in Tsakhur the default gender is gender iv. It appears with predicates that lack an absolutive (Kibrik 1999b: 367), and it also appears when complement-​taking predicates agree with their complement clauses (section 35.10). If the clause contains more than one absolutive argument the copula agrees with the A and the lexical verb with the P. This is called a biabsolutive construction, and can be interpreted as an alignment split triggered by aspect and information structure (see example (5)).

 

854   Diana Forker There are a few known exceptions to the rule that only the absolutive can trigger agreement that need further investigations, e.g. in Sanzhi Dargi an ergative-​marked agent can exceptionally trigger the agreement of the copula if the agent is emphasized (Sumbatova 2013; Forker in press). Similar cases have been reported for Akusha Dargi (van den Berg 1999: 173) though the pragmatic implications seem to be different from Sanzhi Dargi. In the Tsova-​Tush potential construction a potential agent in the dative controls gender/​number agreement on intransitive verbs (Holisky and Gagua 1994: 198).

35.3.2 Person Agreement Nakh–​Daghestanian languages that have person agreement are Dargi languages, Lak, Tabasaran, Tsova-​Tush, Udi, and to a lesser extent Hunzib, Akhvakh and some Avar dialects. Overviews can be found in Helmbrecht (1996) and Schulze (2007). What concerns the morphology of person marking, in some languages there are suffixes (e.g. partially in Dargi, Hunzib, Akhvakh), whereas in other languages person markers are enclitics (e.g. partially in Dargi, Lak, Udi, Tabasaran). The person enclitics in Dargi, Lak, and Udi can be floating off and attach to an item that expresses term focus (see Kazenin 1998, 2002; Harris 2002). Some systems show a lot of syncretism, and often there is no overt marking for third person (e.g. Tsova-​Tush, Tabasaran, partially Dargi). The extent to which person marking is attested throughout the verbal paradigm and whether and when it is obligatory greatly varies from language to language. In languages with limited person marking based on participles only a very limited part of the verbal paradigm allows for it. Person marking may be conditioned by clause type and is usually absent in subordinate clauses. In Akhvakh it depends on the illocutionary force type (assertion vs. question) such that the system expresses egophoricity (Creissels 2008). The standard rule is agreement with the subject, independently of grammatical role or case marking. The systems of Zakatal and Kusur Avar (Sumbatova 2013: 402–​403), Tsakhur (Schulze 2007: 143–​150), Akhvakh (Creissels 2008) and Hunzib (van den Berg 1995: 76–​86) show some similarities that are probably due to their common path of development from participles. In Udi (Harris 2002) and Tsova-​Tush (Holisky and Gagua 1994; Kojima ms), though the basic rule is also subject agreement, the systems are more elaborate because they have markers for different cases according to the semantic role of the subject. These markers originate from personal pronouns, which is also the case for Tabasaran. All languages except for Tabasaran index only one argument (Harris 1994; Bogomolova 2012; Babaliyeva 2013). In Tabasaran, subject agreement is obligatory for all argument types (S in intransitive clauses, A or EXP in bivalent clauses). Agreement with a second argument is optional and possible only in combination with a first person subject. Its occurrence depends, among other things, on the information structure. In

 

Ergativity in Nakh–Daghestanian    855 general, various case-​marked arguments and adjuncts can be indexed on the verb (e.g. possessors, experiencers, beneficiaries, stimulus, goal). (2) Tabasaran (Bogomolova 2012: 107) uzu uvu-​x-​na ʁ-​uš-​un-​za-​vu-​x-​na 1sg 2sg-​apud-​lat pfv-​go.out-​pst-​1sgagentive-​2sg-​apud-​lat ‘I went to you.’ Subject agreement is sometimes realized in close connection with agentivity, in the sense that either only agentive verbs show agreement (e.g. Akhvakh) or there are different markers used for agentive and non-​agentive arguments of intransitive clauses, e.g. Southern Tabasaran (Bogomolova 2012), Tsova-​Tush (4a, b). Exceptions from the subject agreement rule are Lak and Dargi varieties, and for these languages the origin of the person markers is unclear. Lak person agreement mostly follows an ergative pattern like the one used for gender/​number agreement (Helmbrecht 1996; Kazenin 1998). There are some complications that lead to deviations from this rule such as accusative patterns, the neutralization of agreement or default agreement with the third person. The deviations usually depend on the information structure, e.g. on focus marking or agent topicalization by means of biabsolutive constructions (cf. Kazenin 1998, 2013; Schulze 2011). Mehweb and Kubachi Dargi also show subject agreement. In the other Dargi varieties person hierarchies play a significant role (cf. van den Berg 1999; Sumbatova 2011). The rules are: if both arguments are third person, then the third person marker is used and this marker may exhibit gender/​number agreement with the absolutive argument. In clauses with one SAP argument, this argument will be indexed by a person marker. In clauses with two arguments that are both SAPs, the rules diverge: Southern Dargi varieties such as Icari, Sanzhi (3a, b), Qunqi, and Kaytag always index the second person independently of whether it is the subject or the object. In Akusha, Ashti, and Urakhi only the object can trigger agreement and in Chirag only the subject (A). (3) a. Sanzhi Dargi dam u či-​w-​igu-​tːe 1sg.dat 2sg spr-​m-​see.ipfv-​2sg ‘I see you.’ b. at du či-​w-​igu-​tːe 2sg.dat 1sg spr-​m-​see.ipfv-​2sg ‘You see me.’ Thus, if we want to keep things simple we can say that gender/​number agreement is almost exclusively ergative. Person agreement is predominantly accusative (i.e. with the subject-​like argument) except for Lak and many Dargi languages, and there are many exceptions and complications to this generalization.

 

856   Diana Forker

35.4  Dependent Marking: Case Nakh–​Daghestanian languages have ergative case marking. This statement deserves some comments. First, the languages have a basic morphological opposition found with all items that can bear case marking. The opposition is between direct stems, which are identical with the absolutive, and oblique stems. The opposition is also found in modification within the NP, e.g. demonstratives, adjectives and numerals may distinguish two forms according to the case marking of the head noun, or the genitive has two allomorphs. Oblique stems are commonly formed by adding a suffix to the direct stem or to the root. All cases other than the absolutive are formed by adding the appropriate case suffix to the oblique stem, though in a number of languages the ergative is identical to the oblique stem. All proper and common nouns as well as most pronouns inflect for ergative case. However, in a number of languages personal pronouns are defective, e.g. in Hinuq, Hunzib, Bezhta, Godoberi, Tabasaran, Budukh, Udi and Lak no personal pronouns distinguish absolutive from ergative. The basic function of the ergative is to mark the agent of transitive verbs including inanimate agents and causers, but in many Nakh–​Daghestanian languages it additionally expresses other functions such as instrument, cause or patient in the antipassive construction. Tsova-​Tush has an active/​stative (also called fluid split-​intransitive) case assignment system with intransitive verbs. With third person subjects there is no choice, the argument obligatorily appears in the absolutive (4a, b). With first and second person subjects there is variation. A group of verbs always require the subject in the absolutive, another group requires only ergative subjects, and the largest number of verbs allows for absolutive or ergative marking (see Holisky 1987 for details). According to Holisky and Gagua (1994: 194), “the case choice depends on both the semantics of the verb and the speaker’s beliefs about the situation in which it occurs.” (4) a. Tsova-​Tush (Holisky 1987: 105) (as) daħ japx-​jail-​n-​as 1sg.erg pv undress-​aor-​1sg.erg ‘I got undressed.’ b. (so) xe-​n=mak qac’-​u-​sO 1sg tree-​dat=on hang-​prs-​1sg ‘I am hanging in a tree.’ In Udi, we can find a similar phenomenon that is not restricted to first and second person (see Schulze 1994: 492 for an example with an NP subject). In other languages of the family split-​intransitive alignment is restricted to a limited set of diachronically compound verbs that have a petrified object argument. They formally look like

 

Ergativity in Nakh–Daghestanian    857 intransitives, but originate from transitive verbs and have kept their case marking. Examples are Ingush nab.j.u ‘sleep’ (nab ‘sleep’ + j-​u ‘do’ whereby j-​ is a frozen gender agreement prefix that is governed by nab, see Nichols 2008), some verbs in Lezgian (Haspelmath 1993: 178–​180, 284–​286), and onomatopoetic verbs in Hunzib (van den Berg 1995: 124) and Bezhta, e.g. lalaƛal ‘shout’ (lala + transitive verb of speech). Further common grammatical cases are genitive and dative, and less frequently instrumental or affective.1 Additionally, all have spatial cases, ranging from a handful to very elaborate systems with more than 50 cases. In most if not all Nakh–​Daghestanian languages the spatial cases also fulfill grammatical functions, i.e. they mark arguments of extended intransitive and ditransitive verbs. In a few languages, individual case suffixes contain additional gender markers that agree with the absolutive argument, e.g. the in-​essive in Dargi or the affective in Andi. For an overview of the Nakh–​Daghestanian case systems, see Kibrik (2003) and Daniel and Ganenkov (2008). The case marking of arguments strongly depends on the semantic roles, often with many-​to-​many relationships between cases and roles. For instance, the ergative marks agents of transitive verbs, but also other roles such as instruments. In turn, the semantic role of experiencer can be marked by various cases even within one and the same language (cf. Ganenkov 2006; Comrie and van den Berg 2006; Comrie et al. in press). For a restricted number of verbs we can state that S=P vs. A, i.e. they follow an ergative pattern. But there are monovalent verbs that do not take an absolutive-​marked argument and instead require other cases. Similarly, there are bivalent verbs whose arguments are not marked by absolutive and ergative. To this latter group belong extended intransitive verbs and affective verbs (section 35.2). Furthermore, in some languages diachronically trivalent compound verbs with a petrified direct object synchronically appear to be bivalent because they require only two arguments, one in the ergative and another one in the dative or a spatial case (cf. Nichols 2011: 420–​421 for an Ingush example). Thus, I  agree with Kibrik’s (1997, 2001b) conclusion that Nakh–​Daghestanian languages belong to the so-​called ‘role-​ dominated’ languages (Foley and Van Valin 1984: 123) in which the marking of arguments is semantically motivated.2

35.5  Valency Changing and the Lexicon 35.5.1 Biabsolutive Constructions Chechen, Ingush, all Tsezic languages, Avar, Lak, some Dargi varieties, Andic languages such as Godoberi, Bagvalal and Akhvakh, and the Lezgic languages Archi and Tsakhur 1 

The affective is only used for marking experiencer arguments of certain affective verbs. But I refrain from Kibrik’s (2001) analysis of Daghestanian languages as being mono-​pivotal and analyze them as pivotless. 2 

 

858   Diana Forker have biabsolutive constructions (Forker 2012; Gagliardi et al. 2014). These constructions are available with transitive verbs in the imperfective aspect and are characterized by absolutive marking of both A and P. The lexical verb show gender/​number agreement with P and the copula agrees with A (5). (5) Chechen (Molochieva and Witzlack-​Makarevich 2008) so wazha-​sh b-​u’-​ush v-​u 1sg.abs apple.abs-​pl pl-​eat-​cvb m-​cop ‘I (masc.) am eating apples.’ In biabsolutive constructions A is generally topicalized whereas P is pragmatically demoted. They can be interpreted as an alignment split based on aspect: in the perfective aspect case marking and agreement follow the ergative pattern, whereas in the imperfective aspect a neutral pattern is possible (see Coon & Preminger, Chapter 10, this volume, for similar alignment splits and a somewhat different approach).

35.5.2 Causatives, Antipassives, and the Like Nakh–​Daghestanian languages typically have productive valency-​increasing causative constructions and various valency-​decreasing processes. Causativization is realized through verbal derivation via suffixes and periphrastically by means of auxiliary change or by adding a causative auxiliary (cf. Nichols 1982; Comrie 2000; Daniel, Maisak, and Merdanova 2012). Causativization usually adds one ergative argument. Causees with causativized transitive base verbs bear a spatial case, the dative (Udi) or the instrumental (Bezhta). With affective verbs causativization may simply change the case marking of the experiencer from dative to ergative (Kibrik 2001: 1417; Forker 2013a: 512–​513; Forker in press), though the addition of an ergative argument is also attested. There are occasional instances of causatives that do not lead to the introduction of a new argument but only change the meaning of the verb or do not alter the construction at all (e.g. van den Berg 1995: 108 on Hunzib; Forker 2013a: 506, 510–​511, 516–​517 on Hinuq; A.A. Kibrik 1996: 128–​135 on Godoberi; Mel’čuk 1988: 235 on Lezgian). The productivity of individual causativization processes varies from language to language and can further depend on factors such as aspect or transitivity of the base verb, e.g. in Ingush causativization by means of the suffixes -​d.u (direct causatives) and -​d.eit (double causatives) almost never applies to transitive verbs (Nichols 2011: 484–​485) (see also Kibrik 1994: 359 on Archi; Ljutikova 2001: 390–​391 on Bagvalal for restrictions concerning the causativization of transitive verbs). In any case, we can say that causativization of bivalent and trivalent verbs never affects the object (P) which can perhaps be taken as a weak indicator of accusativity. Nakh–​Daghestanian languages normally do not have passives, but other valency-​ decreasing processes. An antipassive is found in Dargi varieties, Bezhta, Tsez, Hinuq, Hunzib, Avar, and Godoberi. Nakh–​Daghestanian antipassives differ considerably from

 

Ergativity in Nakh–Daghestanian    859 canonical antipassive constructions because they either do not make use of verbal derivation (Dargi) or they are also available for intransitive verbs (all other languages). Only in Bezhta the direct object can optionally be demoted to an oblique argument typically marked by the instrumental or less frequently by the inter-​essive case (Comrie and Khalilova 2013). In Dargi varieties the demoted object obligatorily appears as an oblique in the ergative case. In all other languages, the direct object is obligatorily deleted. In contrast to the causative, the antipassive is not very productive and generally only available with a limited number of verbs, which is at least partially due to its semantics. It commonly has an iterative or habitual meaning (see Polinsky, Chapter 13, this volume, for a detailed account of antipassive constructions). Other examples of detransitivization constructions are potential and/​or involuntary agent constructions, exterior force construction, and the undesirable action construction (Ganenkov et al. 2007). Similarly to what has been said about the causative construction we can state that P is unaffected by all valency-​decreasing processes except for the antipassive. A is demoted to an oblique argument. The treatment of S varies: in some constructions, intransitive verbs cannot participate at all (e.g. anticausative in Lezgian). If they can participate, they keep their argument type (S) and case marking, but change their semantic role. The valency-​changing constructions all have a strong semantic motivation and cannot be viewed as a purely syntactic means to manipulate the argument structure.

35.5.3 Lexicon Some Nakh–​Daghestanian languages have a class of compound verbs that have incorporated object-​like arguments. In some languages, a number of these can be said to be semantic intransitives but preserving their ergative case marking pattern or syntactic ditransitives but semantic transitives with a frozen object. In any case, it is never the subject that is incorporated, but the process applies only to the object. In many Nakh–​Daghestanian languages there are a number of labile verbs. It seems that the number of P-​labile (S=P) verbs always outranks the number of A-​labile (S=A) verbs and that some languages have only P-​labile verbs (e.g. Ingush, Lezgian). Nichols (2008) takes this as an indication for ergativity. Creissels (2014) in his discussion of Akhvakh and Avar criticizes the traditional notions of A-​lability and P-​lability and as too broad, but he also recognizes that Nakh–​Daghestanian languages have small classes of verbs exhibiting what he calls ‘argument structure modifying P-​lability.’

35.6 Imperatives Addressees of imperatives can be subjects of intransitive and canonical transitive verbs, and depending on the language also subjects of affective verbs, though not all languages

 

860   Diana Forker allow for the latter. In contrast, the absolutive arguments of transitive and affective verbs can never be the addressee of the imperative. Addressees are often left unexpressed, but in many languages they can overtly appear in the clause without adding any special emphasis. In Tsez, the antecedent in the absolutive can bind a reciprocal pronoun in the ergative and vice versa (section 35.8). Yet only the ergative and never the absolutive can function as addressee of an imperative (6), which indicates that case seems to be the relevant factor in the determination of addressee imperatives. (6) Tsez (Arsen Abdulaev, p.c.) mežä /​ *meži sidä sis b-​ex-​ur-​o! 2sg.pl.erg 2sg.pl one.erg one hpl-​die-​caus-​imp ‘You kill each other!’ Though the addressee of imperatives is often included in investigations of syntactic alignment patterns, it is rather a semantic criterion that does not have any relation to syntactic alignment. According to Dixon (1994: 131), it will always follow the accusative pattern independently of possible ergative traits in other parts of the grammar. More precisely, the referent of the addressee must be conceived as being able to exercise at least some control over the action denoted by the verb. This might explain why in a number of Nakh–​Daghestanian languages experiencers cannot function as addressees in imperatives -​experiencers often lack control over their experiences (cf. Comrie et al. in press). Similarly, in Bagvalal some subjects of intransitive verbs such as b-​ic’a ‘melt’, b-​uwã ‘be enough’ cannot be conceived as addressees because the event denoted by the predicate is beyond the control of the subject referent (Dobrushina 2001: 322–​323). To the contrary, in Lezgian certain possessors (Haspelmath 1993: 149) or arguments bearing spatial cases (D. Ganenkov, p.c.) can serve as addressees of imperatives because the referent can be interpreted as having control over the action.

35.7  Reflexive Constructions Reflexive constructions in Nakh–​Daghestanian are expressed through reflexive pronouns and never through marking on the verb. Like any other nominal the reflexive pronouns trigger gender/​number agreement on the verb if they occur in the appropriate position. They usually have morphologically complex variants formed by reduplication. The simple pronouns occur predominantly in non-​local reflexivization including logophoric contexts and can even express coreference across clausal and sentence boundaries. In some languages, they are also allowed in clause-​bound reflexivization. In contrast, complex reflexives are always clause bound. One part of the complex pronoun

 

Ergativity in Nakh–Daghestanian    861 bears the case suffix appropriate to the role in the clause. The case marking of the other part varies depending on the language and on the construction. In all languages for which data is available local reflexives cannot be bound by genitive noun phrases, but they can be bound by quantified antecedents, which shows that reflexivization involves variable binding and not coreference. Nakh–​Daghestanian reflexive constructions exhibit a very peculiar behavior because they allow reflexive pronouns to be the experiencers and in some languages even the agents of their clause. By this I mean that reflexive pronouns can appear in the ergative or the dative, bound by an absolutive antecedent (7b). In the following I will call this phenomenon reversal of case marking. Among specialists of Caucasian languages, the conspicuity of Nakh–​Daghestanian reflexive constructions has been noticed for some time (cf. Kibrik 1997, 2003; Ljutikova 1997, 1999; Forker 2014). (7) a. Icari Dargwa (Sumbatova and Mutalov 2003: 39; R. Mutalov, p.c.) Murad-​il cin-​na ca-​w w-​alχχ-​a=ca-​w Murad-​erg refl-​gen refl-​m m-​feed.ipfv-​prog=cop-​m ‘Murad earns his own living.’ (Lit. ‘Murad feeds himself.’) b. Murad cinna cinni w-​alχχ-​a=ca-​w Murad refl.gen refl.erg m-​feed.ipfv-​prog=cop-​m ‘Murad earns his own living.’ This includes direct and indirect objects, i.e. reflexive pronouns serving as indirect objects can be bound by antecedents that serve as direct objects and vice versa (8a, b). (8) a. Chechen (Molochieva and Witzlack-​Makarevich 2008) as Muusa shaa-​shie-​na kyzga-​n chu goiti-​ra 1sg.erg Musa red-​refl-​dat mirror-​dat into show-​wpst ‘I showed Musa to himself in the mirror.’ b. as Muusa-​na (shaa)-​shie kyzga-​n chu goiti-​ra 1sg.erg Musa-​dat red-​refl mirror-​dat into show-​wpst ‘I showed Musa to himself in the mirror.’ According to the data in Kibrik (2003), in very few languages reflexives can occur with extended intransitive verbs, bear absolutive case and be bound again by an antecedent marked by a spatial case (cf. Kibrik 2003: 650–​651 for Chamalal). The examples are taken from different languages because the extent to which reversed case marking is possible depends on the individual languages. In general, complex reflexives more easily allow for it than simple reflexives. It seems that only in some Andic languages (Tindi, Chamalal, Bagvalal, and Godoberi) simple reflexive pronouns are allowed to switch the case with the antecedent. Affective verbs more easily allow for the reversal than canonical transitives. In most languages extended intransitive verbs

 

862   Diana Forker are excluded, though Chamalal, Tindi, and partially Bagvalal seem to be exceptions (Kibrik 2003). The above data is enough to show that reflexive constructions do not indicate syntactic ergativity in Nakh–​Daghestanian languages and that there is no straightforward syntactic explanation for the reversed case marking. Instead, Forker (2014) proposes an alternative explanation along the following hierarchy of semantic roles (9). (9) agent < experiencer, non-​canonical agent < stimulus < patient < goal The closer two semantic roles are on the hierarchy, the more semantic properties they share. Semantic roles that are more similar to each other can more easily reverse their case marking because of the similarity and because the referents of the two arguments are identical.

35.8  Reciprocal Constructions Reciprocal constructions are typically expressed through reduplication of some variant of the numeral ‘one’ such forming a reciprocal pronoun. Reciprocal pronouns are usually clause bound. One part of the pronoun takes the case marking that is required by its function in the clause and the other part either does not inflect or copies the case of the controller or takes a frozen case suffix. The reciprocal pronouns also allow for reversed case marking (10a, b) (cf. Comrie et al. 2011; Yamada 2013). (10) a. Chamalal (Zaynab Alieva, p.c.) Madina=la ʡali=la sedʷ-​ɬa se-​bi idal-​eːda Madina=add Ali=add one.obl.pl-​dat one-​pl love-​prs ‘Madina and Ali love each other.’ b. Madina-​ɬa =la ʡali-​ɬa=la sedʷ-​ɬa se-​bi idal-​eːda Madina-​dat=add Ali-​dat=add one.obl.pl-​dat one-​pl love-​prs ‘Madina and Ali love each other.’ In some languages, the reversal is even obligatory. For example, in Bezhta reciprocal constructions with canonical transitive (11a, b) and affective verbs only reversed case marking is allowed. (11)

a. Bezhta (Zaira Khalilova, p.c.) kid=na öžö=nä sidi hos b-​iƛ’e-​yo girl=and boy=and one.erg one hpl-​kill-​pst ‘The girl and the boy killed each other.’

 

Ergativity in Nakh–Daghestanian    863 b. *kibba=na öždi=nä sidi hos b-​iƛ’e-​yo girl.erg=and boy.erg=and one.erg one hpl-​kill-​pst ‘The girl and the boy killed each other.’ In Tsez, Khwarshi, Avar and Icari Dargi reversed case marking is strongly preferred with affective verbs and a grammatical option with transitive verbs. In contrast, Hinuq allows it only for affective but not for transitive verbs. Thus, there is considerable variation. Again, extended intransitive verbs are normally excluded. The possibility of reversed case marking is also available between non-​subject arguments. This means that indirect objects can bind direct objects and vice versa (12a, b). (12) a. Standard Avar Zaʔira-​ca Madina=gi Šamil=gi co caz-​da Zaira-​erg Madina=and Shamil=and one one.obl.pl-​loc suratal-​da r-​ix̂i-​za-​r-​una picture.obl-​loc pl-​see-​caus-​pl-​aor ‘Zaira showed Madina and Shamil to each other on the picture.’ b. Zaʔira-​ca Madina-​da=gi Šamil-​da=gi co cal Zaira-​erg Madina-​loc=and Shamil-​loc=and one one.pl surat-​al-​da r-​ix̂i-​za-​r-​una picture-​obl-​loc pl-​see-​caus-​pl-​aor ‘Zaira showed each other to Madina and Shamil on the picture.’ In short, reciprocal constructions do not indicate that Nakh–​Daghestanian languages show syntactic ergativity (or accusativity). The bipartite morphology of the reciprocals and the semantic roles of the arguments probably play a role in allowing for the unusual case patterns.

35.9  Conjunction Reduction Though many Nakh–​Daghestanian languages have a borrowed conjunction wa, its use is not common for what we would consider to be the equivalent of a European clausal conjunction. Instead, converb constructions with a non-​specialized/​general converb are used (van den Berg 2004). From a syntactic point of view, these converb constructions are often somewhere in between subordination and coordination (cf. Kazenin and Testelec 2004; Kibrik 2007; Polinsky 2007; Creissels 2010; Forker 2013b), but semantically they represent conjunction (in contrast to the specialized converb constructions that express temporal, causal or other adverbial subordinate clauses).

 

864   Diana Forker What concerns coreference constraints in these converb constructions, a lot of data and material can be found in Kibrik (2003, 2007). In addition to the above-​mentioned literature, individual accounts can be found in Comrie et al. (2012), Daniel (2013) and in some grammars. The parameters of variation that determine coreference are: • valency classes of verbs in combination with case marking and semantic roles of arguments • order of the clauses (converb clause before main clause or vice versa or center embedding of the converb clause into the main clause) • referential types of the arguments: NP, pronouns, reflexive pronouns, zeros In addition, in languages with agreement this can sometimes be used to rule out ambiguities. Although shared arguments are very common with non-​specialized/​general converbs, there is no Nakh–​ Daghestanian language in which these converbs grammatically require their subject (or other arguments) to be shared with the finite verb. Constellations that are judged as unacceptable in elicitation might be perfectly grammatical in the right context within a natural text. In general, arguments whose referents the speaker assumes to be known by the hearer are left implicit such that often none of the clauses contains occurrences of the shared arguments. If there are overt arguments, they commonly occur in the main clause. Furthermore, adverbial clauses more commonly precede the finite clauses than vice versa, but the reverse order is also found. A frequent strategy is to have the co-​referent overt argument in sentence-​initial position, but syntactically belonging to the main clause. So, we have center embedding with anaphora in terms of linear order, but cataphora in terms of constituent structure (13). (13)

Bagvalal (Kazenin 2001a: 564) Maħammadi [_​i Anǯi-​ɬa awal=la b-​ihi-​w-​o] Mahammad erg Makhachkala-​loc house=and n-​buy-​m-​cvb ongi=da w-​iʁi there=emph m-​stayed ‘Mahammad bought a house in Makhachkala and stayed there.’

A constellation forbidden in elicitation in some language (cf. Comrie et al. 2012 for examples) and for which I do not have any corpus examples is zero anaphors in main clauses preceding potentially coreferential nouns in converbal clauses. If a preceding converb clause contains a pronoun, this usually cannot be coreferent with a subsequent full NP, at least in elicitation. This means that pronominal cataphora is mostly excluded. However, under certain circumstances this constraint can be overridden (see Forker 2013b for more information). What concerns argument types, the most frequently attested situation in texts is that coreferent arguments are subject-​like (S, A or EXP). Usually the overt argument

 

Ergativity in Nakh–Daghestanian    865 precedes the zero (14). Occasionally the second occurrence of the coreferent argument is a full NP, a pronoun, or a reflexive instead of a zero. (14) Kryz (Authier 2009: 321) [S (NP) = A (zero)] [q’usa-​sbari wa-​b-​xir-​ci] _​i nima mislaˁat yi-​ryu ki … old-​hpl pv-​hpl-​gather-​cvb erg such advice(f) do-​prs.f that ‘The elders gathered and decided that …’ Another less common possibility is the reverse order, i.e. the zero precedes the overt argument that can be an NP, a pronoun, or a reflexive. (15) Lak (Schulze 2011: 32) [A (zero) = S (NP)] [jala _​i qun-​ma-​ssa daˁwatt=gu b-​iču-​nu] ččitui then erg big-​iii-​adjz prayer=and iii-​spreadcvb cat haž-​l-​in tʼajla b-​ukk-​un b-​ur Hajj-​obl-​spr.lat straight iii-​go.out-​cvb iii-​cop ‘Then, having spoken out a strong prayer, the cat went out directly for the Hajj.’ Coreference between subject and object is less often found in texts, but nevertheless judged to be grammatical in all languages investigated by Kibrik (2003, 2007). Judging from the data in Kibrik (2003), a zero S preceding an overt P is rather exotic, but admissible (16). (16) Bezhta (Kibrik 2003: 607) [S (zero) = P (NP)] [_​i j-​õq’o-​na] öž-​di kidi j-​äƛel-​lö abs ii-​come-​cvb boy-​erg girl(ii) ii-​beat-​prs ‘The girl comes and the boy hits her.’ In some languages, this constellation requires to be expressed by means of some other converb that is distinct from the general converb (see also Forker in press). For instance, the general converb in Lak has the suffix -​nu and it is used for expressing S=A or A=S. If S=P/​ STIM, then the posterior converb -​kun ‘when, after’ is used (17a). But if the order is changed and the transitive clause comes first, then the zero can precede the pronoun (17b). (17) a. Lak (Kibrik 2003: 479) [S (zero) = A/​P (pronoun)] [_​i/​j Ø-​uːk’u-​kun] nai gʷa j Ø-​atta-​w abs i-​come-​post 1sg.erg 3sg.m i-​beat-​1 ‘After (he/​I) came, I (masc.) beat him.’ b. Lak (Kibrik 2003: 479) [P (zero) = S (pronoun)] [gʷa-​na-​l _​i Ø-​ttu-​nu] lagu-​ra nai 3sg.m-​obl-​erg abs i-​beat-​cvb leave-​1sg 1sg ‘He beat me and I (masc.) left.’

 

866   Diana Forker If the NP is positioned first, then the following coreferent argument is rather a pronoun or even a full NP (18) than a zero which might perhaps be seen as a weak indication for an S/​A pivot. (18)

Budukh (Kibrik 2003: 590) [P (NP) = S (NP)] [gǝdǝ-​rǝ riži naʕrǝ.süʔür-​na] riži wüxü-​ǯü boy-​erg girl(v) call.v-​cvb girl(v) leave.v-​aor ‘The boy called the girl and the girl left.’

The only language that has been shown to differ from this pattern is Ingush. According to Nichols (2008), the shared argument is usually in the same case (absolutive). She writes that younger speakers sometimes allow a shared A in the finite clause, but conservative speakers do not, and that consequently the syntax of clausal conjunction is ergative (see also Nichols 2011: 539). In sum, presumably there is a pragmatic preference for S=A/​A=S for most Nakh–​ Daghestanian languages, with the exception of Ingush. In contrast, S=P/​P=S is less frequent and perhaps slightly more marked, e.g. through specialized converbs as in (17a) or overt arguments instead of zero as in (18). Still, we have to acknowledge that it is not really possible to establish whether we deal with syntactic constraints and can therefore use conjunction reduction as an indication of a syntactic pivot or whether we rather have contextually licensed pro drop.

35.10  Control Constructions Control constructions have been extensively studied in Kibrik (1987, 2003) and this section heavily builds on Kibrik’s material. Before beginning to examine the control patterns in more detail a few caveats are in order. Verb forms commonly used in the embedded clauses of complement constructions are infinitives, purposive verb forms, participles, converbs, and masdars. The verb form occurring in the complement clause partially determines the morphosyntactic and semantic properties of the whole construction (cf. Forker 2016). With respect to gender/​ number agreement in complement constructions, many Nakh–​Daghestanian languages have two options: long-​distance agreement and local agreement (see e.g. Haspelmath 1999; Polinsky and Potsdam 2001; Polinsky 2003; Serdobolskaya 2009). In local agreement, the complement clause itself controls the agreement on the matrix verb. When we have long-​distance agreement, the absolutive argument of the complement clause controls the agreement (19). (19) Khwarshi (Khalilova 2009: 385) dilʲ l-​iyoːq’ /​ b-​iyoːq’ [ise bertin-​no-​ƛ’o keč’i b-​og 1sg.lat iv-​know iii-​know he.erg wedding-​obl-​spr song(iii) iii-​well

 

Ergativity in Nakh–Daghestanian    867 b-​ez-​dow-​ɬar] iii-​take-​ptcp-​nmlz ‘I know that he sings well at weddings.’ If both options are available, there can be pragmatic differences between them. Yet the basic rule is the same as for gender/​number agreement within the clause, i.e. the agreement is triggered by the constituent in the absolutive (which in case of local agreement is a nominalized clause). Thus, one can, in principle, interpret this as an indication of ergativity as has been done by Nichols (2008) for Ingush. Budukh, Tsakhur, Hinuq, Tsez, and perhaps more languages exhibit backward control with the verb ‘begin’ and Rutul (20) and Tsakhur also with ‘must’ and Tsez with ‘can.’ Backward control means that the overt controller appears in the embedded clause, bearing the case assigned by the embedded verb. Nevertheless, the matrix verb agrees with the controller. On the surface this looks like agreement with an argument not bearing the absolutive, but it has been argued that the matrix clause contains a covert argument in the absolutive case controlling the agreement (cf. Polinsky and Potsdam 2002). (20) Rutul (Kibrik 2003: 553) [gadije-​ri rɨš r-​et-​as] _​i mažbur j-​iʔi boy(i).obl-​erg girl(ii) ii-​beat-​inf abs must i-​aux ‘The boy must beat the girl.’ Occasionally, complements with ‘begin’ are interpreted as raising constructions (Manning 1995 on Lezgian; Polinsky and Potsdam 2002 on Tsez). Serdobolskaya (2009) claims that long-​distance agreement shows the same semantic and pragmatic properties as raising. However, I do not think that it is possible for Nakh–​Daghestanian languages to identify raising constructions that can be clearly differentiated from complement control. Thus, I  treat all complement constructions together and analyze them with respect to their control properties. Some languages allow for optional coreference of the infinitive complement with matrix predicates such as ‘want’ (e.g. Hinuq, Bezhta, Avar, Chamalal, Tindi, Lak, Archi) or somewhat rarer ‘be afraid’ (e.g. Lezgian, Tabasaran). This means that the infinitival clause contains arguments overtly marked for absolutive, ergative or dative/​affective/​ spatial case (20). In other languages, no infinitive can be used, but another non-​finite verb form appears instead (e.g. the simultaneous converb in Chechen, the aorist converb in Sanzhi Dargi and Lezgian). For example, with the verb ‘be afraid’ infinitival complements in Agul show obligatory control. If the case-​marked masdar is used the zero can acquire an indefinite reading in addition to the possible coreferent interpretation (Kibrik 2003: 525–​526). This means that in some cases it is the form of the embedded verb that determines whether we deal with a control construction (obligatory coreference) or not and infinitives have a tendency to appear in the former.

 

868   Diana Forker Modal and phasal verbs, if they are bivalent, must share their nominal argument with an argument in the complement clause. Other complement-​taking predicates such as verbs of liking, fearing, knowledge, etc. have optional coreference. For modal and phasal verbs, usually only subject-​like arguments can be controlled (21a); object-​like arguments are excluded (21b). (21)

a.

Agul (Kibrik 2003: 521–​522) cʷuj-​i-​si /​ cʷuj-​i-​ʁˁ-​asi [_​i či uta-​s] xisu brother-​obl-​dat brother-​obl-​inter-​el erg sister beat-​inf can.fut ‘The brother can beat the sister.’

b. *čij-​i-​si [cʷuj-​i _​i uta-​s] sister-​obl-​dat brother-​erg abs beat-​inf (‘The sister can be beaten by the brother.’)

xisu can.fut

However, Kibrik (2003:  491–​493) presents examples from Chirag Dargi that have object controllees with the matrix predicates b-​uχi ‘can’ and čaq-​b-​arχʷi ‘begin’ that allowed for absolutive-​marked Ps and STIM (22a) in addition to the subject-​like controllees, but dative-​marked goals are excluded (22b). (22) a.

Chirag Dargi (Kibrik 2003: 492–​493) ruccei čaˁqurχʷ-​le [ucci-​le _​i r-​iχχˁi] sister begin-​prs brother-​erg abs f-​guard.inf ‘The sister begins to be guarded by the brother’

b. *ruccei čaˁqurχʷ-​le [ucci-​le _​i sister begin-​cvb brother-​erg dat (The sister begins to be hit by the brother.)

b-​aqˁi] n-​hit.inf

With other complement-​taking predicates that have optional control subject-​like controllees are frequent and usually obligatorily deleted in the complement clause. If they are expressed by means of a reflexive or personal pronoun, then the interpretation is normally emphatic (cf. Forker 2013a: 646). When an object-​like argument is controlled, then there are two frequently occurring strategies to establish the right coreference: reflexive pronouns (23a) and/​or the use of a subordinate verb form that is different from the one used with controlled subjects (23b). (23) a. Hunzib (van den Berg 1995: 130) ož-​di-​ii Ø-​at’ [kid-​bo-​l žui heh-​á] boy-​obl-​dat i-​want girl-​obl-​erg 3sg.refl hit-​inf ‘The boy wants the girl to hit him.’

 

Ergativity in Nakh–Daghestanian    869 b. Tsakhur (Ljutikova and Bonč-​Osmolovskaja 1999: 518) ičiːi qaˁrq’ɨn-​na [gad-​eː (jiǯi) geːt-​a=wɨ] girl fear.pf-​aa boy-​erg refl.sg beat.ii-​ipfv=compl ‘The girl is afraid that the boy will beat her.’ If the controlled object is omitted, the resulting clause may have ambiguous interpretations of either coreference or reference to a third party. If in this sentence the reflexive pronoun is used, then this results in unequivocal coreference between the matrix subject and the embedded object (24). (24) Anzuq Avar (Kibrik 2003: 459) jaša-​ɬu-​jei [waša-​s _​i/​j /​ jii ƛ’ap’-​ži] j-​aƛ’-​un j-​ona-​ru girl-​obl-​dat boy-​erg abs refl.sg beat-​inf ii-​want-​cvb ii-​cop-​neg ‘The girl does not want the boy to beat her.’ (her=the girl/​someone else) However, some verbs may behave exceptionally. For instance, in Godoberi idaɬi ‘want’ allows for object control if the embedded controllee is expressed by a reflexive pronoun. In contrast, the verb ‘fear’ only allows for subject control (Haspelmath 1996: 196). In Bezhta, not only reflexive pronouns but also demonstrative pronouns that regularly serve as third person pronouns may occur in object-​controlled complement clauses with ‘want’ (25). (25) Bezhta (Kibrik 2003: 604) ist’i-​li Ø-​at-​na-​j [do hugii /​ žüi Ø-​äƛ’el-​äl] brother-​dat i-​want-​cvb-​cop 1sg.erg 3sg.m 3sg.refl i-​beat-​inf ‘Brother wants me to beat him.’ In Archi (Kibrik 2003: 566) and Khinalugh the object can apparently be omitted without resulting in ungrammaticality or ambiguity (26). (26) Khinalugh (Kibrik 1994: 402, Kibrik 2003: 577) rišu [gadi _​ χäšsxirsä-​š] ɨnqʰ-​qo-​mä girl.dat boy.erg abs push.ii.ipfv.ptcp-​poss fear-​aux-​ind ‘The girl is afraid that the boy will push (her).’ The controllers are again subject-​like arguments (S, A, EXP). But with a few verbs also syntactic possessors in the genitive that semantically serve as experiencers can be controllers (27).

 

870   Diana Forker (27) Lezgian (Haspelmath 1993: 358) Selim-​a-​ni rik’e-​laj [_​i penžer aq’al-​iz] alat-​na Selim-​obl-​gen heart.obl-​spr.abl erg window close-​inf fall.off-​aor ‘Selim forgot to close the window.’ (Lit. It fell off from Selim’s heart to close the window.) In sum, we notice a general tendency for syntactic accusativity. Subject control is expressed through obligatory deletion of the coreferent argument in the embedded clause and is available with various argument types (S, A, EXP). Object controllees are either prohibited (21b), (22b) or expressed through reflexive pronouns (23a), (24), (25) and/​or a deviant verb form in the complement clause (23b). We also saw some exceptions to this general rule that, however, only point to neutral alignment, never to clear instances of syntactic ergativity. It is sometimes necessary to make fine-​grained distinctions between predicate classes. This means that affective and occasionally intransitive predicates differ from transitive clauses, but the difference has a functional explanation. In fact, differences between affective and transitive predicates and their arguments are also found in other parts of the grammar (e.g. verbal inflection, causativization), and we saw instances with imperative formation and reflexivization/​reciprocalization earlier. Here I will only mention a few examples. For instance, the Hinuq verb -​iq-​‘can, be able’ does not allow for embedded affective verbs (Forker 2013a: 603). Similarly, in Chamalal and Archi the verb ‘begin’ cannot have complement clauses headed by affective verbs. In Hunzib, Tsakhur, Budukh, and Agul complement clauses of ‘begin’ containing affective verbs differ from those that are headed by intransitive or transitive verbs. For all these examples a semantic explanation is readily available: the complement-​taking predicates apparently require that the embedded subject referent must be able to control the action which is not to be expected for experiencers. Similarly, Akhvakh and Bagvalal distinguish between undergoer S arguments and experiencers on the one hand and agentive S and A arguments on the other hand with ‘begin’ (Akhvakh) and ‘fear’ (Bagvalal). Kibrik (1987: 147) analyzes this as active alignment strategy. Finally, in a few languages intransitive and extended intransitive verbs are sometimes singled out and treated differently in complement constructions. This concerns complements with ‘can’ in Hinuq (k’wezi -​iq’-​) and Bezhta (-​aq-​) or the treatment of G in Akhvakh and Chirag Dargi. These examples have no obvious functional explanation and the reason for the diverse behavior is probably the case marking of the embedded arguments (absolutive vs. everything else). Finally, with the Khinalugh verb ‘can’ all three basic valency patterns (transitive vs. intransitive vs. affective) behave differently in their morphosyntactic expression in complement clauses (Kibrik 2003: 574). There is generally not enough information about object-​control verbs and they have thus been omitted. The available data suggests that they do not differ from subject-​ control, e.g. the controllee can easily be S or A.

 

Ergativity in Nakh–Daghestanian    871

35.11 Summary There are some more constructions that are taken as indicators of syntactic alignment, but do not give any results for Nakh–​Daghestanian and have therefore not been discussed here, namely relativization (cf. Daniel and Lander 2010), quantifier floating, and word order. Table 35.1 summarizes the surveyed constructions. We can state that two of the three coding properties show morphological ergativity. The third property, person agreement, follows an accusative pattern in the majority of languages that have person agreement, but Lak has an ergative pattern and Tsova-​Tush and Southern Tabasaran have active/​stative alignment. There is weak indication of syntactic accusativity in the valency (changing) systems (increasing and decreasing operations, lability), conjunction reduction and complement control. Word order and relativization are neutral, and traits of active/​stative alignment can be found in a few languages in person agreement and case marking. There are numerous restrictions on semantic roles that have functional motivations and pop up in reflexivization, reciprocalization, imperatives, and occasionally in complement constructions. In fact, case marking can also be explained through semantic roles and the meaning of the case suffixes. In a few cases the morphological opposition between

Table 35.1 A summary of the surveyed constructions Construction

Ergative

Comments

Gender/​number agreement

yes

case marking

Person agreement

only Lak

even in Lak only partially because depends on information structure; stative/​active traits in some languages

Case

yes

depends on semantic predicate class, clause type, aspect, etc.

Valency changing processes (biabsolutive, causative, antipassive, etc.)

no

partially neutral or accusative

Relativization

no

neutral

Imperative

no

Reflexivization, Reciprocalization

no

Conjunction reduction

no

perhaps weakly accusative

Control constructions

no

accusative tendency

Quantifier floating

no

Word order

no

neutral

Lexicon

?

maybe P-​labile verbs indicate ergativity

 

872   Diana Forker unmarked absolutive and all other cases is decisive (e.g. quantifier floating, some complement constructions).

Acknowledgments This chapter is dedicated to the memory of Aleksandr E.  Kibrik, the father of modern Caucasiology. I am grateful to Dmitry Ganenkov, Timur Maisak, and Kevin Tuite for helpful comments on an earlier version of this chapter, and to Denis Creissels for pointing out relevant references. The chapter was written during my time as a Feodor Lynen fellow at James Cook University. I thank my host Alexandra Aikhenvald for providing me with a stimulating environment and the Humboldt Foundation for financial support.

Abbreviations i-​v, gender agreement markers; A, more agent-​like argument of a transitive verb; aa, attributive, other genders; abl, ablative; adjz, adjectivizer; apud, location ‘next to’; aor, aorist; aux, auxiliary; b, gender b; caus, causative; compl, complementizer; cop, copula; cvb, converb; dat, dative; dem, demonstrative; emph, emphatic; exp, experiencer argument of an affective verb; f, feminine; gen, genitive; erg, ergative; excl, exclusive; hpl, human plural; ind, indicative; inf, infinitive; interel, interelative; ipfv, imperfective; j, gender j; lat, lative; loc, locative; m, masculine; n, neuter; neg, negation; nmlz, nominalizer; NP, noun phrase; obl, oblique stem; p, more patient-​like argument of a transitive verb; pf, perfect; pfv, perfective; pl, plural; poss, possessive; post, posterior converb; prog, progressive; prs, present; pst, past; ptcp, participle; pv, preverb; red, reduplication; refl, reflexive; s, subject of an intransitive verb; sg, singular; spr, localization ‘on’; stim, stimulus argument of an affective verb; wpst, witnessed past.

 

Chapter 36

Ergativi t y i n Neo-​A ra ma i c Geoffrey Khan

36.1 Introduction Ergative constructions are found in many modern spoken dialects of Aramaic (hereafter referred to as Neo-​Aramaic). Aramaic is a Semitic language with an exceptionally long documented history. It is first attested in written form in inscriptions datable to approximately 1,000 bce and is still used as a spoken vernacular language by various minority communities in the Middle East. Ergativity is not a characteristic feature of Semitic languages.1 In the course of its more than 3,000 years of attested history, the Aramaic language has undergone major changes in its structure and it has developed in numerous dialectal varieties. Ergativity is a feature only of certain subgroups of Aramaic dialects. The contemporary vernacular dialects of Neo-​Aramaic are generally classified into the following subgroups: (i) Western subgroup spoken in Maʿlula and various other villages in the region of Damascus. (ii) Ṭuroyo subgroup, spoken in Ṭūr ʿAbdīn and in the village of Mlaḥso in south-​ eastern Turkey west of the Tigris river. (iii) North-​Eastern Neo-​Aramaic (NENA), spoken east of the Tigris river in south-​ eastern Turkey, northern Iraq, and western Iran. (iv) Mandaic, spoken in the city of Ahwāz and Khorramshahr in south-​western Iran Ergativity is found only in the NENA and Ṭuroyo subgroups. The dialects of these subgroups have been in close contact for many generations with Iranian languages that exhibit ergativity2 and it is clear that ergativity in these Neo-​Aramaic dialects is 1  Some scholars have reconstructed the existence of ergative alignment in ancient phases of Semitic, though such theories are controversial (Hasselbach 2013: 55–​65). 2  See Haig (Chapter 20, this volume).

 

874   Geoffrey Khan contact-​induced. The contact of Aramaic on the eastern periphery of Semitic with Iranian languages has very deep historical roots, and there is, indeed, evidence of embryonic ergative-​type constructions already in eastern Aramaic dialects in the Achaemenid period (5th century bce) (Folmer 1995: 376–​380; Gzella 2004: 184–​194). The ergativity constructions of Neo-​Aramaic exhibit considerable diversity across the various dialects of the NENA and Ṭuroyo subgroups. This applies in particular to NENA, which contains a large cluster of dialects, approximately 150 in number. The diversity of NENA has been brought about both by the spread of the dialects over a wide, largely mountainous, geographical area and by social divisions between speakers of different religious communities. The major communal division of NENA is between the dialects spoken by Christians and those spoken by Jews. This communal division, as we shall see, relates to differences in ergative constructions. In this chapter a concise overview will be attempted of the wide diversity of ergativity in Neo-​Aramaic. A feature of ergativity in Neo-​Aramaic is that in a large proportion of the dialects it deviates from canonical forms of ergativity that have been identified in other languages. It will be argued here that the major reason for the existence of these non-​canonical forms of ergativity is that the convergence of the verbal syntax of eastern Aramaic with the ergative constructions of Iranian was only partial. Neo-​Aramaic exhibits various splits in the distribution of ergativity, the most conspicuous one being based on the morphological stem of verbal forms, in that it is restricted to verbal forms derived historically from the passive participle, expressing the past perfective or present perfect. This is a feature of ergativity in Iranian languages, with which the Neo-​ Aramaic dialects have converged in this respect. The majority of other verbal forms in the NENA and Ṭuroyo groups are derived historically from the active participle or infinitive and these do not exhibit ergativity. In the current state of the dialects it would not be strictly true to label these latter verbal forms ‘imperfective’, since they now may express both perfective and imperfective in some contexts (e.g. when expressing the future or irrealis). The split is essentially based on the different historical alignment of the passive participle, on the one hand, and the active participle and infinitive, on the other, rather than on a neat contemporary split between perfective and imperfective aspect. Another general feature of ergativity in Neo-​Aramaic is that it is manifested in suffixes on the verb, which cross-​reference arguments in the clause, rather than in case-​marking of arguments themselves. This differs from the situation in many ergative Iranian languages, which have ergative case-​marking of arguments.

36.2  NENA Dialects The NENA dialects may be categorized into three main groups with regard to the profile of the ergativity that they exhibit. These will be labelled ‘Split-​S’, ‘Extended-​Ergative’, and ‘Dynamic–​stative’ dialects respectively, following the terminology used in Doron and Khan (2012). Figure 36.1 shows a map of the NENA dialect area.

 

Ergativity in Neo-Aramaic    875 A R M E N IA TURKEY

Salamas Bohtan

IRAN

Hertevin Ashitha

Urmi

Solduz Barwar Šeno Amedia Rustaqa Dohok ab Sablag Bokan Z Ruwanduz er Qaraqoš R iv Arbel Saqez Zakho

e

e

SY R IA

IR AQ

Rive r ig r T

R

iv

Eup

h r ates

100 km

Bijar

Sanandaj Qarah Hasan

is

er

Sulemaniyya

Tikab

Khanaqin

Qasr-e Shirin Kerend

Figure 36.1  Map of the NENA dialect area

36.2.1  Split-​S Dialects This group displays a familiar type of ergativity in that, in forms derived historically from passive participles, the subject of intransitive verbs (hereafter S) aligns with the object of transitive verbs (hereafter O) with regard to morphological case-​marking on the verb. The subject of transitive verbs (hereafter A), on the other hand, has a different marking on the verb. The term ‘Split-​S’ refers to the fact that the alignment of S with O is restricted to unaccusative S. An agentive S generally aligns with A (as will be seen). Split-​S dialects include Jewish dialects on the eastern periphery of the NENA area in western Iran (e.g. J. Kerend, J. Sanandaj, J. Bijar, J. Saqəz, J. Bokan, J. Tikab)3 and in locations in north-​eastern Iraq close to the Iranian border (e.g. Sulemaniyya, Ḥalabja, Khanaqin) As noted, the grammatical relations of verbal arguments are marked by suffixes attached to the verb. Two basic types of verbal suffix exist in Neo-​Aramaic, which will be called here D-​suffixes and L-​suffixes respectively. D-​suffixes are historically direct suffixes agreeing in number, gender, and person with the nominative subject of a clause in the original nominative–​accusative alignment system of Aramaic. L-​suffixes are 3 

Henceforth Jewish dialects are prefixed by the letter J. and Christian dialects are prefixed by C.

 

876   Geoffrey Khan historically prepositional phrases consisting of the dative preposition l-​and a pronominal element. When these suffixes co-​occur in a verb, the D-​suffix is always placed before the L-​suffix in all verbal forms, both those derived historically from the active participle (glossed for the sake of convenience IMPF = imperfective, though see the caveats concerning the label ‘imperfective’—​section 36.1) and past perfective and perfect verbs derived from the passive participle (glossed PERF) (1)

a. Vimpf—​D-​suffix—​L-​suffix b. Vperf—​D-​suffix—​L-​suffix

The L-​suffix developed historically from a prepositional phrase which became a clitic. It has indeed retained some features of clitic status, such as its regular peripheral position. Many NENA dialects, however, also have clitics that are less morphologically integrated than L-​suffixes, so the term L-​suffix is preferred here to L-​clitic. In the contemporary Neo-​Aramaic dialects, an inversion of agreement of these two sets of suffixes with grammatical arguments has developed across IMPF and PERF verbs. For IMPF verbs, a D-​suffix cross-​references the A  argument and an L-​suffix cross-​references the O argument, whereas in PERF verbs, this is reversed, a D-​suffix cross-​references the O argument and an L-​suffix cross-​references the A argument. In intransitive clauses with both IMPF and PERF verbs a D-​suffix cross-​references the S argument: (2) a. Vimpf—​D-s​ uffix(A)—L ​ -s​ uffix(O) b. Vperf—​D-​suffix(O)—​L-​suffix(A) c. Vimpf—​D-​suffix(S) d. Vperf—​D-​suffix(S) The forms of the suffixes in J. Sanandaj are as follows (Khan 2009): (3) J. Sanandaj D-​suffixes 3ms -​∅ 3fs -​a 3pl -​i 2ms -​et 2fs -​at 2pl -​etun 1ms -​na 1fs -​an 1pl -​ex

L-​suffixes -​le -​la -​lu -​lox -​lax -​lăxun -​li -​li -​lan

 

Ergativity in Neo-Aramaic    877 The D-​suffixes are morphologically less marked than the L-​suffixes. Historically the 3rd person D-​suffixes contain an expression only of number and gender, the person marker of the 3rd person being zero. In the case of the 3ms form, also the number and gender marking is zero. The 2nd and 1st person D-​suffixes have developed from a coalescence of number, gender and person markers. The L-​suffixes, on the other hand, all contain pronominal elements that syncretize person, number and gender, as well as a relational element l-​, which is historically a dative preposition. The alignment shown in (2) for PERF verbs is canonically ergative also in terms of the morphological markedness of the agreement to the A argument. The following examples show how S aligns with A in clauses with IMPF verbs but aligns with O in clauses with PERF verbs. In each case the common marker is the 3fs D-​ suffix on the verb: (4) J. Sanandaj IMPF a. brat-​i barux-​ăwal-​i garš-​a-​lu daughter-​my friend-​pl-​my pullIMPF-​D.3FS-L​ .3PL ‘My daughter (A) pulls my friends (O).’ b. brat-​i samx-​a daughter-​my standIMPF-​D.3FS ‘My daughter (S) stands.’ PERF c. barux-​ăwal-​i brat-​i gərš-​á-​lu friend-​pl-​my daughter-​my pullPERF-​d.3fs-​l.3pl ‘My friends (A) pulled my daughter (O).’ d. brat-​i smix-​a daughter-​my standPERF-​D.3FS ‘My daughter (S) stood up.’ D-​suffixes are used only on verbs with non-​agentive S arguments, i.e. unaccusative. Unergative intransitive verbs, with agentive S arguments, take L-​suffixes with PERF verbs in this group of dialects, which justifies referring to them as Split-​S according to the terminology of Dixon (1994: 71). The division between the categories of agentive and unaccusative S as manifested by the different suffixes is somewhat arbitrary and, indeed, differs slightly across the dialects of this group. The majority of intransitive verbs that are coded as having an agentive S express the emission of sounds: (5) kalba nwəx-​le dog barkPERF-​L.3MS ‘The dog barked’

 

878   Geoffrey Khan A transitive PERF construction with L-​suffixes as in (4c) must be regarded as an ergative rather than a passive, although the morphological stem is a passive participle in origin. This is for various reasons: • The construction is unmarked in terms of text frequency and productivity. Also it is not derived syntactically from a less complex active construction. • Whereas an A argument is obligatorily marked on the verb by an agreement suffix, an O argument is not obligatorily cross-​referenced by a suffix on the verb. Notably indefinite O arguments generally do not have agreement suffixes on the verb, as seen in (6), where the indefinite A argument is marked by an L-​suffix but the indefinite O is left without agreement on the verb. This situation indicates that the A is the most salient referent in the clause and has the status of syntactic subject, whereas the O has the status of syntactic object, which exhibits differential object marking, and not the subject of a passive: (6) J. Sanandaj xa naša qṭəl-​le a man killPERF-​L.3MS ‘A man killed many dogs.’

raba kalwe many dogs

• The A is a core argument, which is obligatorily present and has verb agreement. It is more tightly integrated into the clause structure than the agent-​phrase of a passive, which is typically omitted and does not have verb agreement (Comrie 1988: 16). The A exponent on the verb cannot be dropped. So, although the morphological coding of the A of PERF and IMPF transitive clauses is different, there are grounds for identifying both on syntactic grounds to have the status of subject. Further arguments for the subject status of the A argument in ergative PERF constructions include: • The order of constituents of PERF constructions such as (4c) is the same as that of corresponding IMPF constructions such as (4a). • The A argument in a construction such as (4c) should be regarded as being in its basic position and not extracted by topicalization out of an adjunct phrase that leaves a resumptive pronominal trace in the L-​suffix. This is shown by the fact that the A with a cross-​referencing L-​suffix may be a non-​referential (7a) or downward entailing (7b) argument, which are not pragmatically appropriate for topicalization, or interrogative (7c), e.g. (7) a. hič-​kas baxt-​ăke gərš-​a-​le nobody woman-​def pullPERF-​D.3FS-​L.3MS ‘Nobody pulled the woman.’

 

Ergativity in Neo-Aramaic    879 b. baṣor naše baxtăke gərš-​a-​lu few people woman-​def pullPERF-​D.3FS-​L.3PL ‘Few people pulled the woman.’ c. măni baxtăke gərš-​a-​le? who woman-​def pullPERF-​D.3FS-​L.3S ‘Who pulled the woman?’ • The O argument may be a reflexive pronoun co-​referential with the A in both IMPF and PERF constructions. This would not be possible if the O were grammatical subject: (8) J. Sanandaj IMPF a. barux-​ăwale noš-​u qaṭl-​i-​la friend-​pl self-​their killIMPF-​D.3PL-​L.3FS ‘The friends (A) kill themselves (O).’4 PERF b. barux-​ăwale noš-​u qəṭl-​a-​lu friend-​pl self-​their killPERF-​D.3FS-​L.3PL ‘The friends (A) killed themselves (O).’ These dialects, therefore, are morphologically ergative but not syntactically ergative. This is further demonstrated by the fact that the default subject of the second clause of conjoined constructions such as (9a) is the A and not the O of the first clause, in that the A is the salient subject, or pivot to use Dixon’s terminology. To induce an interpretation whereby the subject of the second clause is the O referent of the first clause a marked construction must be used with an added independent pronoun (9b), which signals that there is an unexpected shift of subject: (9) J. Sanandaj a ʾo-​baxta brat-​i diy-​a-​la the-​woman daughter-​my hitPERF-​D.3FS-​L.3FS ‘The woman hit my daughter and left’ b

ʾu pliṭ-​a and leftPERF-​D.3FS

ʾo-​baxta brat-​i diy-​a-​la ʾu ʾo pliṭ-​a the-​woman daughter-​my hitPERF-​D.3FS-​L.3FS and she leftPERF-​D.3FS ‘The woman hit my daughter and she (my daughter) left’

A feature that reflects the morphological ergativity of the PERF verbal constructions as distinct from the nominative–​accusative morphological coding of IMPF 4 

Note that the stem of the reflexive pronoun is feminine singular.

 

880   Geoffrey Khan constructions is the fact that the O expressed by the D-​suffix of PERF verbal constructions is restricted to the 3rd person. A PERF verb cannot express a 1st or 2nd person O with a D-​suffix (10a). This contrasts with IMPF transitive constructions which have no restrictions on person in suffixes (10b): (10) a. grəš-​∅-​lox gərš-​a-​lox gərš-​i-​lox *gərš-​ex-​lox *gərš-​et-​lan

‘You (ms.) pulled him’ ‘You (ms.) pulled her’ ‘You (ms.) pulled them’ *‘You (ms.) pulled us’ *‘We pulled you (ms.)

b. garš-​et-​e (< garš-​et-​le) garš-​et-​a (< garš-​et-​la) garš-​et-​u (< garš-​et-​lu) garš-​et-​an (< garš-​et-​lan) garš-​exi-​lox (< garš-​ex-​lox)

‘You (ms.) pull him’ ‘You (ms.) pull her’ ‘You (ms.) pull them’ ‘You (ms.) pull us’ ‘We pull you (ms.)’

1st and 2nd person objects of PERF verbs must be expressed by independent accusative phrases: (11)

grəš-​lox pullPERF-​L.2MS grəš-​lan pullPERF-​L.2MS

ʾəl-​an ACC-​1PL ʾəl-​ox ACC-​2MS

‘You (ms.) pulled us’ ‘We pulled you (ms)’

This restriction demonstrates firstly that the D-​suffixes in (10a) are not the syntactic subject in a passive construction, otherwise there would be no restriction on person as in (10b). The restriction of the person of O has arisen from a mismatch between the syntactic status of the O suffixes in PERF verbs as object and the unmarkedness of their morphological coding. As previously noted, D-​suffixes are morphologically less marked than L-​suffixes but object is a marked function for 1st and 2nd person pronouns, more so than for 3rd person pronouns. This is a reflection of Silverstein’s (1976) hierarchy for the split between accusative and ergative syntax on the basis of the property of arguments, which may be represented as follows: (12)

1st/​2nd person 3rd person proper nouns human common inanimate pronouns pronouns nouns Accusative > < Ergative

This reflects the fact that 1st and 2nd person pronominal objects are more likely to be accusative than 3rd person pronominal objects in languages that exhibit such splits.

 

Ergativity in Neo-Aramaic    881

36.2.2 Extended Ergative Dialects The majority of NENA dialects exhibit a generalized use of ergative L-​suffixes to mark the subject not only of transitive PERF verbs and agentive PERF intransitives, but also unaccusative PERF intransitives. Examples are given here from the C. Barwar dialect spoken in northern Iraq (Khan 2008). (13) C. Barwar IMPF a. xawr-​ăwaθ-​i garš-​i-​la friend-​pl-​my pullIMPF-​D.3PL-​L.3FS ‘My friends pull my daughter.’

brati-​i daughter-​my

b. brati qem-​a daughter-​my riseIMPF-​D.3FS ‘My daughter rises.’ PERF c. xawr-​ăwaθ-​i griš-​a-​la friend-​pl-​my pullPERF-​D.3FS-​L.3PL ‘My friends pulled my daughter.’

brati-​i daughter-​my

d. brati qim-​la daughter-​my risePERF-​L.3FS ‘My daughter rose.’ This involves the use of the marked L-​suffixes to cross-​reference both the A and the S argument (13c–​d), whereas the A  and S of the IMPF verbs are cross-​referenced by the unmarked D-​suffixes (13a–​b). Dixon (1994: 63–​67) prefers to designate the use of marked coding for A and S as a ‘marked nominative system’. His earlier term ‘extended ergative’ (Dixon 1979: 77) is more appropriate for the NENA dialects such as C. Barwar since there is evidence that the transitive constructions such as (13c) exhibit some ergative properties. One such property is the fact that in extended ergative dialects such as C. Barwar the marking of the A on the PERF verb may occasionally be dropped, as in (14): (14) qṭil-​∅ gawṛa killPERF-​D.3MS man ‘They (impersonal) killed the man’ In such constructions, the agent is unspecified (rendered in the translation of 14 by impersonal ‘they’). In an active nominative–​accusative construction the agent phrase

 

882   Geoffrey Khan is typically fully integrated into the clause structure and is obligatory whereas in passive constructions it is frequently omitted. According to Comrie (1988: 18–​19) ergative constructions lie in between these poles in that the agent phrase in the construction may occasionally be omitted in some languages. The remaining argument in the clause after the drop of the A in such constructions in NENA is syntactically the object and is not the syntactic subject of a passive construction. This is seen in (15), in which the indefinite O argument has no agreement on the verb. As noted, syntactic objects exhibit differential object marking, in that in most cases indefinite objects are unmarked. The placement of the O argument after the verb is also typical of syntactic objects since the C. Barwar dialect is predominantly AVO: (15)

prim-​∅ ʾərwe slaughterPERF-​D.3MS sheep.PL ‘They (impersonal) slaughtered sheep’               (Khan 2008: 750)

This shows that the A argument is more integrated than in a passive construction even when dropped and can be considered to be present at some underlying level of the derivation. It should be noted that the L-​suffix exponent of an S argument cannot be dropped on intransitives verbs. A further ergative property of transitive clauses such as (13c) in extended ergative dialects is that in many dialects of this group, including C. Barwar (Khan 2008: 282), the D-​suffix marking the O is restricted to the 3rd person, as is the case in split-​S dialects like J. Sanandaj (10–​11). When the O is a 1st and 2nd person pronoun, it has to be expressed through an independent accusative phrase external to the verb. Following the principles of the Silverstein hierarchy, which were discussed earlier, this implies that the third person D-​suffix is a morphologically unmarked object and the L-​suffix a marked subject, which is a feature of ergative constructions. Some extended ergative dialects do allow a 1st or 2nd person pronominal O to be expressed by D-​suffixes in PERF constructions. One such dialect is C. Urmi: (16) C. Urmi Object 3ms ɟrəš-​∅-​lə 3fs ɟriš-​a-​lə 3pl ɟriš-​e-​lə 2ms ɟriš-​ət-​lə 2fs ɟriš-​at-​lə 2pl ɟriš-​itun-​lə 1ms ɟriš-​ən-​lə 1fs ɟriš-​an-​lə 1pl ɟriš-​ax-​lə

‘He pulled him’ ‘He pulled her’ ‘He pulled them’ ‘He pulled you (ms.)’ ‘He pulled you (fs.)’ ‘He pulled you (pl.)’ ‘He pulled me (m.)’ ‘He pulled me (f.)’ ‘He pulled us’

Close investigation of these constructions in C. Urmi, however, reveals that the configuration of pronominal suffixes on PERF verbal forms in this dialect is not totally

 

Ergativity in Neo-Aramaic    883 without constraint. Although in principle the D-​suffixes can be used in any person, in practice speakers are less comfortable using 1st and 2nd person D-​suffixes than 3rd person D-​suffixes and exhibit a greater tendency to use some kind of repair strategy such as an independent accusative phrase when the object is 1st and 2nd person than when it is 3rd person. Furthermore, speakers express a definite preference for forms with 1st or 2nd person D-​suffixes that have 3rd person L-​suffixes, as in the paradigm in (16), than for forms with 1st or 2nd person D-​suffixes that have 1st or 2nd person L-​suffixes. This scale of preferences can be represented thus: (17) 3rd D + 3rd L > 1st/​2nd D + 3rd L > 1st/​2nd D + 1st/​2nd L The scale of preferences for A can also be correlated with the Silverstein hierarchy (12), in that 3rd person pronoun subjects favour morphological markedness more than 1st and 2nd person pronouns. A 1st and 2nd person subject expressed by a morphologically marked L-​suffix is, therefore, more semantically marked than a 3rd person L-​suffix. A  verbal construction that has an O expressed by a 1st or 2nd person D-​ suffix and an A expressed by a 1st or 2nd person L-​suffix is, therefore, doubly marked semantically. With regard to the typology of ergativity, extended ergative dialects can be regarded as exhibiting a hybrid system in that the alignment of arguments reflected by verbal suffixes is that of nominative–​accusative systems but the morphological markedness of these suffixes is the reverse of canonical nominative–​accusative systems. It shares with canonical ergative systems the unmarkedness of O and exhibits splits characteristic of ergative systems with regard to the semantic nature of O.

36.2.3 Dynamic–​Stative Dialects A small group of NENA dialects exhibit a split in the inflection of the PERF verbal stem that relates to aspect, in that verbs with L-​suffixes agreeing with the subject express past perfective, with a focus on the dynamic action, whereas verbs with D-​suffixes agreeing with the subject express a resultative perfect with focus on the resultant state of a prior action. In some dialects of this group, such as J.  Urmi, the D-​suffix paradigm is used for the resultative perfect only of unaccusative intransitives: (18) J. Urmi a. barux-​ăwal-​i brat-​i gərš-​a-​lu friend-​pl-​my daughter-​my pullPERF-​D.3FS-​L.3PL ‘My friends pulled my daughter.’ b. brat-​i qəm-​la daughter-​my risePERF-​L.3FS ‘My daughter rose.’

 

884   Geoffrey Khan c. brat-​i qim-​a daughter-​my risePERF-​D.3FS ‘My daughter has risen.’ In some dynamic–​stative dialects, such as C. Bohtan, D-​suffixes are used on all PERF verbs to express the resultative perfect, both those with unaccusative subjects and transitive or unergative verbs. Verbs with L-​suffixes are used with all verbs to express the perfective past. There are, thus, parallel paradigms: (19) C. Bohtan (Fox 2009) a. dynamic perfective vPERF-​l-​SUFFIX 3ms grəš-​le ‘He pulled’ 3fs grəš-​la ‘She pulled’ 3pl grəš-​lā ‘They pulled’ b. 3ms 3fs 3pl

vPERF-​l-​SUFFIX qəm-​le ‘He stood up’ qəm-​la ‘She stood up’ qəm-​lā ‘They stood up’

resultative perfect stative vPERF-​d-​SUFFIX vPERF-​d-​SUFFIX griš ‘He has pulled qim-​∅ ‘He has stood up’ griša ‘She has pulled qim-​a ‘She has stood up’ griši ‘They have pulled qim-​i ‘They have stood up’

The transitive resultative perfects in Bohtan express pronominal objects with L-​suffixes (Fox 2009: 56), as is the case with IMPF verbal stems. The split exhibited by dynamic–​stative dialects could be correlated with the common tense/​aspect split in split-​ergative languages whereby ergative marking is most likely to be found in clauses with past tense and perfective aspect. Such splits are, however, generally between past perfective and imperfective/​future (Dixon 1994: 97–​101). It is more likely that the split in question is related to differences in transitivity in the broad sense of the term proposed by Hopper and Thompson (1980), who regard the punctuality and dynamism of the action as properties of transitivity. The split would, therefore, be a split in transitivity based primarily on the transitivity properties of punctuality and dynamism rather than argument structure (Khan 2007a).

36.3  More on the Resultative Perfect A more common means of expressing the resultative perfect in NENA dialects is by combining the copula with a resultative participle. The resultative participle is derived historically from the passive participle, as is the case with the PERF verbal stem, but has an ending that is characteristic of nouns:

 

Ergativity in Neo-Aramaic    885 (20) C. Barwar PERF verbal stem ms griš-​∅ fs griš-​a pl griš-​i

Resultative Participle griš-​a griš-​ta griš-​e

Noun sus-​a ‘horse’ sus-​ta ‘mare’ sus-​e ‘horses’

The formation of the perfect with a resultative participle and copula is widely attested in the extended ergative group of dialects, for both transitive and intranstive verbs. The crucial point to note is that in these dialects the alignment of the perfect is nominative–​ accusative, in that the participle and the copula agree with both the transitive and intransitive subject, e.g. (21) C. Barwar a. ʾo-​gawṛa qima-​ile that-man risepart.ms-​COP.3MS ‘The man has risen’ b. ʾay baxta qimta-​ila that-woman risepart.fs-​COP.3FS c. ʾo-​gawṛa griša-​ile that-man pullpart.ms-​COP.3MS ‘The man has pulled a woman’ d. ʾay-​baxta grišta-​ila that-woman pullpart.fs-​COP.3FS ‘The woman has pulled a man’

baxta woman gawṛa man

In dialects with such constructions the pronominal direct object of the transitive perfect and the object agreement pronoun cross-​referencing a definite object argument is expressed by accusative L-​suffixes (22a) or adnominal dependent suffixes5 (22b). This variation of suffixes reflects the varying morphosyntactic status of the resultative participle as verbal or nominal across the dialects. This contrasts with the PERF stem which always has the morphosyntactic status of a verb (although historically a participle): (22) C. Barwar a.  ʾo-​gawṛa griša-​ile-​la   that-​man pullpart.ms-​COP.3MS-​L.3FS   ‘The man has pulled her’

5 

Such adnominal suffixes are attached only to nouns. In other contexts, they are used to express possessive pronominal elements on nouns, e.g. C. Urmi bet-​o house-​3FS ‘her house’.

 

886   Geoffrey Khan ʾo-​gawṛa griša-​ile-​la that-​man pullpart.ms-​COP.3MS-​L.3FS ‘The man has pulled that woman’

ʾay-baxta that-woman

C. Urmi b. ʾo-​+ɟora ɟriš-​o-​ilə that-​man pullpart.ms-​DEP.PRO.3FS-​COP.3MS ‘The man has pulled her’

ʾo-​+ɟora ɟriš-​o-​ilə that-​man pullpart.ms-​DEP.PRO.3FS-​COP.3MS ‘The man has pulled that woman’

ʾe-​baxta that-​woman

As we have seen, some dynamic–​stative dialects, such as C. Bohtan, have a parallel system of forming a nominative–​accusative perfect by D-​stem suffixes on the verbal stem rather than by participle and copula. In both C. Bohtan and extended ergative dialects with the copula type of perfect, such as C. Barwar, the nominative–​accusative perfect contrasts with the extended ergative perfective. Some dynamic–​stative dialects, such as J. Urmi, exhibit a mixed system, whereby the intransitive perfect is expressed by the PERF stem with D-​suffixes and the transitive perfect by the participle and copula, both transitive and intransitive perfective being expressed by ergative inflection: (23) J. Urmi a. Intransitive perfect: qim-​a risePERF-​D.3FS ‘She has risen’ b. Transitive perfect:

grəšt-​a (< grəšta-​ila) pullPART.FS-​COP.3FS ‘She has pulled’

In some Jewish extended ergative dialects, there are isolated uses of the PERF stem of unaccusative verbs inflected with D-​suffixes. This is the case, for example, in J. Arbel, in which it appears to be restricted to the maximally stative verb p-​y-​š ‘to remain’. In its attested uses it has the function of a resultative perfect, e.g. (24) či-​hulaʾ-​e la piš-​i NEG-​jew-​PL NEG remainPERF-​D.3PL ‘No Jews have remained in it’

gaw inside

(Khan 1999: 284–​285)

 

Ergativity in Neo-Aramaic    887 Most of the Split-​S dialects on the eastern periphery of the NENA area in western Iran express perfects by combining the resultative participle with the copula, but, unlike other dialects discussed so far, the participle and the copula in transitive clauses agree with the object and not with the subject (Khan 2009:  90–​92, 323–​326), e.g. (25) J. Sanandaj a. baxt-​ăke qimta-​ya woman-​the risePART.FS-​COP.3FS ‘The woman has risen’ b. ʾo-​gora baxt-​ăke grəšta-​ya that-​man woman-​the pullPART.FS-​COP.3FS ‘The man has pulled the woman’ It can be seen that the A argument in the transitive construction in (25b) has no explicit agreement marking on the verb in the form of an ergative L-​suffix. When there is no A argument nominal in the clause, the subject can be interpreted as 3rd person of any gender or number: (26) baxt-​ăke grəšta-​ya woman-​the pullPART.FS-​COP.3FS ‘He/​she/​they has/​have pulled the woman’ Such transitive perfect constructions with zero subject can be used only when the subject is 3rd person. They cannot be used with 1st and 2nd person subjects. The PERF verbal form with L-​suffixes is used with 1st and 2nd person subjects to express both the perfective and the perfect: (27) a. 3rd person subject: perfective ʾo gərš-​a-​le he pullPERF-​D.3FS-​L.3MS ‘He pulled her’ perfect ʾo grəšta-​ya he pullPART.FS-​COP.3FS ‘He has pulled her’

 

888   Geoffrey Khan b. 2nd person: perfective/​perfect ʾāt gərš-​a-​lox you pullPERF-​D.3FS-​L.2MS ‘You pulled her/​You have pulled her’ c. 1st person: perfective/​perfect ʾana gərš-​a-​li I pullPERF-​D.3FS-​L.1S ‘I pulled her/​I have pulled her’ Furthermore, a 1st or 2nd person object cannot be expressed by an agreement suffix on either the participle or PERF verbal base, but must be expressed by an accusative phrase external to the verb. The perfect construction with a copula is available for subjects of all persons in unaccusative clauses, the unaccusative S argument agreeing with the participle and copula: (28) J. Sanandaj 3ms zila-​ye 3fs zilta-​ya 2ms zila-​yet 2fs zilta-​yat 1ms zila-​yena 1fs zilta-​yan

‘He has gone’ ‘She has gone’ ‘You (ms) have gone’ ‘You (fs) have gone’ ‘I (ms) have gone’ ‘I (fs) have gone’

If the object in a perfect construction with a copula is indefinite, the participle and the copula do not agree with the object but remain in the default masculine singular form and do not agree with any constituent in the clause. This is clear in (29b), in which the subject is plural and the object is feminine singular but the participle and copula are masculine singular: (29) J. Sanandaj a. ʾo-​gora baxta gərša-​ye that-​man woman pullPART.MS-​COP.3MS ‘The man pulled a woman’ b. ʾo-​naše baxta gərša-​ye those-​people woman pullPART.MS-​COP.3MS ‘Those people pulled a woman’ When the verb is unergative, and is treated as transitive (see (5)), the copula and the participle are likewise in the default masculine singular form and do not agree with any constituent in the clause:

 

Ergativity in Neo-Aramaic    889 (30) J. Sanandaj a. tat-​i šəhla-​ye father-​my coughPART.MS-​COP.3MS ‘My father has coughed’ b. baxt-​i šəhla-​ye wife-​my  coughPART.MS-​COP.3MS ‘My wife has coughed’ c. ʾo-​naše šəhla-​ye those-​people coughPART.MS-​COP.3MS ‘Those people have coughed’ In the J. Sanandaj dialect an irrealis perfect may be formed by replacing the copula by the paradigm of the irrealis form of the verb hwy ‘to be’. In the irrealis perfect construction in transitive clauses the participle and copula agree with the object, as in the indicative perfect, but, unlike in the indicative perfect, the subject is marked by an ergative L-​suffix. Since there is explicit subject marking, the construction can be used with subjects of all persons (Khan 2009: 92–​94): (31) a. ʾo-​gora baxt-​ăke grəšta-​hawya-​le that-​man woman-​the pullPART.FS-​be.IRREALIS.3FS-​L.3MS ‘That man may have pulled the woman’ b. ʾāt baxt-​ăke grəšta-​hawya-​lox you (ms) woman-​the pullPART.FS-​be.IRREALIS.3FS-​L.2MS ‘You (ms) may have pulled the woman’ c. ʾana baxt-​ăke grəšta-​hawya-​li I woman-​the pullPART.FS-​be.IRREALIS.3FS-​L.1S ‘I may have pulled the woman’ One apparent anomaly of J. Sanandaj and related Split-​S dialects is that the past copula, which predicates a permanent or contingent state in the past, is inflected with ergative L-​suffixes: (32) J. Sanandaj 3ms ye-​le 3fs ye-​la 3pl ye-​lu

‘he was’ ‘she was’ ‘they were’

This will be discussed in the section on the proposed historical development of ergativity in Neo-​Aramaic. The omission of L-​suffixes in the indicative perfect

 

890   Geoffrey Khan may have arisen due to the fact that the placement of L-​suffixes after the present copula would have resulted in forms that are identical or close in form to those of the past copula.

36.4  More on Objects As we have seen in various places, there are restrictions on the occurrence of objects expressed by morphologically unmarked absolutive D-​suffixes of the PERF stem of verbs. In many dialects absolutive D-​suffixes are restricted to the third person. Dialects that allow 1st and 2nd person absolutive D-​suffixes, moreover, often exhibit some further restrictions. It should be noted that the dialects that allow 1st and 2nd absolutive D-​suffixes are all extended ergative dialects that have a predominant AVO word order. 1st and 2nd absolutive D-​suffixes are not attested in the split-​S dialect group, which have a predominant AOV word order. Many dialects have developed strategies for avoiding the expression of objects by absolutive D-​suffixes altogether. Several dialects express the pronominal object of PERF ergative verbs by attaching a second L-​suffix after the L-​suffix that expresses the ergative subject. This is an assimilation to the nominative–​accusative strategies of IMPF verbs, which mark the accusative pronominal object by L-​suffixes (see 2a), e.g. (33) J. Urmi a. grəš-​le-​le pullPERF-​L.3MS-​L.3MS

‘He pulled him’

b. grəš-​le-​li pullPERF-​L.3MS-​L.1S

‘He pulled me’

c. grəš-​le-​lox pullPERF-​L.3MS-​L.2MS

‘He pulled you (ms.)’

Since the object suffix is a marked accusative (L-​suffixes being marked and D-​suffixes unmarked) there is no restriction on person as it conforms to the alignment profile of the Silverstein hierarchy. In J. Urmi absolutive 3rd person D-​suffixes are also used on PERF verbs, e.g. (34) J. Urmi a. gərš-​a-​le pullPERF-​D.3FS-​L.3MS ‘he pulled her’

 

Ergativity in Neo-Aramaic    891 b. gərš-​i-​le pullPERF-​D.3PL-​L.3MS ‘he pulled them’ Some dialects do not allow any absolutive D-​suffixes at all and express all pronominal objects with marked L-​suffixes. This applies, for example, to the C.  Bohtan dialect (Fox 2009:  53). In some dialects that have this pronominal object-​marking strategy, adjustments are made to the form of the L-​suffix that expresses the ergative subject. A noteworthy adjustment of this kind has been documented in the C. Hertevin dialect (south-​e astern Turkey) by Jastrow (1988: 61). In this dialect, the L-​suffixes remain unadjusted if the subject is third person but are adjusted by giving them the ending of D-​suffixes if the subject is 1st or 2nd person, e.g. (35) C Hertevin a. PERF stem + L-​suffix (ERG) ḥze-​le ‘he saw’ ḥze-​lox ‘you (ms) saw’ ḥze-​li ‘I saw’ b. PERF stem + L-​suffix (ERG) + L-​suffix (ACC) ḥze-​le-​li ‘he saw me’ ḥze-​let-​ti (< ḥze-​let-​li, cf. 2ms D-​suffix –​et) ‘you saw me’ ḥze-​len-​ne (< ḥze-​len-​le, cf. 1ms D-​suffix –​en) ‘I saw him’ The result is the conversion of the marked L-​suffixes of the 1st and 2nd person subject to suffixes that resemble the form of unmarked subject D-​suffixes. This development can also be explained by the Silverstein hierarchy, in that 1st and 2nd person pronominal subjects have a greater tendency to be unmarked than 3rd person subjects. The effect is to shift the morphological coding of suffixes closer to the canonical nominative–​accusative coding rather than the anomalous morphological marking of both subject and object with two L-​suffixes as we find in J. Urmi (33). The C. Hertevin dialect does not use unmarked absolutive D-​suffixes to express pronominal objects on PERF stem verbs and relies entirely on marked L-​suffixes to mark pronominal objects, as in the IMPF system. Some dialects avoid absolutive coding and anomalous coding of subject and object by two marked L-​suffixes by replacing the PERF-​stem by a form containing the IMPF-​stem with a past tense prefix when the verb has a pronominal object. This is the case, for example, in C. Qaraqosh:

 

892   Geoffrey Khan (36) C. Qaraqosh a. grəš-​lə pullPERF-​l.3ms ‘he pulled’

kəm-​garəš-​∅-​lə PST6-​pullIMPF-​D.3MS-​L.3MS ‘he pulled him’

b. grəš-​la pullPERF-​l.3fs ‘she pulled’

kəm-​garš-​a-​li PST-​pullIMPF-​D.3FS-​L.1S ‘she pulled me’

c. grəš-​lux pullPERF-​l.2ms ‘you (ms.) pulled’

kəm-​garš-​ət-​li PST-​pullIMPF-​D.2MS-​L.1S ‘you pulled me’

Definite nominal object arguments in NENA dialects generally have some kind of morphological object marking in the clause, both where the verb has a IMPF stem and also where it has a PERF stem. As we have seen, one type of object marking is in the form of pronominal agreement suffixes on the verb, either accusative type L-​suffixes or absolutive D-​suffixes. Another type of object marker is a preposition on the direct object argument itself. This is typically a preposition that is derived historically from a dative marker and is used predominantly with definite direct objects with human referents. In some dialects, the marking of object arguments in this way excludes pronominal agreement argument on the verb. This is the case, for example, with C. Urmi. Contrast (37a–​b) with pronominal agreement marking of the object with (37c–​d) in which there is marking of the object argument: (37) C. Urmi Pronominal agreement on verb a. ʾaha-​+ɟora ɟriš-​a-​lə this-​man pullPERF-​D.3FS-​L.3MS ‘This man pulled that woman’

ʾay-​baxta that-​woman

b. ʾaha–​baxta bət-​ɟarš-​a-​le this-​woman FUT-​pullimpf-​D.3FS-​L.3MS ‘This woman will pull that man’

ʾo-​+ɟora that-​man

Marker on object argument c. ʾaha-​+ɟora ɟriš-​lə k̭a-​day-​baxta this-​man pullPERF-​L.3MS OBJ-​that.OBL-​woman ‘This man pulled that woman’ d. ʾaha–​baxta bət-​ɟarš-​a this-​woman FUT-​pullIMPF-​D.3FS ‘This woman will pull that man’ 6 

k̭a-​do-​+ɟora OBJ-​that.OBL-​man

The particle kəm-​converts the IMPF stem to past perfective. For a recent discussion of its usage and historical origin, see Fassberg (2014).

 

Ergativity in Neo-Aramaic    893 In other dialects, pronominal object agreement on the verb is maintained when a prepositional marker is attached to an object argument. This is the case, for example, in J Urmi: (38) J. Urmi Pronominal agreement on verb a. o-​gora baxta gərš-​a-​le that-​man woman pullPERF-​D.3FS-​L.3MS ‘That man pulled the woman’ b. o-​baxta gora garš-​a-​le that-​woman man pullIMPF-​D.3FS-​L.3MS ‘That woman will pull the man’ Marker on object argument and pronominal agreement on verb c. o-​gora əl-​baxta gərš-​a-​le that-​man OBJ-​woman pullPERF-​D.3FS-​L.3MS ‘That man pulled the woman’ d. o-​baxta əl-​gora garš-​a-​le that-​woman OBJ-​man pullIMPF-​D.3FS-​L.3MS ‘That woman will pull the man’ The double marking of objects in J. Urmi results in constructions such as (38c), in which an object argument has object accusative marking but is cross-​referenced by an absolutive D-​suffix on the verb. This form of split between argument case-​marking and agreement is typologically highly unusual, the opposite, ergative case-​marking with accusative agreement, being more widely documented (Manning 2006:  216). It has apparently arisen due to assimilation of the syntax of PERF verb clauses with that of IMPF verb clauses.

36.5  Ṭuroyo and MlaḤso The inflection of PERF verbal stems by ergative L-​suffixes is found in the Ṭuroyo and Mlaḥso dialects, which form a separate subgroup of Neo-​Aramaic. Ṭuroyo expresses 3rd person pronominal objects of transitive verbs by absolutive D-​suffixes as in many NENA dialects, e.g. (39) Ṭuroyo a. grəš-​∅-​li pullPERF-​D.3MS-​L.1S ‘I pulled him’

 

894   Geoffrey Khan b. griš-​o-​li pullPERF-​D.3FS-​L.1S ‘I pulled her’ c. griš-​i-​li pullPERF-​D.3PL-​L.1S ‘I pulled them’ 1st and 2nd person are expressed by accusative L-​suffixes after the ergative subject L-​suffix: (40) Ṭuroyo a. grəš-​le-​lux pullPERF-​L.3MS-​L.2MS ‘He pulled you (ms.)’ b. grəš-​lux-​li pullPERF-​L.2MS-​L.1S ‘You pulled me’ The profile of Ṭuroyo with regard to pronominal object-​marking is, therefore, similar to a NENA dialect such as J. Urmi (33–​34). Mlaḥso, by contrast, does not allow any absolutive D-​suffixes to express pronominal objects and all pronominal objects are expressed by accusative L-​suffixes or separate accusative prepositional phrases (Jastrow 1994: 54–​56), as is the case in some NENA dialects, such as C. Bohtan. A point of difference between Ṭuroyo and NENA is that in transitive clauses neither IMPF nor PERF verbs have pronominal suffixes agreeing with a definite object argument (Hemmauer and Waltisberg 2006: 32, 35) and as a result there is no absolutive agreement of PERF verbs: (41) g-​nəšq-​o FUT-​kissIMPF-​D.3FS ‘She will kiss the man’

ʾu-​zlām ART.MS-​man

(42) nšəq-​le ʾi-​aṯto kissPERF-​L.3MS ART.FS-​woman ‘He kissed the woman’ In Ṭuroyo a group of past perfective verbs have their subject expressed by D-​suffixes rather than L-​suffixes, e.g. (43) 3ms 3fs 3pl

daməx-​∅ damix-​o damix-​i

‘He went to sleep’ ‘She went to sleep’ ‘They went to sleep’, etc.

 

Ergativity in Neo-Aramaic    895 These include intransitive unaccusative verbs and also transitive experiential verbs such as šaməʿ ‘he heard’, ʾaḏəʿ ‘he knew’ (Jastrow 1985: 71; Furman and Loesev 2014). Intransitive unergative verbs are expressed by ergative L-​suffix inflection (e.g. šʿile ‘he coughed’). In the related dialect of Mlaḥso the equivalent verbal form with D-​suffixes is used as a resultative perfect, the ergative inflection being used to express the past perfective of all verbs. As far as can be established from the documentation of this now extinct dialect by Jastrow (1994), it is used with the same range of lexical verbs as in Ṭuroyo, i.e. those with unaccusative and experiential subjects. In both Ṭuroyo and Mlaḥso it is formed from a historically different morphological stem from that of the PERF ergative verb, i.e. the verbal adjective *CaCCīC rather than the passive participle *CCīC. This is different from the stem of unaccusative PERF verbs in split-​S NENA dialects such as J.  Sanandaj, which is derived historically from the passive participle CCīC.

36.6  Historical Background It is clear from the survey presented in this chapter that ergative constructions in Neo-​ Aramaic exhibit considerable diversity across the numerous dialects. All dialects are split ergative, with ergativity found only in verbs with the PERF stem or resultative participles. The ergative constructions include a type that conforms to split-​S morphological ergativity and an assortment of hybrid variations in which there are differing degrees of levelling with the nominative–​accusative morphosyntax of IMPF-​stem verbal forms. These hybrid systems exhibit the alignment of argument cross-​referencing but not the morphological markedness of cases characteristic of nominative–​accusative systems (i.e. the extended-​ergative dialects), morphological markedness of cases based on transitive properties of dynamicity and punctuality rather than argument structure (i.e. the dynamic–​stative dialects), various degrees of reduction of the distribution of the unmarked absolutive marking of the object in transitive clauses, the acquisition by the ergative L-​suffixes of syntagmatic properties of nominative subject suffixes, notably their ability to take accusative L-​suffixes, and, in the case of dialects like C. Hertevin, even assimilation of the ergative L-​suffixes to the morphological form of nominative subject D-​suffixes. In this concluding section a proposal for how the various constructions developed historically will be presented in brief outline.7 Within NENA, the majority of dialects belong to the extended ergative type. Most dialects of this group exhibit more archaisms in grammar and lexicon than the dialects of the split-​S and dynamic–​stative groups.8 These archaisms of the extended 7  A more detailed account with a full description of the extant historical evidence can be found in Khan (forthcoming). 8  For the general innovative nature of the Split-​S and dynamic stative groups, see Khan (to appear). For a different account of the historical development of ergativity in Neo-Aramaic see Coghill (2016)

 

896   Geoffrey Khan ergative dialects reflect a lesser degree of convergence with non-​S emitic languages in contact, notably Kurdish, than is the case with the dialects of the other groups. One syntactic archaism, for example, is the preservation of the predominant AVO word order of earlier Aramaic in most extended ergative dialects, whereas the split-​S and dynamic–​stative dialects are predominantly AOV through convergence with the verb-​final syntax of Kurdish and, in the case of the dialects of north-​ western Iran, with Azeri Turkish. It is likely that the extended ergative profile of the dialects is also an archaism vis-​à-​vis the other dialects and, although hybrid in typology, is not a transitional stage of diachronic development from the canonical poles of ergative and accusative within Aramaic itself. Rather it first appeared in Aramaic as a hybrid system. The ergative L-​suffixes had their origin in phrases with the preposition l-​, which in earlier Aramaic, in addition to the dative, was used to express a range of relationships including the beneficiary, experiencer, affectee, possessor, and agent (Bar-​Asher 2007; Bar-​Asher Siegal 2011). Through contact with Iranian, which exhibited ergative alignment based on the passive participle and oblique subject,9 constructions developed in eastern Aramaic that consisted of a passive participle and an oblique subject expressed by the preposition l-​. These replaced the original active perfective verbal forms of Aramaic, which had nominative subject suffixes, most likely functioning originally as resultative perfects alongside the perfective forms and then developing into perfectives and completely replacing the older forms according to the typologically common pathway of diachronic change (Goldenberg 1992). The convergence with Iranian was only partial, since the semantic range of the preposition l-​was initially retained. The earliest attestations of the constructions are predominantly found with verbs with experiential subjects with a perfect function, e.g. ‘to hear’ (šmiʿ li ‘I have heard’) and there are also early examples of unaccusative verbs with the affectee subject in intransitive verbs (mhallaḵ li ‘I have walked’), again mostly with a perfect function (Nöldeke 2001: 219; Khan 2004: 93). Eventually a generalized marking of the subject of all perfective verbs with L-​suffixes developed. Goldenberg (1992: 119) holds that in both the transitive and the intransitive the participle had passive function, the subject of the passive of the intransitive being the ‘inner object’ of the verb (i.e. ‘a going was gone by me’). An alternative, or at least concomitant, factor conditioning the generalization of the marking of the subject by l-​ may have been the phenomenon of constructional persistence, i.e. the continuing existence of the formal and semantic framework of a particular construction throughout the history of a language.10 This is reflected by the fact that the nominative suffixes of the original active perfective transitive and intransitive verbs were replaced symmetrically by the L-​suffixes:

9 

For ergativity in Old and Middle Persian see Haig (2008) and Jügel (2015). For the phenomenon of constructional persistence in the development of Iranian languages see Haig (2004: 55–​57). 10 

 

Ergativity in Neo-Aramaic    897 (44) Syriac NENA

qṭal-​t killPST-​D.2MS qṭil-​lox killPERF-​L.2MS ‘you killed’

dmex-​t sleepPST-​D.2MS dmix-​lox sleepPERF-​L.2MS ‘you slept’

Transitive perfective verbs had more properties of passive constructions at an earlier period than they do in the present state of the dialects. This is shown by the fact that early Neo-​Aramaic texts from Iraq datable to the seventeenth century, such as the Jewish Nerwa homilies published by Sabar (1984) and the Christian texts studied by Mengozzi (2012), have perfective passive verbs (45a). Moreover, in these texts perfective verbs exhibit features such as 1st and 2nd person D-​suffixes expressing pronominal O referents (45b) and the attachment of the l-​ agentive marker to nominal agent arguments (45c): (45) a. griš-​a pullPERF-​D.3FS ‘She was pulled’ b. griš-​ax-​lu pullPERF-​D.1PL-​L.3PL ‘They pulled us’ c. griš-​ax l-​arye pullPERF-​D.1PL L-​lion ‘We were pulled by the lion’ Such constructions are obsolete in most of the extended ergative dialects in areas where these texts were written. The passive (45a) has come to be reinterpreted as an active ergative with drop of the subject as in (15) in C. Barwar.11 1st and 2nd person pronouns, which would be tolerated as subjects in a passive construction, are eliminated in many dialects since they have become to be interpreted as syntactic objects in an ergative construction. Likewise (45c) is eliminated when the A becomes interpreted as the syntactic subject in an ergative construction. The reinterpretation of the pronouns expressed by unmarked D-​suffixes as objects of ergative constructions rather than subjects of passive constructions gave rise to various repair mechanisms for forms conflicting with the Silverstein hierarchy, such as the kəm-​garəšlə constructions in (36), which are rare in early texts (Mengozzi 2012:  33). Likewise repair mechanisms such as (35) in C. Hertevin developed when the A referent came to be interpreted as an ergative subject rather than an oblique agent adjunct. These are harmonizations of the ergative constructions with the nominative accusative 11 

See Gutman (2009) for a discussion of this historical development in Jewish Zakho and related dialects.

 

898   Geoffrey Khan constructions of IMPF clauses.12 It is important to note, however, that they were conditioned by the presence of ergativity and so can be taken to be diagnostics of the development of ergativity. The development of perfective stems with D-​suffixes, as is found in Split-​S and dynamic stative dialects is likely to be a later innovation, triggered by a greater degree of convergence to the ergative alignment of Kurdish in the dialects in question.13 This is likely to have begun as a means of expressing the resultative perfect, especially in highly stative verbs, as we see in J. Arbel (24), and then developed into a perfective, as in split-​ S dialects, or extended to the perfect of transitive verbs, as in dynamic–​stative dialects such as Bohtan (19).14 It should be noted that a split-​S type of alignment is strictly not a canonical type of ergativity with complete identity of S and O, since it is dependent on the semantic properties of the S (cf. Donohue and Wichmann 2008). Even this group of NENA, therefore, fell short of developing fully canonical ergativity. The development of resultative perfects with the copula is likewise a later development due to convergence with Kurdish. In Kurdish such constructions have ergative alignment, so perfect constructions such as those found in the Split-​S dialect J. Sanandaj reflect a closer convergence with Kurdish than is the case with the nominative–​accusative aligned constructions in other dialects (Khan 2007b: 204–​205). The tolerance of 1st and 2nd person objects in some extended ergative dialects but not in split-​S and dynamic stative dialects may also reflect a lower degree of approximation to canonical ergativity in the former group and be a relic of an earlier period when the construction was a passive.15 The construction of verbs with D-​suffixes in Ṭuroyo and Mlaḥso on the western periphery of the NENA area is clearly an independent innovation from that of the NENA split-​S dialects on the eastern periphery. This is shown by the fact that the Ṭuroyo and Mlaḥso forms are based on a stem of different historical origin and they also include transitive experiential verbs, unlike the Split-​S dialects. The anomalous form in the split S-​dialects of a past copula inflected with L-​suffixes (32) can be explained as being a vestige from an earlier period of generalized marking of L-​ suffixes. The shift to D-​suffixes was blocked by the fact that this would have made the past copula very similar in form to the present copula.

12 

See the discussion in Mengozzi (2005).   Kurdish dialects exhibit a more canonical type of ergativity than NENA extended ergative dialects, but it should be noted that also in Kurdish dialects there is some degree of hybridity in alignment between an ergative and accusative system (Haig 2004, 2008; Mengozzi 2005). 14  In some dialects, it seems to have developed in embryonic form in the intransitive but was replaced completely by the compound perfect with copulas. It appears sporadically, for example, in the Jewish Nerwa texts of the seventeenth century (Goldenberg 1992) and the Christian texts of similar date from the Mosul plain studied by Mengozzi (2012) but has been replaced by the compound construction in modern dialects of the area. In some modern dialects of Iraq such as J. Arbel and J. Koy Sanjak (Mutzafi 2004: 105) it remains but is very rare. 15  The innovative nature of intransitive perfectives in the Split-​S dialects was recognized by Goldenberg (1992). Other scholars, by contrast, have regarded such forms as archaic, e.g. Hopkins (1989), Mutzafi (2014), Barotto (2014a, 2014b), Coghill (2016). 13

 

Ergativity in Neo-Aramaic    899

Acknowledgments I am grateful to the anonymous reviewer, and to the editors of this volume for their valuable comments, which have helped to improve it in several places.

Abbreviations 1, first person; 2, second person; 3, third person; acc, accusative; art, article; cop, copula; d, d-​suffix; def, definite; f, feminine; fut, future; impf, imperfective; l, l-​suffix; m, masculine; neg, negative; part, participle; perf, perfective; pl, plural; pst, past; s, singular.

 

Chapter 37

Ergativit y i n A fri c a Christa König

37.1 Introduction 37.1.1 Typical Features of African Case Languages Before going into ergativity in Africa, it may be useful to give a brief overview of case marking in Africa, providing some salient features of case systems in Africa. Since ergativity is related to marked nominative (abbreviated as MNOM) the latter will be outlined as well. Whenever I use the term “case” in this chapter, then this is not to be understood as an abstract syntactic category but rather as a morphological means of marking functional relations. The following is based on my understanding of case as outlined in König (2008b), where I have given atypological overview of case in Africa taking the following definition of case as a basis: A case system is an inflexional system of marking nouns or noun phrases for the type of relationship they bear to their heads. Inflexional systems are expressed by affixes, tone, accent shift, or root reduction; adpositional systems are included only in so far as they encode core participants such as S [the intransitive subject], A [the transitive subject], and O [the object]. (König 2008a: 5)

According to König (2008a), grammaticalized case shows the following distribution in Africa: Of the 1,000 to 2,000 different languages, roughly 100 are case languages, only one really an ergative language, and a handful of additional ergative languages showing split systems. Two third of all case languages (sixty-​four languages) belong to the marked nominative type (also called extended-​ergative by Dixon 1994). Nominative-​accusative languages are the second frequent with less than one third belonging to this type (thirty languages).

 

Ergativity in Africa   901 Marked nominative is a mixture of the two basic types nominative-​accusative and ergative absolutive. It shares the alignment pattern with the nominative-​accusative case type in that A is treated like S and simultaneously different than O. And it shares with an ergative absolutive system that the functionally marked form of the case system is the one covering A. "Functionally marked" is the case which is morphologically marked; if any case is derived then it is the functionally marked one. The functionally marked case is not the one used for citation, it occurs in a smaller range of functions compared to the functionally unmarked one. Marked nominative languages are nearly a salient feature of Africa, they are the most widespread within Africa, they hardly occur outside of Africa, even though some recent studies suggest they are more numerous outside Africa than has previously been assumed (see Handschuh 2014). Handschuh includes under the label marked-​S language systems with marked nominatives or marked absolutives. Apart from East Africa they appear as well in North-​Western America and the Pacific region. Tone as an exclusive marker for case is perhaps the most striking feature of African case languages: It is not found elsewhere in the world, and it appears in marked nominative systems only. The following prediction can therefore be made: If a language uses exclusively tone to express case then it is a marked nominative language of Africa. Nearly one third of all marked nominative languages use exclusively tone to mark case. Marked nominative occurs particularly in the Surmic and Nilotic languages of Nilo-​ Saharan, and in the Amazigh, Cushitic, and Omotic languages of Afroasiatic (see König 2006, 2008a: 192f.). An overlap of marked nominative and ergative appears in Nilo-​ Saharan only. So far no Afroasiatic marked nominative language has been claimed to show ergative features.

37.1.2 No Case before the Verb Case in Africa works mostly after the verb. Before the verb, the case opposition is neutralized in nearly all African case languages, which means that all core participants appear in one case only and this is typically the functionally most unmarked one. I have proposed this feature under the slogan “no case before the verb” (see König 2008a: 240–​273). The “no case before the verb” split works regardless of the case system, regardless of whether verb initial or verb second as the basic constituent order of the language (verb final languages are excluded for obvious reasons). In ergative systems it is the absolutive which appears pre-​verbally for S, A and O. Historically and/​or synchronically, this neutralization of case is triggered by pragmatic functions, such as topic and/​or focus marking. Both appear in a preverbal slot still exhibiting features of the construction from which the topic and focus clause are derived. That is, a syntagma of a former copula clause, “it is X” or “as far as X is concerned” followed by a relative clause of the kind “which/​who does Y,” or by a main clause “X does Y” where the connection of the two clauses is appositional. The X participant still shows the case form required in its source

 

902   Christa König construction. In copula clauses of the kind “it is X,” the nominal predicate “X” appears in the functionally most unmarked form, which equals in marked nominative systems the accusative and in ergative systems the absolutive (see König 2008a: ch. 5).

37.1.3 The Discovery of Ergativity in Africa—​ a Historical Outline Perhaps the first striking feature of ergativity in Africa is its absence. For a long time, Africa was seen as a continent without ergativity. Handbooks on African languages claim there is no ergativity in Africa. So does Gregersen (1977: 63) in his compendium on African languages, and so does Creissels (2000: 236) in his chapter on typology in another handbook on African languages edited by Heine and Nurse (2000): But I am aware of no African language with intransitive subject markers identical to the object markers and different from the transitive subject markers, that is with a system of subject and object markers following an ergative pattern. (Creissels 2000: 236)

The world atlas of language structure (WALS) also shows no ergative dot in Africa on Map 98 presenting the alignment of case marking in 268 languages of the world (see Comrie 2005 in Haspelmath et al. 2005) (eds.). Andersen 1988 was the first to mention ergativity for the Western Nilotic language Päri. Shilluk the only full-​fledged ergative language in Africa discovered so far was no undescribed language when Miller and Gilley (2001) claimed it to be ergative. How come then that the ergative system of Shilluk has not been discovered before by the different scholars working on the language, such asWestermann (1912:  78), Kohnen (1933: 136), Tucker and Bryan (1966: 424–​425), or Buth (1981: 85–​86)? They all claim that the Shilluk people “tend to speak in the passive.” Miller and Gilley (2001: 48f.), however, doubted this analysis and postulated instead that the so-​called passive clause is not a passive but an active clause with an omitted agent and the so-​called preposition, which introduces the agent in a passive clause, is an ergative case marker. They argued that Kohnen might not have distinguished between two prepositions which differ only in their ATR (advanced tongue root) quality: yī [+ATR] and yī [-​ATR], the first one being a preposition used to introduce peripheral participants such as destination and the second being the ergative case marker. Structurally examples (1) and (2) are similar, example (3) by Tucker and Bryan example is more likely to be interpreted as (4). Shilluk (Northern Lwoo, West-​Nilotic, Nilo-​Saharan) (1) a-​pwot yi yan. PAST-​strike by me ‘He has been struck by me.’

Shilluk (Kohnen 1933)

 

Ergativity in Africa   903 (2)

(3)

(4)

bʊ̄ l á-c̍ wɔ̄l! Bol past.e-​call.tr ‘Col called Bol.’

yī erg

cʊ̄ l. Col

yáá cwɔ̂l I call ‘I was called (by…).’ yá á-c̍ wōl! 1.SG past.e-​call.tr ‘X called me.’

Shilluk (Miller & Gilley 2001: 49) Shilluk (Tucker and Bryan 1966: 427)

yī … erg …

Shilluk (Miller and Gilley 2001: 50)

The Jebel language Gaahmg(Nilo-​Saharan) might be a further example of a language where ergativity has not been discovered since the construction concerned was interpreted as a passive construction (see 37.2.3.1) A second reason may lie in the fact that ergative languages were believed to be absent in Africa. The first mention of ergativity in the literature (Andersen 1988; Miller & Gilley 2001)  led to the discovery of four languages with ergative features, all being closely genetically related: The Northern Lwoo languages, a subbranch of the Western Nilotic languages belonging to the Nilo-​Saharanphylum, all spoken also in an adjacent area of the Southern Sudan and Ethiopia. In addition to Shilluk, the other NorthernLwoo languages are Päri, Anywa and Jur-​Luwo. In the period following the publication of König (2008a), an additional five languages were claimed to be ergative. It seems since then that ergativity appears to have become a “fashionable” feature to look for. The other Northern Lwoo languages are split languages with either a marked nominative or an ergative absolutive system. The same case marker works as a nominative in certain clause types but in others as an ergative. To sum up, reasons for the rare occurrence of ergativity in Africa may lie in a combination of factors. As the historical overview shows, ergativity obviously has not been something that scholars were prepared to look for, as the Shilluk example shows. Also, the increasing number of ergative languages “discovered” particularly since 2008 might be in accordance with this observation. Since manyAfrican languages are still poorly described it is very likely that the amount of languages with ergative features to be discovered will increase in the future. Nevertheless, the main reason for the few ergative languages found so far may lie in the fact that principally there are not very many. Since ergativity in Africa often is closely related to marked nominative, with the latter being a characteristic feature of African case languages one may wonder whether this is a coincidence. Since 2008, four, or even five new candidates for ergative languages have been proposed. Dimmendaal (2014:  15)  presents a map with ten languages, all spoken in Eastern Africa stretching from the Northern Sudan, over the Southern Sudan, into Ethiopia, covering the Nilo-​Saharan and Niger Congo language phyla. In addition to

 

904   Christa König the already mentioned Northern-​Lwoo languages Shilluk, Päri, Jur-​Luwo (called Luwo by Dimmendaal) and Anywa there are the Surmic languages Majang and Tennet, the Koman language Uduk (classified as an isolate by Dimmendaal), the Jebel language Gaahmgof the Nilo-​Saharan phylum,and Tima (Katla), a Kordofanian language of the Niger-​Congo phylum. However, some of the languages proposed by Dimmendaal (2014) do not clearly seem to be ergative languages (see 37.2.3).

37.2  Case Studies In the following I will present case studies of different types of ergative languages, starting with the full-​fledged system in 37.2.1, over different types of split systems in 37.2.2., and finally presenting a few languages where it is not entirely clear whether they are ergative or not in 37.2.3.

37.2.1 A Full-​Fledged Ergative Language: Shilluk So far, there is only one full-​fledged ergative language in Africa and that is Shilluk, a Northern-​Lwoo language (Western Nilotic, Nilo-​Saharan). By “full-​fledged” I mean a system which basically works homogeneously throughout the language without any splits. According to Miller and Gilley (2001) the ergative case marker is expressed by the preposition yī plus a preceding downstep [!]. Shilluk has a syntactic split in what Miller and Gilley call independent clauses and dependent clauses. As is illustrated in clauses (5) to (8), the only participant which is encoded for case is A; S and O are always left morphologically unmarked—​ that is, the ergative takes a case marker and the absolutive is morphologically unmarked. Shilluk has in most clauses also an ergative constituent SV/​OVA order. If A is expressed nominally before the verb, the ergative case marker no longer appears on A. However, A has to be taken up as a pronoun after the verb, which shows the ergative case (see ­example 7). Shilluk (Northern Lwoo, West Nilotic, Nilo-​Saharan) Independent clause (5) twɔ́ŋ á-​gūt. Twong past.e-​drive.AP ‘Twong drove (a stake/​stakes) for tethering cows.’ (6) ʊ́ gīk á-​k̍ ēl! yī buffalo past.e-​spear.tr erg ‘Onyoti speared the buffalo.’

ʊɲótī. Onyoti

SV

Shilluk

(Miller and Gilley 2001: 42) OVA

Shilluk

(Miller and Gilley 2001: 45)

 

Ergativity in Africa   905 (7)

ʊ̄ ɲótī ʊ́ gīk á-​k̍ ēl! Onyoti buffalo past.e-​spear.tr ‘As for Onyoti, he speared the buffalo.’

yī erg

έn. A OVAPP Shilluk 3.SG.A (Miller and Gilley 2001: 45)

Miller and Gilley (2001) distinguish between three different types of dependent clauses, two behave like independent clauses with regard to the ergative case marker. The third one, called sequential clause, is different. The constituent order is verb initial, with basic VOA/​VS-​order. The transitive subject, A in sequential clauses, is also marked by the ergative (see ­example 8). Dependent clause (8) ā-​kwāāɲ gīncâm ` yī ɲīmε̄n. SQ-​take.tr food erg sister ‘… and then the sister took the food.’

VOA

Shilluk

(Miller and Gilley 2001: 57)

In focus clauses, the verb gets a suffix ​–​a, and the focused participant appears after the verb. A focused object clause appears in an AVO structure, where no ergative case is used (see 9). If A is focused, the constituent order is OVA, and an ergative case is used (see 10). (9)

ɲan d̯ájɔ̀ á-​r̍ ākk-​à` person female past.e-​grind.tr.REP-​FOC ‘The woman chose to grind the durra.’

byέl̀. AVO Shilluk grain.PL (Miller and Gilley 2001: 36)

(10) byέl á-​rākk-​à` yī ɲān d̯ájɔ̀. Shilluk grain.PL past.e-​grind.tr.REP-​FOC erg person female ‘The woman (not someone else) ground the durra.’  (Miller and Gilley 2001: 36) There is no data available for an intransitive focus clause; therefore it remains unclear how S is encoded in a focus clause. In sum, Shilluk is a pure ergative language, in all clause types the preposition yī plus downstep functions as an ergative case marker. Only focused object clauses with a preverbal A do not get any case marking. The case neutralization before the verb, which I have referred to as ‘no case before the verb’ fits into the general behavior of all verb initial or verb medial case languages in Africa and into the particular behavior of the other Northern Lwoo languages presented here, as well. Additional ergative features in Shilluk are present with regard to the constituent order in the independent clauses which is OVA and SV, that is, an ergative pattern with regard to cross-​reference.

 

906   Christa König

37.2.2 Split Systems 37.2.2.1 Marked Nominative/​Ergative Splits The first split presented in this section concerns languages with a marked nominative/​ergative split, meaning that the languages concerned either show an ergative or a marked nominative case system(but see also Coon and Preminger, Chapter 10, this volume for an alternative perspective on split ergative). The split is triggered by clause types. Interestingly, the case marker of both systems is identical, that is, the nominative of the marked nominative systems functions as an ergative in clauses which show an ergative absolutive system. These types of split systems occur in the genetically closely related Northern Lwoo languages, Päri, Anywa and Jur-​Luwo, a subbranch of the Western Nilotic. Many of the Nilotic languages are case languages. Of the three branches, Eastern, Southern Nilotic are throughout marked nominative and nearly always verb initial. In Western Nilotic many are no (longer) case languages. Among the few Western Nilotic languages with case there are the Northern Lwoo languages with marked nominative/​ergative split systems. The latter are verb medial, as most other Western Nilotic languages are. A special case is provided by Western Nilotic Dinka, a verb-​initial marked nominative language with a highly grammaticalized pre-​verbal topic slot, so that nearly all clauses appear with V2 order. Päri (Northern Lwoo, West-​Nilotic, Nilo-​Saharan)  I would like to exemplify the ergative system of Päri (Table 37.1), one of the three split ergative languages. All three, Päri, Anywa and Jur-​Luwo, are closely related Western Nilotic languages, belonging to the Northern Lwoo subbranch of Western Nilotic. The ergative case marker is expressed by a suffix -​ì with the allomorphs -​ì, -​ı, -​i, or -​ε,̀ -​e (Andersen 1988: 294). The syntax of Päri, as in Shilluk, is divided into two patterns, one runs under the label ‘NP-​initial’ clauses (called independent clauses by Miller & Gilley 2001), the other one under the label ‘VP-​initial’ clauses (called dependent clauses by Miller & Gilley 2001). Clauses which are subsumed under the NP-​initial type differ from the VP-​initial type in various respects, among them being case marking, cross referencing of pronouns on the verb, and word order. Relevant for our purpose here is the fact that NP-​initial clauses Table 37.1 Case inflexion in Päri. (Andersen 1988: 294, 297) Absolutive/​Accusative

Ergative/​Marked nominative

Meaning

Ùbúr mʌ̀ʌn

ùbúrr-​ì mʌ̀ʌn-​ì

Ubur woman

tɔ́ŋŋ

tɔ́ŋŋ-ì

spear

ʡáan1

ʔáan ̀-​ì

I

 

Ergativity in Africa   907 show an ergative system, as examples (11) to (13) may illustrate, whereas VP-​initial clauses show a marked nominative system (see ­examples  16 and 17). The same case marker functions in NP-​initial clauses as an ergative case encoding A, whereas in VP-​ initial clauses it functions as a nominative, encoding S and A. Note that in the glosses the case suffix -​ì is consistently glossed as ERG, irrespective of whether it serves as an ergative or a nominative. Päri (Northern Lwoo, West Nilotic, Nilo-​Saharan) NP-​initial clauses—​ergative system (11) ùbúr á-​túuk’. SV Päri Ubur COMP-​play ‘Ubur played.’ (Andersen 1988: 292) (12) jòobì à-​kèel buffalo COMP-​shoot ‘Ubur shot the buffalo.’

ùbúrr-​ì. Ubur-​ERG

O V A-​ERG

Päri (Andersen 1988: 293)

The only participant which gets a case inflexion is A, while S and O are left unmarked. The absolutive is morphologically zero. As soon as A  is placed before the verb, the ergative marker is no longer used, A appears in the absolutive form, and A has to be cross referenced on the verb by a suffix, cf. (13). Päri shows also the “no case before the verb” split. (13) ùbúr joobì á-​kèel-​é. Ubur buffalo COMP-​shoot-​3.SG.A ‘Ubur shot the buffalo.’

A O V-​A1

(14) á-​kwʌ̀ʌŋ-​ó. 1.SG.S-​swim+M-​SUF ‘I am swimming.’

S-​V

(15) á-​yáŋg`-​ì 1.SG.O-​skin+M-​2SG.A ‘You will knife me.’

Päri (Andersen 1988: 294) Päri (Andersen 1988: 296)

yàŋg-​ɔ.́ skin+M-​SUF

O-​V-​A

Päri (Andersen 1988: 297)

NP-​initial clauses show also an ergative alignment with regard to cross referencing (see 14 and 15). VP-​initial clauses show a marked nominative pattern since the same case marker, the suffix –​ì, encodes S and A, but never O. O is left unmarked. Therefore, the ergative marker of NP-​initial clauses serves as a nominative marker in VP-​initial clauses 1  The symbols A, S, O and V are used to show the constituent order of the example presented. If placed after a hyphen, affix order on the word level is signaled. The latter is marked only when indicating a case alignment pattern of cross referenced bound pronouns.

 

908   Christa König (see 16 and 17). As in NP-​initial clauses, before the verb, the core participants appear in the morphologically unmarked form (see 18). Päri -​verb initial clauses—​marked nominative (16) pìr ŋɔ̀ ì pʌ̂ʌr cícʊ̀-​ε̂. matter what LINK jump man-​ERG ‘Why did the man jump?’

V S-​ERG

Päri

(Andersen 1988: 318)

(17) pìr ŋɔ̀ ì cʊ̀ɔl yí ɲìpɔǹd`-​ὲ.        V O A-​ERG Päri matter what LINK call 3.SG.O child-​ERG ‘Why did the child call her?’ (Andersen 1988: 319) (18) pìr ŋɔ̀ ɲìpɔǹd`-​ɔ2̀ dháagɔì cʊ̀‑ɔl-έ AOV-​A OPRON Päri matter what child woman LINK call–​3.SG.A gɔ̀. 3SG.O ‘Why did the child call the woman?’ (Andersen 1988: 319) (19) pʌ́ʌyy-​ú`! jump-​2.PL ‘Jump!’

V-​S

Päri

(Andersen 1988: 317)

Päri has the following features: It is a split ergative/​marked nominative language with an opposition of two cases, where the same case marker serves as an ergative in NP-​initial and focus clauses (not presented here, see König 2008a: 99–​102), and as a nominative in verb initial clauses. Since the nominative is morphologically and functionally marked as opposed to the functionally/​and morphologically unmarked accusative, this represents a marked nominative system. Before the verb, the case inflexion is neutralized. In addition, Päri exhibits an ergative constituent order with an OVA, SV order in NP-​initial clauses. Jur-​Luwo, another closely related Northern Lwoo language, shows a similar case profile as Päri. The case marker is the suffix -​ê, which serves as an ergative case in NP-​initial clauses and as a nominative in VP-​initial clauses (König 2008a: 114–​117). Anywa (Northern Lwoo, West-​Nilotic, Nilo-​Saharan)  In Anywa, the third closely related language, also Northern Lwoo, there is a suffix -​Cì (with its allomorphs -​ε,̀ and O), which is related to the Päri case suffix -​ì. Reh (1996) analyses -​Cì in Anywa as a definiteness marker, which basically appears only after the verb. In NP-​initial clauses it is only A which can be marked by it, S and O are always left unmarked. As soon as A is placed before the verb, it no longer takes -​Cì (see 22)—​once more in accordance with the ‘no case before the verb’ neutralization principle. The NP initial clauses show also an ergative alignment with regard to cross-​referencing (see 21 and 22). 2 Suffix -​ɔ̀ remains unglossed by the author cited.

 

Ergativity in Africa   909 Anywa (Northern Lwoo, West Nilotic, Nilo-​Saharan) -​NP initial clause (20) wàaŋì3 lwʌ̀ʌr. SV Anywa grandmother.his be.afraid ‘His grandmother is afraid.’ (Reh 1996: 311) (21)

gε̄n-​ū-​tèén ɲìlàal-​lì. 3.PL.O-​FUT-​cook.PD.BEN child-​DEF ‘The child will cook for them.’

(22) ó dìmó tɔ̌ŋ wʌ̀nní son.mN Dimo spear.mN uncle.his ā-​kwáa-​ε̌. PAST-​ask.for-​3SG.A ‘Dimo’s son asked for his uncle’s spear.’

O-​V A

Anywa (Reh 1996: 190)

A O V-​A

Anywa

(Reh 1996: 311)

Verb initial clauses are fiddly in the sense that a nominal A after the verb hardly ever occurs.This is due to the fact that two core participants nominally expressed are not allowed after the verb, therefore (25) is excluded. The default way would be to place the nominal A before the verb and cross reference it on the verb, see(24). Nevertheless, it holds that the only participants which could host -​Cì are S and A (see 23), O is left unmarked. Verb initial, MNOM restricted to DEF (23) n-​ā-​òo jɔ́ɔwwì[...] C-​PAST-​come people.DEF4 ‘When the people come [...]’ (Reh 1996: 318) (24) ɲìlàál n-​ā-​kéel líεε kī child C-​PAST-​hit.3.SG.A elephant OBL ‘After the child had speared the elephant, [...]’ (25)

But *n-​ā-​kéel ɲìlàál-​lì C-​PAST-​hit.3SG.A child-​DEF tɔ̌ŋ, [...] spear

líεε elephant

V S.DEF

Anywa

tɔ̌ŋ, [...] spear

AVO

Anywa

kī OBL

*V A O

(Reh 1996: 315) Anywa

(Reh 1996: 315)

As argued in König (2008a: 108f) on the basis of Andersen (1988 and 2000), the definiteness suffix -​(C)ì of Anywa is a case marker but restricted to definiteness.5 The case function covered by -​(C)ì is either a nominative in a marked nominative system with 3 

Emphasis in bold, here and elsewhere, is mine. It is unclear whether in this example the definite marker -​(C)ì can be morphologically seperated from the preceding noun. 5  Concerning the relationship between case-​marking and definiteness see König 2008a: 224–​239. 4 

 

910   Christa König verb initial clauses, or an ergative in a ergative absolutive system with NP-​initial clauses. Anywa is a spilt ergative/​marked nominative language with case being restricted to definiteness. Ergativity Restricted to Clauses with Focused or Topicalized Object Constructions  There are a few genetically unrelated languages with ergative features in restricted clause types, namely clauses with a pragmatically marked object, which in some languages is focused objects, in othersis topicalized objects; in these pragmatically marked clauses, the transitive subject, A, gets a formal marking mostly after the verb which can be interpreted as an ergative marker. A fairly clear instance of this phenomenon is present in Uduk, where according to Killian, ergativity appears in clauses with topicalized objects, a less clear case is Tima, where in accordance with Dimmendaal, ergative marking appears in clauses with focused objects. Tima (Katla, Kordofanian, Niger-​Congo)  Dimmendaal (2009, 2010b)  claims Tima to be a language with ergative features. Tima, is a poorly known language spoken in the Nuba mountains of the Sudan, the genetically most closely related language is Katla. Tima’s wider genetic affiliation is still unclear (see Dimmendaal 2009: 331). Greenberg (1963) groups it with Katla into one of the five subgroups within the Kordofanian branch of Niger-​Congo (called Niger-​Kordofanian by Greenberg). The ergative case marker is expressed by a preposed nasal Nwhich precedes the transitive subject, A. In the pronominal system tone, in addition to the nasal, is used to express the ergative case. According to Dimmendaal (2010b), Tima is a split ergative language. Ergativity is restricted to focus clauses with focused objects (see 26). The rule “no case before the verb” does not apply to Tima. Tima (Niger-​Congo, Kordofanian) (26) káɓʋ̀h-​έ kʌ́lùk ŋ̀-​khamis meat-​FOC 3.PL.eat ERG-​Khamis ‘Khamis is eating a piece of meat.’

OVA

Tima

(Dimmendaal 2009: 348)

If A is a pronoun, the ergative case marker fuses with the pronoun to a verbal clitic, like-​nʌ́ for first person singular A in example (27), with the corresponding form -​ɗʌ in clauses without focus. Tima has three sets of pronouns, one independent, one expressed by verbal clitics in non-​focus clauses, and one expressed by verbal clitics with preverbal focus. In the latter, the ergative case marker fuses with the pronominal clitic. (27) ɪ́mmɔ́ŋ-​ɛ́m̀-​ mímíí-​nʌ́ fish-​FOC 1.SG-​cook-​ERG.1.SG ‘I am/​was cooking fish.’

Tima (Dimmendaal 2009: 349)

 

Ergativity in Africa   911 The ergative marker appears also in clauses with what Dimmendaal calls “verbal focus” (28): (28) ɲ̀cʌ́-ɗ ​ ə́k-​áá-​ŋàŋ-​nʌ́ PROG-​1.SG-​hit-​INST-​2.SG.O-​ERG.1.SG ‘I will hit you with it!’

Tima (Dimmendaal 2009: 350)

Tima shows a lot of freedom with regard to constituent order, intransitive subjects, S, occur always pre-​verbally (see 29), transitive subjects, A, can occur in a variety of positions such as AVO, OVA, AOV, so that Tima is a language with a basic SV, AVO order. (29) cɪ́ɪ́ɗɪ́ à-​hhɪ̀kɛ́r bɛ̀ɛ̀ thorn PRED-​sharp IDEO ‘The thorn is extremely sharp.’

SV

Tima (Dimmendaal 2009: 346)

Dimmendaal’s observes: Pre-​verbal agents are also possible with transitive predications [...]; the agent does not receive any formal marking in that case.2010: 236

This statement does not seem to hold with regard to his own examples. There are examples where the ergative marker appears pre-​verbally, as in examples (30) and (31): (30) m̀-​pɨ́nʌ̀-​wʌ̀ kúɗù ERG-​3.SG-​FOC catch ‘(S)he did/​caught it.’ (31)

ɲ-​ɪ̀yɛ́mɛ́ ú-​kùɗú-​í ERG-​who REL-​catch-​Q ‘Who cought it?’

SV

Tima (Dimmendaal 2009: 349)

SV

Tima (Dimmendaal 2009: 349)

From the examples presented it remains unclear, first, whether the ergative case appears only post-​verbally and, second, whether it is restricted to A. There seem to be examples where the ERG case marker possibly appears with S (see ­example 32):6 (32)

kʋ́-​m̀-​mwɔ́ɔ́k-​nɔ́k-​nɔ́-ɔ​ ́ɔ̀ŋ NEG-​1.SG-​drink-​ERG.1.SG-​NEG ‘I don’t drink/​I am not drinking.’

Tima (Dimmendaal 2009: 346)

Intransitive subjects are always placed pre-​verbally. 6 

Mary Laughren (p.c.) rightly observes, however, that this is conceivably not an intransitive sentence.

 

912   Christa König Dimmendaal argues that the ergative marker in Tima has emerged from an instrumental marker. The instrument is expressed by a preposition, which is a proclitic homorganic nasal N (see 33). (33) táàn kɨ ̀cɨ́mbʌ́rí ŋ̀-​kɨ́wáà hit child PREP-​stick ‘(S)he hit the child with a small stick.’

Tima (Dimmendaal 2010b: 237)

The peripheral agent in passive constructions is introduced by a preposition expressed by a nasal as well (see 34): (34) kɨcɨmbʌ́rí-​li í-​tá↓àn-​á t ̼áŋ ŋ̀-​kɨwáà m̀-​pɨ́nʌ̀ Tima child-​FOC REL-​hit-​PASS PREP-​stick PREP-​3.SG ‘The child was beaten with a small stick by him/​her.’     (Dimmendaal 2010b: 238) Dimmendaal (2010b: 236f.) points out that in his view the ergative has not been developed out of the agent marker used in passive constructions, particularly because the agent is not an obligatory participant in the passive construction and the passive has its own marker. The agent in passive constructions is also expressed by a preceding nasal. From a grammaticalization perspective it would be very likely that both markers, the instrumental, and the agent marker used in passive constructions, are of the same origin (see Narrog 2014).7 Uduk (Koman, Nilo-​Saharan)  According to Killian (2015), Uduk is a split ergative language: In topicalized object constructions with the topic being placed clause initially, A is marked by the ergative post-​verbally (see 36). Pre-​verbally there is no case inflexion (see 35). The basic constituent order in Uduk is AVO, SV (see 37). Topic constructions are quite common. The topicalized participant appears in preverbal position. The intransitive subject, S, nearly always appears in topic position; if it appears after the verb it also does not take the ergative marker. In topicalized object clauses (with O being topicalized),A appears after the verb. The transitive subject, A, gets an ergative marker when used after the verb. In preverbal position all participants appear in the unmarked absolutive. Since AVO/​SV is the default constituent order, the ergative marking is restricted to object topic clauses. The basic constituent order is SV/​AVO, and in topicalized clauses it changes to TOP VAERGO. The topicalized participant appears pre-​verbally. According to Killian (2015) the topic position has to be filled nearly obligatorily: (35) wat̀híʔ ʼcíṯh cwá mò man cut:PFV tree COMP ‘The man cut down a tree.’

7 

AVO

Uduk (Killian 2015: 180)

Note that according to Mary Laughren (p.c., see also Dench 1982) there is an opposite directionality from ergative to agent marker in Australian languages. This issue is in need of further research.

 

Ergativity in Africa   913 (36) tāshá wòʼc mà snake bite:IPV ERG.CL2 ‘The dog bit the snake.’ (37)

áʼdī wú-​ʼd s/​he die:PFV-​3sg ‘He died.’

ʼká dog

OVA

Uduk

(Killian 2015: 180)

mò COMP

SV

Uduk

Case marking in Uduk is conflated with gender marking. Two genders appear in the language, called class I and class II. Gender is expressed by clitics preceding the noun. Table 37.2 shows the gender/​case forms in Uduk. Table 37.2 Gender and case in Uduk. (Killian 2015) A in AVO

O in AVO

A in OVA

O in OVA

Class I Class II

ø à

Ø ā, à

-​(N) ā/​à -​(N) mā/​mà

ø à

Case

ABS

ABS

ERG

ABS

37.2.3 Questionable Cases of Ergativity 37.2.3.1 Gaahmg (Jebel, Nilo-​Saharan) Gaahmg (also called Gaam or Ingassana) (Jebel, Nilo-​Saharan) is probably the last (Eastern) Jebel language spoken today to have a basic AVO, SV word order. Stirtz (2012:  281)  claims that subjects (and objects) are not marked for case. Dimmendaal (2014) argues that what has been called an agented passive by Stirtz(2012) can be interpreted as an ergative. The agented passive has the following properties: The agented passive clitic decreases the valency of the clause by demoting the agent to non-​ argument status, although an explicit expression of the agent is still required. In clauses with agents encoded by post-​verbal constructions, an agented passive clitic= E᷇ /​=ÉĒ, which agrees with a genitive agent in number, is attached to the verb stem. (Stirtz 2012: 284)

Only postverbal A gets the genitive marker, preverbal A doesnot (see 38 and 39), and S has to occur pre-​verbally. In Dimmendaal’s interpretation the genitive case functions as an ergative case marker. Since in agented passive constructions it is obligatory to express the agent in this type of agented passive construction, it could reflect the birth of an ergative system (Dimmendaal 2014: 9). (38)

ɲǝ́m-​ǝ̄n-​s-​ aggaàr(-​ɛ̀) VA Gaahmg break.CAUS-​AP-​COMP-​PAS.A hunter.GEN ‘Something was broken by a hunter making someone break it.’   (Stirtz 2012: 215)

 

914   Christa König (39) ɟɛ̄n ɲām-​sá guld̪ūn person break-​COMP branch.DEF ‘The person broke the branch.’

AVO

(40) guldū̪ n ɲam᷄ -​s=ɛ̄́ ɟɛn̂ branch break-​COMP=PAS.A person.GEN ‘The branch was broken by the person.’

OVA

Gaahmg (Stirtz 2012: 215) Gaahmg (Stirtz 2012: 215)

As Stirtz (2012: 284) points out, the agented passive construction is used for focused objects which appear pre-​verbally, and the A has to appear post-​verbally (see 41–​43). The encoding of post-​verbal A is rather complex: A can either be encoded by the general preposition ɛ́ (see 41), by the genitive (43) or, if it is a pronoun, by the prefix d̪-​ (42). (41) a. gaar cúə́ … nām-​án=ɛ᷇ ɛ́ pork sweet /​nam/​eat-​CONT.P=PAS.A GP káɛ-​́ gg=a witchdoctor-​PL=DEF ‘Sweet pork … was being eaten by witchdoctors.’ b. gààr cúə́ … nām-​án=ɛˊɛ̄ pork sweet /​nam/​eat-​CONT=ERG kááɛˊggà sorcerer.ERG.PLDEF ‘Sweet pork the sorcerers were eating.’

Gaahmg

(Stirtz 2012: 285)

ɛ́ GP

Gaahmg

(Stirtz 2013: 2)

(42) ɟāām kəə́ m̀ -​s=ī d̪-ɛ​ ɛ́ n̄ someone /​káam/​bothered.CAUS-​COMP=PAS.A PP-​3sO wá. not ‘No one was bothered by it.’ (Stirtz 2012: 285)

Gaahmg

(43) nāms ɲáɔ-​́ s=ɛ᷇ food /​ɲaw/​need-​COMP=PAS.A ‘Food is needed by the hunter.’

Gaahmg

āggāar hunter.GEN

(Stirtz 2012: 285)

The genitive triggers additional tone change. Stirtz (2013) himself redefines his agented passive of Stirtz (2012) as an ergative construction. The passive translation is converted into an active one (compare 41a and 41b), the agented passive suffix of the verb is now called an ergative enclitic. The tonal change of the noun is called ergative case by him (see Stirtz 2013: 2) The general preposition ɛˊ is not taken as part of the ergative encoding but is still called a general preposition which among other functions appears with an ergative case marking (see Stirtz 2013: 6, table 2), and for post-​verbal pronouns he considers the prefix d̪-​plus tonal changes as being ergative.

 

Ergativity in Africa   915 If Stirtz(2013) interprets the agented passive construction as an ergative it remains unclear whether all three markers should or should not count as ergative, namely the general preposition ɛ́, the prefix d̪-​for pronouns, and genitive expressed by tone change whereby the LH pattern of the noun aggáár ‘hunter’ changes to a ML tone melody āggāar only or tone change plus the suffix -​ɛ̀.

37.2.3.2 Majang (Surmic, Nilo-​Saharan) There have been several attempts to claim that Majang is ergative (Randal 2000: 74), Joswig (2011) in his brief grammatical sketch of Majang is one of the latest. On the basis of Unseth (1989a, 1989b) it has to be classified as marked nominative. According to Joswig, nouns appear in two different tonal forms (compare 44 to 45). The nominal predicate appears in the absolutive form (see 46). Unfortunately there are no glosses or detailed descriptions of the case system. (44) malɛ íɗì tɔɔn. beat man.ERG child.ABS8 ‘The man beats the child.’ (45)

(46)

malɛ tɔn íɗít. beat child.ERG man.ABS ‘The child beats the man.’ sɛɛn agalt. ^ 3.SG.S thief.ABS ‘He is a thief.’

Majang (Joswig 2011: 30) Majang (Joswig 2011: 30) Majang (Joswig 2011: 22)

If Majang is an ergative language, there would be at least some nouns where according to Joswig’s description the absolutive form is longer than the ergative form, as with the noun ‘man’: íɗít man.ABS and íɗì man.ERG. This however would be in need of explanation. If any form in the opposition absolutive vs. ergative is morphologically marked it should be the ergative and not the absolutive. Joswig (2011: 31) presents a set of “subject pronouns” and a set of “object pronouns.” For second and third person singular there are however different forms for S and O: iin second person singular S, A; sɛɛnfor third person singular A, S; ììnfor second person singular O, and cɛ̀ɛ̀n for third person singular O. Second and third person singular pronouns do not follow an ergative pattern but an accusative one. The picture of Majang remains somewhat puzzling.

37.2.2.3 Tennet (Surmic, Nilo-​Saharan) According to Randal (2000),Tennet is a marked nominative language with the nominative being marked by a suffix -​i and the accusative being left unmarked. The basic word order is VAO/​VS. In complement clauses, S has to appear pre-​verbally in an SV 8 

Glosses added by the present author.

 

916   Christa König order (see 48) in the morphologically unmarked form. In a corresponding transitive complement clause the VAO order remains (see 47) and A appears in the morphologically marked form. (47) órông Lɔwɔ́r-​i kákát Lɔhɑ́m-​i want Lowor-​NOM spear.SBJ Loham-​NOM ‘Lowor wants Loham to spear the bull.’

áríz bull.ACC

Tennet (Randal 2000: 72)

Comparing transitive and intransitive complement clauses, the only participant which gets marked is A. Therefore the resulting pattern is an ergative one. On this basis, Randal (2000: 72) claims Tennet to be ergative in complement clauses. But as pointed out in König (2012), although this observation is correct, there is an alternative interpretation of the situation, which seems to be more plausible. Taking into account that in Tennet as in all other marked nominative languages in the area the split condition “no case before the verb” applies, there need not to be an ergative system at work, there is just case neutralization. If there is no further evidence for the claim that Tennet is ergative in certain subordinate clauses, the ergative hypothesis seems to remain controversial. Since for Dimmendaal (2014) the supposedly ergative feature of Tennet is crucial for his historical reconstruction of ergativity in Africa, his entire reconstruction appears to be built on shaky grounds. The following sentences of Tennet (48a and 48b) serve as a basis for claiming that Tennet shows ergative case marking in certain subordinate clauses. In complement clauses S and O appear in the unmarked accusative (compare 48a and 48b), and A in the marked nominative (see 48a), therefore an ergative pattern appears according to which S and O are treated equally and A differently in the only morphologically marked case form. (48) a órông Lɔwɔ́r-​i kα̌kα̌t Lohám-​ı want Lowor-​NOM spear.SBJ Loham-​NOM Lowor wants Loham to spear the bull. b órông Lɔwɔ́r-​i Lɔhɑ́m want Lowor-​NOM Loham ‘Lowor wants Loham to come.’

kíkíya come.SUBJ

α̌rız. Tennet bull.ACC9 (Randal 2000: 72) Tennet (Randal 2000: 72)

While claims such as the one above by Dimmendaal suggest that Tennet as well has ergative features, they are not supported by my analysis. As argued in König (2008a) and König (2012), it is very likely that Tennet and Majang are both marked nominative and that the ergative feature postulated by Dimmendaal is an epi-​phenomenal effect and, hence, can well be explained by the “no case before the verb” neutralization principle in a marked nominative system. More evidence is needed to establish whether these two languages really are ergative rather marked nominative. 9 

Glossed as áríz by the author cited.

 

Ergativity in Africa   917

37.3  Ergative Features 37.3.1 Constituent Order Ergativity in Africa is closely related to constituent order. Table 37.3 gives an overview of the ergative languages and the constituent order in which the ergativity appears. Languages which are listed with an initial question mark indicate that an ergative analysis is problematic. As can be seen there, nearly all ergative languages are verb medial and the constituent order in which ergative marking appears is OVA, SV. The constituent order itself is ergative. Since all languages in Table 37.3, as all verb medialand verb initial case languages in Africa in general, neutralize case before the verb, a shift in a marked nominative language with a VAO/​VS order to an OVA/​SV order would automatically result in an ergative system. The only core participant which is able to get a case coding is A; O and S in preverbal position can only get the neutralized form. In Tima it is not completely clear whether case neutralization applies to the preverbal position. Table 37.3 Overview of ergativity in Africa

Nilo-​Saharan: Shilluk Anywa Päri Jur-​Luwo ?Majang ?Tirma-​Chai ?Tennet ?Gaahmg Uduk Niger-​Congo: Tima

West-​Nilotic West-​Nilotic West-​Nilotic West-​Nilotic Surmic Surmic Surmic Jebel Koman

Constituent order in which ERG appears

No case before the verb (NCBV) Basic order

Case

OVA/​SV OVA/​SV OVA/​SV OVA/​SV

NCBV NCBV NCBV NCBV NCBV NCBV NCBV NCBV NCBV

verb middle verb middle verb middle verb middle verb initial verb initial verb-​middle verb middle

ERG ERG/​MNOM ERG/​MNOM ERG/​MNOM MNOM MNOM MNOM ERG/​none ERG/​none

?

verb middle

ERG/​none

OVA/​SV OVA/​SV

Katla OVA/​SV Kordofanian

The constituent order SV and OVA is crosslinguistically a rare phenomenon. A basic OVA word order contradicts Greenberg’s first universal constituent order according to which [i]‌n declarative sentences with nominal subject and object, the dominant order is almost always one in which the subject precedes the object. (Greenberg 1963: 61)

 

918   Christa König It contradicts also the findings of Mallinson and Blake, made about ergative languages, according to which “in practically every ergative language A precedes O” (Mallinson and Blake 1981: 123). It is very unlikely that OVA reflects historically any basic order. The unusual constituent order of the Western Nilotic languages in question thus is in need of an explanation. Andersen (1988) and Reh (1996) agree that OVA10 is an older constituent order which later on led to AOV—​a development for which Andersen (1988) claims the following sequence of word order changes: VAO > OVA > AOV Ergativity is primarily found in the Nilo-​Saharan family, there is only one instance of a possibly Niger-​Congo language showing ergative features, and that is Tima, the Kordofanian language spoken in the Nuba mountains. Figure 37.1 presents a simplified model of the Nilo-​Saharan language family focusing on the branches which are of relevance with regard to ergativity: Ergative languages appear in the Koman branch with Uduk, showing an ergative system in focused object clauses, and in the East Sudanic branch. A more detailed view of case in East Sudanic is given in Figure 37.2. Nilo-Saharan Satellite-Core

Core

East Sudanic

Satellite Koman

Berta

Uduk

Figure 37.1.  Selection of Nilo-​Saharan language branches Source: Bender (1996)

The region in the Southern Sudan and adjacent areas in Ethiopia appear to be one center of ergativity in Africa. The languages concerned are genetically closely related, namely the Northern Lwoo languages of Western Nilotic. It is not yet clear whether the Surmic family of the Nilo-​Saharan phylum has a second concentration of ergative languages but, as argued above, Majang and Tennet are not really clear cases of ergativity. Ergativity appears basically in the region which is the primary region for case languages in Africa in general. Figure 37.2 illustrates the closeness of ergativity and marked nominative in Nilotic. Ergativity is marked in bold and by capitals, marked nominative in italics. Figure 37.2 gives an overview of case languages in East Sudanic, focusing on ergative languages and on marked nominative languages. With regard to the latter only a selection is 10 

They use S instead of A. In order to be consistent, I have changed S to A.

 

Ergativity in Africa   919

East Sudanic

Jebel North East

GAAHMG

Surmic

South East

Tennet Majang SHILLUK

Southern West

Nilotic

East

South

Northern Lwoo Dinka Maa

PÄRI ANYWA JUR-LUWO

Teso Kalenjin Datooga

Figure  37.2  A  selection of case languages in East Sudanic (Nilo-​Saharan) (italics indicate marked nominative; italics, bold caps indicate marked nominative-​vergative; bold caps indicate ergative)

presented. As can be seen in Figure 37.2, Nilotic languages show a high concentration of case languages. Eastern and Southern Nilotic languages are throughout marked nominative languages, with verb initial constituent order, where case is exclusively expressed by tone. Western Nilotic languages are throughout verb medial languages, the Southern Lwoo languages (such as Luo, Lango, Acholi, Kumam, Labwor, or Jopadhola) have no case inflection whereas at least some of the Northern Lwoo languages have case inflections. Generally speaking, the Western Nilotic languages differ from Eastern and Southern Nilotic, being verb medial languages. Some of them are marked nominative (indicated in italics in Figure 37.2), some of them are marked nominative-​ergative (indicated in italics and bold and capitals), and one is ergative (indicated in bold and capitals). Surmic languages present an additional branch of Nilo-​Saharan where many case languages are found. All are marked nominative. I have listed two of them, Tennet and Majang, in Figure 37.2, since they have also been claimed in the literature to possess ergative features. All languages of Figure 37.2 obey the split ‘no case before the verb’ principle. Surmic and Nilotic languages are both subbranches of South East Sudanic. They both are spoken in East Africa, in adjacent areas, its core being located in the Sudan, Southern Ethiopia, Northern Kenya, and Eastern Uganda.

37.3.2 The emergence of ergativity The following has to be seen as a preliminary attempt to make suggestions about the emergence of ergativity, being aware that there are neither historical records nor

 

920   Christa König appropriate synchronic descriptions in support of the hypotheses proposed. Given this situation, it does not seem empirically sound to formulate any reasonable hypothesis on possible ergativity in the hypothetically set-​up Proto-​Surmic or Proto-​Eastern Sudanic (a much higher node within Nilo-​Saharan). I therefore will restrict myself to languages where ergativity seems to be a securely established feature, ignoring alternative accounts, such as Dimmendaal (2014), for the reasons mentioned. As has been argued in König (2012: 35–​39), there are two hypotheses about the origin of ergativity in split marked nominative/​ergative languages:  Either marked nominative was first and ergative absolutive second or the other way round.For the Northern Lwoo languages Anywa, Päri, and Jur-​Luwo,the ergative marker must be assumed to go back to a definite marker, and the grammaticalization went from definite marker to marked nominative to ergative (see Table 37.4). Since main clauses tend to reflect historically younger and subordinate clauses historically older stages, it is most likely that in Anywa, Päri and Jur-​Luwo the development has taken place from nominative to ergative; note that—roughly speaking—ergative marking appears in main clauses and marked nominative marking in subordinate clauses in these languages (further see König 2012: 36). In Northern Lwoo, the development proposed here is from nominative to ergative on the level of case forms and from a marked nominative to an ergative absolutive system on the level of the case alignment system. The change of the case system is triggered by a constituent order change from VAO/​VS to OVA/​SV. This interpretation is in line with the word-​order changes claimed by Anderson and Reh and discussed in section 37.3.1, with VAO being historically older than OVA (König 2008a, 2012: 35). In Shilluk, it is most likely that the ergative case marker goes back to a preposition introducing peripheral, instrument-​like participants or the agent in a passive clause (further see König 2012: 37–​38).Worldwide, one main source for ergative markers are agent markers which encode peripheral agents in passive clauses (see Anderson 1977; Table 37.4. The emergence of a case marker in Anywa, Päri and Jur-​Luwo. (König 2012: 36) Proto Anywa, Päri & Jur-​Luwo

Anywa

 

Päri

 

Päri

-​*C1̀  

> -​C1̀    

>  

-​ì Jur-​Luwo

>  

low-​extra low Jur-​Luwo

 

 

 

>

-​ê

 

-​ê

Definite marker

 

(i) Definite marker (ii) Case marker MNOM/​ERG with definite nouns only

 

Case marker MNOM/​ERG with nouns definite & indefinite

 

Case marker MNOM/​ ERG personal pronouns (kinship terms)

 

Ergativity in Africa   921 Givón 1980; Lehmann 1982; Garrett 1990; Harris & Campbell 1995: 243–245; Narrog 2014; and Aldridge, Chapter 21, and McGregor, Chapter 19, this volume). The grammaticalization of the ergative case in Shilluk would be in line with Reh’s (1996) analysis for the emergence of the odd OVA order in the genetically closely related language Anywa (see also König 2008a). A third origin of ergative case might be the genitive case as postulated by Dimmendaal (2014) for Gaahmg). A  development from genitive to ergative is well established in grammaticalization theory (Narrog 2014: 81); a paradigm example is provided by the Sino-​Tibetan language Ladakhi (Palancar 2002: 229). The ergative claims with regard to Surmic in general and subordinate clauses in Tennet in particular are of a different nature (see 37.2.3). The supposedly Tennet ergative has been taken as an argument to claim that the verb initial word order in Nilotic and Surmic is an innovation (Dimmendaal 2014: 11). On the basis of the observations made in the preceding paragraphs, such a claim must be treated with care. And much the same applies to the claim that proto-​Surmic, or even Eastern Sudanic were ergative languages (Dimmendaal 2014), given the controversial information on ergativity in Tennet, Majang, and Gaahmg (see König 2012: 38–​39). If in fact all marked nominative languages of Eastern Sudanic go back to an earlier ergative language, as implied by Dimmendaal (2014: 16), then one might wish to know why virtually all languages underwent a shift from ergativity to full-​fledged marked nominative languages while none of the many languages has retained a full-​fledged ergative case format. König (2012: 35–38), by contrast, favors the hypothesis of an evolution from marked nominative to ergative. This analysis is based on the situation in the Northern-​ Lwoo languages, which is also suggestive of a development from marked nominative to ergative. The main argument underlying this hypothesis is that subordinate clauses are more likely to reflect the historically older stage than main clauses (König 2012: 40–​42). For the ergative case markers in the Northern Lwoo languages, two sources have been claimed: First, a definiteness marker and, second, an oblique agent marker either in a passive construction or as the extension of a peripheral participant, such as an instrument (see e.g. Narrog 2014).

37.4 Conclusion Ergativity in Africa is an attractive topic, it exhibits features which are unique and of crosslinguistic interest. But only Shilluk can be considered to be a full-​fledged ergative language. Five to seven more languages (depending on the particular position one wishes to adopt) are split ergative languages of two kinds: Either they have a split ergative/​no-​case, or a split ergative/​marked nominative system.

 

922   Christa König Unique features of African ergativity are: • First, its co-​occurrence with marked nominative, as in the three Nilotic languages showing a marked nominative/​ergative absolutive split. • Second, its neutralization before the verb. The ‘no case before the verb’ principle is not restricted to ergativity in Africa, rather, it is also found generally in verb initial and verb medial case languages. • Third, its association with an unusual constituent order.Nearly all ergative languages show their ergativity in the odd OVA/​SV order. • Fourth, its occurrence in pragmatically marked constructions only. Often ergativity is related to the pragmatic marking of topicalized or focused objects. Since the latter appear pre-​verbally, and A then moves after the verb whereas S remains in its pre-​verbal position, an ergative pattern arises. The strong relation between ergativity and word order might be a reflection of its young age and an indication of an early stage of grammaticalization. The result is an ergative/​no-​ case split where ergative case marking appears in focused or topicalized object constructions while elsewhere there is no case inflexion. This appears to be apply to the Koman language Uduk, the Kordofanian language Tima, and perhaps also to Gaahmg. Two contrasting scenarios have been proposed for the emergence of ergativity. Whereas Dimmendaal claims that the East African Surmic shifted from ergative to marked nominative, König (2008a) favors a hypothesis in the opposite direction—​a position defended in this chapter. Main sources for ergative case are definiteness markers, as in the Northern Lwoo languages, prepositions for peripheral participants such as instruments or agents in passive constructions, as in Shilluk, and the genitive case, as in Gaahmg. Due to lack of historical records and appropriate synchronic descriptions, e.g. of the Surmic languages Tennet and Majang, reconstructions on ergativitymust remain tentative. But considering the interest that ergativity and related phenomena have found more recently among students of African languages, it is hoped that this situation will change in the future.

Abbreviations 1, first person; 2, second person; 3, third person; A, transitive subject function; ABS, absolutive; ACC, accusative; AP, anti-​passive; BEN, benefactive; C, subordinating conjunction; CAUS, causative; CL2, class I gender; CL2, class II gender; CONT.P, past continuous; COMP, completive (Andersen, Killian); COMP, complementizer (Miller & Gilley); DEF, definite; DEM, demonstrative; E, evidential; ERG, ergative; GP, general preposition; IDEO, ideophone; INST, instrument; IPV, imperfective; FOC, focus; FUT, future; LINK, linker; mN, modified noun (form); M, masculine; NEG, negation; MNOM, marked nominative; NOM, nominative;

 

Ergativity in Africa   923 NCBV, no case before the verb; O, transitive object function; OBL, oblique; PP, prepositional prefix; PASS, passive; PAS.A, agented passive; PAST, past tense; PREP, preposition; PRED, predicate; PFV, perfective; PD, patient deleted; PL, plural; Q, question marker; REL, relative marker; REP, repetitive; S, intransitive subject function; s, singular; SG, singular; SQ, sequential; SUBJ, subjunctive; SUF, suffix; V, Verb; TR, transitive.

 

Chapter 38

Ergativ i t y i n Tibeto-​B u rma n Shobhana Chelliah

38.1 Introduction The availability of high-​quality descriptions of Tibeto-​Burman languages in recent years has greatly increased our understanding of the case systems of this large family. In this chapter, I review the literature on morphological case alignment and differential marking in Dolakha Newar, Chintang, Tibetan, Meitei, and Burmese, using these to illustrate four patterns of core argument marking in Tibeto-​Burman. The first pattern is exemplified by languages like Dolakha Newar and Chintang which exhibit ergative alignment in a straightforward way as defined in seminal writings on the topic, e.g. Silverstein (1976) and Dixon (1994). That is, intransitive subjects (S) and transitive objects (P) exhibit similar morphological marking while transitive subjects (A), pattern differently. Pattern 2 is exemplified by Tibetan varieties where there is a strong tendency for marking along ergative lines (A takes ergative marking, S takes absolutive (zero), and P may take zero or other marking). However, while there is a high probability of A being marked differently from S, when factors congruent with transitivity, such as affectedness of the P, are weak (cf. the transitivity hypothesis of Hopper and Thompson 1980), an A may take zero marking like S. As will be seen in section 38.3.1, the strength or weakness of transitivity is only one factor determining the distribution of A marking in Tibetan. Information status, contrastiveness and temporal and spatial distance from the scene of telling are also significant; for example, A may be unmarked like S if it refers to temporally close events. Pattern 3, is illustrated by Burmese which has accusative alignment, i.e. P patterns differently from transitive and intransitive subjects. Here, information structure (topicality, contrastiveness, and theticity) is predictive of when “subject” marking, as it is traditionally called in Burmese linguistics, occurs. Many of the functions

 

Ergativity in Tibeto-Burman    925 of subject marking in Burmese are strikingly similar to those found for A and S marking in Tibetan. Pattern 4 is exemplified by Meitei, which falls somewhere between Tibetan and Burmese with respect to A and S marking. A marking is preferred with volitional activities, so transitivity becomes relevant in as much as transitivity and volitionality overlap. A marking is also more often observed in telic and punctual clauses with highly affected or individuated patients—​again bringing to mind Hopper and Thompson’s catalog of consequences of reduced and heightened transitivity. However marking of A with transitive clauses is not obligatory; rather, it is determined by the same factors of information and discourse structuring seen for Tibetan and Burmese. The four patterns described here could be further expanded because, as LaPolla (1995a) states, Tibeto-​Burman core argument case marking patterns fall on a continuum. On one end of the continuum A and S marking is “fully systematic” and determined by transitivity, e.g. ergative alignment in Dolakha Newar and languages of the Kiranti group like Chintang. The other end of the continuum is typified by “non-​ systematic” case marking as exemplified by languages like Burmese where functional features such as information structure are relevant for A and S marking. In the middle of the continuum are languages like Tibetan and Meitei where both information structure and transitivity play a role in whether A and S are marked. The presence or absence of case marking motivated by information or discourse structuring has been called ‘optional ergative marking’ or ‘pragmatic ergative marking’ (McGregor 2009, 2010; Hyslop 2010; DeLancey 2011). Recent treatments use the term differential agent or differential subject marking (Jenny and Hnin Tun 2013). Differential marking refers to a patterning where prominent arguments take overt marking but non-​prominent arguments in the same roles take zero marking (Aissen 2003b: 436).1 Prominence is assessed on the strength of features such as animacy, definiteness, volitionality, and kinesis (Haspelmath 2005). I adopt the term differential marking recognizing that factors triggering Tibeto-​Burman differential marking are unique in that they are determined by information and discourse structure. The chapter is organized as follows. Section 38.2 provides an overview of the Tibeto-​ Burman language family. Section 38.3 provides descriptions of A and S marking in four languages of the family. In section 38.4 I review factors which overlap with the distribution of A marking: past time reference, transitivity, volitionality, and the need for disambiguation. I evaluate to what extent these factors independently or in aggregate predict the distribution of A marking. In section 38.5 I review a set of parallel issues: the historical evolution of core argument marking in Tibeto-​Burman, data sources for the description of case, and the place of Tibeto-​Burman case in alignment typology. Section 38.6 provides a summary and conclusion. 1  In some practices, ‘optional ergative marking’ is reserved for alternations between a case marker and zero while differential marking is for alternations between two different case markers (McGregor 2009: 497).

 

926   Shobhana Chelliah

38.2  The Tibeto-​Burman Languages Tibeto-​Burman languages are spoken south of and across the Himalayas. Going west to east, Tibeto-​Burman languages can be found in Pakistan and Kashmir, Bangladesh, Himachal Pradesh, Uttarakhand, the seven states of Northeast India, Nepal, Bhutan, Sikkim, Tibet, China, Burma, Thailand, Laos, and Vietnam. There are approximately 250 Tibeto-​Burman languages in South Asia and at least 200 more in China and Southeast Asia.2 While reconstruction at the higher subgroup levels is ongoing, there are some accepted lower level subgroupings. I provide an overview of these subgroupings in Table 38.1 based on Burling (2003) and Genetti (2015). Table 38.1 List of Tibeto-​Burman language groups with notes on location, number of languages, and representative grammatical descriptions Groupinga

Location and number

Karenic

Burma and Thailand; four major branches, 25–​30 varieties Lolo-​Burmese China, Burma, Thailand, Laos, Vietnam; four major subgroups, Naxi, Burmish, Jinuo, Loloish rGyalrongic Sichuan; three branches Qiangicb Sichuan and Yunnan, 9 varieties Nungish Yunnan, Myanmar Bodish Tibet, Sikkim, Bhutan; 141 Tibetan varieties divided into seven subgroups

Tamangic

Nepal

Western Himalayish Kiranti

Himachal Pradesh, Jammu and Kashmir, India–​Nepal border Nepal, 29 languages

Tani Kachinicd

Northeast India, Arunachal Pradesh Northeast India, Tripura, Manipur, Meghalaya, Assam, Nagaland, northern Burma, China Northeast India, Assam

Bodo-​Koch

References to case and NP marking in representative languages Eastern Kayah Li (Solnit 1997), Kayan (Manson 2010) Burmese (Soe 1999), Lhaovo (Sawada 2012) Nyagrong-​Minyag (Suzuki 2012) Dulong (LaPolla 2003b) Qiang (LaPolla 2003b) Bunan (Widmer 2015), Kurtöp (Hyslop 2010), Lamjung Yolmo (Gawne 2013) Ladakhi (Western Archaic), Zanskar (Western Innovative) Lhasa Tibetan (Central) (references in section 38.3)c Manange (Bond, Hildebrandt, Dhakal 2013a) Darma (Willis 2011) Belhare (Bickel 2003), Chintang (Stoll and Bickel 2013) Apatani (Sun 2003) Jingpho (Dài and Diehl 2003) Atong (van Breugel 2014)

2  LaPolla (2014) notes that the postulated numbers for Tibeto-​Burman languages should be accepted with healthy skepticism. Chinese linguistic surveys tend to lump different languages together while in India, each variety is often given a distinct name. Thus, it is difficult to determine the exact number of Tibeto-​Burman languages.

 

Ergativity in Tibeto-Burman    927

Groupinga

Location and number

Konyak Ao Angami-​ Pochuri Zeme

Northeast India, Arunachal Pradesh Northeast India, Nagaland Northeast India, Nagaland, and Manipur Northeast India, Nagaland, and Manipur Northeast India, Manipur, Mizoram, Meghalaya, Bangladesh, and Burma Nepal Northeast India, Manipur

Kukish Newarice Unaffiliated languages

References to case and NP marking in representative languages Chang (Coupe 2011a,b) Mongsen Ao (Coupe 2008, 2011a) Angami (Annamalai and Giridhar 1991; Giridhar 1980) Inpui (Waikhom 2014) Khumi (Peterson 2011); Daai Chin (So-​Hartmann 2009) Dolakha Newar (Genetti 2007) Meitei (Chelliah 1997)

a

The table excludes Sinitic. It also excludes unclassified languages and extinct languages such as Pyu which was spoken in Burma but was extinct by the twelfth century. b Chirkova (2012) argues that Qiangic is better considered a typological rather than genetic grouping. c Zeisler (p.c. 2015) believes that Ladakhi should refer to all varieties spoken in Ladakh and this would include Zanskari. Also, she notes that the so-​called innovative varieties are conservative in some grammatical and phonological features. d In Burling’s classification (2003), Jingpaw is part of the Luish group, which includes languages like Kadu and Sak. e Genetti explains that there are 19 postulated groups comprised of one language or a small set of languages. Newaric is one of these groups.

Many of the languages spoken in the East-​Himalayish region, e.g. languages from the Kiranti branch (Belhare, Chintang, Kham, Chepang, Limbu, Dumi, Sunwari) and Newaric branch (Dolakha Newar) exhibit systematic ergative alignment as do some languages in Northeast India (e.g. Mizo (Kukish) and Chang (Konyak)). Many languages are said to have an ergative marker but it is not clear if they exhibit ergative alignment, e.g. the Gyalrong languages (Cogtse Gyarong (Nagano 2003) and Caodeng rGyalrong (Sun 2003). For many other languages, case alignment is neither straightforwardly ergative nor accusative; rather, information and discourse structure play a major role in A and S marking instead of, or in addition to, transitivity. It is still to be determined exactly how prevalent these systems are in Tibeto-​Burman, but the consensus is that these unusual case marking types are predominant in the languages of Northeast India, Burma, and Bangladesh and probably beyond (LaPolla 1995a; DeLancey 2011).

38.3 Exemplars This section provides examples of case marking in five well documented languages. Data from Tibetan, Meitei, Burmese illustrate non-​syntactically determined A and S marking. Examples from Dolakha Newar and Chingtang illustrate ergative alignment.

 

928   Shobhana Chelliah

38.3.1 Tibetan The Tibetan languages can be divided into seven groups ranging from Northern Pakistan eastward across the Tibetan plateau to communities in the Sichuan and Yunnan Provinces of China. Bielmeier et al. (2008) labels these seven groups as Western Archaic, Western Innovative, Central Tibetan, Southern, Northern Kham, Eastern Kham, and Eastern Amdo. There has been extensive discussion about case marking in the Central Tibetan language, Lhasa Tibetan (DeLancey 1984a, 1984b, 1990; Saxena 1989; Agha 1990, 1993; Tournadre 1991; Zeisler 2004, 2007). Zeisler (2012) also provides an in-​depth description of case for the Western Archaic Tibetan language, Ladakhi. Due to the availability of these sources, my discussion will be on Lhasa Tibetan and Ladakhi, specifically the Kenhat variety from Upper Ladakh. Case is indicated in Tibetan via enclitics or postpositions, depending on how these case markers are analyzed.3 I will call them case markers for ease of exposition. As will be shown in what follows, Tibetan does not exhibit ergative alignment, however, in works like Zeisler (2012), ‘ergative’ and ‘absolutive’ are used as case marker labels. When citing Zeisler’s work, I use quotes to remind readers of this. Others, like Tournadre (1991) use ‘agentive’ instead of ‘ergative,’ which is also unsatisfactory since, as we will see, the marker can occur on non-​agentive NPs. Based on an investigation of case marking patterns with monovalent, bivalent, and trivalent predicates of different semantic classes, Zeisler (2012) concludes that case marking in Ladakhi and other Tibetan varieties is predictable on the basis of semantic verb classes. As shown in Table 38.2, reproduced from Zeisler (2012: 81), agents of less transitive verbs (e.g. where patients are only partially affected) tend to occur without the ‘ergative’ marker while for verbs higher on the transitivity scale the ‘ergative’ marker is used. Thus, while with 07 and 08 sentence patterns (see Table 38.2 for sentence types) the ‘ergative’ marker is often used, it can be omitted when agent volitionality is weak, e.g. no individuated object. This means that, as seen in (1) where the possible nonoccurrence of the ‘ergative’ marker is indicated with parentheses, core argument marking for 07 verbs can pattern like 03 verbs. (1)

palaŋ-​(e) petse-​a bearak cow-​(agn) calf-​dat/​all call.out-​prs ‘The cow is calling out for its calf (as I hear).’ (Zeisler 2012: 90, Kenhat variety)

In fact, even at 08, and certainly for frames 01–​07, variability in marking for both A and S is possible on pragmatic grounds characterized by Zeisler as ‘distance.’ The ‘ergative’

3  Classical Tibetan has eight overt case markers and a zero absolutive. These are genitive, agent–​ instrumental, locative, dative–​allative, allative–​purposive, ablative I, ablative II, and comitative. Modern Tibetan varieties have different degrees of reduced inventories (Zeisler 2012: 78).

 

Ergativity in Tibeto-Burman    929 Table 38.2 Ladakhi case frames for 1-​place, 2-​place, and 3-​place predicates Type

Case marking combinations

Semantic characteristics of the verb

1-​place predicate

01

‘abs’

_​_​

change and motion

2-​place predicate

02

‘abs’

‘abs’

03a

‘abs’

~loca

03b 04 05 06 07

~loc=top ‘abs’ ‘abs’ aesb ‘erg’

‘abs’p abl com abs dat/​all

predication, transformation (reflexive agents) affection, oriented motions, position, change into existence get out, move away contact, separation possession

08

‘erg’

‘abs’

3-​place predicate 09a 09b

‘erg’ ‘erg’

10a 10b 11a

‘erg’ ‘erg’ ‘erg’

11b

‘erg’

directional activity, focused attention non-​directional activity, transformation

perception (seen in Western Tibetan), possession dat/​all ‘abs’ give, bring-​type I ‘abs’ ~loc bring-​type II, deposit, transformation into abl ‘abs’ take away-​type 1 ‘abs’ abl take away-​type 2 ‘abs’ com join, mix, separate, exchange, (‘object’ oriented) com ‘abs’ exchange (‘subject-​oriented comitative marker)

Source: adapted from Zeisler (2012). a The tilde indicates variation in locative case in Old and Classical Tibetan.

b The aes abbreviation stands for ‘aesthetive’ which is similar to an experiencer (Zeisler 2012: 78).

marker tends not be used when an event is spatially close (in front of one’s eyes); temporally close (right now, today); or emotionally close (is the action of known/​specific persons, is one’s own action, is an expected, unexciting, neutral action, is a non-​contrastive framing of the action). On the other hand, the use of a case marker indicates ‘distance,’ i.e. physical distance (happened away from speaker); temporal distance (happened in the past); or emotional distance (an event or involvement of participants that is unfamiliar or surprising, or is a contrastive framing of an action). It is this default distribution of ‘ergative’ marking with transitive verbs and past time reference that led DeLancey in earlier work to characterize Tibetan case as split ergative (DeLancey 1984a, 1990). I characterize this patterning of the ‘ergative’ as differential marking based on distance which, as defined here, includes contrastiveness and unexpectedness. The marking seems to place the entire proposition in focus; similar uses of agentive marking to create thetic statements are discussed for Meitei and Burmese (see sections 38.3.2 and 38.3.3).

 

930   Shobhana Chelliah It is not clear how robust ‘ergative’ marking is for the strongly transitive sentence frames 09, 10, or 11. Zeisler states that while there is a tendency for zero marking and ‘ergative’ marking to alternate at the, “middle ranges of semantically lowered transitivity … the exact cut-​off point [for possible variation] has yet to be established.” Zeisler (p.c. 2015) emphasizes that ‘ergative’ marking for transitive verbs is standard, non-​marking is marked; conversely, non-​marking is standard for intransitive verbs, and ‘ergative’ marking is marked. Lhasa Tibetan also exhibits differential marking. Examples (2–​5) illustrate that transitivity is not fully predictive of when the ‘ergative’ will mark the transitive subject. In fact, as seen in (2)–​(4) both transitive and intransitive subjects can take ‘ergative’ marking. (2) Lopsang-​(gis) lug cig Lopsang-​agn sheep a ‘Lopsang killed a sheep.’

bsadsong killperf

(Saxena 1989: 426)

(3) nga-​(s) stag bsad-​kyi-​yin I-​ agn tiger kill-​fut/​voluntary ‘I will kill the tiger.’

(Saxena 1989: 427)

(4) tshiring-​(kyi) chi song Tshiring-​ agn went perf ‘It is Tshiring who went.’

(Agha 1990: 107)4

A and S marking forces a contrastive focus reading. Agha (1990: 109) explains it in this way: “The noun bearing E/​I [ergative/​instrumental] case specifically identifies the referent whose existence is already presupposed from context (but whose identity may be in question).” The conversational example in (5) is an illustration of this contrastive focus use of agentive marking. In this exchange, interlocutors A and B know that someone from a group of possible agents has seen the item under discussion, a cup. Speaker B identifies the agent. (5) A: su-​s dkar-​yol mthong who-​erg/​inst cup-​abs/​nom see ‘Who saw the cup?’

song perf

B: bkra-​shi-s dkar-​yol thong song proper.name-​ erg/​inst cup-​abs/​nom see perf ‘Tashi (is the one who) saw the cup.’

(Agha 1990: 69)

This can be contrasted with the situation in (6) where no A marking occurs with the same verb thoŋ ‘see,’ because there is no selection from a set of possible agents. To the question, ‘What did [Pma] see?’ the answer is: 4 

I’ve used the same transcription system for the Saxena and Agha examples for ease of comparison.

 

Ergativity in Tibeto-Burman    931 (6) pad-​ma proper.name-​ abs/​nom khong spri-​u he-​ abs/​nom monkey

red-​pas ... alee... auxiliary-​yes.no.question oh gdong-​pa dmar-​po sugu ring-​po face red tail long

khetsha-​po cik thong song strange one see perf ‘[You mean] Pma? … Oh, [I see] … he saw a strange long tailed, red faced monkey.’ (Agha 1990: 70) It should be noted that differential marking is not observed to the same extent in all Tibetan varieties—​ transitive subjects almost systematically take agentive marking in Shamskat Ladakhi and intransitive subjects very rarely take agentive marking in Zanskari, a related variety spoken in Jammu and Kashmir. Classical Tibetan texts, devised on the basis of Classical Tibetan grammatical rules and inspired by Sanskrit grammatical systems, approximate systematic ergative alignment; however, texts predating Classical Tibetan grammars exhibit variability in A and S marking (Takeuchi and Takahashi 1995; Tournadre 2014). Furthermore, the predictability of case marking for A and S varies from text to text (Zeisler 2015 p.c.). When marking on A or S does occur the extra meaning of ‘distance,’ as described above, holds. See also Vollman (2008) for distribution of the ‘ergative’ in other varieties of Tibetan, such as Shigatse Tibetan, Kyirong, and Drokpa.

38.3.2 Meitei The Tibeto-​Burman language Meitei is primarily spoken in Manipur state, Northeast India.5 As observed for Tibetan, there is a default correspondence between A and S marking and verb valency and verb aspectual classification (following Vendler 1967). However, as in the case of Tibetan, Meitei A and S marking has additional meaning: in Meitei, case marking on A or S indicates a (contrastive) topic or places a proposition in broad focus. Example (7) provides an elicited Meitei example. The default constituent order is Agent-​ Patient-​Verb. Semantic role markers are enclitics which can occur with NPs or with clauses. For example, the locative =də on an NP indicates location (‘at,’ ‘on,’ or ‘in’) and on a clause 5 

Meitei is sometimes spelled Meetei and is also known as Meiteilon and Manipuri. The following analysis of the Meitei agentive is based on fieldwork conducted over a 20-​year period with native speakers, culminating in a study I conducted with native Meitei linguists Harimohon Thounaojam and Chungkham Yashawanta Singh in 2007. It takes into consideration data from translations, especially sentences that I created using verbs from different semantic verb classes based on Levin (1993), grammaticality judgements and substitution exercises based on naturalistic language data, including conversations, traditional narratives, and dialogues created in role playing activities, “Pear Story” (Chafe 1980) retellings, and monologues.

 

932   Shobhana Chelliah indicates temporal or sequential subordination (‘when that happened’ or ‘after that’). See example (15). (7) ə́y=nə má=pu kaw-​í I=agn him=pat kick-​nhyp ‘I kicked him.’ In non-​elicited speech, patients are marked only when individuated. Inanimate patients often occur with a numeral or a determiner and are rarely marked by =pu ‘patient’ (Chelliah 2009). The Agent marker in Meitei is =nə. The label ‘agent’ covers A (i.e. the logical subjects of two-​ or three-​place predicates) or S (i.e. the logical subject of a one-​place predicate.) The marker =nə is called an agent marker because it almost always occurs on A or S, but perhaps would more appropriately be called a topic marker for reasons that become clear in what follows. As summarized in Table 38.3, the Meitei system is similar to Tibetan in that the verb aspectual semantics provides some predictability to A and S marking. In the Table 38.3 Verb classification and Meitei agent marking Verb aspectual classification

Verb mood

Morphological marking

Achievement (e.g. Realis caused X to laugh, hit X, kill X, break X, enter X, walk to X) and inform, tell, or say.a

In elicitation =Agent takes =nǝ, In discourse: other focus markers or a determiner may be used

Achievement (e.g. hit Irrealis (future, imperative, X, kill X, break X, enter desiderative, uncertain, X, walk to X) and interrogatives) inform, tell, and say.

In elicitation: A and S occurs with =nǝ or zero marking. In discourse: In backgrounded clauses, A and S need not occur with =nǝ; may occur with a determiner or quantifier. In foregrounded clauses, A and S may occur with =nǝ and/​or with =di ‘delimitative’ and/​or =su ‘also.’ A and S may also occur with a determiner or quantifier.

Activity (e.g. walk, sing, dream, sleep, cry) State (e.g. know X, want X, be afraid of X, be X, be on X)

In elicitation: S takes zero In discourse: In backgrounded clauses, A and S need not occur with =nǝ; but may occur with a determiner or quantifier. In foregrounded clauses, A and S may occur with =nǝ and/​or with =di ‘delimitative’ and/​or =su ‘also.’ A and S may also occur with a determiner or quantifier.

a A common feature in Tibeto-​Burman is for A/​S to take agent marking with ‘say’ and ‘tell’ (e.g. in Khumi

(Peterson 2012), Mongsen Ao (Coupe 2011a), Neumhphuk Singpho (Morey 2012) and Na (Lidz 2011).

 

Ergativity in Tibeto-Burman    933 default, especially in data from translation, verbs of Achievement are more likely to have logical subjects marked by the agentive or some type of focus or determiner marking. Logical subjects of Activities, or States are more likely to occur with no marking.6 This distributional pattern is easily disturbed by needs of information structuring. The Meitei agentive marker is exploited for creating statements with broad focus (thetic statements) or indicating (contrastive) topic. These focus and topic functions are described in what follows. A common function of =nə marking is to highlight an entire proposition, fitting straightforwardly into the categories provided in Sasse (1996: 49) and co-​occurring with low presuppositionality for A and S; the use of agentive marking may place an entire proposition in focus, similar to what was seen for Tibetan. Such constructions have the discourse function of pointing to a surprising or unexpected event as in (8)–​(10); providing background information as in generic or existential statements in (11)–​(12); or providing an explanation as in (7). In example (8), =nə ‘agent’ indicates an unexpected or noteworthy event since Tomba is a vegetarian. In (9), agent marking occurs because the activity of draining a pond is culturally noteworthy. In (10) there is an expectation that S is not expected to have gone to bed for some reason, e.g. it is still early.    (8) tomba=nə chá čá-​í Tomba=agn meat eat-​nhyp ‘Tomba ate meat.’    (9) mə́khóy=nə pat=tu they=agn pond=ddet ‘They drained the pond.’

čit-​thok-​lək-​í drain-​out-​dist-​nhyp

(10) má=nə túm-​khi-​lə=e he=agn sleep-​still-​perf=asrt ‘He’s gone to bed.’ Agentive marking is the default with existential statements and generics. This falls under Sasse’s descriptive functions for thetic utterances which refer to natural phenomena or habitual activities (49). See Coupe (2008) for a similar distribution of the agentive or generic statements in Mongsen Ao. 6  Other focus marking refers specifically to =di ‘delimitative’ a general contrastive marker; NP=di indicates that a referent is the only alternative to other possible referents. The delimitative can occur on any semantic role and can combine with =nə). The determiners =si ‘proximate’ and =tu ‘distal’ can also be used as topic markers on the agent with or without the agentive, e.g., masinə ‘this one, the one discussed recently.’ The marker =su ‘also’ can be used instead of or in addition to the agent marker to indicate additive focus which adds to alternatives already introduced in previous discourse. See Chelliah (2009) for examples in Meitei and Konnerth (2014) a discussion of additive focus in the Tibeto-​Burman language Karbi.

 

934   Shobhana Chelliah (11)

hɪndu-​síŋ=nə lukun-​si Hindu-​pl=agn sacred.thread-​pdet ‘Hindus wear the sacred thread.’

tháŋ-​í wear-​nhyp

(12)

učék-​siŋ-​nə sul atiya=tə pay-​í bird-​pl=agn air sky=loc fly-​nhyp ‘Birds fly in the open air.’

A second function for agentive marking is to indicate topicality, with topic defined as the referent that is framed by the speaker as under discussion and is a familiar anchor in the discourse, such that the topic can felicitously appear in the frame “As for … ” as in, “As for Jane, she’s writing a paper this week.” This framing of the referent could also put Jane in contrast with others who are, it is implied, not writing papers. In Meitei, when there is a contrast provided through context between Jane and other non-​ contending referents, the NP Jane will be interpreted as a contrastive topic. Examples (13) and (14) illustrate contrastive topics. Several examples of (non-​contrastive) topics are provided in (16). (13)

ə́-č​ a=bə layrík=tu=nə lum-​í att-​big=nmlz book=ddet=agn heavy-​nhyp ‘The big book is heavy (the others are not).’

(14) ə́y=nə ə́-p​ hə-​pə nú-p​ á-n ​i I=agn att-​good=nmlz person-​male=cop ‘I am a good boy (the others are not).’ There are also less preferable topics, i.e. those that are not yet active in the discourse and likely not known by the addressee and therefore require more effort to track on the part of the listener (Lambrecht 1994: 165). Examples are topics which occur in the introduction to narratives as in example (15) from the introductory sentence to a Pear Story retelling.7 In these instances there is no broader meaning of surprise, unexpectedness, or cultural noteworthiness as in (7)–​(10) and no contrast as in (13) and (14). (15)

nóŋ-​mə ni-​pa ə́-​mə=nə naspati day-​one person-​man att-​one-​agn pear ‘One day a man was plucking pears.’

hék=pə pluck-​nmlz

lak-​li came-​prog

Devoid of context, examples (7)–​(14) are ambiguous between broad focus (thetic), topic, and contrastive topic readings. In narratives, the most common interpretation of an agentive NP is topic. In elicitation, with two and three-​argument predicates, the first translation 7 

See Ozerov (2014: 4–​17) for a useful review of the relevant literature and motivation for separating given and new information values from topicality.

 

Ergativity in Tibeto-Burman    935 given often has broad focus interpretation (a thetic statement). With one-argument predicates, the default interpretation is contrastive topic. Such multi-​functionality for a nominal marker has also been reported for Tamang where the same marker can indicate contrastive focus or broad focus depending on the context (Mazaudon 2003: 312) and Chantyal where the same marker can indicate focus or topic (Noonan 2003: 320). While it is almost always the case that Meitei=nə marking occurs on A or S, in Patient-​ Agent-​Verb word order as in (16), the clause initial NP is always a contrastive topic. The example supports the analysis that =nə primarily indicates information structure and not agentive since the =nə marked NP here is a patient. The second NP, if =nə marked, is a secondary topic (in the sense of Nikolaeva 2001). The glossing of this example again reflects the problematic nature of consistently annotating Meitei =nə and the equivalent marker in other Tibeto-​Burman languages (see Chelliah and Hyslop 2011). (16) mə́-​hák=nǝ ə́y=nǝ ə́-​yúk=tǝ erpor=tǝ 3P-​self-​agn(c.top) I= agn(s.top) att-​early=loc airport=loc thil-​lǝm-​le dropped.off-​inevd-​perf ‘I dropped her (as opposed to others) off at the airport earlier.’ Finally, consider (17) which provides a portion of a monologue where the function of =nə is not clear; it does not mark all topics or only first mention, reintroduced, or contrastive topics. Rather, =nə appears to have the function of creating peaks or tension points in the narrative reminiscent of Sasse’s characterization of the “sudden event” or interruptive function of thetic constructions (1996: 34). Pavel Ozerov finds similarly for Burmese that an established topic can be marked repeatedly to unpack a single event as sequenced, separate actions (p.c. 2015). (17) a. ə́y imphal=dəgi kolkota=də hannə lak-​í I Imphal=abl Kolkata=loc first come-​nhyp ‘I first came from Imphal to Kolkata.’ b. ə́-​dú=dəgi ə́y mənipur bhəbən=də əhiŋ ə́mə lék-​í att-​ddet=abl I Manipur house-LOC night one spent-​nhyp ‘Then I spent a night at Manipur House.’ … (Omitted: ‘Early in the morning I went from Manipur House to the Netaji Subhashchandra airport in Kolkata again.’) c. mə́-​phəm=du=dəgi ə́y=nə mumbai pháw=bə lak-​í nm-​place=ddet=abl I Mumbai up to=nmlz came-​nhyp ‘From there I came to Bombay.’ … (Omitted: ‘In Bombay it was 12: 15.’) d. ə́-​dú=dəgi ə́y intǝrnesnǝl erpor=tə tǝrminal=tu=tə cəŋ-​í att-​ddet=ablI international airport=loc terminal=ddet=loc enter-​nhyp ‘Then I entered the international airport terminal.’ … (Omitted: ‘There I stayed in a waiting room from two in the afternoon to ten at night.’)

 

936   Shobhana Chelliah     e. numidaŋ  puŋ    tə́ra  khə́rə  hal-​lək-​lə          kan=də day     time ten   some  reach-​ dist-​perf  time=loc ə́y ə́y=ki cecaŋ pót cəy cek təw=bə həw-​í. I I-​poss papers thing stuff check do=nmlz start-​nhyp ‘When it was around ten o’clock, I started checking my things in. … (Omitted ‘The trip from Bombay to Zurich is a very long nine-​hour trip.’) f. ə́y=nə ə́-​yuk=tə jurik=tə puŋ tə́ruk=kə I=agn att-​early=loc Zurich=loc time six=ass minit     əmə=tə    yəw-​lək-​í minute one=loc  reach-​dist-​nhyp ‘I arrived at Zurich in the morning at six o’clock.’ g. mə́-​phəm=tu=tə ə́y=nə púŋ məri=muk ləy= lə́gə nm-​place=ddet=loc I=agn time four=only stay=after ə-​ yúk=ki      puŋ   təra=kə    minit     niphuməŋa=tə att-​early-​poss  time ten=ass  minute  forty-​five= loc dəlas=ki=dəmək=tə           ə́y  jurik      tha-​thok-​lək-​í Dallas-​poss=precise=loc  I   Zurich left-​ out-​dist-​nhyp ‘After I’d stayed there for just four hours, I left from Zurich for Dallas at 10: 45.’ Finally in Meitei, as in several other Tibeto-​Burman languages (Chelliah and Hyslop 2011, 2012), there exists a contrastive focus marker polysemous with the agentive marker. See Chelliah (2009) for an account of the historical development of contrastive focus from the agentive in Meitei. As seen in example (18), the highlighted =nə functions to identify a referent from a set of possible referents. Such a sentence could be used to correct inaccurate identification by an interlocutor—​a good test for contrastive focus readings (Chafe 1976). (18)

ə́y=pu=nə khóŋ=nə čə́t-​nə-​həl-​lə-​í I=pat=agn(c.foc) foot=adv go-​means-​caus-​perf-​nhyp I            by foot     caused to go ə́-​tu=gə         má=nə        gari=tə        lak-​lə-​í att-​ddet-​ass  he-​agn(top) vehicle-​ loc  go-​perf-​nhyp and     he        car     went ‘It was I who was forced to go on foot while he went by car.’

To summarize, Meitei A and S marking show default correspondence with verb aspectual semantics in translation data but reviewing natural speech data it appears that this correspondence may be epiphenomenal (addressed further in section 38.4). In natural speech data, =nə marking is used for information structuring, marking topic, contrastive topic, broad focus, and contrastive focus. Bringing the Meitei description in line with Tibetan, this use of =nə marking can also be called differential marking. As will

 

Ergativity in Tibeto-Burman    937 be seen in the next section, the Burmese has similar differential marking with a similar multi-​functional A and S marker.

38.3.3 Burmese Burmese subject marking has been discussed extensively, e.g. Okell (1969), Thurgood (1978), Wheatley (1982), Johnson (1992), Sawada (1995), Soe (1999), Jenny and Hnin Tun (2013), and Ozerov (2014). Relevant findings are that Burmese exhibits both morphological and syntactic accusative alignment.8 Oblique cases are marked by semantic role markers (Soe 1999: 75). Core grammatical relations of A/​S and P are indicated through prosodic cues, word order, and morphological marking. The default order is Agent-​Patient-​Verb. For both A/​S and P, a lowering of pitch with pause can be used to delineate the boundaries of the A/​ S or P constituent, which may be accompanied by overt marking: -​ká for A/​S and -​ko for P. Another option for A/​S and P is no marking with or without characteristic prosody. A/​S marking has a topicalizing function, the effect of which been described as “foregrounding” (Jenny and Hnin Tun 2013), shining a “spotlight” (Soe 1999), or “bringing attention to the constituent” (Ozerov 2014). An example of a topic is given in (19) and a contrastive topic in (20).9 The following conventions are used for the Burmese examples: A/​S marking is glossed as sub for ‘subject’; the genitive is treated as homophonous but distinct from the subject marker although the two are clearly historically related. (19) ɕé-s​ hõu-​ká lu-​ká … nauʔ-​ká lu-​n̥ə-​jauʔ-​ká … shãù -​thà-t​a front-​most-​gen man-​sbj back-​gen man-​two-​clf-​sbj cover-​aux-​real.nmlz ɕé-s​ hõu-​ká       lu-​ká front-​most-​gen  man-​sbj ‘The one furthest to the front … the two men in the back … those covered ones furthest to the front, etc.’               (Wheatley 1982: 176)10 (20) tɕənɔ-​ká ko-​nãdá=pa. θu-​ká 1-​sbj brother-​proper.name=pol . 3-​sbj ‘My name is Kou Nanda, his name is Kou Tin Maun.’

8 

ko-​tĩ.mãu=pa brother-​proper.name=pol       (Wheatley 1982: 170)

Jenny and Hnin Tun (2013) identify Agent/​Subject as syntactic pivots in conjunction reduction and other constructions but Source/​Goal/​Patient are not pivots. 9  However, not all contrastive topics are marked by -​ká (Jenny and Hnin Tun: 718). It is not clear what factors allow for a non-​marked contrastive topic. 10  Thanks to Pavel Ozerov for converting all Burmese examples to phonetic transcription.

 

938   Shobhana Chelliah The absence of marking has been analyzed as continuing topic and -​ká as new topic (Thurgood 1978) or a reintroduced topic (Wheatley 1982). However, Ozerov notes that a referent already in the spotlight can, in a seemingly unmotivated way, occur with -​ká marking, presumably to indicate the beginning of a new discourse event or to introduce a comment outside of the storyline (94). In (21), which appears to be a standard opening strategy for narratives, the -​ká marked constituent is “in the spotlight” but because it does not refer to an established referent, we may think of it as a weak topic. See discussion in (Jenny and Hnin Tun (2013: 715). (21)

tɕũ.tɔ-​ká ʔə.nauʔ-​bədʑinìjà-​pji.nɛ-​m̥a tɕì.pjĩ -​́ khɛ́-t​ ɛ́ θu=pa 1-​sbj West-​Virginia-​state-​loc grow-​distal-​real\rel person=pol [Beginning his story: ] ‘I grew up in West Virginia.’ (Soe 1999: 100)11

Jenny and Hnin Tun (2013) list additional functions of -​ká marked constituents: • Disambiguation: In main clauses, A and S marking is not triggered by the need for disambiguation. Rather the P is marked, indicating that it is not A. This type of marking is called ‘anti-​agentive’ marking by LaPolla (1995a) following Comrie (1975). • Broad focus: A and S marking can be used to create a focused proposition similar to Meitei thetic statements (see section 38.3.2). • Main versus Subordinate:  In complex clauses, marking only the A/​S of the main clause, but not the A/​S of the subordinate clause is preferred but not categorically so. • List enumeration: The subject marker can be used to highlight selected items on a list. • Distance from V: A/​S is more likely to be marked when they are sentence initial rather than preverbal. The farther the A/​S is from the verb, due to intervening subordinate clauses for example, the more likely it is to be marked. Ozerov finds that the characteristic prosody delineating A and S creates an “information package” that indicates to the hearer that there is more to come in the final information package which contains the V. The effect of marking P is exactly the opposite; it places P in an information package which separates P prosodically from the V, which is a focal portion of the clause. This defocuses the P. It is the unmarked P that is salient for the discourse and could be a secondary topic (Ozerov 2015). A second, rarer subject marker, -​ha has also been characterized as topic marker indicating a previously discussed or an otherwise anchored topic. Compare (22) with (23) and (24). As seen in (25), -​ha can also be used to mark culturally salient knowledge. 11 

Some of Soe’s examples are from Wheatley (1982).

 

Ergativity in Tibeto-Burman    939 (22) da-​(ká) ko-​wĩ.phe=pa this-​(sbj) brother-​proper.name=pol ‘This is Ko Win Pe.’ (Used to answer the question, “What’s his name?”) (Soe 1999: 97) (23)

da-​(ha) ko-​wĩ.phe=pɛ̀ this-​sbj (disc.top) brother-​proper.name=emph ‘This is Ko Win Pe.’ (Used to connect Ko Win Pe’s name to the person when he arrives at the scene where he has been previously discussed.) (Soe 1999: 97)

(24) tɕənɔ-​(ha) ko-​nãdá=pa 1-​ sbj (disc.top) brother-​proper.name=pol ‘I’m Kou Nanda’ (when introducing yourself) (25)

kamá.bõu-​(ha) ʔələgà=phɛ̀ sensual.pleasures-​ sbj (disc.top) useless=emph ‘Sensual pleasures are useless indeed.’

(Wheatley 1982: 169)

(Soe 1999: 114)

But, as with -​ka, -​ha can also introduce new topics (Ozerov 93). For this reason, Ozerov treats both -​ka and -​ha as discourse-​level constituent delimiters; once constituents are set apart in information packages through -​ka or -​ha marking, those packages can be used for a variety of information and discourse organizing purposes. To summarize, Burmese exhibits accusative alignment with the unique differential marking patterns similar to those seen for Tibetan and Meitei. The occurrence of marking cannot be characterized syntactically. Information structure is a significant factor. The main subject marker -​ká marks topicality, marking given, new, or contrastive.

38.3.4 Dolakha Newar Dolakha Newar (Newaric group, Nepal) is markedly different from the three languages discussed thus far in that it has verb agreement (agreement between person, number, and honorific status between A or S and V) and it exhibits ergative alignment. Transitive subjects take the -​n ‘ergative’ and intransitive subjects take the zero absolutive. Transitivity is defined mainly on the basis of verb inflection: if the verb is 3rd person past, the transitive form would carry the inflection -​ju and the intransitive form would carry -​a.12 Another factor mentioned is argument structure; monovalent verbs do not occur with an ergative NP (but there are exceptions as we will see). Semantic factors typically

12 

These defining characteristics for transitivity are for the affirmative conjugation. The negative, imperative, prohibitive, and optative conjugations behave differently (Genetti 2007: 108).

 

940   Shobhana Chelliah associated with transitivity such as affectedness of patient, individuating of patient, or agent volitional action are not used to demarcate transitivity from intransitivity in Dolakha Newar. Defined on the basis of verb inflection, subjects of low transitivity can be instigators of action. The following is a list of examples of verbs that occur in ergative constructions: ‘raid,’ ‘read,’ ‘tie,’ ‘catch,’ ‘steal,’ ‘bite,’ ‘look,’ ‘give,’ and ‘say,’ as well as, ‘fly,’ ‘be cold,’ ‘boil’ (e.g. a liquid), ‘ascend,’ ‘fall’ (as of ripe fruit from a tree), ‘weep,’ ‘heal,’ ‘die,’ and ‘shrink.’ This is not a split-​ergative system along aspect or person–​ergative alignment can be found with present, past, and future tense and with all persons. A small class of verbs can be inflected as 3rd person past tense transitive or intransitive. In (26) the verb ‘peel’ inflects like an intransitive to give a nonvolitional, unaccusative reading. In (27) ‘peel’ inflects like a transitive to give a volitional reading. This is fluidity in valence and not differential marking. (26) [āme mati=e tika]np-​s hur-​a 3sgen forehead=gen tika peel-​3spst ‘The tika on his/​her forehead peeled off (of its own accord).’  (Genetti 2007: 109) (27) [āmun] np-​a [tika]np-​o 3serg tika ‘She peeled off the tika.’

hur-​ju peel-​3spst

(Genetti 2007: 109)

Variations worth noting involve experiencers where either the dative or ergative may be used in limited instances. With dative marking as in (28) the reading is non-​volitional but in (29) the agent has more control over the action. (28) [siha᷈=ta]exp besna rak lion=dat very anger ‘The lion became very angry.’

yer-​a come-​3spst

(29) [siha᷈=n] besna rāk yeŋ-​an lion=erg very anger do-​part ‘The lion became very angry …’

(Genetti 2007: 299)

(Genetti 2007: 299)

Finally, (Genetti 2007: 451 and 2011: 30) reports that in sequences of subordinate clauses where the sequence of verbs is intransitive-​transitive or transitive-​intransitive, either verb can determine case marking. Thus in (30) either the intransitive ‘stay’ or transitive ‘cook’ can determine case marking, similarly in (31) either ‘kill’ or ‘go’ can determine the case on A.13 13  In Tibetan verb serialization as well, in V1 V2 sequences which relay the linear order of events, the first verb determines case marking. But in other cases it may be either V1 or V2 (see Tournadre 2010: 15). However, if a clause is purposive or modifying, case is usually governed by the following main verb. Pragmatics may also disrupt this pattern (Zeisler: 2012: 81).

 

Ergativity in Tibeto-Burman    941 (30) bũ=e dani-​n dyābā yeŋ-​an field-​(erg) owner-​(erg) money take=part ‘The owner of the field took the money and left.’ (31) wɔ(-​ɔ̃) duku syan-​a-​a he-​(-​erg) goat kill-​ptcpl-​nf ‘He killed the goat and left.’

ona go-​3spst

won-​ɔ go-​past

The languages of the Kiranti group also exhibit systematic ergative alignment. In a series of articles investigating the nature of ergativity in Belhare and Chintang (e.g. Bickel and Nichols 2001; Bickel 2003; Stoll and Bickel 2013), it is shown that these languages also exhibit syntactic ergativity where transitivity and ergative patterning are central to the organization of the grammar. Differential marking is highly constrained. In Chintang for example, A marking is obligatory with nouns, deictics, and numerals but is almost always optional with 1st and 2nd person pronouns. For 1st and 2nd person pronouns, factors which may trigger ergative marking are: deontic modality (if A should/​should not perform an action); the register (archaic registers controlled by older speakers); and disambiguation (Schikowski et al. 2012: 4). The differential marking found in these languages is of the familiar kind noted in Haspelmath (2005) and not the unique differential marking based on information structure noted for Tibetan, Meitei, and Burmese.

38.4  Temporal Reference, Transitivity, Volition, and Disambiguation Thus far, agent marking in Meitei, Tibetan, and Burmese has been characterized as unique differential marking with information structure (topicality and focus) determining the use of agentive marking. In this section, I will discuss the consistent overlap between four additional parameters and the occurrence of the agentive: temporal reference, transitivity, volitionality, and the need for disambiguation.

38.4.1 Temporal Reference and Agent Marking Ergative constructions have traditionally been thought to arise from passives which in turn are associated with perfective aspect, i.e. a deverbal adjective or past participial (Harris and Campbell 1995:  245–​246). (See Haig (2008) on alignment in Iranian for an alternative analysis.) Split ergativity along aspectual lines is thought to be due to this historical development. However, there is no evidence of a passive to ergative historical development in Tibeto-​Burman. No passive morphology exists and neither is there evidence of agreement morphology for many of the languages in

 

942   Shobhana Chelliah question. In the discussion of Tibetan and Meitei provided above, it is noted that agent marking correlates with past time reference. If the agentive marker is not distributed according to a split ergative system, what accounts for this connection between past reference and agent marking? DeLancey (1981) attributes the overlap between perfectivity and agent marking in Lhasa Tibetan to “viewpoint.” In perfective clauses the focus of attention is on the end of the event, which involves the patient. Ergative or agentive marking serves to reorient attention to the agent, which the speaker achieves by using an overt agent NP and case marking. With imperfective clauses the focus of attention is already on the beginning of an event (and the agent) and therefore no case marking is necessary to highlight that referent. This functional explanation for the distribution of the agentive mimics split ergativity along aspectual lines. Zeisler provides a functional explanation for agent marking in Tibetan control verbs, using a spatial conception of time (2004: 516–​517).14 Recall that with present time reference, it is most usual to find no agent marking, and with past time reference, agent marking is most common. She postulates that when speakers talk about the here and now, the identity of the agent is clear and therefore no marking is necessary. Farther from the temporal center, as with past time reference, the identity of the subject is less clear and here agent marking is needed.

38.4.2 Transitivity and Volitionality Descriptions of Meitei often state that agentive marking occurs with transitive but not intransitive verbs (e.g. Bhat and Ningomba 1997). As has already been shown in section 38.3.2, this is not accurate. Instead, reminiscent of the principles governing the distribution of the ergative in Sacapultec Maya discourse (Du Bois 1987b), the significant overlap of agent marking with transitivity in Meitei seems to be an outcome of discourse factors of foregrounding (e.g. activities moving plot forward) and backgrounding (e.g. scene setting) so that arguments are marked as topic or focus when foregrounding is needed. The foregrounding function is so central in Khumi (Kuki-​Chin), that Peterson (2011) elects to use the term ‘foreground’ instead of ‘agentive.’ Thus, as Coupe (2011b) observes for other Tibeto-​Burman languages, transitivity is another functional category rather than a categorical grammar-​organizing parameter. The overlap of agentive marking with transitivity is most likely discourse based and requires further study. It has been suggested that Meitei agentive marking occurs on only and all volitional activities (e.g. de Hoop and Malchukov 2008, Bhat and Ningomba 1997). However, as seen in section 38.3.2, while many volitional clauses occur with agentive marking, not all do. In addition, as discussed in Chelliah (2009: 390–​392), there are many examples

14 

Tibetan control verbs refer to those actions that can be performed volitionally. Control verbs take different set of inflectional suffixes than non-​control verbs.

 

Ergativity in Tibeto-Burman    943 of agentive marking in non-​volitional clauses as in (32). In this clause, the highlighted referent is accidentally drowning. (32) mə-​náw nu-​pi=nə  púkhrí=tə  i-​rák=nə=rə́gə nm-​small person-​fem=agn pond=loc water-​power.over=inst=after mə-​təy=nə ú=nə ú=nə  upay 3p-​relative.of.opposite.sex=agn see=inst see=inst means  ləy-​tə=nə  yeŋ=dúnə  ləy-​í be-​neg=inst see=ing    be-​nhyp ‘While the sister-​in-​law was drowning in the pond with only the brother-​in-​law to see it, there would be no means for her to be saved (since he could not touch her).’ Similarly, DeLancey (1990) provides examples of non-​volitional activities with agent marking in Lhasa Tibetan as in (33) and (34). Example (34) even has an evidential that indicates the speaker did not have (or pretends to not have had) control or volition over the activity. (33) nga-​s khong-​gi ming I-​agn he-​gen name ‘I will forget his name.’

brjed-​kyi-​red forget-​fut/​involuntary

(34) nga-​s dkar-​yol bcag-​song I-​agn cup broke-​ perf/​involuntary ‘I accidentally broke the cup.’

(DeLancey 1990: 274)

(DeLancey 1990: 274)

For Mongsen Ao, Coupe observes that predicates of bodily functions such as ‘cough’ and ‘cry’ can take agent marking when agents unexpectedly exert control over the activity (Coupe 2008: 159). Here, it is not only volitionality of the action that is being indicated by agentive marking. Rather, agentive marking indicates the unexpected effort taken to perform an action that is usually performed accidentally. These Ao constructions can be understood as thetic statements with interruptive discourse functions (Sasse 1996: 48–​49).

38.4.3 Disambiguation De Hoop and Malchukov (2008:  567–​569) state that there are two functions for case marking. The first is the identifying function which provides some information about the the semantic role of an NP. The second is the distinguishing function which disambiguates which NP is agent and which is patient. They say, “If one argument is case-​marked, this already suffices for the purpose of disambiguation … Thus, from the distinguishing perspective, there is no need to case-​mark both arguments. Neither would it be necessary to case-​mark the one and only argument of a one-​place (intransitive).”

 

944   Shobhana Chelliah An unquestionable need for disambiguation is when overtly realized agent and patient are equal with respect to animacy and sentience and the constituent order is Patient-​ Agent-​Verb rather than the more common Agent-​Patient-​Verb. In this instance, either the Agent (as in Khumi) or the Patient (as in Lhaovo) requires semantic role marking (Peterson 2011, Sawada 2012, respectively). LaPolla (1995a) provides additional examples. While disambiguation is an outcome of agent marking, it is generally agreed that a high level of clause-​level ambiguity is tolerated for most Tibeto-​Burman languages. Context fills in the necessary details (see Johnson 1992 on Burmese for example). As well, it is noted that agent marking occurs where there is no ambiguity observed at all or that even with agentive or patient marking there can be ambiguity on the semantic role of the participants (see Lidz (2011) for a discussion of ambiguity with non-​canonical constituent order and Patient ellipsis). Again from Burmese we learn that in some situations, the Patient must be marked even if the Agent is marked (Soe 1999: 99). Thus, while agentive and patient marking serve to disambiguate referent roles, this is not the only or main function of these markers.

38.4.4 Summary There are correlations between the occurrence of the agentive marker and these four factors: temporal reference, transitivity, volitionality, and the need for disambiguation. None of these in isolation is predictive of whether agentive marking will or will not occur. However, it should be possible to set up a series of statements, perhaps stated as constraints as in de Hoop and Malchukov (2008), to show the probabilistic occurrence of the agentive given these four factors when combined with the information structuring and discourse factors discussed in section 38.3.

38.5  Additional Topics of Interest The following are still-debated questions for Tibeto-​Burman case marking. How does data source (translation versus natural data) affect analysis of case? What is the historical source and development trajectory of Tibeto-​Burman case? How does Tibeto-​ Burman core-​argument marking expand existing typologies of case alignment? And, on a related note, what is the appropriate gloss for A and S markers in Tibeto-​Burman? These questions are briefly dealt with in this section.

38.5.1 Data Sources Readers accessing the Tibeto-​Burman literature will notice the curious difference between early and later descriptions of agentive marking. Early characterization of case

 

Ergativity in Tibeto-Burman    945 in Tibeto-​Burman was handicapped by the use of introspection, translation, and back translation so that the resulting data showed the influence of normative behavior related to careful speech and prescriptivism (Bohnemeyer 2015; Chelliah 2016). Differences between elicited responses in interview situations and naturalistic speech have been widely reported for Tibeto-​Burman languages, especially with respect to NP relational morphology (Chelliah and Hyslop 2011). In the case of Meitei there is an unstated norm that “the subject marker is =nə” even though speakers will readily agree that the marker is not consistently used with A or S. This idea may stem from a strong Indic-​based, prescriptive, grammatical tradition (Chelliah 2016). This, along with a tendency to topicalize A  when providing sentence translations, results in increased occurrence and illusorily consistent agentive marking. Data collected more recently includes conversational and narrative data which are much less homogeneous with respect to nominal marking than earlier data sets. More such data is needed and questions related to information structure need to be carefully formulated and further investigated.

38.5.2 Historical Development In LaPolla’s study of nominal relational morphology in 150 Tibeto-​Burman languages, he found that a single ergative morpheme is not reconstructable for Proto-​Tibeto-​Burman (1995a, 2004). He concludes that the ergative morpheme was a later development for the family and sees A and S marking as evolving from a local case. The progression postulated for agentive marking is ablative > instrumental > agentive and for patient marking the postulated progression is locative > anti-​agentive. It is assumed that at an early stage of development, languages with semantic role marking included pragmatically determined agent and patient (or anti-​ergative) marking. In later stages of development, the distribution of the agentive became predictable on the basis of transitivity and gave rise to systematic ergative alignment for some languages (see further discussion in LaPolla 1995b). At the same time, some languages have not progressed to systematic alignment but show correlations between transitivity and/​or past time reference and A and S marking in a way that mimics, but often falls short of, systematic alignment since the correlations are true in some but not all relevant constructions, e.g. in Mongsen Ao, causative verbs take marked agents but other transitive verbs have arguments marked under unique differential marking (Coupe 2011b). Tibetan and Meitei also show some correlation between transitivity and A marking. Finally other languages, such as Burmese, also have not moved in the direction of systematic ergativity but instead exhibit nominal relational marking based on information and discourse structure with no correlation with transitivity. Tibetan and Meitei also show some corre­lation between transitivity and A marking.15 15  Zeisler (p.c. 2015) postulates the opposite progression for Tibetan since Old Tibetan exhibits relatively standard ergative alignment but several younger varieties show variation based on information structure.

 

946   Shobhana Chelliah If the analysis of differential marking for Tibetan, Meitei, and Burmese can be replicated for other languages of the family—​and it certainly appears that it can—​then another scenario for the development of case should be considered. That is, local cases evolved into semantic role markers and from there to information structuring morphemes and remain there to constitute a new type of nominal marking system (Coupe 2008, 2011b; Chelliah 2009). There is a strong possibility that many languages follow the A and S marking pattern seen for these three information structuring languages. See, for examples, descriptions for Tamang Tamangic, Tshangla (Bodic), Hakha Lai (Kukish) found in Thurgood and LaPolla (2003) and the languages discussed in Chelliah and Hyslop (2011, 2012).

38.5.3 Typology Comrie (2013a) reviews these well-​ known alignment possibilities:  Neutral; Nominative–​ Accusative; Ergative–​ Absolutive (including split ergative); Tripartite; Active–​Inactive. To this one could add the Philippine type system (Austin, Blake, and Florey 2001). However, to date, there has not been much coverage in the typological literature on languages that exhibit no straightforward alignment system and exhibit both differential agent and differential patient marking as seen for the non-​ergative languages discussed here. In addition, the languages discussed here also enrich current understanding of differential marking based on a differential marking hierarchy as formulated by Haspelmath (2005) and reproduced in (40). (40) Animacy Scale: Human > Animate > Inanimate Definiteness Scale: Personal pronoun > Proper name>Definite NP>Indefinite specific NP> Nonspecific NP Differential Object (Patient) Marking is thought to be more likely for Patients when they fall in the left area of the scale where they are less patient-​like. Differential Subject (Agent) Marking is thought to be more likely for Agents when they fall on the right side of the scale, where they are less Agent like. In the Tibeto-​Burman cases discussed here, however, the opposite is true—​i.e. the more agent-​like As are more likely to be marked but ultimately, information and discourse structure are central to whether marking is present or not.

38.6 Conclusion When trying to characterize the case alignment systems of Tibeto-​Burman we can state that ergative alignment is alive and well in the family as seen in Newar and the

 

Ergativity in Tibeto-Burman    947 Kiranti languages. Some languages exhibit accusative alignment as illustrated by Burmese.16 However, the family also includes languages which exhibit neither accusative nor ergative alignment. In these languages, A and S marking cluster around features that are reminiscent of ergative marking (high transitivity, past time reference), so, at first glance and with limited data, they appear to be leaky ergative systems. McGregor (2009: 501) states that “no languages consistently show ergative patterning across the board in all domains.” Would it then be possible to say that all non-​accusative Tibeto-​ Burman languages have ergative alignment but exhibit differing amounts of variability in case marking. To do this, one would have to greatly weaken the definition of ‘ergative,’ for example, accepting that the ‘ergative’ marks an intransitive S performing a non-​volitional activity, as seen in Meitei and Lhasa Tibetan. It seems more reasonable to accept as a typological subclass, languages which differentially mark A and S on the basis of information structure and to allow for multi-​functional morphemes that can mark topics, contrastive topics, broad focus, and contrastive focus with disambiguation being a useful outcome and intended interpretation being context and discourse bound.

Acknowledgments I am thankful to the editors of this handbook, and to Balthazar Bickel, Nancy Caplow, Alec Coupe, Jack Du Bois, Randy LaPolla, William O’Grady, Pavel Ozerov, Willem de Reuse, Chungkham Yashwanta Singh, Harimohon Thounaojam, Bettina Zeisler, and the audience at the 48th meeting of the Berkeley Linguistics Society for insightful comments and suggestions. This research was partially funded by the University of North Texas through the Charn Uswachoke International Development fund. Many thanks to Somi Roy for discussion of the Meitei data. I’m indebted to my students Josh Yamane and Tyler Utt for help with data coding, analysis, and discussion. All errors are mine.

Abbreviations 2sgen, second person singular genitive; 3serg, third singular ergative; 3spst, third singular past; abl, ablative; abs, absolutive; aes, aesthetive; after, after Ving; agn, agentive; att, attributive; aux, auxiliary; c.foc, contrastive focus; cls, classifier; c.top, contrastive topic; com, comitative; cop, copula; dat, dative; ddet, distal determiner; dist, distal; disc.top, discourse topic; emp, emphatic; erg, ergative; gen, genitive; inev, indirect evidence; inst, instrumental; loc, locative; nhyp, nonhypothetical; nmlz, nominalizer; nom, nominative; part, participle; pat, patient; pdet, proximate determiner; perf, perfect; pl, plural; poss, possessive; pres, present; prog, progressive; ptcpl, participial; rel, relative; sbj, subject; top, topic.

16 

Garo (Bodo-​Konyak-​Jingpho), the Tani languages, and Lahu and Akha (Lolo-Burmese) are also described as accusative with topicalization (see relevant chapters in Thurgood and LaPolla 2003).

 

Chapter 39

T he ergative i n Wa rl pi ri : a case st u dy Mary Laughren

39.1 Introduction Warlpiri syntax and morphology has been extensively documented in published works, including Hale (1973a, 1976a, 1981a and b, 1982, 1983), L. Nash (1982, 1996), Simpson (1988, 2005, 2007, 2009), Laughren (1989, 1992, 2001, 2010), Hale et al. (1995), Legate (2001, 2008), in PhD dissertations some of which draw extensively on Hale’s Warlpiri field notes, e.g. D. Nash (1980, published 1986), Larson (1982), Levin (1983), Simpson (1983, published 1991), and Legate (2002).1 In addition to sections of these works, studies that focus on case in Warlpiri include Levin (1983); Laughren (1989, 1992); Bittner and Hale (1996a, b); Legate (2008, Chapter 6, this volume); Bavin (2013, Chapter 25, this volume). Warlpiri has been frequently cited as a morphologically or superficially ergative language with nominative–​accusative syntax but ergative–​absolutive morphology. It has been contrasted with Australian languages such as Dyirbal and Kalkutungu claimed to be syntactically or deeply ergative (e.g. Dixon 1972, 1979; Blake 1976, 1979b; Levin 1983; Marantz 1984; Mylne 2000; Legate 2008). This chapter offers a detailed description of ergative case-​marking in Warlpiri, the syntactic behavior of ergative-​marked DPs and their distribution in both finite and non-​finite clauses.2 Where relevant, comparison

1  David Nash maintains an extensive bibliography of works on Warlpiri language at www.anu.edu.au/​ linguistics/​nash/​aust/​wlp/​wlp-​lx-​ref.html 2  I use ‘ergative’ herein to name the morphological suffix which marks the subject DP in a transitive finite clause (as opposed to an unmarked subject) and DPs in a range of non-​subject functions. In this I depart from the terminological usage which distinguishes homophonous case suffixes on the basis of the syntactic and/​or semantic function of their host.

 

The ergative in Warlpiri    949 with languages of the Dyirbal and Kalkutungu type will be made, as well as with other Australian languages. This introductory section provides some general background on Australian languages: their genetic classification and typological characteristics, the distribution of ergative case-​marking languages, a brief overview of syntactic categories which interact with ergative case-​marking and the range of functions marked by morphological ergative case. Relevant patterns of morphological case-​marking are also described.

39.1.1 Genetic and Typological Groupings Australian languages have been classified into two major groups: the Pama–​Nyungan language family (Blake 1990a; Bowen and Atkinson 2012; chapters in O’Grady et al. 1966; Bowern and Koch (eds.) 2004; Simpson 2014) which includes most of the languages spoken over the continent, and a set of non-​Pama–​Nyungan language families confined to Northern Australia (see chapters in Evans (ed.) 2003).3 O’Grady (1979) redefined Pama–​Nyungan languages as “Nuclear Pama–​Nyungan” a subset of his “extended Pama–​Nyungan” group. Warlpiri belongs to the Ngumpin–​Yapa subgroup of Pama–​ Nyungan languages (McConvell and Laughren 2004). The extended Pama–​ Nyungan versus non-​ Pama–​ Nyungan divide corresponds to a typological distinction between “dependent-​ marking” or “suffixing” versus “head-​marking” or “prefixing” languages (Capell 1962; Nichols 1986). However, many Australian languages are of a mixed type. Many dependent-​marking Pama–​Nyungan languages have a set of bound pronouns hosted by the verb or an auxiliary which encode subject and/​ or object grammatical functions—​ mostly in a nominative–​ accusative pattern—​in addition to the case-​marking morphology on argument DPs whose functional features are marked by a co-​referring clitic. Some of the non-​Pama–​Nyungan head-​marking languages are also dependent-​marking, and exhibit ergative–​absolutive case-​marking.4

39.1.2 Distribution of Ergative Case-​Marking The distribution of ergative case-​marking in Australian languages is almost diametrically opposed to what is found on the African continent (see König this volume, Chapter 37; Nichols 1993) in that the vast majority of Australian languages exhibit an ergative–​absolutive pattern of case-​marking which distinguishes the transitive subject DP (A) in an active voice main clause by marking it—​typically with a suffix—​from the morphologically unmarked intransitive subject DP (S)  and transitive object DP

3 

4 

This classification is disputed by some linguists, notably Dixon (1997, 2002). Blake (1979b: 303) lists non Pama–​Nyungan languages of this type.

 

950   Mary Laughren (O) (Blake 1977, 1979b, 1987b; Dixon 1979, 1980, 1994, 2002; and chapters in Dixon (ed.) 1976).5 The presence in a very restricted set of Australian languages of an unmarked nominative subject and marked accusative object system of case-​marking represents an innovation, accompanied by the loss of bound pronouns. These languages are confined to some members of the Ngayarda subgroup of Pama–​Nyungan languages spoken in the Pilbara area of Western Australia (Dench 1982, 2006) and some members of the Tangkic group of non-​nuclear Pama–​Nyungan languages spoken in northwestern Queensland (Evans 1995; Klokeid 1976a, b, 1978; McConvell 1981).6 The nominative–​accusative languages have also innovated a contrast between active and passive voice in finite clauses, a feature lacking in ergative–​absolutive languages. There are also languages in which the functions of the ergative case suffix have been extended to mark discourse prominence (McGregor 1992, 2007, 2010; Pensalfini 1999; Gaby 2008a, 2010; Meakins 2009; McGregor and Verstraete 2010; Meakins & O’Shannessy 2010; Rumsey 2010; Verstraete 2010; Schultze-Berndt, Chapter 44, this volume) in some functioning as a marked nominative suffix (Meakins 2015). In Jingulu, the ergative case suffix marks both syntactic and discourse functions; in the latter use it marks either subject or object DP (Pensalfini 1999). A subset of ergative–​absolutive languages in northeastern Australia display an alternate case-​marking array with the same set of transitive verbs:  the subject DP is unmarked, while the object DP is marked: by a Dative as in Kalkutungu (Blake 1979a) and Yukulta (Keen 1983), Locative as in Yidiny (Dixon 1977) or either Dative or Ergative as in Dyirbal (Dixon 1972)  case suffix. Named the “antipassive” voice by Silverstein (1976), it is also marked in languages such as in Dyirbal, Kalkutungu and Yidiny on the lexical verb, whereas in Yukulta it is marked by the choice of auxiliary and bound pronominal forms (McConvell 1976; Keen 1983) in addition to case-​marking. As well as interacting with the voice system, ergative marking of subjects in Australian languages interacts with functional categories such as aspect, mood, animacy, and a person/​number hierarchy. See Polinsky (Chapter 13, this volume) for a cross-​linguistic study of the antipassive. In languages of the Warlpiri type, verbs, with few exceptions, have a fixed pattern of subject case-​marking—​at least in finite clauses. Verbs fall into distinct classes depending on whether, in finite clauses, they occur with a marked ergative subject DP or an unmarked (absolutive) subject DP. The only variation in case-​marking in finite clauses is limited to the non-​subject arguments of some transitive verbs which may alternate between an unmarked or a dative marked object.7 A common tendency of Australian languages is towards free word order ( Hale 1981a, 1982, 1983; Dixon 1980, 2002; Blake 1983), with the order of DPs relative to 5 

For convenience, I follow Dixon in using the symbols A, O, and S to refer to an active voice transitive subject (A), an intransitive (S) and a direct object (O) respectively, without subscribing to his theory that these represent primitive linguistic categories (Dixon 1979). 6  See also Legate (Chapter 6, this volume). 7  Other Ngumpin–​Yapa languages are also of this type.

 

The ergative in Warlpiri    951 the verb and to each other—​at least in finite clauses—​b eing mostly pragmatically determined (Swartz 1991; Simpson 2007; Mushin & Simpson 2008). Thus word order is often not a reliable indication of hierarchical syntactic structure in these languages. In non-​finite subordinate clauses, however, word order does interact with case (Hale 1973b) and is more constrained, to be discussed in reference to Warlpiri in section 39.3.8.

39.1.3 Case-​Marking Patterns Dyirbal and Warlpiri exemplify another characteristic which distinguishes Australian languages, including some closely related ones. Dyirbal only has a set of ‘free’ pronouns, as opposed to Warlpiri which has both ‘bound’ and ‘free’ pronouns. While Warlpiri bound pronouns and Dyirbal first and second person free pronouns exhibit a Nominative–​Accusative case-​marking pattern, Warlpiri free pronouns, like other DPs, exhibit an Ergative–​Absolutive pattern of case-​marking.8 Like Dyirbal, many languages exhibit split case-​marking between non-​pronominal and free pronominal DPs (Silverstein 1976). In languages such as Guugu-​Yimidhirr (Haviland 1979), the case split extends to third person pronouns which combine with nouns to form DPs with specific reference. In (1) the nominative pronoun combines with both the ergative-​marked noun in a transitive subject DP and an unmarked noun in an intransitive subject DP; the marked accusative pronoun combines with the unmarked noun in the object DP. These facts argue for the distinction between syntactic and morphological case (Goddard 1982; Legate 2008).9 (1)

Guugu-​Yimidhirr a. Nyulu bidh-​al gudaa wagi naaybu-​unh. 3sg.nom child-​erg dog cut knife-​erg ‘The child cut the dog with a knife.’            (Haviland 1979 #202) b. Bidha nyulu-​ugu wagi-​idhi naaybu-​unh. child 3sg.nom-​emph cut-​refl.past knife-​erg ‘The child cut himself with a knife.’             (Haviland 1979 #203) c. Wanhdhu nhangu wangarr guugu-​ungu binaal-​gurra-​y? who.erg 3sg.acc whiteman language-​purp teach ‘Who taught the whiteman the language?’        (Haviland 1979 #255)

8  Warlpiri free first and second person singular pronouns derive from historical ergative forms and are sometimes unmarked when construed with a transitive subject. In the bound pronoun paradigm, there is syncretism between accusative and dative forms except for third person singular: accusative is unmarked, dative is marked (=rla). 9  Several chapters in this volume address ‘split ergativity,’ e.g. Coon and Preminger (Chapter 10); Laka (Chapter 7); Mahajan (Chapter 4); Nash (Chapter 8); Woolford (Chapter 9).

 

952   Mary Laughren In Warlpiri bound pronouns occur only in finite clauses. The relationship between bound pronouns and co-​referring DPs has been analyzed in different ways. An approach which treats the bound pronouns as agreement markers construed with argument DPs (Simpson 1991; Legate 2002) contrasts with the approach which treats the bound pronouns as arguments and the co-​referent DPs as adjuncts predicated of the pronominal argument (Jelinek 1984; Laughren 1989; Meakins 2014b). Yet a third approach attempts to unite these two seemingly diametrically opposed positions by positing phonologically null arguments which relate to both the bound pronouns as agreement markers and to adjoined DPs, so that the functional features such as person, number, gender, and syntactic case are marked by the bound pronouns, while the non-​functional or ‘encyclopedic’ features are encoded in the adjoined phrases (Baker 1996; Pensalfini 2004). Free pronouns, if also present, tend to reinforce the salience of their referent. In Australian languages, the ergative case morpheme is generally a suffix, although pronouns and determiners may have distinct case forms, including a distinct ergative form, as do some nouns in a more restricted set of languages. Many languages have more than one ergative suffix. Which variant is used is determined by phonological and/​ or semantic properties of the host. In some languages, including Dyirbal and Kalkutungu, there is obligatory case concord: all words in a DP in the scope of a particular case must be marked by the appropriate case form. In southern Western Desert languages such as Yankunytjatjara (Goddard 1986)  and Ngaanyatyarra (Glass & Hackett 2003)  only the final word in a complex phrase is case-​marked. Warlpiri exhibits both of these case-​marking patterns, since the only constraint on case-​marking in Warlpiri is that the final word in a complex phrase host the case suffix.10

39.2  Warlpiri Verbs and Case-​M arking in Finite Clauses In Warlpiri, ergative case-​marking is restricted to verbal clauses.11 With the exception of one verb, janka ‘burn’ (and its synonym kampa), Warlpiri verbs fall into one of two distinct categories depending on whether they have ergative-​marked DP subjects or unmarked DP subjects in finite clauses.12 The class of ergative-​marked subject verbs contains three subtypes with some verbs featuring in more than one subtype. 10 

Variation in Warlpiri case-​marking is illustrated in (17). Ergative case is restricted to verbal clauses in most Australian languages, however Bell (2003) documents its use in Badjala (south-​east Queensland) on nominal predicates in nominal finite clauses to distinguish stage-​predicates from unmarked individual predicates. 12  When the subject is a source of heat acting on some object, the subject DP is marked ergative; where the subject undergoes the effect of heat, the DP is unmarked. Where an agent-​subject uses a source of heat to affect a patient-​object, a different verb is used, namely purra. In restricted respect or avoidance registers, however, some identical verb forms are used as either transitive or intransitive verbs (Laughren 2001). 11 

 

The ergative in Warlpiri    953 The ‘transitive’ subtype divides into two classes: (i) those with agentive, typically animate, subjects that may employ some means or manipulate some instrument, or a body part—​which can be overtly expressed by an ergative case-​marked DP—​to carry out an activity, achievement, or accomplishment; (ii) those with inanimate direct cause or instrument subjects.13 A: transitive: ergative-​marked subject and unmarked object (i) subject = agent (animate) (ii) subject = instrument or direct cause (inanimate) B: conative: ergative-​marked (agent) subject and dative-​marked object C: ditransitive: ergative subject, dative object (recipient) and unmarked complement (theme) Verbs with an unmarked subject can be divided into two subtypes: D: intransitive: unmarked subject E: semi-​transitive: unmarked subject and dative-​marked object Many intransitive verbs also belong to the semi-​transitive category, which I distinguish from intransitive verbs with a dative-​marked complement which does not have object properties (to be discussed in what follows), or an intransitive verb used with a dative adjunct such as an applicative which may be benefactive or malefactive (Hale 1982; Simpson 1991). Unmarked cognate objects are also found in clauses with ergative-​ marked (2e) or unmarked subject verbs (3c).14 (2) a. Transitive (agent subject) Ngarrka-​ngku wawirri luwa-​rnu (makiti-​rli).15 man-​erg kangaroo shoot-​past (rifle-​erg) ‘The man shot the kangaroo (with a rifle).’ b. Transitive (direct cause/​instrument) Wirnpa-​ngku ngarrka luwa-​rnu. lightning-​erg man shoot-​past ‘Lightning struck the man.’ c. Conative Ngarrka-​ngku=rla wawirri-​ki warru-​rnu. man-​erg=3dat kangaroo-​dat seek-​past ‘The man looked for a kangaroo.’ 13  Fauconnier and Verstraete (2010) provide an insightful discussion of animacy effects with regard to ergative-​marked subjects. 14  See Austin’s (1982) survey of cognate objects in Australian languages. 15  The ergative case suffix has four allomorphs whose distribution is partly lexically, but mostly phonologically determined: -​ngku/​i, -​rlu/​i. The choice of vowel is subject to vowel harmony constraints. Many Warlpiri speakers under 60 use only the -​ngku/​i forms, in some dialects reduced to -​ngu/​i.

 

954   Mary Laughren d. Ditransitive Ngarrka-​ngku=rla kuyu yungu karnta-​ku. man-​erg=3dat meat gave woman-​dat ‘The man gave the meat to the woman.’ e. Cognate Object Ngarrka-​ngku kuntul-​pungu (kuntulpa). man-​erg phlegm-​struck (phlegm) ‘The man coughed.’ or ‘The man coughed up phlegm.’ The English translation given for (2b) is somewhat odd without a priming context, as the preferred expression would be the passive counterpart in which the affected person is expressed as the subject: ‘The man was struck by lightning.’ Unlike English which favors linking animate and/​or active moving participants to the transitive subject, Warlpiri strictly links the patient to the object relation so that what produces an effect on the patient is linked, as if by default, to the subject role. There is no pleonastic subject. While the grammatical English sentence ‘The tree struck the car’ is appropriate if a tree fell on the car, it is inappropriate if the car drove into a standing tree. In Warlpiri, however, the same sentence is appropriately used of both situations: Watiya-​rlu mutukayi pakarnu (tree-​erg car struck) ‘The tree (moving/​standing) struck the car.’ In (2e) the case-​marking on the subject is ergative irrespective of whether the “cognate object” kuntulpa, cognate with the preverb kuntul-​compounded with the inflecting verb pungu, is expressed or not. “Cognate objects” function as complements of the verb, but are not raised into the higher object position within the syntactic hierarchy. In the so-​called intransitive use of transitive verbs in which the logical object is non-​overt with a generic non-​specified reference, e.g. He shoots, she eats, etc., the subject DP is still marked ergative. Most of the verbs of the kuntul-​pinyi class, which denote either a deliberate act, or a spontaneous physical reaction to some uncontrolled stimulus, have an intransitive counterpart in which the inflecting verb in the compound is intransitive e.g. Ngarrka kuntul-​karri-​ja (man pleghm-​stand-​past) ‘The man coughed.’ These tend to express only involuntary acts. As illustrated in (3a), intransitive verbs may express voluntary, controlled acts, or involuntary ones; however semi-​transitive verbs typically presuppose an agent-​like subject as in (3b). (3) a. Intransitive Ngarrka/​rarralykaji/​jurlpu wangka-​ja. man/​motorcar/​bird speak-​past ‘The man spoke.’/​‘The car sounded (e.g. sound of car’s motor).’/​‘The bird chirped.’ b. Semi-​transitive Ngarrka=rla wangka-​ja karnta-​ku. man=3sg.dat speak-​past woman-​dat ‘The man spoke to the woman.’

 

The ergative in Warlpiri    955 c. Cognate Object Ngarrka yimi wangka-​ja. man language speak-​past. ‘The man spoke (in) language.’ As illustrated by (3a and b) the same verb may alternate between the intransitive and semi-​transitive use. Not all intransitive verbs may do so. Stative verbs such as stance verbs (e.g. sit, stand, lie), may have a dative-​marked complement, which does not function as an object; that is, it cannot control the null subject of a non-​finite clause of the type to be discussed in section 39.3. To function as the object, a DP must either raise from complement of VP (i.e. sister of V) to a higher specifier position or be directly linked to it. Some verbs participate in a regular alternation between the transitive (ergative–​ absolutive) and conative (ergative–​dative) case array.16 Unlike the syntactically determined use of the antipassive voice in Dyirbal-​type languages, this formal alternation is always associated with regular semantic contrasts in Warlpiri (Laughren 1988). (4)

a. Ngarrka-​ngku wawirri luwarnu. man-​erg kangaroo shot ‘A man shot a/​the kangaroo.’ b. Ngarrka-​ngku=rla-​jinta wawirri-​ki man-​erg=3dat.​ext kangaroo-​dat ‘A man shot at a/​the kangaroo.’

luwarnu. shot

c. Ngarrka-​ngku wawirri nyangu. man-​erg kangaroo saw ‘A man saw/​looked at a/​the kangaroo.’ d. Ngarrka-​ngku=rla wawirri-​ki nyangu. Man-​erg=3sg.dat kangaroo-​dat saw ‘The man looked for the kangaroo.’         (Synonymous with (2c)) In many Australian languages the alternation between unmarked O and marked Odat in sentences equivalent to Warlpiri (4c and d), is additionally marked by the presence of an ergative case-​marked subject with O but an unmarked subject in clauses with Odat.. Warlpiri ergative subject verbs, on the other hand, maintain their ergative-​marked subject in finite clauses irrespective of the case of their object (O or Odat), as well as in reflexive/​reciprocal clauses. Voice and diathesis alternations, marked by verbal morphology and case-​marking of both subject and object in languages such as Dyirbal and Kalkutungu, are marked in

16 

Bittner and Hale (1996b: 557–​559) view the Warlpiri conative as an expression of antipassive voice.

 

956   Mary Laughren Warlpiri by the form of the non-​subject enclitic pronoun in the auxiliary complex, as well as by an alternation in the case-​marking of the logical object as shown in (4).17 The case of the subject remains unchanged. Similarly, the reflexive/​reciprocal construction is marked by use of an anaphoric non-​subject pronoun enclitic =nyanu (except in first person singular subject clauses in which the first person object clitic pronoun is used) whose reference must be bound by the subject.18 The anaphor -​nyanu may be construed with an object (O or Odat), a dative-​marked complement or adjunct (Hale 1973a, 1982; Hale et al. 1995; D. Nash 1996; Simpson 1991). In Warlpiri, both ergative-​marked subjects (5a and c) and unmarked subjects (5b) bind the argument construed with the anaphoric enclitic. (5) a. Ngarrka-​ngku=lu=nyanu luwarnu. man-​erg=3pl.s=anaph shot ‘The men shot themselves/​each other.’19 b. Karnta ka=lu=nyanu wangkami. woman aux=3pl.s=anaph talk ‘The women are talking to themselves/​each other.’ c. Ngarrka-​ngku=lu=nyanu karli yungu. man-​erg=3pl.s=anaph boomerang gave ‘The men gave each other boomerangs.’ The sentences in (5) also show that the same pronominal subject enclitic form is construed with either an ergative-​marked (5a and c) or unmarked subject (5b). Similarly the same set of pronominal enclitic forms are construed with unmarked or dative-​marked objects, with the exception of the unmarked third person singular.20 Third person singular subject and object construed with an unmarked DP is phonologically null, whereas a dative-​marked object is construed with the pronominal enclitic =rla (as is a dative complement or adjunct), or the extended dative enclitic -​rlajinta in ‘frustrated conative’ clauses (‘act to achieve x, but fail’).21 Warlpiri enclitic pronouns combine in an auxiliary complex with morphemes which mark aspect and mood features of finite clauses; they are not found in non-​finite clauses. 17  D. Nash (1996: 130–​136) discusses variation in the form of the pronominal object enclitic in conative clauses. 18  Older speakers also use the second person singular non-​subject clitic =ngku in reflexive imperative clauses, rather than the anaphoric -​nyanu used in other constructions. 19  With a different analysis of what =nyanu is construed with, (5a) can mean “The men shot (it) for each other/​themselves” or “The men shot each other’s/​their own.”) 20  This syncretism of accusative and dative pronominal enclitic forms derives from the loss, in pre-​ Warlpiri, of accusative forms, and their replacement by dative ones. 21  The extended dative enclitic has other uses not discussed herein (Hale 1973a; Simpson 1991; D. Nash 1996).

 

The ergative in Warlpiri    957

39.2.1 Non-​Subject Marking Uses of Ergative Case Morpheme in Warlpiri 39.2.1.1 Instrument As in many other Australian languages, the ergative suffix marks DPs in several non-​subject roles in Warlpiri. Verbs with an ergative-​marked subject, also employ the ergative suffix to mark a DP with an instrument or means function. In Warlpiri, the ergative-​marking of instrument DPs is only grammatical if the subject is also marked ergative, as in (6a). In intransitive or semi-​transitive clauses an instrument or means adjunct is marked by the comitative suffix as in (6b). In transitive clauses, the instrument may also be expressed by a comitative-​marked DP, but the comitative phrase must also be marked by the ergative, as in (6c). (6) a.

b.

Kuyu ka ngarrka-​ngku paji-​rni junma-​ngku. meat aux man-​erg cut-​npast knife-​erg ‘The man is cutting meat with a knife.’ Ngarrka ka wapa-​mi jukati-​kirli. man aux walk-​npast walking-​stick-​com ‘The man is walking with a walking-​stick.’

b’. *Ngarrka ka wapa-​mi jukati-​kirli-​rli/​ jukati-​rli. man aux walk-​npast walking_​stick-​com-​erg/​walking_​stick-​erg c.

Kuyu ka ngarrka-​ngku paji-​rni meat aux man-​erg cut-​npast ‘The man is cutting meat with a knife.’

junma-​kurlu-​rlu. knife-​com-​erg

c’. *Kuyu ka ngarrka-​ngku paji-​rni junma-​kurlu. meat aux man-​erg cut-​npast knife-​com-​erg ≠‘The man is cutting meat with a knife.’ The sentences in (6a and c), express the same meaning by different means, although they do not have the same range of meanings. Both are syntactically (and semantically) ambiguous: (6c) is ambiguous between an instrument reading which modifies the VP (‘cut meat by means of a knife’) and a comitative nominal expression predicated of the subject (‘The man with a knife is cutting meat/​The man is cutting meat, the one with a knife’). On an analysis of (6a) where junma-​ngku is predicated of the subject, ‘knife’ would be interpreted as the man’s name (i.e. qualifying the ergative subject). Where the instrument phrase is the subject of a transitive clause, only the bare ergative-​marked phrase is acceptable as shown in (7).

 

958   Mary Laughren (7) a. Junma-​ngku ka kuyu paji-​rni. knife-​erg aux meat cut-​npast ‘The knife is cutting the meat.’ or ‘(He/​she) is cutting the meat with a knife.’ b. Junma-​kurlu-​rlu ka kuyu paji-​rni. knife-​com-​erg aux meat cut-​npast ‘(The one) with a knife is cutting the meat.’ /​‘(He/​she) is cutting the meat with a knife.’ ≠ ‘The knife is cutting the meat.’ The ability to directly assign an instrument thematic role and to assign the ergative case to the instrument DP is a property of transitive verbs in Warlpiri. What motivates the ergative-​marking of the subject of transitive verbs and the role of the ergative case-​ marking will be discussed after examining a greater range of data including non-​finite clauses.

39.2.2.2 Body Part Associated with Subject In clauses with an ergative-​marked subject, the (body) part of the subject’s referent that causes the event to occur, typically by making contact with the affected object, must also be expressed by an ergative-​marked DP as in (8a). In an intransitive or semi-​transitive clause, the relevant body part DP is unmarked, as is the subject DP.22 Unlike instrument phrases in intransitive clauses, body part expressions do not require the comitative to assign their thematic role, as shown in (8a and b). The active body part is assigned the case of the subject, but does not form a unified constituent with it. (8) a. Kula=lpa purlka-​ngku nya-​ngkarla milpa-​ngku/​*milpa-​kurlu-​rlu. neg=impf old.man-​erg see-​irr eye-​erg ‘The old man can’t see with (his) eyes.’ b. Wangka-​mi ka=rlipa lirra/​*lirra-​kurlu. speak-​npast aux=12pl.s mouth ‘We speak with (our) mouth.’

39.2.2.3 Circumstantial Adjuncts Circumstantial adjuncts such as expressions of time, place, and manner with clausal scope are also marked ergative in a transitive or ditransitive clause, as in (9b). In intransitive or semi-​transitive clauses such as (9a) they are not marked. (9) a. Yama-​ngka/​pulya/​jalangu ka=lu ngarrka shade-​loc/​quiet/​now aux=3pl.s man ‘The men are sitting in the shade/​quietly/​now.’ 22 

nyina-​mi. sit-​npast

See Hale (1981b) for the syntax of body part expressions in Warlpiri, and Laughren (1992).

 

The ergative in Warlpiri    959 b. Yama-​ngka-​rlu/​pulya-​ngku/​jalangu-​rlu ka=lu karli jarnti-​rni. shade-​loc-​erg/​quiet-​erg/​now-​erg aux=3pl.s boomerang carve-​npast ‘They are carving boomerangs in the shade/​quietly/​now.’ Scopal distinctions are marked by the presence or absence of ergative case-​marking as in the contrast between (10a and b).23 (10) a. Kurdu-​ngku maliki wajili-​pu-​ngu ngurra-​kurra-​rlu. child-​erg dog run-​strike-​past camp-​allat-​erg ‘The child chased the dog right up to the camp.’   (child and dog run to camp) b. Kurdu-​ngku maliki wajili-​pu-​ngu ngurra-​kurra. child-​erg dog run-​strike-​past camp-​allat ‘The child chased the dog to(wards) the camp.’     (i.e. dog runs to camp) Although Warlpiri conative clauses have an ergative-​marked subject and dative-​ marked object, the circumstantial adjunct may be in an unmarked case where it designates the area of the activity, as opposed to the location of the referent of the dative-​ marked object or ergative-​marked subject. These scope contrasts, marked by the case of the adjunct, are illustrated in (11). (11)

a. Kuyu-​ku ka=nyanu=rla nya-​nyi purlapurla-​rlu=ju nguru-​wana-​rlu. meat-​dat aux=anaph=dat.EXT look-​npast kite-​erg=top sky-​perl-​erg ‘The kite looks for its meat (as it flies) through the sky.’      (Hale’s field notes)

b. Nyampu ka=rna=rla warru nya-​nyi watiya-​ku here aux=1sg.s=3dat around look-​npast wood-​dat yungu=rna   rdilyki-​rdilyki-​paka-​rni. comp=1sg.s  broken-​broken-​hit-​npast ‘I’m looking around here for wood that I can chop up.’   (Hale’s field notes) c. Yankirri-​ki=lpa=rna=rla nya-​ngu yurdi-​ngka-​rlu, emu-​dat=impf=1sg.s=3dat look-​past top_​of_​tree-​loc-​erg ya-​ninja=rni-​kirra, ngapa-​kurra-​ku. go-​inf=hither-​objcomp water-​objcomp-​dat ‘I was up in a tree looking out for an emu coming to the water hole.’ (Hale’s field notes) In (11a) the ergative marking on the adjunct signals that the location of the Kite is ‘in the sky’ although the prey may not be in the same location. In (11b) the absence of ergative marking on the adjunct nyampu ‘here’ indicates the location of the searching activity but only the potential location of what is sought. The ergative case on yurdi-​ngka-​rlu in (11c) 23 

Sentences (10a and b) from Granites, Hale, and Oddling-​Smee (1976).

 

960   Mary Laughren marks an adjunct that specifies only the location of the subject, whereas the dative marking on ngapa-​kurra-​ku marks scope only over the dative-​marked object yankirri-​ki.

39.2.2.4 Case-​Agreement and Secondary Predicates Case-​agreement operates in Warlpiri to allow the insertion of nominal and non-​finite secondary predicates into a finite clause (Hale 1982; Simpson 1991, 2005; Laughren 1992). The nominal predicate mata ‘tired’ is marked ergative in (12a) like the subject it is predicated of, while in (12b) it is unmarked, as is the body-​part expression wanarri-​jarra linked to the intransitive subject. (12)

a. Wati-​ngki ka ngalngal-​kiji-​rni mata-​ngku. man-​erg aux pant-​throw-​npast tired-​erg ‘The man is panting, (being) tired.’ b. Wanarri-​jarra ka=rna nyina mata —​yi=rna parnka-​ja wurnturu. leg-​two aux=1sg.s sit tired comp=1sg.s run-​past far ‘I’m sitting tired (in) both legs, as I’ve run a long way.’ (Hale’s field notes)

39.2.2.5 Ergative-​Marked Complement At least one Warlpiri verb, pina-​yinyi ‘teach/​show,’ has an ergative-​marked complement shown in (13a). This verb contrasts with the synonymous pina-​mani which has a dative-​ marked complement (13b). The object of both verbs refers to the person who acquires knowledge.24 (13)

a. Karnta-​ngku=ju ngaju pina-​yu-​ngu woman-​erg=1sg.o 1sg knowledgeable-​give-​past

Warlpiri-​rli. Warlpiri-​erg

b. Karnta-​ngku=ju ngaju pina-​ma-​nu Warlpiri-​ki. woman-​erg=1sg.o 1sg knowledgeable-​make-​past Warlpiri-​DAT ‘The woman taught me Warlpiri.’ The distribution of ergative case-​marked subjects in Warlpiri finite clauses differs from that found in languages such as Dyirbal and Kalkutungu where ergative case-​marking of the subject is restricted to active voice transitive clauses with an unmarked object. In Warlpiri, the ergative marking of both subject and non-​subject DPs is lexically determined in the sense that it is linked to the choice of verb. As will be seen in 39.3, where the distribution and functions of the ergative case in non-​ finite clauses are examined, ergative-​marking in non-​finite clauses is more restricted than in finite transitive clauses.

24 Although pina-​yungu is based on the ditransitive ‘give’ verb yinyi, the former’s diathesis differs from that of the latter (cf. (2d)). However, ‘give’ verbs in Australian languages including Dyirrbal and Waanyi express their arguments in the same way as pina-​yungu in (13a).

 

The ergative in Warlpiri    961

39.3 Ergative Case-​M arking in Dependent Non-​Finite Clauses In Warlpiri, the null subject of a class of dependent non-​finite clauses is obligatorily controlled by an argument of the verb in the matrix finite clause, irrespective of the transitivity of the verb in the non-​finite clause. The null subject aligns with the A or S argument, never with the O or Odat argument. Thus the null subject in these non-​finite clauses behaves like PRO in comparable English non-​finite clauses.25 PRO represents the phonologically null realization of a subject to which abstract case is not assigned because of the absence of the case-​assigning category; PRO is anaphoric in that its reference must be controlled by a DP outside the non-​finite clause, but inside the matrix finite clause.26 It is mainly on the basis of these constructions that the claim that Warlpiri is a morphologically ergative, as opposed to a syntactically ergative, language has been based. Depending on the complementizer which heads the non-​finite clause, the identity of PRO’s controller is indicated by the case suffix on the embedded non-​finite clause as in (14) and (15), and/​or the form of the complementizer suffix marking the non-​finite predicate.

39.3.1 Control by Matrix Object The phonologically null subject of the non-​finite clauses in (14)—​shown in square brackets—​ is obligatorily controlled by the object of the matrix clause, whether unmarked or dative-​marked.27 (14) a. Wati-​ngki karnta nyangu [miyi(-​kirra) man-​erg woman saw food(-​objcomp) ‘The man saw the womani PROi cooking food.’

purra-​nja-​kurra].28 cooking-​inf-​objcomp

b. Wati=rla karnta-​ku wangkaja [miyi(-​kirra-​ku) purra-​nja-​kurra-​ku].29 man=3dat woman-​dat spoke food(-​objcomp-​dat) cook-​inf-​objcomp-​dat ‘The man spoke to the womani PROi cooking food.’ 25   For more detailed accounts of Warlpiri embedded non-​finite clauses see Hale (1976a, 1982); Levin (1983); Simpson and Bresnan (1983); Nash (1986); Simpson (1988, 1991); Laughren (1992); Hale et al. (1995); Legate (2002, 2008). 26  PRO subjects can also be quantifier bound, e.g. ‘To serve is to do good.’ 27  The complementizer (-​kurra/​kirra) is homophonous with the allative post-​positional suffix. See Simpson (1988) for an analysis of homophonous case and complementizer suffixes in Warlpiri, and Nordlinger (1998b and 2000) in a range of Australian languages. See Austin (1981a) for a survey of inter-​ clausal co-​reference constructions in Australian languages. 28  The morphemes in parenthesis in the Warlpiri examples are optional. 29  The dative case-​marking on the infinitive whose subject is controlled by a dative object as in (14b and d) is not always present in Hale’s recordings from 1959 and 1966 to 1967; it has been generalized in contemporary Warlpiri.

 

962   Mary Laughren c.

Wati-​ngki karnta nyangu [wapa-​nja-​kurra]. man-​erg woman saw walk-​inf-​objcomp ‘The man saw the womani PROi walking.

d. Wati=rla karnta-​ku wangkaja [wapa-​nja-​kurra-​ku]. man=3dat woman-​dat spoke walk-​inf-​objcomp-​dat ‘The man spoke to the womani PROi walking.’

In (14a and b) it is the null subject of the infinitive transitive verb purra-​nja-​ ‘cook’ which is controlled by the object in the matrix finite clause. In (14a) the controller in the matrix finite clause is the unmarked object karnta ‘woman’ whereas in (14b) it is the dative marked object karnta-​ku ‘woman-​dat.’ In (14c and d) it is the null subject of the intransitive non-​finite verb which is controlled. In (14c) the controller is the unmarked object of the matrix clause, while in (14d) it is the dative marked object. As argued by Simpson (1991), it is the ability of an unmarked or dative-​marked DP in a finite clause to control the reference of the phonologically null subject of an embedded non-​finite clause of this type that provides the most solid test of objecthood in Warlpiri. It serves to distinguish the object from both complements and dative-​marked adjuncts.

39.3.2 Control by Matrix Subject 39.3.2.1 Control of Subject of Contemporaneous Event Clause The subject of the matrix clause controls the reference of PRO in the embedded non-​ finite clause in (15), as signaled by the form of the complementizer -​karra and the case-​ marking on the complementizer-​headed non-​finite phrase, which agrees with the case of its subject’s controller.30 (15)

a. Wati-​ngki karnta nyangu [kuyu(-​karra-​rlu) purra-​nja-​karra-​rlu]. man-​erg woman saw meat(-​scomp-​erg) cook-​inf-​scomp-​erg ‘The mani saw the woman (while) PROi cooking meat.’ b. Wati=rla karnta-​ku wangkaja [kuyu(-​karra) purra-​nja-​karra]. man=3dat woman-​dat spoke meat(-​scomp) cook-​inf-​scomp ‘The mani spoke to the woman (while) PROi cooking meat.’ c. Wati-​ngki karnta nyangu [parnka-​nja-​karra-​rlu]. man-​erg woman saw run-​inf-​scomp-​erg ‘The mani saw the woman (while) PROi running.’

30 

Speakers under 60 years of age have replaced -​karra in these constructions with -​kurra/​-​kirra, so that the identification of the controller depends solely on the case agreement.

 

The ergative in Warlpiri    963 d. Wati=rla karnta-​ku wangkaja [parnka-​nja-​karra]. man=3dat woman-​dat spoke running-​inf-​scomp ‘The mani spoke to the woman (while) PROi running.’ As in (14) in which PRO is controlled by the matrix object, PRO may correspond to either A (15a and b) or S (15c and d). Its controller may be an ergative-​marked subject (15a and c) or an unmarked one (15b and d). The non-​ finite clauses in (14) and (15) are headed by a complementizer. The complementizer-​headed phrase may contain either a nominal or verbal predicate. If in (14a) the sentence were to terminate before the non-​finite verb, then the phrase miyi-​ kirra ‘food-​objcomp’ would have ‘an activity involving food’ interpretation, something like English ‘fooding’ converted into a dynamic activity predicate from the substantive noun.31 This contrasts with the allative interpretation of the same suffix; in another context miyi-​kirra would mean ‘to(wards) the food’ in which the DP miyi is directly governed by the allative. The activity interpretation of miyi-​kirra in (14a) derives from the unmarked aspectual head which governs the NP, so that what the complementizer selects (or introduces) is an aspect phrase with [+dynamic, -​telic] features. This type of complementizer-​headed predicate must be embedded in a finite clause, it cannot operate as an independent predicate in a finite clause, meaning something like ‘(X) be fooding,’ nor can the combination of nominal object and verb within a complementizer phrase such as miyi purranja-​kurra or miyi-​kirra purranja-​kurra seen in (14a) constitute an independent predicate in a finite clause. Compare the grammatical nominal sentence in (16a) with either an individual (wiri ‘big’) or stage (mata ‘tired’) predicate, with the ungrammatical (16b) with the activity predicate headed by an activity complementizer used in (14) and (15). (16) a.

Karnta —​ mata/​wiri. woman tired/​big ‘The woman (is) tired/​big.’

b. *Karnta(-​ngku) —​ miyi+purranja-​kurra/​karra. woman(-​erg) food+cook-​objcomp/​scomp ≠ ‘The woman (is) cooking food.’ Warlpiri nominal finite clauses are unmarked for TAM features, but as in other finite clauses, pronominal clitics mark person and/​or number features of the unmarked subject and, if present, a dative-​marked complement. The semantic relationship between the unmarked subject and the unmarked nominal predicate in a finite nominal clause is typically one of equivalence, property specification, or state. The subject of nominal finite clauses is assigned unmarked case, including phonologically null pro in the

31 

Vowel harmony dictates that following /​i/​the form is -​kirra, while it is -​kurra elsewhere.

 

964   Mary Laughren Chomskyan framework. This contrasts with the phonologically null logical subject of the non-​finite clauses in (14) and (15), which is obligatorily controlled by a subject or object DP of the matrix clause. This class of non-​finite clause which specifies the time of the matrix clause event as contemporaneous with the event it refers to is dynamic and is not equivalent to a stative nominal predicate, and hence cannot constitute an individual-​referring expression:  ngapa+ngarninja-​kurra/​karra cannot mean ‘the one who is drinking water.’32 As an alternative to (14a) in which the objcomp is hosted only by the verb, which is preceded by its logical object, the object DP may also host the objcomp suffix in which case it may precede or follow the infinitival phrase, or even be discontinuous with it. This structure follows the pattern of case-​marking in complex nominal phrases mentioned in section 39.1.3; any word in a complex phrase may be case-​marked, but only the final word of the phrase must be case-​marked. However, there is an additional constraint, in that the infinitive must host the complementizer and any other case-​marking associated with its function within the matrix clause. Co-​referent phrases marked by an identical case may be discontinuous, as may those with a shared syntactic function. Despite their independent syntactic status, the time of the activity referred to by each of the complementizer-​marked phrases is tied to that of the matrix event as contemporary with it, and these phrases can be interpreted as referring to elements of a single activity. How each aspect of the activity is related to the other is determined by the meaning of the predicates embedded in each of the complementizer phrases. The fact that the unmarked logical object of the infinitive verb can only occupy the position immediately preceding the verb indicates that it is embedded, along with the verb, in the objcomp-​headed phrase. Legate (2008) argues that in this configuration the logical object of the verb is assigned structural (or ‘abstract’) accusative case, which is unmarked, as in a finite clause. However, unlike the unmarked object in a finite clause—​which can occupy any position within the clause—​the unmarked logical object of a verb in a non-​finite clause can only occupy the immediate preverbal position. An alternative interpretation of the data in (14) and (15) is that the bare object of the non-​finite transitive verb is not assigned case by the infinitival verb, but as a function of being governed by the complementizer which assigns case to the complex predicate. When the complementizer suffix is hosted by both elements of the VP as in (14a), then distinct activity predicates are created, e.g. miyi-​kirra ‘food-​ing,’ purranja-​kurra ‘cook-​ ing (something).’ The semantic relation between these, both predicated of the same DP in the matrix clause, is interpreted as one of mutual modification: the ‘fooding’ is specified as ‘cooking’ (not ‘eating’ or ‘gathering’ or something else), while the ‘cooking’ is specified as ‘vegetable food-​cooking’ (not ‘meat-​cooking’ or ‘tea-​cooking’). This parallels multiple co-​indexed DPs which may constitute elements of a single phonological phrase in which both elements are case-​marked as in (17a), or one formed by 32  Warlpiri has a set of dynamic nominal predicates which have a different syntactic distribution to stative nominal predicates. As noted by Simpson (1991: 125–​129) they do not function as depictive or attributive predicates, but rather as event modifying adverbs, like the non-​finite clauses in (14) and (15).

 

The ergative in Warlpiri    965 an unmarked DP and its case-​marked modifier in (17b), or which may constitute independent discontinuous phonological phrases as in (17c). (17)

a.

Ngulya-​ngka jinta-​ngka kala-​lu pu-​ngu yurturlu-​patu warrarna=ju. burrow-​loc one-​loc past-​3.pl.s kill-​past heap-​several skink=top ‘They would kill several piles of skinks in the one burrow.’     (Hale’s field notes)

b.

Luurr-​pinyi kapu-​jana yurapiti panu-​jarlu Jampijinpa-​rlu ngulya jinta-​kurra. enter-​make fut-​3pl.o rabbit many-​very J-​erg burrow one-​allat ‘Jampijinpa will chase a big number of rabbits go into the one burrow.’ (Hale’s field notes)

c.

Kala wilypiri-​rla=ju yi=ka=pala nyina jinta-​ngka=lku. but hollow-​loc=top comp=aux=3du.s sit one-​loc=cs ‘But then in the tree-​hollow they both perch in the same one.’    (Hale’s field notes)

39.3.2.2 Control of Subject of Prior Event Non-​Finite Clause Unlike the non-​finite clauses in (14) and (15) which refer to an event contemporaneous with that referred to by the matrix clause, the non-​finite clauses in (18) refer to an event which precedes that of the matrix finite clause. The PRO subject, whether A or S, is controlled by the subject of the matrix finite clause, whether A or S. The non-​finite clause is headed by a complementizer (glossed loccomp) homophonous with the locative case suffix, but which like the complementizer use of the allative suffix in 39.3.1, imposes a dynamic aspectual interpretation, in this case of a completed accomplishment or achievement.33 These ‘prior event’ non-​finite clauses are not subject to case-​agreement with the controller DP.34 (18) a. [Ya-​ninja-​rla]=nganpa karnta wangkaja (nganimpa-​ku). come-​inf-​loccomp=13PL.O woman spoke (13pl-​dat) ‘The woman came and spoke to us.’ (Lit. PROi on coming, to us the womani spoke.) b. [Wawirri(-​rla) luwa-​rninja-​rla] ngarrka kulpaja ngurra-​kurra. kangaroo(-​loccomp) shoot-​inf-​loccomp man returned home-​allat ‘The man shot the kangaroo and then returned home.’ (Lit. PROi on shooting the kangaroo, the mani returned home.) c. [Ya-​ninja-​rla]=nganpa nyangu karnta-​ngku. come-​inf-​loccomp=13pl.o saw woman-​erg ‘The woman came and saw us.’ 33  Like the ergative suffix, the locative has two allomorphs, -​ngka and -​rla. Most speakers only use the latter on infinitive verbs, irrespective of their moraic property (Hale 1976b). 34  See Dench and Evans (1988) for a fine-​grained classification of non-​finite clause types in Australian languages.

 

966   Mary Laughren d. [Wawirri(-​rla) luwa-​rninja-​rla] ngarrka-​ngku (wawirri) purraja. kangaroo(-​loccomp) shoot-​inf-​loccomp man-​erg (kangaroo) cooked ‘The man shot the kangaroo and then cooked it/​(the kangaroo).’

e. [Luwa-​rninja-​rla] purra-​ja. shoot-​inf-​loccomp cook-​past ‘Having shot (iti) (he) cooked (iti/​j).’

The sentences in (18) show the same pattern of obligatory co-​reference as those in (15); the subject of the finite clause, irrespective of its case, controls the A or S subject of the non-​finite clause. In (18e), a typical Warlpiri sentence, the arguments of both verbs are phonologically null, but they are syntactically active. The PRO subject of the non-​ finite clause luwarninja-​rla is obligatorily co-​referent with the ergative pro subject of the matrix clause containing the transitive finite verb purraja ‘cooked.’ The phonologically null pro object of purraja can be interpreted as having the same referent as the phonologically null object pro of the finite verb, but, as predicted by Binding Theory (Chomsky 1980), disjoint reference is also possible.

39.3.3 Morphological versus Syntactic Ergativity It is mainly on the basis of the co-​reference relationship between the subject of non-​ finite clauses discussed thus far and an argument of the finite clause in which they are embedded that Warlpiri has been classified as a morphologically ergative language, since A and S are both realized as PRO. This has been contrasted with the constraints on co-​ reference in languages classed as syntactically ergative, in which a phonologically null argument in a dependent clause co-​referent with a DP in the main clause is restricted to an unmarked, or absolutive argument, irrespective of whether it is S or O as in the Kalkutungu sentences in (19) and (20) respectively. For the reference of the null subject of a clause with a transitive verb to be bound to that of an argument of the main clause in these languages, the dependent clause must be in the antipassive voice in which the subject is unmarked, as in a clause with an intransitive verb. In (19a) the matrix S pirlapirla ‘child’ controls the reference of the null S in the antipassive non-​finite clause thuku-​u lha-​yi-​nyin ‘hitting the dog.’ In (19b) the unmarked O of the matrix clause marapayi ‘woman’ is co-​referent with the null subject of the intransitive participial clause ingka-​ tyin ‘going’ and of the antipassive clause pirlapirla-​a wathinti-​yi-​tyin ‘carrying the child.’ The null constituent in the dependent clause is indicated by ‘e’ in the translations of (19) and (20). (19) a. Pirlapirla unuwani-​nha [thuku-​u lha-​yi-​nyin]. child rejoice-​past dog-​dat hit-​ap-​prtc ‘The childi was happy [ei hitting the dog].’         (Blake 1979a: #4.20)

 

The ergative in Warlpiri    967 b. Nga-​thu nanya marapayi [ingka-​tyin] [pirlapirla-​a wathinti-​yi-​tyin]. 1sg-​erg see:past woman go-​prtc child-​dat carry-​ap-​prtc ‘I saw the womani [ei going], [ei carrying the child].’ (Blake 1979a: #3.7) In (20) in which the understood O in the dependent transitive clause is co-​referent with an argument in the main clause, the overt A of the dependent transitive clause is marked ergative. (20) a. Yurru ingka-​nha [natha-​ngku nanyi-​ntyaaya]. man go-​past nurse-​erg see-​purp ‘The mani went [ei to be seen by the nurse].’        (Blake 1994: ch. 6, #23) b. … kaanta-​nha [nga-​thu lha-​nyin=ka] … leave-​past 1sg-​erg hit-​prtc=foc ‘(Shei) left [ei having been hit by me…]’           (Blake 1979a: 152) c. Ngkuma-​yi ngayi=ka thumpaki-​i [pa-​yi kupangurru-​thu uthiyakapi-​nyin-​ku]. seek-​Tind 1sg.s=foc tobacco-​dat that-​erg old_​man-​erg lose-​prtc-​dat ‘I’m looking for the tobaccoi [ei lost by that old man.]’    (Blake 1979a: #4.47)

In section 39.3.4 I examine Warlpiri non-​finite clauses which do not have an obligatory PRO subject. It will be seen that, like the expression of the covert PRO subject seen in (14), (15), and (18), an overt subject DP may also be expressed in a uniform way, irrespective of the transitivity of the verbal predicate. Differences in the expression of the overt subject of a transitive, as opposed to intransitive, verb, in a non-​finite clause will be seen in section 39.3.5.

39.3.4 Non-​Finite Clauses with a Covert or Overt Subject 39.3.4.1 Obviative Clauses The obviative is yet another complementizer-​headed non-​finite clause which may contain a dynamic nominal or verbal predicate.35 It also specifies the time of the matrix clause event as contemporaneous with that of the non-​finite clause. The subject of an obviative clause may be co-​referent with a dative-​marked complement of the finite verb, as in (21a), or a dative-​marked adjunct in the finite clause, as in (21b), but not a dative object which is in a higher specifier position than the complement (cf. (14b and d)). Thus, its reference is always disjoint from that of the subject or object. 35  The obviative complementizer is constructed on the form of the locative case suffix which has variant forms -​ngka/​-r​ la depending on the phonological or lexical features of the host; when hosted by an infinitive, most speakers employ only the -​rlarni form, irrespective of the phonological properties of the host.

 

968   Mary Laughren (21)

a.

Pardarni ka=rna=ngku (nyuntu-​ku) [ngapa(-​ngkarni) nga-​rninja-​rlarni].36 wait aux=1sg.s=2sg.dat 2sg-dat water(-​obvcomp) drink-​inf-​obvcomp ‘I am waiting for you, while (you are) drinking water.’

a’. Pardarni ka=rna=ngku (nyuntu-​ku) [ngapa-​ngkarni]. wait aux=1sg.s=2.sg.dat (2sg-​dat) water-​obvcomp-​obvcomp ‘I am waiting for you, while (you are involved with) water (i.e. ‘watering’).’ b. Yama-​ngka ka=rnalu=jana jurnta nyina, [karli(-​ngkarni) paka-​rninja-​rlarni]. shade-​loc aux=1pl.s=3pl.dat apart sit boomerang chop-​inf-​obvcomp ‘We’ll sit apart from themi in the shade while (theyi are) chopping (wood for) boomerangs.’                     (Hale’s field notes)

An obviative clause may also have an overt subject not co-​referent with a DP of the main clause as in (22). Irrespective of whether the verb embedded in an obviative complementizer phrase is transitive or intransitive, the overt subject is marked by dative case.37 (22) a. Pirli-​ngka nyina-​ka nyuntu, [ngaju-​ku walya-​ngka nyina-​nja-​rlarni]. rock-​loc sit-​imp 2sg 1sg-​dat ground-​loc sit-​inf-​obvcomp ‘You sit on the rock, while I sit on the ground.’        (Hale’s field notes) b. Nyuntu=wiyi ya-​nta, [ngapa-​ku wanti-​nja-​rlarni]. you=first go-​imp water-​dat fall-​inf-​obvcomp ‘You go first, while it’s raining.’              (Hale’s field notes) (Lit. … while water is falling.) In (22a and b) no DP in the matrix clause is co-​referent with the subject of the embedded non-​finite clause. The overtly expressed logical subject of each of the intransitive -​rlarni-​headed predicates is marked dative. The dative-​marked subject in (22a) is external to the complex predicate walya-​ngka nyina-​nja-​rlarni and is separated from the infinitive by its locative complement. As shown in (23), the dative-​marked subject of the transitive verb may either follow (23a) or precede (23b) the complementizer-​ headed predicate.

36  Pardarni ‘wait’ is an intransitive verb with a dative-​marked complement. Intransitive stance verbs may also operate as ‘wait for/​stay till’ meaning verbs, taking a dative-​marked complement which cannot control the subject of an objcomp clause, but which may be co-​referent with the phonologically null subject of an obviative clause. 37  I have recorded at least one instance of a speaker using an ergative-​marked subject of a -​rlarni clause containing a transitive verb but Hale’s extensive field notes have no instances of this. Legate (2008) discusses this variation.

 

The ergative in Warlpiri    969 (23) a. Kuyu nga-​nja, [ngapa+nga-​rninja-​rlarni ngaju-​ku]. meat eat-​imp water+drink-​inf-​obvcomp 1sg-​dat ‘Eat the meat, while I drink water.’              (Hale’s field notes) b. Kuyu nga-​nja, [ngaju-​ku ngapa+nga-​rninja-​rlarni]. ‘Eat the meat, while I drink water.’              (Hale’s field notes) The subject may also be phonologically null pro whose reference cannot include the addressee as in (24). (24) Kuyu nga-​nja, [ngapa+nga-​rninja-​rlarni]. meat eat-​imp water+drink-​inf-​obvcomp Eat the meat, while pro (≠ you) drink water. ‘Eat the meat, while (I/​we/​they) drink water.’          (Hale’s field notes) The overt dative-​marked or pro subject of the obviative clause which must be disjoint in reference from a DP in the matrix clause has a contrastive focus discourse function. The object of the transitive infinitive may be unmarked—​if immediately before the verb. As exemplified by ngapa ‘water’ in (23) and (24), the object and infinitive form a single phrasal constituent which is governed by the complementizer: [[ngapa ngarninja]-​rlarni]. Alternatively, the object may be marked by the same complementizer suffix as the verb, following the pattern observed in the non-​finite constructions with an obligatory PRO subject in section 39.3.1: [ngapa-​ngkarni] [ngarninja-​rlarni]. Unlike the non-​ finite clauses headed by objcomp, scomp, and loccomp, an obvcomp-​headed clause may contain a nominal predicate as in (25) without necessarily implying an activity, however the predicate must be dynamic in the sense that only a stage predicate reading is possible. (25) a. Panu-​kari-​rli=li kurdu paka-​rnu, [kirda-​nyanu-​ku lawa-​ngkarni]. many-​other-​erg=3pl.s child hit-​past father-​anaph-​dat absent-​obvcomp ‘Other people hit the child [(while) his father (was) away].’   (Hale’s field notes) b. Kala=rla jurnta nyina-​ja karnta-​ku=ju [jarda-​ngkarni] yankirri=ji. past=3dat away sit-​past woman-​dat=top sleep-​obvcomp emu=top ‘The emu sat apart from the womani [while (shei was) sleeping].’  (Children’s book)

The non-​finite clauses headed by one of the complementizers discussed thus far share a common feature: they do not admit an unmarked or ergative-​marked subject DP. The subject of an obviative clause may be controlled PRO, whereas the subject of the other non-finite clauses can only be PRO. These clauses are strictly active voice with a dynamic aspect. They cannot constitute an individual modifying predicate, or be converted into an individual-​referring expression. Hence, they cannot constitute

 

970   Mary Laughren the predicate of a nominal finite clause, or be adjoined to or embedded in one. Some of these clauses do have some nominal-​like features: objcomp and scomp clauses are case-​marked showing agreement with the controller of PRO, in the manner of nominal secondary predicates. The Warlpiri infinitival clauses described thus far cannot substitute an ergative-​ marked A for PRO. Thus, the sentences in (26) are ungrammatical on the analysis given, since O cannot be controlled while A must be. (26) a. *Karnta=rna nyangu [(wati-​ngki) _​_​ pakarni-​rninja-​kurra] woman=1sg.s saw man-​erg _​_​ hit-​inf-​objcomp ≠ ‘I saw the womani [PROi being hit by the man]/​the man (is/​was) hitting ei.’ b. *Purla-​ja=rna=rla karnta-​ku [(wati-​ngki) _​_​ pakarni-​rninja-​kurra-​ku] shout-​past=1sg.s=3dat woman-​dat man-​erg _​_​ hit-​inf-​objcomp-​dat ‘≠I called out to the womani [PROi being hit by the man]/​the man (is/​was) hitting ei.’

The ungrammatical nature of (26a and b) would seem to contrast with the grammatical Kalkutungu sentences in (20). As will be seen, Warlpiri also has non-​f inite clauses in which either a phonologically null subject or object may be co-​referent with a DP in the matrix finite clause. The phonologically null argument of the infinitive verb may correspond to a case-​marked or unmarked constituent: A, S, O, or Odat.

39.3.4.2 Purposive Clauses Purposive clauses are headed by a complementizer homophonous with the dative case morpheme.38 Like the other complementizers, the purposive complementizer governs either a nominal (27a) or infinitival clause (27b–​d). The purposive clauses in (27) are adjuncts, not selected as an argument of the finite verb. The understood subject of the purposive clause in (27a–​c) is co-​referent with the subject of the matrix clause. In (27d), however, there is disjoint reference between the matrix subject and the understood subject of the purposive clause. (27) a. [Kuyu-​ku] ka=lu ya-​ni. meat-​purp aux=3pl.s go-​npast ‘They are going for meat.’ (i.e. to get meat) b. Yinya kapi=rna ya-​ni, [watiya(-​ku) paka-​rninja-​ku]. there fut=1sg.s go-​npast tree(-​purp) chop-​inf-​purp ‘I’ll go over there, to chop down the tree.’           (Hale’s field notes) 38 

Like the allative and locative suffixes, the homophonous purposive/​dative suffix can be analyzed as a single morpheme marking multiple syntactic functions. As a complementizer, it heads a clause and is involved in case assignment and case-​marking.

 

The ergative in Warlpiri    971 c.

Yi=ka wapa-​mi jaala yangka —​ [warlu(-​ku) yarrpi-​rninja-​ku]. comp=aux walk-​npast up_​and_​down like fire(-​purp) stoke-​inf-​purp ‘As (he) walks up and down like to stoke the fires.’         (Hale’s field notes)

d. Jaarl-​karri ka=nyanu, ngula=ju yangka murrumurru, block-​stand aux=anaph that=top like sore [pingka-​ngku marnpi-​ rninja-​ ku].           (Hale’s field notes) gently-​erg touch-​inf-​purp ‘One protects oneself like when one is in pain, so as to be touched gently.’

In (27d), the null subject of the purposive clause is quantifier bound having an arbitrary interpretation, while the null object of the transitive infinitive marnpi-​rninja ‘touch’ in the purposive clause is co-​referent with the subject of the semi-​transitive finite verb jaarl-​karri ‘be in way of.’ The adverbial nominal pingka ‘gently’ which modifies the transitive infinitive verb is in the ergative case, as it would be in a finite clause with the same verb. Unlike the unmarked object or complement of a non-​finite verb, the ergative-​marked modifying adjunct cannot be further marked by the complementizer (*pingka-​ngku-​ku). The inner structure and form of these non-​finite purposive adjunct clauses is the same as that of clauses which serve as the semantic complement of a matrix verbal or nominal predicate as shown in (28). (28) a. Ngarrurnu=ju, [yujuku-​rla(-​ku=ju) nyina-​nja-​ku=ju]. told=1sg.o hut-​loc(-​purp=top) sit-​inf-​purp=top ‘(He) told me [to sit in the hut].’               (Hale’s field notes) b. Jamulu-​ngarri-​rni ka=lu=jana yapa jinta-​kari no_​more_​than-​tell-​npast aux=3pl.s=3pl.o person one-​other [wirlinyi-​ki     [turaki-​kirli-​rli   ka-​nja-​ku]]. day_​trip-​purp  truck-​com-​erg  take-​inf-​purp ‘They tell other people [(they will) take (them) out for the day with a truck], but then don’t.’ (Warlpiri Dictionary: jamulu-​ngarrirni) c. Pina=jala         ka=rna     nyina, knowledgeable=FOC  aux=1sg.s  sit.npast [marlu-​ku=ju [kurlarda-​rlu=ju panti-​rninja-​ku=ju]]. kangaroo-​purp=top spear-​erg=top pierce-​inf-​purp=top ‘I do know [(how) to spear a kangaroo].’         (Hale’s field notes) In (28a) the intransitive infinitival verb and its locative-​marked post-​positional (PP) complement are jointly embedded in the non-​finite clause, both in the scope of the purposive complementizer. If placed immediately before the infinitive verb, the complementizer suffix on the locative complement can be omitted, a behavior similar to that

 

972   Mary Laughren already observed with the unmarked object of non-​finite verbs. The understood subject of the non-​finite clause is co-​referent with the object of the matrix clause. In (28b) the ergative-​marked instrument phrase turaki-​kirli-​rli ‘with a truck’ immediately precedes the transitive verb. Like the ergative-​marked manner adjunct pingka-​ngku ‘gently’ in (27d), the ergative-​marked instrument DP cannot be further marked by the purposive complementizer (*turaki-​kirli-​rli-​ki). The adverbial noun wirlinyi ‘daytrip’ is more distantly related to the verb as a circumstantial adjunct within the scope of the purposive complementizer which both assigns its case and marks it. A similar structure is observed in (28c) in which the logical object marlu ‘kangaroo’ is marked by the complementizer and is separated from the infinitive by the ergative-​marked instrument phrase kurlarda-​ rlu ‘spear’ in a closer structural relationship to the verb it modifies than the purposive-​ marked logical object phrase. Both the purposive-​marked adjunct wirlinyi-​ki in (28b) and the purposive-​marked object marlu-​ku in (28c) could be postposed to the inner complementizer phrase containing the infinitive verb and the ergative-​marked instrument phrase. Unlike the ergative-​marked instrument phrases in (28b and c), an ergative-​marked manner adjunct in a non-​finite purposive clause with a transitive verb may immediately precede (as seen in (27d)) or follow the verb it modifies. This is illustrated by muurlpa-​rlu ‘carefully’ in (29) in which the same speaker explains the meaning of palkarni ‘scarce.’ (29) a. Palkarni, … ngula yangka palkarni marda-​rninja-​ku muurlpa-​rlu. scarce that like scarce have-​inf-​purp carefully-​erg ‘“Palkarni” that’s something scarce to be held on to carefully.’ b. … Kala muurlpa-​rlu marda-​rninja-​ku —​ palkarni.   Well carefully-​erg have-​inf-​purp scarce   ‘Well (it is) to be carefully held onto—​(being) scarce.’  (Hale’s field notes) As will be seen, this relative syntactic freedom of an ergative-​marked manner adjunct differs not only from that of an ergative-​marked instrument, but also of an ergative marked subject in a non-​finite clause. In (30a) the logical subject of the transitive infinitive in the purposive complement to the adjectival nominal pirrjirti is overt, while its object is gapped, co-​referent with pirli yalumpu ‘that rock,’ the subject of the nominal finite clause. Although jakurr-​mani is a transitive verb, its subject is not marked ergative, but is by a suffix homophonous with the purposive complementizer. In (30b) the body part expression taka ‘hand’, which is the relevant part of the referent of the understood subject of the transitive infinitival clause, is also marked like the subject in (30a), and unlike the ergative-​marked instrument phrase in (28b and c). In a finite clause corresponding to the purposive clause in (30b), the body part would be in the same case as the subject, namely ergative (cf. 39.2.2.2).

 

The ergative in Warlpiri    973 (30) a. Pirli yalumpu pirrjirti, [ngaju-​ku=ju jakurr-​ma-​ninja-​ku=ju]. rock that heavy 1sg-dat=top manage_​to_​lift-​inf-​purp=top ‘That rocki (is) (too) heavy [for me to lift _​_i​].’         (Hale’s field notes) b. Miri=lki=rnalu panti-​rni, [taka-​ku=lku marta-​rninja-​ku]. hollow=cs=1pl.s pierce-​npast hand-​dat=cs hold-​inf-​purp ‘We can then gouge out a hollowi (in shield) [to hold _​_​i (with the) hand].’ /​ ‘We can then gouge out a hollowi (in shield) [for the hand to hold _​_i​].’ (Hale’s field notes) The case-​marking of the logical subject of the transitive verb in (30a) and the body part associated with the subject in (30b) contrasts with the ergative-​marked instrument DP in (28b and c). This reflects the difference in their syntactic relationship to the verb. The subject is external to the infinitival phrase, as shown in (31b), whereas the instrument is internal to it. Unlike the non-​finite clauses with subjects obligatorily controlled by a specified matrix DP examined in sections 39.3.1 and 39.3.2 the subject of a purposive clause can be controlled by any argument of the matrix clause, or, as with obviative clauses, it can be projected as a dative-​marked DP. Other arguments can be gapped, with the purposive clause predicated of its ‘filler.’ It can also constitute the predicate of a nominal clause with a DP topic subject (the ‘filler’) which binds the reference of the gapped argument in the purposive clause. In (31a) the logical subject of the purposive clause is co-​referent with the subject of the nominal clause. In (31b) the logical object of the verb in the purposive clause is co-​referent with the subject of the finite sentence. The overt subject of the purposive clause is expressed by the dative-​marked phrase ngarrka-​ku (cf. the dative-​marked subject of obviative clauses in (22) and (23)). In (31c) the subject of the finite nominal sentence is co-​referent with the understood instrument associated with the verb in the purposive clause, of which the understood subject is quantifier bound. (31) a. Ngarrka, [watiya(-​ku) paka-​rninja-​ku]. man tree(-​purp) chop-​inf-​purp ‘The mani (is) PROi to chop down the tree.’ b. Watiya, [ngarrka-​ku [paka-​rninja-​ku]] tree man-​dat chop-​inf-​purp ‘The treei (is) for the man to chop down _​_i​.’ c. Watiya, [ngarrka(-​ku) paka-​rninja-​ku] tree man(-​purp) chop-​inf-​purp ‘The treei (is) to hit the man with _​_​i.’ Because dative and purposive suffixes are homophonous, it could be argued that the overt logical subject ngarrka-​ku in (31b) is governed by the purposive. However,

 

974   Mary Laughren while the suffix is obligatory on the subject of the purposive clause, purposive-​ marking is optional on the pre-​verbal object in the purposive-​headed phrase in (31a and c). Like the overt subject of obviative-​headed predicates, the overt subject of a transitive verb in a purposive clause, as in (31b), and in (30a), is not marked by ergative case, but by dative, in the same way as the subject of an intransitive verb. The subject of both obviative and purposive clauses can also be null, either controlled or arbitrary PRO. Both obviative and purposive clauses have an active voice reading. Thus, Warlpiri purposive clauses with an overt ‘agent’ subject and gapped object as in (30a) and (31b) differ from the Kalkutungu purposive in (20a) in which the overt subject is ergative-​marked. The next section examines non-​finite clauses which form stative predicates whose subject (A or S) or object (O or Odat) can be gapped.

39.3.5  Stative Predicates: Associative and Privative Infinitive phrases can be converted into stative predicates headed by complementizers such as the associative -​warnu and privative -​wangu. These are predicated of an external subject co-​referent with either the logical subject (A or S) or logical object (O or Odat) of the infinitive. When the object of the infinitive is ‘gapped,’ there is interesting variation in the way that the overt logical subject of the infinitival predicate is expressed.

39.3.5.1 Associative Clause An associative phrase is marked by the suffix -​warnu which forms a stative predicate in which either a nominal or infinitival predicate is embedded. It functions somewhat like an English perfect participle phrase with either an active (having VERBed) or passive voice (having been VERBed) interpretation. With a nominal predicate, it has either a ‘having been NOMINAL’ or ‘associated with (event involving) NOMINAL’ interpretation. When the logical object of the infinitival predicate is gapped, the logical subject may be expressed in a number of ways. The first is shown in (32a) in which it is expressed as an unmarked DP which must occupy the position immediately before the infinitive verb with which it forms a single intonational phrase (indicated by ‘+’). In the alternate expression in (32b and c) the agent-​subject is embedded in an independent associative phrase which may precede or follow the infinitive. (32d) is ungrammatical since unmarked kurdu follows the infinitive and hence is not within the scope of the dative-​marked complementizer and hence not assigned case. Like other stative nominal predicates, an associative predicate agrees with its external subject (co-​referent with the gapped argument of the infinitive) in case, which in (32a-​d) is the dative object karli-​ki in the matrix clause.

 

The ergative in Warlpiri    975 (32) a. Wayirni ka=rna=rla nyanungu-​ku karli-​ki search aux=1sg.s=3sg.dat 3-​dat boomerang-​dat [kurdu+wuruly-​ kiji-​ rninja-​ warnu-​ ku].         (Hale’s field notes) child+hidden-​throw-​inf-​assoc-​dat ‘I’m looking for that boomerangi [_​_i​ thrown away out of sight by the child].’ b. …

kurdu-​warnu-​ku child-​assoc-​dat

c. …

wuruly-​kijirninja-​warnu-​ku

kurdu-​warnu-​ku.

d. …

*wuruly-​kijirninja-​warnu-​ku

kurdu.

wuruly-​kiji-​rninja-​warnu-​ku. hidden-​throw-​inf-​assoc-​dat

Warlpiri finite clauses are restricted to active voice. The non-​finite associative clause in (32) operates like a passive voice clause in which the logical subject of the embedded infinitive is internal to the projection of the verb, while its logical object is external to it, as the subject of the associative clause. When both the infinitive and its logical subject are expressed as associative-​marked constituents as in (32b and c), their order is free and they may also be discontinuous—​in the manner of complex nominal phrases with case-​marked elements, and as observed in relation to the complementizer-​headed constituents described in previous sections. Furthermore, either of the -​warnu-​headed phrases in (32b and c) could serve as the predicate. If the infinitive, then the reference of its logical subject would be arbitrary (or contextually established); if the noun, then the nature of the child’s association with the boomerang would not be specified (‘the one the child had something to do with’) and in fact the child could be interpreted as either the agent or patient of an event involving a boomerang, depending on the underlying syntactic analysis. In being marked by the complementizer, rather than being dative-​marked like the subject of an obviative or purposive-​marked predicate, the logical subject of the associative clause is ‘object’-​like in (32a–​c), in that it is internal to the complementizer phrase (cf. (38) with an internal object and external subject). In the Kalkutungu sentences in (20) in which the logical object is gapped, the overt A is marked by ergative case, as predicted if ergative case is lexically assigned within the maximal lexical projection of the verb as argued by Woolford (1997). In Warlpiri, the logical subject of a (di)-​transitive infinitive verb in an associative phrase may also be expressed as an ergative-​marked DP on condition it immediately precede the infinitive as in (33a). This same distribution pattern has previously been observed with the ergative-​marked instrument phrases in purposive clauses as seen in (28b and c). (33)

a. Lakurr-​lakurr-​karrimi ka=lu pirli walya-​ngka, manu watiya pile-​pile-​stand aux=3pl.s stone ground-​loc or wood yangka    [yapa-​ngku+paka-​rninja-​warnu,  warlkurru-​jangka]. that_​same  person-​erg+chop-​inf-​assoc    axe-​src ‘Stones can stand piled up on the ground, or wood chopped by someone, (resulting) from an axe.’           (Warlpiri Dictionary: lakurr-​lakurr-​karri-​mi entry)

 

976   Mary Laughren b. *… paka-​rninja-​warnu yapa-​ngku … hit-​inf-​assoc person-​erg c. *… [yapa-​ngku-​warnu paka-​rninja-​warnu, warlkurru-​jangka]. person-​erg-​assoc chop-​inf-​assoc axe-​src d.

Yapa-​ngku watiya paka-​rnu warlkurru-​rlu. person-​erg wood chop-​past axe-​erg ‘Someone chopped the wood with an axe.’

The ergative-​marked DP agent in (33a) cannot be post-​posed to the -​warnu headed infinitive phrase as in (33b) unlike the ergative-​marked manner adjunct in (29a), nor can it be incorporated into an independent -​warnu phrase, hence the ungrammatical status of (33c). Like the ergative-​marked instrument phrase in (28b and c), the ergative-​ marked subject DP must immediately precede the infinitive, as must the unmarked subject in (32a). The instrument warlkurru ‘axe’ in (33a) cannot also be expressed as an ergative-​ marked phrase, as in the corresponding finite clause in (33d). Rather it forms an independent nominal predicate headed by the ‘source’ post-​position -​jangka meaning ‘from, caused by, affected by, as a result of ’ which assigns its thematic role as well as its case. Only one ergative-​marked DP is permitted in these non-​finite clauses, unlike in the corresponding finite clause. As shown in (32b and c) and in (36) the logical subject in an associative clause may alternatively be expressed by a -​warnu-​marked phrase. Alternative ways of expressing the direct cause inanimate subject in an associative clause are shown in (34). The non-​finite clauses in (34a and b) can be compared with the corresponding finite clause in (34c), in which the ergative-​marked subject could alternatively be clause initial or final. (34)

a.

Watiya-​ku ka=rna=rla warrirni, [wirnpa-​ngku+luwa-​rninja-​warnu-​ku]. tree-​dat aux=1sg.s=3sg.dat seek lightning-​erg+strike-​inf-​assoc-​dat ‘I am looking for the treei [_​_i​ struck by lightning].’          (Hale’s field notes)

b. Watiya-​ku ka=rna=rla warrirni, [wirnpa-​jangka-​ku], [luwa-​rninja-​warnu-​ku]. lightning-​src-​dat strike-​inf-​assoc-​dat ‘I am looking for the tree, for (the one affected) from the lightning, for (the) struck (one).’                          (Hale’s field notes) c.

warrirni, [ngula wirnpa-​ngku luwa-​rnu]. that(one) lightning-​erg strike-​past ‘I am looking for the tree that the lightning struck.’        (Hale’s field notes) Watiya-​ku ka=rna=rla

While the ergative-​marked subject wirnpa-​ngku in (34a) forms a single constituent with the associative-​marked phrase, the postpositional phrase wirnpa-​jangka-​ku in (34b) is

 

The ergative in Warlpiri    977 syntactically (and phonologically) independent of the associative phrase—​each phrase may be in either order, or may be discontinuous. In (35a) the associative clause containing the ergative-​marked subject is placed immediately before the auxiliary complex, a well-​attested diagnostic of its single phrase status. This contrasts with what is observed in finite clauses in which the verb fails to integrate into any phrase with a DP, be it case-​marked or not. In (35b) the syntactic independence of the discontinuous source phrase warna-​jangka and the associative phrase pajirninja-​ warnu is demonstrated. (35) a. [Warna-​ngku+paji-​rninja-​warnu] ka=rna nyina-​mi. snake-​erg+bite-​inf-​assoc aux=1sg.s sit-​npast ‘I have been bitten by a snake.’            (Hale’s field notes) (Lit. ‘Ii am sitting [_​_i​ (having been) bitten by a snake]).’ b. [Warna-​jangka] ka=rna nyinami, [paji-​rninja-​warnu]. snake-​src aux=1sg.s sit-​npast bite-​inf-​assoc ‘I have been bitten by a snake.’            (Hale’s field notes) (Lit. [From snake] Ii am sitting, [_​_i​ (having been) bitten].) Alternatively, both the agent and instrument DPs that would be ergative-​marked in the corresponding finite clause can be projected in an associative-​ headed phrase as in (36) in which the speaker explains the meaning of the word karalypa ‘smooth.’ (36) Karalypa, ngula=ji yangka kuja=ka ngunami pirli smooth, that=top that comp=aux lie stone puturrputurrpa-​ wangu, manu yangka kuja=ka    karli      manu rough_​surface-​priv    or      that    comp=aux  boomerang or kuturu ngunami puturrputurrpa-​ wangu yapa-​warnu club  lie      rough_​ surface-​ priv    person-​assoc rii-​pi-​nja-​warnu      junma-​warnu, yangka puturrputurrpa-​ wangu. smooth-​make-​inf-​assoc  knife-​assoc      that     rough_​surface-​priv ‘Smooth, that (is) like a stone with no protruding bits on its surface, or like a boomerang or club without any rough notches on its surface having been smoothed down by someone with a knife, that (is) without any protrusions (on its surface).’               (Warlpiri Dictionary: karalypa) Thus far the logical subject of a transitive infinitive in an associative clause has been expressed in one of three ways:  (i)  by an unmarked or ergative marked DP placed immediately before the infinitive, (ii) by a DP hosting the associative suffix which is syntactically independent of the infinitive phrase, and (iii) by an independent phrase marked by a semantic case suffix such as the source suffix -​jangka

 

978   Mary Laughren in (34b) and (35b), which assigns the thematic role and case to the agent (or instrument) DP it governs. In (37a) in which the ergative-​marked instrument karli-​ngki ‘boomerang-​erg’ in the initial finite clause is qualified by the second finite clause (in parentheses), the agent-​subject of the latter, ngajuku-​purdangka ‘my sibling,’ is marked ergative as required in a transitive finite clause.39 In (37b) the animate logical subject of the non-​finite stative predicate is expressed in an independent phrase headed by the “possessive” semantic case -​kurlangu ‘belonging to, created by, associated with.’ The infinitive verb is embedded in an associative phrase. Both the agent and infinitive phrases are marked ergative in (37b) in agreement with their external subject karli-​ngki co-​referent with the logical object of the infinitive verb. It is this pattern of case-​agreement with the external subject that allows these constituents to be treated as either independent phrases or as a merged single phrase, and which reflects the logical relations. (37) a. Kapi=rna=ngku    karli=ngki     luwarni, futcomp=1sg.s=2.sg.0  boomerang-​erg  hit [ngaju-​ku-​pirdangka-​rlu ngula ngurrju-​ma-​nu]. 1sg-​dat-​sib-​erg that good-​make-​past ‘I will hit you with a boomerang that my brother made.’   (Hale’s field notes) (Lit. … my brother that made.) b. Kapi=rna=ngku karli=ngki luwarni, futcomp=1sg.s=2.sg.o boomerang-​erg hit [ngaju-​ku-​pirdangka-​kurlangu-​rlu]  [ngurrju-​ma-​ninja-​warnu-​rlu]. 1sg-​dat-​sib-​poss-​erg        good-​make-​inf-​assoc-​erg ‘I will hit you with a boomerang made by my brother.’    (Hale’s fieldnotes) (Lit. … my brotheri’s _​_j​ , [_​_​i having made _​_j​]) The choice of semantic case, or post-​position on the agent DP construed with the logical subject of the non-​finite verb, for example, -​jangka in (35b) as opposed to -​kurlangu in (37b), is semantically determined. The subject of the infinitive embedded in an associative phrase may also be gapped. In (38a) the understood logical subject is co-​referent with the subject of the matrix clause. The logical object of the transitive non-​finite verb is incorporated as the unmarked noun yapa ‘person’ immediately preposed to the verb. Alternatively, the patient-​object yapa can be directly incorporated into an associative phrase and host the associative suffix -​warnu; the resultant constituent yapa-​warnu may be pre-​or post-​posed to the infinitive as in (38b) and (38c) respectively, or can be discontinuous with the infinitive.

39 

See Hale (1976a) for his analysis of the Warlpiri finite relative clause, and Larson (1982).

 

The ergative in Warlpiri    979 (38) a. Malamala, parnkaparnka, yurrurnturu, ngula ka=rnalu jinta-​jarri yangka —​ that aux=1pl.s one-​inch that_​same [yapa+wajawaja-​ma-​ninja-​warnu]. person+lose-​do-​inf-​assoc ‘Malamala, parnkaparnka and yurrurnturu (ceremonies) (are) when wei come together, like ei having lost someone (in death).’         (Hale’s field notes)

b. … yapa-​warnu wajawaja-​maninja-​warnu. person-​assoc lose-​do-​inf-​assoc c. … wajawaja-​maninja-​warnu yapa-​warnu. d. *… wajawaja-​maninja-​warnu yapa. As noted above, associative clauses in which the object is gapped have a passive-​like interpretation. When the subject is gapped, the associative clause has an active interpretation. The subject of a transitive or intransitive verb in an associative clause may be gapped. In this, Warlpiri differs from Dyirbal and Kalkutungu in which only the unmarked object or intransitive subject, including the unmarked subject of an antipassive clause, may be gapped. Just as either ergative or unmarked subjects of infinitival verbs may be gapped in associative clauses, both unmarked and dative-​marked objects can be gapped. In (39a) the dative object of the infinitive in the associative phrase is co-​referent with yapa ‘person’ the unmarked subject of the matrix clause, of which the associative clause is predicated. (39) (from Warlpiri Dictionary: mangily-​wangkami) a. Kurnta-​jarri-​mi=lki ka=lu ngula=ju yapa shame-​inch-​npast=cs aux=3pl.s that=top person [mangily-​mangily-​wangka-​nja-​warnu]. protest-​protest-​speak-​inf-​assoc ‘Feel ashamed is what peoplei (do) having been protested to _​_​i (over their behavior).’ b. Mangily-​mangily-​wangka-​mi ka=janai yapa-​kui, kulu-​jangka, protest-​protest-​speak-​npast aux=3pl.o person-​dat fight-​src [japirdi-​nja-​warnu-​ku]. threaten-​inf-​assoc-​dat ‘After a fight, (someone) complains to the peoplei [_​_i​ having made threats].’ In (39b) the semi-​transitive verb mangily-​mangily-​wangka is used in a finite clause in which its dative-​marked object yapa-​ku is construed with the plural object enclitic =jana. Since the associative phrase in (39b) is predicated of the dative argument of the finite verb, it too is marked dative. In (39b) it is the subject of the infinitive in the associative clause which is gapped.

 

980   Mary Laughren Associative phrases differ from the non-​finite clauses with PRO subjects whose controller is specified by the complementizer heading the clause containing PRO. Any subject or object argument of the predicate contained in the associative phrase may be gapped. Furthermore, associative clauses may operate as the main predicate in a finite nominal clause. In (40) in which the speaker explains the meaning of yawarra ‘wound’ by referring to its typical causes, he juxtaposes two instrument-​referring DPs embedded in source phrases, followed by an infinitive phrase embedded in an associative phrase. These phrases form a complex predicate or series of related predicates which qualify yawarra. (40) Yawarra yangka [kurlarda-​jangka]=rlangu, [junma-​jangka]=rlangu, wound that spear-​src=eg knife-​src=eg [paji-​rninja-​warnu]. cut-​inf-​assoc‘ A wound (is) like from a spear for instance, or from a knife for instance, (from) having been cut.’                      (Hale’s field notes) Like other stative nominal predicates, an associative phrase can also function as a referential DP which may occupy any argument—​or adjunct—​position in a clause, as well as serving as a predicate. In (41) it qualifies the dative object maliki (or could be analyzed as one of the juxtaposed phrases which constitute the dative object phrase). Similarly, in (41b) the associative phrase is the dative object of ‘give,’ construed with the dative enclitic =rla.40 In (41c) the associative phrase incorporates its subject kurdu, its logical object being gapped. The associative phrase is marked by ergative case signaling that it is the referential subject of the finite clause. Again, there is a series of associative phrases each one modifying the referential scope of the other. (41)

a.

Mari-​jarrimi ka=rna=rla maliki-​ki [warna-​jangka-​ku], karnuru-​ku, sorry-​inch aux=1sg.s=3dat dog-​dat snake-​src-​dat poor_​thing-​dat [paji-​rninja-​warnu-​ku]. bite-​inf-​assoc-​dat ‘I feel sorry for the dogi, for (the one) affected by the snake, for the poor thing, for (the one) having been bitten.’                    (Hale’s field notes)

b. Tiyi=rlangu yangka kaji=lpa=rla yu-​ngkarla manu miyi, manu kuyu, tea=eg that comp=impf=3dat give-​irr or bread or meat [warlu+paka-​rninja-​warnu-​ku]. firewood+chop-​inf-​assoc-​dat ‘If (he) were to give him something like tea, or flour, or meat—​to (the one) having chopped the wood.’                 (Hale’s field notes) 40 

Distinguishing secondary predicate from argument depends on the approach taken to the relationship between bound pronouns and DP, which I don’t address herein.

 

The ergative in Warlpiri    981 c. Ngarrpangarrpa-​mani     ngula=ji     kuja=ka=nyanu “Ngarrpangarrpa-​mani”  that=top  comp=aux=anaph [[kurdu+miyi-​warnu-​rlu]  [wiji-​warnu-​rlu]  [nga-​rninja-​warnu-​rlu] child+food-​assoc-​erg    steal-​assoc-​erg  eat-​inf-​assoc-​erg [ma-​ninja-​warnu-​rlu]] ngarrpangarrpa-​ma-​ni  warlka-​ngku,  manu get-​inf-​assoc-​erg deceive-​do-​npast      lie-​erg      or lani-​ ngki,  paka-rninja-kujaku afraid-​erg   hit-inf-lest ‘Ngarrpangarrpa-​mani is like when [a child who has stolen some food and eaten it] [having taken it] deceives by lying, or out of fear of being beaten.’ (Warlpiri Dictionary: ngarrpangarrpa-​mani) The dative-​marked associative phrases in (41a and b) can be analyzed as constituting (part of) the dative object phrase, while in (41c) the juxtaposed ergative-​marked associative phrases constitute the subject of the finite clause.

39.3.5.2 Privative Clause An infinitive verb may also be embedded in a phrase headed by the privative complementizer -​wangu ‘without.’ Like associative clauses, privative clauses may be stative and may serve as the main predicate in a finite nominal clause; they also allow A, S, O, and Odat to be gapped. Within the non-​finite clause A is expressed as either an unmarked or ergative-​marked constituent in immediate pre-​infinitival position, or is embedded in a separate (but related) privative phrase; Odat maintains its case-​marking and is syntactically independent of the infinitive; O is unmarked in immediate pre-​infinitival position or is incorporated into a privative phrase. Additionally, O may be expressed as a dative-​marked constituent, licensed by negation as in (42a and b).41 The dative-​marked object is syntactically independent of the privative phrase, unlike an unmarked object which must be integrated into the privative phrase. (42) a. [Yi-​nja-​wangu] ka=rna nyina [ngapa-​ku]. give-​inf-​priv aux=1sg.s sit water-​dat ‘I haven’t been given water.’ (Hale’s field notes) (Lit. [ungiven (to) _​_i​] Ii am sitting [of water].) b.

41 

Lawa=ju ngarru-​rnu, [paka-​rninja-​wangu], neg=1sg.o tell-​past chop-​inf-​priv

Hale (1970: 778, 1973b), discusses this use of the dative.

[watiya-​ku=ju]. tree-​dat=top

 

982   Mary Laughren b’. Lawa=ju ngarru-​rnu, [watiya-​ku=ju] [paka-​rninja-​wangu]. ‘He told me not to chop down the tree.’ (Hale’s field notes) (Lit. He told mei not (to) [PROi to be not chopping down of the tree]) Unlike the privative clause in (42a), the one in (42b and b’) is not stative, but dynamic, referring to a potential future event. The logical subject of the infinitive is PRO which is controlled by the dative object of the matrix clause =ju ‘me’. The privative clause in (42b and b’) is not predicated of this object but qualifies lawa the negative complement of ngarrurnu ‘told’. In (42a) the Odat (co-​referent with the subject of the finite verb) is gapped in the privative clause predicated of the co-​referent subject. By contrast, the dative-​marked complement of the semi-​transitive verb parda-​rni ‘wait for’ in (43a) cannot be gapped as in the non-​finite clause in (43b). Only the unmarked subject may be gapped (43c).42 (43) a. Nganimpa ka=rna=ngku=lu nyuntu-​ku parda-​rni. 13pls​ aux=1.s=2sg.dat=pl.s 2sg-​dat wait-​npast ‘We are waiting for you.’ b. *Nyuntu ka=npa nyina [(nganimpa)+parda-​rninja-​warnu]. 2sg aux=2sg.s sit 13pl+      wait-​inf-​assoc ≠‘You have been waited for (by us).’ (Lit. Youi are sitting [(we) having waited for _​_i​]) c. [Nyuntu-​ku parda-​rninja-​warnu] ka=rnalu nyina nganimpa=ju. 2sg-​dat wait=inf-​assoc aux=13pl.s sit 13pl=top ‘We have been waiting for you.’ (Lit. [_​_​i having waited for you] wei are sitting.) In this respect, Warlpiri resembles languages such as Dyirbal and Kalkutungu in which a dative complement cannot be gapped. But these languages, unlike Warlpiri, lack dative-​ marked objects. Privative-​marked infinitival predicates are employed in negative imperative sentences. Since the marker of sentential negation kula is incompatible with the imperative (Laughren 2002) the imperative form of intransitive nyina ‘sit, be, stay’ combines with the privative marked infinitive of an intransitive or semi-​transitive verb as in (44a), while the imperative form of transitive yampi ‘leave’ combines with the privative marked infinitive of a transitive verb as in (44b). The transitive privative phrase is marked ergative like the scomp-​ headed clauses in 39.3.2.1. Thus, the subject of the privative-​marked infinitive is bound by the addressee subject of the imperative verb, while the object of the transitive infinitive is gapped, being co-​referent with the object of the imperative verb. (44) a. Nyina-​ya [wangka-​nja-​wangu]! sit-​impf speak-​inf-​priv ‘Don’t speak!’ 42 

(Lit. (youi) sit [PROi [not speaking]])

The grammaticality judgments in (43b and c) were verified by Helen Napurrurla Morton (June 2015).

 

The ergative in Warlpiri    983 b. Yampi-​ya karnta [paka-​rninja-​wangu-​rlu]. leave-​imp woman hit-​inf-​priv-​erg ‘Don’t hit the woman!’ (Lit. (youi) leave the womanj [PROi [not hitting _​_j​]].) Although in (44b) the logical object of the infinitive in the privative phrase is gapped and is co-​referent with the object of the imperative verb, the privative clause is not predicated of the object. It is an active voice clause, in which the agent is realized as the controlled PRO subject of a dynamic predicate which qualifies the imperative verb (cf. the privative-​marked infinitive in (42b and b’)).

39.3.6 Discussion The non-​finite clauses with an obligatorily controlled subject examined in sections 39.3.1 and 39.3.2 differ in kind from obviative and purposive clauses whose subject may be either controlled or expressed as an overt dative-​marked DP. The associative and privative complementizer-​headed non-​finite clauses may form stative predicates whose external subject binds the reference of a subject or object argument of the infinitive. When the complementizer-​headed phrase is predicated of the logical object of its embedded infinitive, the infinitive’s in situ subject may remain internal to the infinitive phrase and be assigned ergative case if the infinitive verb is transitive, or the subject may be unmarked irrespective of the verb’s transitivity, on condition that it be immediately preposed to the infinitive. I have argued that the abstract case assigned to an unmarked subject or object DP in a finite clause is not assigned in a non-​finite clause. In a non-​finite clause, a bare DP must be in the scope of the complementizer which assigns its case, be it subject or object. It may also be marked by the complementizer, forming a syntactically independent predicate which may be discontinuous with the complementizer-​marked infinitive. In these respects, associative and privative clause types differ from obviative or purposive clauses in which the logical subject, whether PRO or overt DP, must be external to the complementizer-​headed phrase. Whereas both an agent subject and an instrument may be marked by ergative case in a transitive finite clause, only the overt subject may be assigned ergative case in an associative or privative clause with a transitive verb, while only an instrument may be assigned ergative case in a purposive clause. The instrument associated with a transitive infinitive verb in an associative or privative clause predicated of its logical object, can only be expressed external to the infinitive phrase by a DP governed by a semantic case which assigns its role and case, in the manner of an instrument phrase in an intransitive finite clause. The logical subject of an associative or privative clause predicated of an external DP co-​referent with the logical object of its embedded infinitive may also be expressed in this way. Privative clauses may also be dynamic with a PRO subject as in (42b and b’) and (44a and b). Like other complementizer-​headed infinitive clauses which form a dynamic predicate, a dynamic privative clause only has an active voice reading, since only its logical subject is external to the infinitival predicate.

 

984   Mary Laughren Assuming that the obligatorily controlled subject of the non-​finite clauses examined in section 39.3.1–​2 occupies a unique syntactic position irrespective of the transitivity of the verb, and that like the subject of a finite clause, this position is external to the maximal projection of the verbal or nominal predicate, it has been argued that case is not assigned to this position in non-​finite clauses, while in finite clauses a single syntactic case is assigned which has variant morphological forms, ergative-​marked or unmarked, depending on the transitivity of the lexical verb (Levin 1983). Alternatively, it has been argued that the subject must raise out of the maximal projection of the predicate in a finite clause, irrespective of whether it has been assigned ergative case (Woolford 1997). In non-​finite clauses, ergative case can only mark the argument of a transitive verb which is internal to the projection of the verb; an external argument must have a different case-​assigner and marker, independent of the transitivity of the verb such as the dative case in obviative and purposive clauses, or the complementizer in associative and privative clauses. If external to the scope of the complementizer, the overt logical subject can only be expressed if governed by a semantic case (or post-​position). Like Warlpiri clauses with a controlled PRO subject or an overt dative-​marked subject, stative associative and privative clauses which are predicated of an external DP co-​ referent with their null logical subject have an active voice interpretation, and active voice syntactic structure in which the subject dominates the object. When , on the other hand, their logical object is co-​referent with the external subject of the associative or privative clause, these have a passive voice-​like interpretation. The question then arises as to their syntactic status: are these complementizer headed non-​finite clauses structurally active or passive? If passive, the subject of the complementizer-​headed predicate would be PRO linked to the thematic role associated with O or Odat in the corresponding active transitive clause. This would assume that dative case is not assigned to PRO, the same way that accusative case on the object of a transitive clause in English or Latin is not assigned to the subject of a passive clause involving the same thematic role. A passive analysis would assume a phonologically null passive voice feature that would govern the infinitive. A passive syntactic analysis would assume that the agent role could only be linked to an optional adjunct to the non-​finite clause (like the by-​phrase in an English passive clause). However, when the null logical object is co-​referent with the external subject of the associative/​privative stative predicate, the agent role can be assigned to an ergative-​marked DP within the projection of the infinitive verb, or a bare DP within a complementizer or postpositional phrase, as shown in (32) to (36). It is instructive to compare aspects of Warlpiri ergative and dative case-​marking with that found in Panyjima which has overtly marked passive clauses, presented in section 39.3.7.

39.3.7 Warlpiri and Panyjima Ergative-​Marked Agents and Instruments Compared Panyjima is one of the Ngayarda languages of the Pilbara area of Western Australia which have innovated an unmarked nominative subject and marked object (derived

 

The ergative in Warlpiri    985 from the historical dative) case-​marking system, and an active versus passive voice contrast, marked on verbs. In passive clauses, the agent can be expressed by a DP marked by a suffix derived from the historical ergative case (Dench 1982, 2006). This is exemplified by the main clause examples in (45a and b) and dependent clause examples in (46). Warlpiri translations are given in the prime-​marked examples. (45) Panyjima (from Dench 1991:139) a. Ngatha wirnta-​nnguli-​nha marnta-​ku. 1sg.s cut-​pass-​past stone-​erg ‘I got cut by a stone.’ (e.g. I stepped on a sharp stone which cut me). (#15) a’. Ngaju=ju paju-​rnu 1sg=1sg.o cut=past ‘A stone cut me.’

pirli-​ngki. stone-​erg

b. Ngatha wirnta-​nnguli-​nha marnta-​ngarni-​lu (jilya-​ngku). 1sg.s cut-​pass-​past stone-​com-​erg child-​erg ‘I was cut (by a child) with a stone.’ (#16) b’. Ngaju=ju paju-​rnu pirli-​kirli-​rli/​pirli-​ngki (kurdu-​ngku). 1sg=1sg.o cut=past stone-​com-​erg/​stone-​erg child-​erg ‘(The child)/​pro cut me with a stone.’ c. Ngatha wirnta-​rna jilya-​ku 1sg.s cut-​past child-​acc ‘I cut the child with a stone.’ c’. Ngajulu-​rlu=rna paju-​rnu 1sg-​erg=1sg.s cut-​past ‘I cut the child with a stone.’

marnta-​ngarni. stone-​com

kurdu child

pirli-​ngki/​pirli-​kirli-​rli. stone-​erg/​stone-​com-​erg

Comparing the Panyjima sentences in (45a and b) with their Warlpiri translations, the similarities in case-​marking are striking. However, ergative case is only assigned in the Panyjima clauses if the verb is marked passive and the patient/​theme is realized as the subject of the clause. In the active voice (45c) the instrument is not marked ergative as in the passive sentences but projected in a comitative phrase as in Warlpiri intransitive or semi-​transitive clauses, for example (6b). The case-​marking of the subject and object DPs in the Panyjima active voice clause in (45c) has the same morphological form as in a Kalkutungu antipassive clause, i.e. unmarked subject and dative-​marked complement. However, the Panyjima accusative-​marked DP in (45c) has object properties like the Warlpiri dative-​marked DP object in semi-​transitive and ditransitive clauses, as demonstrated by its equivalence with the subject of the corresponding passive clause. Panyjima can be viewed as having simplified the Warlpiri case-​marking system by eliminating the contrast between morphologically marked and unmarked subject and object, innovating a system where the subject is unmarked, while the object is marked. There is a one-to-one match between abstract and morphological case.

 

986   Mary Laughren In Panyjima finite passive clauses, an agent or instrument phrase may be adjoined to the verb phrase if assigned ergative case, corresponding to the ergative-​marked agent or instrument subject in a Warlpiri finite (active) clause. In conjunction with the verb, the ergative assigns the instrument role in (45a) and the agent role in (45b). As in Warlpiri non-​finite clauses, the ergative case cannot directly assign both roles in the same clause, or more precisely in the clause’s predicate. In (45b) the agent role is assigned by the ergative, while the comitative must be used to assign the instrument role, as in the active voice clause in (45c), but in the passive clause in (45b) the comitative phrase must also be assigned ergative case. As discussed in section 39.2, ergative case is a property of a category of Warlpiri verbs; in Panyjima it is a property of passivized verbs. In Warlpiri, the ergative, in conjunction with the verb, assigns the agent/​instrument thematic role, but does not assign case to the subject, only to a non-​subject instrument phrase, adjoined to the VP, as in Panyjima. The subject position, which is external to the projection of the verb, does not differentiate between transitive and intransitive verbs; in finite clauses, the same abstract syntactic case is assigned to the subject, while the overt subject of non-​finite clauses is assigned dative case. The fact that the covert subject of classes of infinitive clauses must be controlled PRO, in a position that is not assigned abstract case, indicates that the ergative-​ marking on the subject of transitive verbs does not represent a case distinct from the case on unmarked subject DPs but is an inherent or lexical case. In Warlpiri non-​finite clauses, only instrument or agent phrases internal to the infinitive phrase are marked ergative. In Panyjima only instrument or agent DPs adjoined to a passive VP are marked ergative. Inherent case is incompatible with structural case. Panyjima has clauses similar in type to Warlpiri associative clauses, with verbs glossed by Dench (1991) as ‘perfect’ forms. In both (46a and b) the perfect clause is predicated of a DP in the matrix clause which controls the reference of the null logical object (patient) which is realized as the PRO subject of a non-​finite passive clause. Like a Warlpiri stative associative clause, the perfect-​marked Panyjima verb agrees in case with its external subject in the matrix clause, but the agent DP is assigned its thematic role and inherent ergative case as an adjunct to the VP headed by the passive perfect-​marked verb and is not within the scope of subject–​predicate case-​agreement (accusative in (46a and b)). (46)

Panyjima (Dench, p.c. 2010) a. Jilya ngarna-​ku mantu-​yu, [[kampa-​rnaanu-​ku] ngunya-​ngku palya-​ngku]. child eat-​pres meat-​acc cook-​pperf-​acc that-​erg woman-​erg ‘The child is eating the meati _​_i​ cooked by the woman.’ b. Jilya ngarna-​ku mantu-​yu, [[kampa-​nnguli-​lha-​ku] ngunya-​ngku palya-​ngku]. child eat-​pres meat-​acc cook-​pass-​perf-​acc that-​erg woman-​erg ‘The child is eating the meati _​_i​ cooked by the woman.’

 

The ergative in Warlpiri    987 c. Warlpiri translation of Panyjima a. and b. Kurdu-​ngku=ka nga-​rni kuyu [[karnta-​ngku+purra-​nja]-​warnu] child-​erg=aux eat-​npast meat woman-​erg+cook-​inf-​assoc ‘The child is eating the meat cooked by the woman.’ c’. [karnta+[purra-​nja]-​warnu] woman+cook-​inf-​assoc c”. [karnta-​warnu] [purra-​nja-​warnu] woman-​assoc cook-​inf-​assoc Panyjima and Warlpiri differ in the syntactic relation between their ergative-​marked ‘agent’ and the verb in the dependent clause. The Warlpiri ergative-​marked agent DP obligatorily pre-​posed to the verb in (46c) is more tightly integrated into the associative-​ marked infinitive phrase than its counterpart in the Panyjima passive perfect clause. Similarly, the alternative expressions of the logical subject in (46c’ and c”) are also embedded in the associative phrase. The Warlpiri ergative-​marked DP is the specifier of the verbal phrase from which the object has been extracted, while the Panyjima ergative-​marked DP is adjoined to the passive verb phrase whose controlled subject also corresponds to the logical object of the verb.

39.3.8  The Ergative in “Light Warlpiri” To conclude, it is enlightening to observe that in the mixed language dubbed “Light Warlpiri” that has emerged from code-​switching between Warlpiri and Kriol, an English lexified creole (Meakins and O’Shannessy 2010; O’Shannessy 2005, 2006, 2013), Warlpiri case forms have been retained. The basic word order is SVO. A new auxiliary consisting of subject morpheme proclitic to a tense/​mood morpheme has replaced the Warlpiri auxiliary and most verbs are Kriol based. While ergative-​marking is mostly absent on the immediate pre-​auxiliary subject DP, it is nearly always present in other positions, particularly in the post-​verbal topic position. This suggests that the ergative is not required to assign the ‘agent’ or ‘instrument’ thematic role of the subject, but that it is required to license a DP adjunct co-​referent with the typically null (pro) pre-​auxiliary transitive subject while in the pre-​verbal subject position unmarked nominative case is assigned. One can envisage a further development where this difference in the marking of the post-​verbal subject DP, depending on the category of the verb, is either reanalyzed as a discourse marker applicable to all post-​verbal subjects, or it is dropped. Either way, the distinction between marked and unmarked subject would be neutralized. For the present, however, it would seem that the distinction between ergative-​marked and

 

988   Mary Laughren unmarked subjects is preserved, even though the traditional Warlpiri verbs have been replaced with English derived forms.

Abbreviations 1, first person; 2, second person; 3, third person; ACC, accusative; ALLAT, allative; ANAPH, anaphor; AP, antipassive; ASSOC, associative; AUX, auxiliary; COM, comitative; COMP, complementizer; CS, changed state; DAT, dative; DAT.EXT, dative extended; DET, determiner; DP, determiner phrase; DU, dual; EG, for example; EMPH, emphatic; ERG, ergative; FOC, focus; FUT, future; HAB, habitual; IMP, imperative; IMPF, imperfective; INCH, inchoative; INF, infinitive; IP, inflectional phrase; IRR, irrealis; LOC, locative; LOCCOMP, locative complementizer; NEG, negative; NFUT, non-​future; NOM, nominative; NP, noun phrase; NPAST, non-​past; O, object; OBJCOMP, object complementizer; OBVCOMP, obviative complementizer; PASS, passive; PAST, past tense; PERF, perfect; PERL, perlative; PL, plural; POSS, possessive; PPERF, passive perfect; PRES, present tense; PRIV, privative; PRTC, participle; PURP, purposive; REFL, reflexive; RSLT, resultative; S, subject; SG, singular; SIB, sibling; SRC, source; SCOMP, subject complementizer; TOP, topic; VP, verb phrase.

 

Chapter 40

E rg ative–​a b s olu t i v e pat terns in Tong a n: an overv i ew Yuko Otsuka

This chapter provides an overview of ergativity in Tongan, a Polynesian language of the Austronesian family, spoken in the Kingdom of Tonga. Tongan shows ergative–​ absolutive (ERG–​ABS) patterns in morphology as well as syntax, treating the subject of a transitive verb differently from the subject of an intransitive verb and the object of a transitive verb (A, S, and O in Dixon’s (1979) terms). Furthermore, the ERG–​ABS pattern is not consistent throughout the language. Noun morphology shows a split between clitic pronouns and other types of nouns. In syntax, the ERG–​ABS pattern comes in at least three distinct types. Whether to consider all of these ERG–​ABS patterns to be manifestations of ergativity is an interesting theoretical question. This issue is addressed at the end of the chapter.

40.1 Basic Morphosyntax The unmarked constituent order in Tongan is VSO. More accurately, Tongan is a predicate-​initial language: not only verbal predicates, but also nominal as well as prepositional predicates occur before the subject noun phrase (NP). Verbs do not inflect for tense or agreement. Instead, tense/​aspect/​mood is expressed by means of sentence-​ initial particles (henceforth TAMs).1 1  In Tongan standard orthography, an inverted apostrophe (‘) is used to represent the glottal stop and a macron, a long vowel. Definiteness is phonologically realized as definitive accent (Churchward 1953), i.e. lengthening of the phrase final vowel (Anderson and Otsuka 2006), and is indicated in orthography by an acute accent (´).

 

990   Yuko Otsuka (1) a. Na‘e kata ‘a pst laugh abs ‘John laughed.’ b. Na‘e langa ‘e pst build erg ‘John built a house.’

Sione. John Sione John

(‘a) abs

e spec

fale. house

VOS is also freely permitted. The alternation between VSO and VOS appears to be governed by discourse–​pragmatic factors. That is, given a context in which one of the arguments is new information and the other isn’t, the NP representing new information must occur in the position immediately following the verb (Otsuka 2005b).2 For example, VSO, but not VOS, is felicitous in response to subject wh-​questions, and vice versa for object wh-​questions. Consider (2), where VOS is rendered infelicitous, indicated by #. The item bearing new information focus is indicated in bold in English translation. (2) Q:   Ko   hai   na‘a  ne     fili     ‘a      Pila?   pred  who  pst   3.sg choose  abs  Pila       ‘Who chose Pila?’ A1:  Na‘e fili   ‘e  Sione ‘a  Pila.   pst   choose    erg  Sione  abs  Pila  ‘Sione chose Pila.’ A2: #Na‘e fili ‘a   pst choose abs   ‘Sione chose Pila.’

Pila Pila

‘e erg

Sione. Sione

2 

Contrary to Otsuka’s (2005b) observation, Custis (2004: 19) argues that the position immediately following the verb is a pragmatic topic. Thus, she reports that VSO, but not VOS, is felicitous in response to ‘what did Mele do?’ and vice versa if the question is ‘What happened to the fish?’ Although Custis does not provide the Tongan sentences corresponding to these questions, the most natural way to ask these in Tongan would be those given in (i) and (ii) here. It should be noted, however, in these questions, the new information is the predicate rather than a particular argument. Custis notes that VSO is a felicitous answer to the subject wh-​questions, but argues that it is because VSO is a default order (and that therefore, it does not undermine her analysis that the position immediately after V is reserved for the topic). Custis doesn’t mention whether VOS or VSO is preferred (or infelicitous) as an answer to an object wh-​question.   (i) Ko e hā e me‘a   na‘a  ne   fai ‘e Mele? pred spec what spec thing  pst 3.sg do erg Mary Lit. ‘What is the thing that Mary did?’ (ii) Ko e hā e me‘a na‘e hoko pred spec what spec thing pst happen Lit. ‘What is the thing that happened to the fish?’

ki to

he spec

iká? fish.def

 

Ergative–absolutive patterns in Tongan    991 Noun phrases are always accompanied by a determiner, specific he/​e or indefinite ha, which in turn is preceded by a case marker or a preposition.3 As seen in (1), the sole argument of an intransitive verb (S) as well as the patient-​like argument of a transitive verb (O) are marked by ‘a, whereas the agent-​like argument of a transitive verb (A) is marked by ‘e. Thus, case morphology on common nouns and proper names shows an ERG–​ABS pattern. The ABS marker is obligatory with proper names, but is optional with common nouns when followed by a specific article. The ERG marker, on the other hand, is always obligatory. Property-​denoting predicates behave like intransitive verbs when used as a predicate, occurring in the position immediately following the TAM without a copula. (3)

‘oku poto ‘aupito prs smart very ‘John is very smart.’

‘a abs

Sione. John

Examples of non-​verbal constructions are provided in (4). Non-​verbal predicates do not require a copula. Prepositional (locative) predicates occur immediately after the TAM, as seen in (4b). Nominal predicates, on the other hand, cannot co-​occur with a TAM. Instead, the sentence-​initial position is occupied by the predicate marker ko.4 (4)

a. Ko e faiako pred spec teacher ‘John is a teacher.’ b. ‘oku ‘i heni prs in here ‘John is here.’

‘a abs

‘a abs

Sione. John

Sione. Sione

40.2 Pronouns Case marking on pronominal arguments presents a more complicated picture. Tongan has two sets of personal pronouns: independent and clitic (Table 40.1). Independent pronouns behave like other NPs in that they occur in any position that an NP can occur and can also be marked by a case marker (although omission is preferred especially 3  The articles in Tongan and their semantic features are discussed in great detail by Macdonald (2014), who convincingly argues that he/​e expresses specificity rather than referentiality. The distribution of the two allomorphs of the specific determiner, he and e, is morphophonologically conditioned: he follows a particle ending in [i]‌or [e] (prepositions ‘i, ki, mei, and ERG marker ‘e) and e follows a particle ending in [a] or [o] (possessive markers ‘a and ‘o, ABS marker ‘a, predicate marker ko, comitative mo). 4 Since ko does not occur in property-​denoting predicates or locative predicates, I do not consider it to be a copula. Rather, I treat it as an item that indicate the predicative use of NPs or that has the function of turning a NP into a predicate. Otsuka (2000) proposes that ko is a type of K head and licenses a case-​ less NPs. This view is partially supported by the fact that ko cannot co-​occur with a case marker, ‘e or ‘a.

 

992   Yuko Otsuka when ABS) or a preposition, as illustrated in (5). They can also occur as a predicate marked by ko. (5) a. Na‘e taa‘i ‘e pst hit erg ‘John hit you.’

Sione John

‘a abs

koe. 2.sg

b. Na‘e tala mai ‘e ia ‘oku pst tell dir.1 erg 3.sg prs ‘He told me that was correct.’ c. Na‘e ‘omai ‘e Sione ‘a pst give.me erg John abs ‘John gave a book to me.’

tonu correct

e spec

tohi book

ki-​ate au. to-​pron 1.sg

Table 40.1 Tongan pronouns 1.sg 1.d.incl 1.d.excl 1.pl.incl 1.pl.excl 2.sg 2.d 2.pl 3.sg 3.d 3.pl

Independent

Clitic

au (ki)taua (ki)maua (ki)tautolu (ki)mautolu koe (ki)moua (ki)moutolu ia (ki)naua (ki)nautolu

ou, u, kua ta ma tau mau ke mo mou ne na nau

a Allomorphs of 1.SG pronoun are governed mainly

by tense/​aspect: u is used after te (future) and kuo (perfective); ku is used after na‘a (past); ou is used after ‘oku (present).

It should be noted, however, that the use of independent pronouns as core arguments (ABS or ERG) as illustrated in (5) is marked and has an effect of emphasis. The following two strategies are used much more commonly to express pronominal core arguments. First, clitic forms are used for A/​S arguments. Clitic pronouns occur immediately after TAM and cause allomorphy in TAM (e.g. na‘e vs. na‘a). Second, pronominal O arguments occur in their independent forms, but tend to occur immediately after the verb (preceding the A-​argument) and without the case marker as in (6c).5 5  The position of such pronominal O is reminiscent of VOS discussed in section 40.1. Obligatory omission of the ABS case marker in this position suggests that these pronouns are not in the same position as the ABS NP in VOS constructions, however.

 

Ergative–absolutive patterns in Tongan    993 (6) a. Na‘a ku pst 1.sg ‘I laughed.’

kata. laugh

b. Na‘a ku ma‘u pst 1.sg get ‘I caught a fish.’ c. Na‘e taa‘i au pst hit 1.sg ‘John hit me.’

‘a abs

e spec

‘e erg

Sione. John

d. *Na‘a ku taa‘i ‘e    pst 1.sg hit erg    Intended: ‘John hit me.’

Sione. John

ika. fish

As illustrated in (6), clitic pronouns show a nominative–​accusative (NOM–​ACC) pattern: only S or A may be realized as a clitic pronoun. Clitic forms may not be used for pronominal O’s, as shown in (6d). This pattern could be regarded as a type of split ergativity. However, despite such a superficially NOM–​ACC pattern, pronominal S and A are distinguished syntactically (see section 40.4.2). Thus, I assume that despite their identical forms, pronominal S and A bear different abstract case, ABS and ERG, respectively.6 It should also be noted that third person singular pronouns (ne and ia) are often omitted when their referents are retrievable from context. Given their discourse dependence, the null allomorph should be understood as a variable bound by the discourse topic, in a way similar to null arguments in Japanese and Chinese (Huang 1984). Unlike these languages, however, the use of topic variable is limited to third person singular in Tongan (cf. Dukes 1996; Ball 2009).7

40.3  Ergative–​Absolutive Patterns in Syntax Ergativity in the sense of ERG–​ABS pattern is not limited to case morphology in Tongan. In the following sections, we examine three syntactic phenomena showing an 6 

Ball (2009) treats these dependent pronouns as inflectional affixes on T (which he treats as auxiliary) and conjunctions such as pea. See also Macdonald 2014 for an alternative analysis of Tongan clitic pronouns. 7  Dukes (1996) observes that topic drop is permissible for non-​third person pronouns in ERG positions. However, his examples of non-​third person topic drop can be better analyzed as a separate phenomenon, i.e. agentless transitive construction (Otsuka 2010b). On the other hand, Ball (2009: 79) argues that only ABS can be phonetically null. Thus, there appears to be variation among speakers as to the constraints on topic drop.

 

994   Yuko Otsuka ERG–​ABS pattern from a descriptive point of view. In the subsequent section, we will consider the theoretical implication of the descriptive observations presented in this section.

40.3.1 Relativization Relativization shows an ERG–​ABS pattern in that the use of gap strategy is limited to S/​ O arguments, as illustrated in (7). (7) a. ki he fefinei [na‘e to spec woman pst ‘to a woman who laughed’

kata _​_​i] laugh

b. ki he fefinei [na‘e fili to spec woman pst choose ‘to a woman whom John chose’

[ABS-​S gap]

‘e erg

Sione _​_​i] John

c. *ki he fefinei [na‘e kai _​_​i ‘a e    to spec woman    pst eat abs spec ‘to a woman who ate a fish’

ika] fish

[ABS-​O gap]

[*ERG gap]

However, relativization of A-​arguments is not completely banned. While the gap strategy results in ungrammaticality, relativization of A-​ arguments is permissible as long as the relative clause contains a resumptive pronoun, as illustrated in (8) (Chung 1978; Otsuka 2000, 2006). In this sense, Tongan is different from other so called syntactic ergative languages such as Dyirbal in which relativization of A-​ arguments is simply not allowed (see in this volume Aissen, Chapter 30 and Erlewine et al., Chapter 16). (8) ki he fefinei [na‘a nei to spec woman pst 3.sg ‘to a woman who (she) ate a fish’

kai eat

‘a abs

e spec

ika] fish

[ERG pron]

A similar constraint is observed in wh-​questions. This is expected, as wh-​questions in Tongan are pseudo-​cleft, whose subject is a headless relative clause. (9) a. Ko hai na‘e kata _​_​? pred who pst laugh ‘who laughed?’               (lit. ‘Who is the one that laughed?’) b. Ko e hā na‘e kai ‘e Sione _​_​? pred spec what pst eat erg John ‘What did John eat?’          (lit. ‘What is the thing that John ate?’)

 

Ergative–absolutive patterns in Tongan    995 c. *Ko hai na‘e kai _​_​ ‘a e ika?   pred who pst eat abs spec fish   ‘Who ate a fish?’            (lit. ‘Who is the one that ate a fish?’) d. Ko hai na‘a ne kai ‘a e ika? pred who pst 3.sg eat abs spec fish ‘Who ate a fish?’ ’         (lit. ‘Who is the one that (he) ate a fish?’)

40.3.2 Coordination Reduction Coordination involving a conjunction pea ‘and (then)’ also shows an ERG–​ABS pattern. Pea conjoins two clauses or what appear to be verb phrases. In (apparent) VP-​ coordination, the second conjunct may contain a gap coreferential with an argument of the first conjunct, as shown in (10). (10)

Na‘e foki ‘a Sionei ki ‘api pea nofo ai _​_​_​i. pst return abs John to home and stay there ‘John went home and stayed there.’

[S = S gap]

The example in (10) is not strikingly unique, as instances of coordination reduction like this are commonly found crosslinguistically. However, careful examination reveals that coordination reduction involving pea is constrained in an ERG–​ABS pattern: the gap and the antecedent must be in the same case. In (11), in which the first conjunct contains two arguments, the gap can only be coreferential with the ABS argument Mele, but not with the ERG argument Sione. (11)

Na‘e fakamatalili‘i ‘e Sionei ‘a Melej pea hola _​_​_​*i/​j. [O/​*A = S gap] pst tease erg John abs Mary and run.away ‘John teased Mary and (*he/​she) ran away.’

Likewise, when a gap occurs as an O, it can be coreferential with the ABS argument in the first conjunct, be it S as in (12a) or O as in (12b). Coreference with the ERG argument is impossible, however. Thus, in (12b), the gap can only be coreferential with the ABS-​ marked argument, Mele, of the first conjunct. (12)

a. Na‘e kata ‘a Sionei pea pst laugh abs John and ‘John laughed and Mary hit (him)’

taa‘i ‘e   Mele _​ _​_​i hit erg Mary

[S = O gap]

b. Na‘e ‘ave ‘e Sionei ‘a Melej ki he palasí [O/​*A = O gap] pst take erg John abs Mary to spec palace.def

 

996   Yuko Otsuka pea fakamolemole‘i ‘e he kuiní _​_​_​*i/​j. and pardon erg spec queen.def ‘John took Mary to the palace and the queen pardoned (*him/​her).’ Interestingly, as with relativization, the ERG–​ ABS pattern observed in pea-​ coordination does not necessarily prohibit ERG to participate in coordinate reduction. Coreference with the A-​argument is permissible as long as the gap itself is an A-​argument. Thus, (13a), in which the gap is an A-​argument, is grammatical as long as it is interpreted as coreferential with the A-​argument Sione of the first conjunct: ‘John took Mary to the market and he bought taro.’ However, it cannot mean ‘John took Mary to the market and she bought taro,’ where coreference is intended between the ERG gap and the ABS argument. Similarly, (13b) is ungrammatical because the intended coreference is between the ERG gap and the ABS-​marked argument.8 (13) a. Na‘e ‘ave ‘e Sionei ‘a Melej ki he māketí [A/​*O = A gap] pst take erg John abs Mary to spec market.def pea fakatau mai _​ _​ i/​*j  ‘a     e       talo and buy     dir.1    abs spec taro ‘John took Mary to the market and (he/​*she) bought some taro.’ b. *Na‘e kata ‘a Sionei pea taa‘i _​_​i ‘a Mele.    pst laugh abs John and hit abs Mary    ‘John laughed and hit Mary.’

[*S = A gap]

It should be noted that this same-​case constraint on coreference is in effect only when the second conjunct contains a gap. When the second conjunct contains an overt pronoun instead of a gap, the pronoun (be it ABS or ERG) can be coreferential with any of the arguments in the first conjunct. Compare (13) and (14). (14) a. Na‘e ‘ave ‘e Sionei ‘a Melej ki he māketí [A/​O = A pron] pst take erg John abs Mary to spec market.def pea nei/​j  fakatau  mai  ‘a     e        talo. and    3.sg  buy      dir.1     abs spec taro ‘John took Mary to the market and he/​she bought some taro.’ b. Na‘e kata ‘a Sionei pea ne i taa‘i ‘a Mele. pst laugh abs John and 3.sg hit abs Mary ‘John laughed and he hit Mary.’

[S = A pron]

8  Contrary to the present observation, Ball (2009: 78–​80) claims that the gap cannot be ERG in pea-​ coordination and suggests that this is due to the general ban on null realization of ERG arguments. My consultants permit null instantiation of the ERG argument in the second conjunct of pea and also as a topic variable when 3.SG (section 40.2).

 

Ergative–absolutive patterns in Tongan    997

40.3.3  ‘o-​Construction Yet another ERG–​ABS pattern is found in constructions involving a complementizer ‘o ‘so that.’ Semantically, ‘o-​constructions have a single-​event reading, with the action or state denoted by the second verb phrase representing the necessary (or expected) result of the action denoted by the first. (15)

a. Na‘e ha‘u ‘a Sionei ‘o pst come abs John comp ‘John came (for him) to stay with us.’

ne i 3.sg

nofo stay

mo with

b. Na‘e hiva mālie ‘a e tamaikí ‘o pst sing beautifully abs spec children.def comp e kau faiakó. spec pl teacher.def ‘The children sang beautifully and the teachers were happy.’

mautolu. 1.pl.excl fiefia ‘a happy abs

When the ‘o-​clause contains a gap, an ERG–​ABS pattern emerges. Constructions involving ‘o are similar to pea coordination (but different from relativization) in that the gap may be ABS (16a, b) or ERG (16c). On the other hand, they are different from pea coordination in that the gap and the antecedent need not to be in the same case. As illustrated in (16c), for example, the ERG gap can be coreferential with the ABS argument of the first verb. (16)

a. Na‘e ha‘u ‘a Sionei ‘o nofo _​_​i mo mautolu pst come abs John comp stay with 1.pl.excl ‘John came to stay with us.’

[S = S gap]

b. Te ui ha‘u ‘o fakalangilangi‘i ‘e Mele _​_​i. fut 1.sg come comp praise erg Mary ‘I will come so that Mary will praise (me).’

[S = O gap]

c. Na‘a naui ha‘u ‘o langa _​_​i ‘a e pst 3.pl come comp build abs spec ‘They came and built a hotel.’

[S = A gap]

hotele. hotel

However, when the first verb is transitive, the gap in the ‘o-​clause must be coreferential with the ABS argument. The ERG argument cannot be the antecedent, no matter whether the gap is ABS (17a) or ERG (17b). In other words, the antecedent is restricted to ABS, showing an ERG–​ABS pattern and the preference for ABS. (17)

a. Na‘e fili ‘e Sionei ‘a Melej ‘o fiefia _​_​_*i/​ ​ j. [O/​*A = S gap] pst choose erg John abs Mary comp happy ‘John chose Mary and (*he/​she) was happy.’

 

998   Yuko Otsuka b.

Na‘e ‘ave ‘e Sionei ‘a Melej ki he māketí [O/​*A = A gap] pst take erg Sione abs Mary to spec market.def ‘o       fakatau  mai _​_​*i/​j  ‘a     e    kumala. comp  buy         dir.1    abs  spec  sweet.potato ‘John took Mary to the market and (*he/​she) bought sweet potato.’

Again, as with pea-​coordination, when ‘o-​clauses contain a pronoun instead of a gap, its antecedent need not be the ABS argument in the matrix clause. A pronoun can be coreferential with the ERG argument of the matrix clause, as shown in (18). This indicates that the Case constraint on coreference is not a constraint on ‘o-​clauses per se, but concerns specifically the null argument in ‘o-​clauses. (18) Na‘e fili ‘e Sionei ‘a Melej ‘o nei/​j fiefia. pst choose erg John abs Mary comp 3.sg happy ‘John chose Mary and (he/​she) was happy.’

[A/​O = S pron]

40.4  Formal Analysis of Syntactic Ergativity in Tongan The various ergative–​accusative patterns Tongan exhibits in syntax are summarized in (19). (19) ERG–​ABS patterns in Tongan syntax a. Relativization: ABS gap, ERG pronoun/​*gap b. coordination reduction involving pea: only ERG-​ERG gap or ABS-​ABS gap c. coreference in ‘o-​construction: only ABS can be the antecedent Syntactic ergativity generally refers to syntactic prominence of ABS. In that sense, only the patterns observed in relativization and ‘o-​constructions fit the bill. Even within these two phenomena, the “prominence” of ABS is manifest in different manners. This leads to two theoretically important questions: (a) whether any or all of these ERG–​ABS patterns should be considered an instance of syntactic ergativity, i.e. syntactic prominence of ABS; and (b) if so, whether there is a single underlying factor, e.g. structural position, that is responsible for the syntactic prominence of ABS in all of these patterns. These two questions will be addressed, following formal analysis of the three phenomena.

40.4.1 Assumptions about Phrase Structure in Tongan The following discussion is couched in a theoretical framework of the minimalist program (Chomsky 1995, 2000, 2001). In this framework, syntactic objects (lexical as well

 

Ergative–absolutive patterns in Tongan    999 as functional items, phrases, and sentences) are regarded as consisting of a set of formal features. Some formal features such as categorial features are interpretable and others such as agreement features (hereafter phi-​features) on functional heads are uninterpretable. Interpretability of a feature is determined based on its relevance to the LF interface. It is assumed that a syntactic object that contains uninterpretable features would be deemed ungrammatical. Syntax is thus understood as a process in which uninterpretable features are eliminated by means of feature valuation (Pesetsky and Torrego 2001). Feature valuation is achieved in Agree, a relation between an uninterpretable feature and the closest matching interpretable feature in its c-​command domain, in which the former receives a specific feature-​value from the latter. In relation to Tongan specifically, I adopt the following assumptions. V-​initial order in Tongan is derived by means of verb raising, cyclic head movement of V-​to-​T-​to-​C (Otsuka 2000, 2005a).9 ERG and ABS are both regarded as structural cases, valuated by T and v, respectively (Levin and Massam 1985; Bobaljik 1993a; Laka 1993b; and Otsuka 2000, 2005a,b).10 In this approach, case checking in transitive constructions is essentially the same in the ERG–​ABS system and the NOM–​ACC system: A receives case from T and O, from v (T-​case and V-​case, respectively).11 The difference between the two systems lies in intransitive constructions, specifically, which structural case is available in intransitive constructions. In the NOM–​ACC system, it is T-​case, whereby S receives the same case as A; in the ERG–​ABS system, it is V-​case, whereby S receives the same case as O. This can be understood as parametric variation in functional heads: accusative languages have two kinds of v’s, one with a case feature and the other without (for unaccusatives); ergative languages have two kinds of T’s, one with a case feature (for transitives) and the other without. A-​arguments are generated in [Spec, v] as standardly assumed, but S-​arguments are generated VP-​internally; in [Spec, V] for unergatives and as V’s complement for unaccusatives, although it is not clear whether the distinction between the two kinds of intransitive constructions exists in Tongan (Otsuka 2000).12 It is also assumed that both T and v have an EPP-​feature, which licenses movement of a relevant DP: the ERG-​marked NP to [Spec, T] and ABS-​marked NP, to a lower [Spec, v], assuming the raised direct object is tucked in below the subject (Richards 1997). The basic phrase structure for transitive constructions in Tongan is provided in (20), where the copy of a moved item is indicated by angular brackets . 9  See Custis 2004, Ball 2009, and Macdonald 2014 for alternative analyses of Tongan verbal constructions. Also see Massam 2000, 2001 for an alternative VP-​raising approach to V-​initial order in Niuean, a Polynesian language closely related to Tongan. 10  This particular approach is proposed to account for Tongan facts and should not be considered to be applicable to all languages that show an ERG–​ABS case alignment. Morphological ergativity is likely to arise due to various different case marking strategies in different languages (see Legate, Chapter 6, this volume). 11  Though from a different perspective, Bittner and Hale (1996a,b) also treat the distinction between marked and unmarked cases as a crucial factor in analyzing the difference between ERG–​ABS and NOM–​ACC systems. 12  See Wiltschko 2006 for a similar proposal for Halkomelem Salish, an ergative language spoken in British Columbia.

 

1000   Yuko Otsuka (20)

CP TP

C-T-v-V DP1

T

[erg]

vP v’

v’

DP2

[abs]

VP



Though there is not enough space to discuss arguments and evidence for each of these assumptions, let us consider one argument in support of T’s EPP-​feature. Based on the observation that VOS is preferred when O represents new information, Otsuka (2005b) has proposed that scrambling in Tongan should be understood as movement to [Spec, T], which is motivated by a new information focus feature (as proposed by Zubizarreta (1998) for subject inversion in Spanish) and an EPP-​feature on T (as proposed by Miyagawa (2001, 2005) for scrambling in Japanese). The former identifies the DP with a new information focus feature and the latter licenses its movement to [Spec, T]. Thus, movement to [Spec, T] is not contingent on case feature valuation. In unmarked contexts, however, it is the ERG-​marked NP that moves to [Spec, T], as it enters Agree relation with T for case feature valuation, whereas the ABS-​argument, without the information focus feature, does not. Binding facts seem to support this analysis. Binding in Tongan are subject to two constraints (Otsuka 2000; Ball 2009). First, ERG can bind ABS, but not vice versa, regardless of the linear order. Second, when the linear order is VOS, ERG cannot bind ABS. This second fact suggests that the linear position entails c-​command relations and that accordingly, ERG c-​commands ABS in VSO constructions.

40.4.2 Relativization In theory, the prominence of ABS over ERG with respect to relativization could be attributed to their relative hierarchical positions. If ABS is structurally higher than ERG in such a way that the former blocks access to the latter, the feature that licenses A-​bar extraction (i.e. the EPP-​feature on C) cannot reach ERG and hence, extraction would be impossible. Such an approach has been proposed by Aldridge (2004), for example,

 

Ergative–absolutive patterns in Tongan    1001 to account for a similar constraint on relativization in two Austronesian languages, Tagalog and Seediq of the Philippines and Taiwan, respectively. In her analysis, ABS is located in the upper [Spec, v] due to the EPP feature on transitive v (and assuming that the additional Spec is created above the existing Spec, contra the tacking-​in approach proposed by Richards (1997), and that there is a hierarchical order among multiple Specifiers of a single head in terms of accessibility). This puts ABS in a position structurally higher than ERG and as a result, C’s EPP-​feature cannot reach ERG to license its raising to [Spec, C]. Such a structural approach does not work for Tongan, however, as there is no independent evidence to support the structural prominence of ABS over ERG. As seen in (20), in unmarked context, ERG is structurally higher than ABS, being in [Spec, T]. ABS does end up higher than ERG when O has new information focus and moves to [Spec, T], yielding VOS. Crucially, however, ABS is not consistently higher than ERG. And from C’s perspective, anything that is in [Spec, T] should be accessible and in fact, closer than the one inside vP. Note also the relevant constraint is not the prohibition of ERG extraction, but the requirement of resumptive pronoun in ERG relatives. It may be better captured in terms of the Spell-​out requirement on the lower copy rather than the accessibility of ERG arguments. Adopting DP movement approach to relativization (Kayne 1994; Bianchi 2000; and especially Tonoike 2008), I consider relativization to involve movement of a DP, which leaves a copy of D inside the relative clause.13 We may understand the resumption requirement on ERG relatives as a condition on spell-​out that a copy of A-​bar movement with a case feature [ERG] be pronounced.14 The assumption that A-​bar movement leaves a copy of D alone explains why the resumptive pronoun in the relative clause must be a clitic (a D head) rather than free pronoun with a case marker (i.e. a DP). The proposed structure for (8) ‘the woman who (she) ate the fish’ is given in (21). The copy of DP1 that is left as a result of movement to [Spec, C] is in [Spec, T]. This will be spelled out as a third person singular pronoun ne due to its case feature [ERG].15

13  The assumption that DP movement leaves a copy of D alone instead of that of the whole DP is due to Tonoike (2008). See Otsuka 2014 for analyses of relative clauses in various Polynesian languages adopting this approach. 14  Ball (2009) reports that his consultants consistently reject null instantiation of ERG arguments in all contexts, including in the second conjunct of pea. Although my consultants do not share this judgement, this seems to suggest that for Ball’s consultants, the requirement to pronounce items bearing ERG applies not only copies of D (due to A-​bar extraction), but also all instances of D. 15  The exact implementation of this spell out needs a slight stipulation. In (21), where T and V reside in C as a result of V-​to-​T-​to-​C movement, with the copy of D in [Spec, T], the spell-​out of the copy would result in T-​V-​CL (*na‘e kai ne) instead of the grammatical T-​CL-​V (na‘a ne kai). One way to achieve the correct order would be to allow DP1 to directly move from its base position, [Spec, v] to [Spec, C] and to have T’s EPP-​feature be checked by the copy of D (by head adjunction) instead of DP movement to [Spec, T], before V-​to-​T raising.

 

1002   Yuko Otsuka (21)

CP DP1 e fefine C-T-v-V TP na‘e kai

T [3.sg, erg]

vP

DP2 e ika [abs]

v’ v’

VP



In this approach, the ERG–​ABS pattern concerning relativization can be seen as a side effect of morphological ergativity: relativization is a case-​sensitive operation.

40.4.3  Pea Coordination To recap, the ERG–​ABS pattern concerning pea-​coordination is such that the gap and the antecedent must match in their case feature. There is evidence to suggest that this case-​matching requirement is a licensing condition on the gap and not a general restriction on coreference between particular types of arguments. That is, ERG and ABS can be coreferential as long as neither of them is phonetically null. When the second conjunct of pea contains a pronoun, coreference between ERG and ABS is allowed. Compare (11) with a null argument, repeated here as (22a), and (22b) with a pronoun. (22)

a. Na‘e fakamatalili‘i ‘e Sionei ‘a Melej pea hola e*i/​j. pst tease erg John abs Mary and run.away ‘John teased Mary and (*he/​she) ran away.’

[O/​*A = S gap]

b. Na‘e fakamatalili‘i ‘e Sionei ‘a Melej pea nei/​j hola. [O/​A = S pron] pst tease erg John abs Mary and 3.sg run.away ‘John teased Mary and (he/​she) ran away.’

Thus, coreference per se is not subject to any constraint. Rather, an ERG–​ABS pattern emerges only when the second conjunct contains a gap. Otsuka (2010a) proposes that the gap in the second conjunct of pea should be understood as a result of a PF deletion operation, namely, deletion under identity; and that “identity” should be defined in terms of featural identity, that is, features with identical values, including case features. In this analysis, a DP in the second clause can be deleted only if it bears the same case feature as that of the antecedent, resulting superficially in the matching-​case

 

Ergative–absolutive patterns in Tongan    1003 condition on a gap and its antecedent. In other words, pea coordination shows an ERG–​ABS pattern because the relevant operation is sensitive to case (as is the case with relativization—​see 40.4.2). Otsuka’s (2010a) analysis of the gap in the second conjunct of pea provides an intriguing argument for the present hypothesis that clitic pronouns bear ERG and ABS rather than NOM and ACC. Recall the distribution of clitic pronouns shows an NOM–​ACC pattern: they can occur only as S or A, but not as O. Nevertheless, clitic pronouns behave the same as other nominals with respect to coordination reduction: coreference between S and A is not possible if one of them is phonetically null. (23)

a. Na‘a ku nofo pea *(u) fafanga ‘a e tamaiki pst 1.sg stay and    1.sg feed abs spec children ‘I stayed and fed the children.’

[*S=A gap]

b. Na‘a ke poto pea fili (koe) ‘e pst 2.sg smart and choose 2.sg erg ‘You were smart and John chose (you).’

[S=O gap]

Sione. John

c. Na‘a mau fili ‘a Sione pea *(mau) fiefia. [*A=S gap] pst 1.pl.excl choose abs John and    1.pl.excl happy ‘We chose Sione and we were happy.’ The prohibition of coreference between S and A is lifted if the gap is replaced with an overt pronoun, as shown in (23a) and (23c). Again, we witness exactly the same behavior as full NPs. If the PF-​deletion analysis is correct, these examples support our present hypothesis that the NOM–​ACC pattern demonstrated by the clitic pronouns does not entail they bear NOM and ACC (cf. section 40.2). If the case feature values on clitic pronouns are [ERG] and [ABS], but never [NOM], the seemingly contradictory facts illustrated in (23) can be readily accounted for. PF-​deletion under featural identity does not apply if the two pronominal forms bear different Case features despite their identical phonological forms, e.g. the first mau [3.SG, ERG] vs. second mau [3.SG, ABS] in (23c).16

16 

Recall 1.SG clitic pronoun have three allomorphs, ku, u, and ou, whose distribution is governed by the preceding item: TAMs (past na‘a, future te, present ‘oku, perfective kuo, subjunctive ke), complementizers (conditional ka, resultative ‘o), and conjunction pea. Ball (2009: 126) argues that such allomorphic variation suggests that bound pronouns in Tongan are actually inflectional affixes on these functional items. He further argues the gap in the second conjunct of pea cannot be an instance of coordination reduction based on the form of 1.SG pronoun following pea: it is always u regardless of the form of the antecedent, thus showing a mismatch in phonological forms, e.g. ku vs. u in (23a). He considers this mismatch to suggest that TAM is absent in the second conjunct of pea; if it were the case that only TAM is deleted in the second conjunct, the remaining clitic should have the same form as that of the first conjunct. This apparent problem can be resolved if we assume that spell-​out (assignment of phonological forms) takes place after the relevant deletion operation. The deletion rule proposed here is not a post-​spell out operation that applies under identity of phonological features; rather, deletion applies under identity of formal features, after which all relevant features will be spelled out.

 

1004   Yuko Otsuka

40.4.4  ‘o-​Construction The ERG–​ABS pattern found in ‘o-​constructions can be seen as prominence of ABS in that only ABS can serve as an antecedent. Otsuka (2011b) argues that the gap in ‘o-​ clauses is a phonetically null equivalent of what Reinhart and Reuland (1993) call SE anaphors, i.e. anaphors that permit non-​local binding (as opposed to SELF anaphors that must be locally bound). That is, the coreference between the gap and the antecedent in ‘o-​clauses is understood as an instance of non-​local binding of SE anaphor. The structure Otsuka (2011b) proposes for ‘o-​construction is given in (24). It is postulated that ‘o is a complementizer and the CP headed by ‘o is adjoined to VP, taking the obligatory single-​event reading to be evidence to suggest that the two verbs are subsumed under the same VP. vP

(24)

VP

DPERG v

VP V

DPABS ‘o

CP ... SE…

The gist of Otsuka’s (2011b) analysis is as follows. A null SE is base generated in an argument position within the lower vP and agrees with v (if generated as O) or T (if generated as A), thereby receiving a case value (ABS or ERG, respectively). Its phi-​features do not receive any value as a result of this Agree operation, because the phi-​features on a functional category (v or T) are by definition uninterpretable. The relevant functional category ends up in C as a result of V-​to-​T-​to-​C movement. In the matrix vP phase, v’s phi-​features Agree with those on the complement of V, assigning in turn ABS to this DP. Additionally, it is proposed that this v continues to search for a matching feature down the tree.17 The matrix v thereby agrees also with the phi-​features on the relevant functional head that is adjoined to the lower C. The series of Agree operations depicted in (24) can be better captured in terms of Frampton and Gutmann’s (2000) idea that feature valuation is in fact feature sharing. That is, coreference between the DP and SE is explained in terms of several heads being identified by a single set of phi-​features through feature sharing, as illustrated in (25), where uninterpretable phi-​features are represented as uΦ and feature-​sharing relation is indicated by a double-​headed arrow. 17  This ability to continue to probe down the tree after finding a goal is argued to be the privilege of a functional head that bears the unmarked Case feature, i.e. T in accusative languages, v in ergative languages (see section 40.4.1). Note that non-​local binding in accusative languages is known to be subject (i.e. NOM)-​oriented.

 

Ergative–absolutive patterns in Tongan    1005 (25)

[vP1 v1 [VP1 V DPABS] [CP ‘o-T-v2 [TP SE [vP2… …]]]]] uΦ

Φ

uΦ

uΦ

In this analysis, the ERG–​ABS pattern in ‘o-​constructions can be seen as a consequence of the syntactic prominence of ABS, that is, the ability of the functional category with a case feature [ABS] to enter into multiple Agree relations. The inability of ERG to bind the null SE in this construction can be independently explained in terms of its position relative to ABS. Even if T may probe down the tree, the matrix vP, being a phase (Chomsky 2000), would block feature-​sharing between T and SE inside the embedded CP.

40.4.5 Theoretical Implications Let us now turn to the two questions raised at the beginning of this section. First, should any or all of these ERG–​ABS patterns be considered an instance of syntactic ergativity, i.e. syntactic prominence of ABS? Summarizing the preceding discussion, the relevant factor in two of the phenomena (relativization and coordination reduction) seems to be the specific case feature value. The ERG–​ABS pattern in relativization is explained in terms of the spell out rule that requires the copy of A-​bar movement with ERG feature must be pronounced. For coordination reduction, it is argued that the relevant PF deletion is contingent on featural identity, including the value of case feature (ERG or ABS). The latter cannot possibly be regarded as a property of ABS per se. As for ‘o-​constructions, it is proposed that the relevant factor is v’s ability to enter multiple Agree relations, a property associated with a functional category that bears the unmarked case feature value. Incidentally, the structural position of ABS is also relevant, but crucially, its advantage is being located below ERG. All things considered, it is reasonable to conclude that the ERG–​ABS pattern arises in all three phenomena because of the difference between ERG and ABS. Not all of them, however, can be undoubtedly interpreted as manifestation of ABS’s prominence over ERG. This simultaneously answers the second question, namely, whether the observed syntactic prominence of ABS is attributable to a single underlying factor in all of the observed ERG–​ABS patterns. There does not seem to be any single factor that is responsible for all three patterns found in Tongan. It is possible to argue that for relativization and ‘o-​ construction, the syntactic prominence of ABS is related to its status as the unmarked case; A-​bar copies with the unmarked case feature (ABS) are not required to be spelled out and the functional head that assigns the unmarked case (ABS) can enter multiple Agree relations. However, such an approach does not explain why ABS demonstrates syntactic prominence over ERG. It simply states that ABS is syntactically prominence because of being ABS.

 

1006   Yuko Otsuka

40.5  Concluding Remarks In this chapter, various ERG–​ABS patterns in Tongan have been discussed. At the level of morphology, case marking on non-​pronominal NPs shows this pattern: A is marked by a prenominal particle ‘e whereas S and O are marked by ‘a. At the level of syntax, three phenomena exhibit an ERG–​ABS contrast: (a) in relative clauses, the gap strategy is limited to ABS and the relativization of ERG requires resumption; (b) coordinate reduction applies only if the gap and the antecedent are in the same case, be it ABS or ERG; and (c) in ‘o-​constructions, only ABS, but not ERG, can serve as the antecedent of the null SE anaphor. These three phenomena, though they all exhibit an ERG–​ABS pattern of some sort, cannot be explained in terms of a single common factor. Tongan facts therefore suggest that syntactic ergativity should be understood as a construction-​ specific phenomenon rather than a language-​specific property that is reducible to a single parameter (Otsuka 2000; Aissen, Chapter 30, this volume). Many have proposed in the literature that syntactic ergativity arises when given operations are sensitive to case (as opposed to thematic roles) (e.g. Campana 1992; Guilfoyle et al. 1992; Murasugi 1992; Manning 1996). Tongan data not only support this view, but also further demonstrate that “sensitivity” to case can manifest in many different ways, thereby highlighting the difficulty of speaking of ergativity as a uniform phenomenon.

Abbreviations 1, first person; 2, second person; 3, third person; ABS, absolutive; ACC, accusative; CL, clitic; COMP, complementizer; DEF, definite; DIR, directional; DP, determiner phrase; ERG, ergative; EXCL, exclusive; FUT, future; NOM, nominative; NP, noun phrase; PL, plural; PRED, predicate; PRON, pronoun; PRS, present; PST, past; SG, singular; SPEC, specific.

 

Chapter 41

ALIGNM ENT ac ro s s Tsim shia ni c Tyler Peterson

41.1 Introduction Dixon (1994) claims that there is no language that manifests a purely ergative case or agreement system: all ergative languages have a split, which is typically conditioned by semantic factors such as animacy or an argument hierarchy, tense, aspect and modality, or syntactic factors such as matrix vs. subordinate environments and various kinds of extraction and displacement operations. Typically, what lies on the other side of this split is a nominative-​accusative alignment. In this chapter I argue that the Tsimshianic languages present us with a case of a family where the individual languages are entirely morphologically ergative in the agreement and pronominal systems. While there is a split in Tsimshianic, conditioned by both clause type and a person hierarchy, the other side of the split is not the expected nominative-​accusative alignment. Rather, other logical groupings of semantic roles are found. One of the splits groups the Object (O) and Agent (A), or what is called a transitive alignment in the Tsimshianic literature. In another split the Subject (S), A, and O are each marked distinctly, or what is called a contrastive (or sometimes tripartite) alignment. It is notable that the transitive and contrastive alignments are highly unusual in the world’s languages. In a third split, A, S, and O are all grouped together in a neutral alignment. The goal of this chapter is to present a description of the agreement patterns and pronoun distribution across Tsimshianic with the aim of explaining these unusual alignments. This is done by undertaking a comparative analysis of the individual languages that make up the Tsimshianic family. The motivation for this comes from the observation that not all of the Tsimshianic languages realize these unusual alignments. The more conservative Tsimshianic languages, Gitksan and Nisga’a, are purely morphologically ergative: the patterns of agreement morphology and pronouns on both sides of the split (conditioned by clause type) are ergative. A more complex picture is found in the

 

1008   Tyler Peterson Sm’algyax and Sgüüxs branches of Tsimshianic, where transitive, contrastive and neutral alignments are found, in addition to an ergative one. I show that this complexity arises from two factors: first, Sm’algyax and Sgüüxs have three sets of agreement morphemes plus one set of independent pronouns. Second, the splits found in Sm’algyax and Sgüüxs are conditioned not only by clause types, but also a person hierarchy. Because these languages have more options for realizing agreement, we find a greater variety of possible alignments—​including the two non-​standard ones mentioned above. I claim that these are in fact expansions of ergativity, as A and S are never grouped in the agreement morphology. The outcome is that the Sm’algyax and Sgüüxs do not have any special, exotic properties that give rise to these unusual alignments: they fall out from the well-​known effects of the person hierarchy. I extend this analysis to the connective system, a notable feature of Tsimshianic. In the Sm’algyax and Sgüüxs branches of Tsimshianic the connectives are determiner-​like morphemes that appear to be sensitive to the semantic role of the NP they represent—​a feature lacking in the Gitksan and Nisga’a connectives. This leads to a variety of complications, including no less than four distinct alignments (nominative, ergative, neutral, and contrastive) in the Sm’algyax and Sgüüxs connective systems. I show how an understanding of the alignments in the agreement and pronouns across Tsimshianic can shed light on this complexity, and that a comparative analysis eliminates the purported alignments in the connectives, thus revealing a fairly standard (and non-​aligned) paradigm of determiners.

41.1.1 Background and Sources The Tsimshianic languages are spoken on the northwest coast of Canada, almost entirely within the province of British Columbia, adjacent areas of the interior, and the southern tip of the Alaska panhandle. There are four linguistic and sociocultural divisions that make up the Tsimshianic family, given in (1): (1) The Tsimshianic Languages (Rigsby 1986; Mulder 1994; Tarpent 1997) Coast Tsimshianic (CT) Coast Tsimshian (Sm’algyax) Southern Tsimshian (Sgüüxs) Interior Tsimshianic (IT) Nisga’a Gitksan The four languages of the Tsimshianic family are sociopolitically distinct entities. However, from a purely linguistic perspective the individual languages in (1) form a continuum, as speakers of one of these languages report at least some degree of mutual intelligibility with the other languages that loosely correlates with geographical proximity, albeit with some asymmetries. For example, many Sm’algyax speakers can understand

 

ALIGNMENT across Tsimshianic    1009 and speak Gitksan, but Gitksan speakers have slightly more difficulties with Sm’algyax. Nisga’a and Gitksan, which are geographically proximate, are very similar; aside from some lexical differences speakers report few or no difficulties in communicating with each other. These observations are supported by the fact that, aside from some lexical, phonological, and morphosyntactic differences all of the four languages share very similar core grammatical properties. Because one of these properties is the morphosyntax of the agreement and connective systems, for the purposes of this chapter I will use the groupings in (1) and analyze Sm’algyax and Sgüüxs together as the CT languages, and the Nisǥa’a and Gitksan as IT.1 I will also freely alternate language examples within these two groups, while noting any relevant language-​specific differences. One of the underlying strategies of this chapter is to use a comparative analysis of CT and IT to shed light on the grammatical features that make up this dialect continuum. In this chapter I make extensive use of what can be considered the foundational literature in Tsimshianic language studies: Tarpent’s voluminous (1987) grammar of Nisga’a, Rigbsy’s (1986) grammar of Gitksan (complementing this with my own fieldwork), Dunn’s (1979a) grammar of Sm’algyax, and Boas’ (1911) original studies on, and wealth of language data in what was known then as ‘Tsimshian.’ I also draw upon many of the insights in Mulder’s (1994) dissertation on the ergative properties of Sm’algyax, even though I make several revisions to her analyses.

41.1.2 Outline Section 41.2 is an outline of the basic morphosyntactic features of the Tsimshianic languages, with an emphasis on the two main clause types in the languages. A comprehensive description and comparative analysis of the Tsimshianic pronominal and agreement system is undertaken in section 41.3. I examine first the IT languages, as they represent a conservative side of Tsimshianic in their manifestation of ergativity. I then turn to the CT languages, which are more complex. Using what we ascertain about IT, I attempt to unravel some of the more puzzling aspects of CT agreement and pronouns and show how the unusual alignments in CT are actually quite regular when viewed at the level of the language family. Section 41.4 examines another area of Tsimshianic grammar, the connectives. Here, too, we find many unusual alignments of semantic and grammatical roles. Again, I show that a comparative analysis coupled with an understanding of IT agreement and connectives can be used to explain these puzzling patterns. Section 41.5 concludes.

1  Within this continuum there are several dialects and varieties, usually corresponding to the different villages and settlements the IT and CT languages are spoken in, and some dialects, such as Kitselas, share properties of both IT and CT. See Mulder 1994 and Brown 2010 for details.

 

1010   Tyler Peterson

41.2  Basic Morphosyntactic Features of Tsimshianic There are two clause types in Tsimshianic. In the IT literature these are called the Independent and Dependent clauses (Rigsby 1986). Generally speaking, Independent clauses have a base VSO word order: the verb stem is the first constituent in the clause followed by the grammatical subject, object and optional indirect object, as in (2a). Dependent clauses are called as such when another constituent, often called a dependent marker (dep.mark), is at the left periphery of the clause, which is then followed then by the verb stem, subject and object, as in (2b). (2) IT (Gitksan) a. Independent clause V S O IO gi’nam-​i-​(t)=hl hanaḵ=hl bilaa ’as na’aa-​t give-​tr-​3sg=con woman=con abalone prep mother-​3sg ‘The woman gave the abalone to her mother.’ b.

Dependent clause dep.mark V S O IO yukw=t gi’nam-​(t)=hl hanaḵ=hl bilaa ’as na’aa-​t imperf-​3sg give-​3sg=con woman=con abalone prep mother-​3sg ‘The woman is giving the abalone to her mother.’

Independent clauses generally tend to have the simplest verbal morphology, but are relatively less common in everyday speech. Dependent clauses are usually found in paratactic and syntactically embedded environments, such as a coordinate and subordinate clauses, or in narratives. Dependent markers do not appear to form a homogeneous semantic or lexical class, as they include negation, circumstantial modals, spatio-​temporal and aspectual particles (such as the imperfective yukw in (2b)), sentential conjunction, conditionals, and other narrative markers. Additionally, any kind of leftward displacement of a postverbal constituent triggers a Dependent clause, such as the relativization, clefting, or focusing a subject, object or indirect object. For example, lexical negation in Tsimshianic is preverbal, and thus triggers a Dependent clause, as in (3b). However, not every preverbal element is a dependent marker; for example, the preverbal future particle tim in (3a) does not trigger a Dependent clause: (3) IT (Gitksan) a. Independent clause V S O tim gi’nam-​i-​(t)=hl hanaḵ=hl bilaa ’as fut give-​tr-​3sg=con woman=con abalone prep ‘The woman will give the abalone to her mother.’

IO na’aa-​t mother-​3sg

 

ALIGNMENT across Tsimshianic    1011 b. Dependent clause dep.mark V S O needii=t gi’nam-​(t)=hl hanaḵ=hl bilaa ’as neg-​3sg give-​3sg=con woman=con abalone prep ‘The woman didn’t give the abalone to her mother.’

IO na’aa-​t mother-​3sg

In the CT literature, Independent and Dependent clauses are called the indicative and the subjunctive, respectively (Boas 1911; Dunn 1979a). For the ease of comparison (and given the fact that the syntax at this level is nearly identical across Tsimshianic), I will use the Rigsby Independent/​Dependent labels in CT. (4)

CT (Sm’algyax) a. Independent clause (Indicative) V S O IO nah k’yilum hana’a bilhaa das noo-​t past give woman abalone prep mother-​3sg ‘The woman gave the abalone to her mother.’        (Mulder 1994: 42) b. Dependent clause (Subjunctive) dep.mark V S O IO yagwa=t k’yilum=da hana’a bilhaa das noo-​t imperf=3sg give=con woman abalone prep mother-​3sg ‘The woman is giving the abalone to her mother.’      (Mulder 1994: 42)

Independent and Dependent clauses are also distinguished verbal agreement and pronoun patterns. In IT Independent clauses such as (2a) the agreement suffix -​t represents the grammatical subject, the woman. This suffixal agreement on the verb stem appears to be lacking in CT in the corresponding example in (4a). I address this difference between IT and CT below. In Dependent clauses such as (2b) and (4b) the verb is preceded by a dependent marker, in this case the imperfective marker (IT yukw; CT yagwa). In Dependent clauses the agreement patterns are more complex, as two agreement morphemes appear: in both IT and CT the 3rd person agreement -​t attaches to the dependent marker, which shifts to representing the grammatical subject (the woman). In IT the agreement suffix -​t on the verb assumes the role of representing the grammatical object (the abalone). CT differs slightly in that there is no agreement with the grammatical object; instead, a different enclitic connective =da appears. I also examine this difference in more detail below. The form and function of the connectives is also conditioned to some extent by the clause types, especially in CT (glossed as con).2 I will hold off on a description of these until section 41.4, which will be greatly aided by developing first a clearer picture of the 2 

In the CT literature the connectives are described as enclitics but notated with a dash: -​da (Dunn 1979a; Mulder 1994). For ease of comparison I use the standard ‘=’ notation for enclitics: =da.

 

1012   Tyler Peterson alternations in agreement morphology between Independent and Dependent clauses. This is the next task.

41.3  Tsimshianic Pronouns and Agreement 41.3.1 Interior Tsimshianic Independent intransitive clauses are the simplest constructions in IT, consisting of a verb stem and an enclitic determiner (the connective; see section 41.4) which marks the argument status of its single DP argument and its status as either a proper noun or common noun. Example (5) shows this in both Gitksan and Nisga’a (Tarpent 1987: 475): (5) limx=t sing=pn.det ‘Mary sings.’

Mary Mary

A pronominal intransitive subject (S) is represented by what is known in the IT literature as a Series III person marker, given in Table 41.1.3 Syntactically, Series III are stand-​alone independent pronouns, and occupy the same position as a nominal argument. However, an important difference between a full DP and the Series III independent pronouns is that the latter do not have an enclitic determiner when they immediately follow the verb. Example (6) is an example of this in both Gitksan and Nisga’a:4 Table 41.1 IT Series III person markers 1 2 3

sg

pl

’nii’y ’niin ’nit

’nuum ’nisi’m ’nidiit

Sources: Rigsby (1986); Tarpent (1987)

3  Throughout this chapter I use the following convention for glossing pronouns and agreement: the person and number followed by a dot and then the Series number. For example, a gloss for Series III 1st person singular is 1sg.III, 2nd person plural is 2pl.III, etc. 4  Tarpent (1987) and Forbes (2013) report that the Series III pronouns can, under certain conditions,

be preceded by a proper noun determiner, such as a coordinate structure: ’nii’y gan=t ’niin (1sg.III coord=pn.det 2sg.III) ‘Me and you.’

 

ALIGNMENT across Tsimshianic    1013 (6) a.

limx ’nii’y sing   1sg.III ‘I sing/​sang.’

b.

*limx=t ’nii’y

As described above, Dependent clauses are triggered by the insertion of a Dependent marker while maintaining the base VSO order. In Dependent clauses the Series II person suffixes, given in Table 41.2, emerge. Table 41.2 IT Series II suffixes 1 2 3

sg

pl

-​’y -​n -​t

-​m -​si’m -​diit

Sources: Rigsby (1986); Tarpent (1987)

The 3rd person Series II suffix co-​occurs with a DP to function as agreement. Example (7) is introduced with the Dependent marker yukw (imperf), which triggers Dependent morphology: the (often optional) 3rd person -​t functions as agreement on the verb with the subject John:5 (7)

yukw=hl prog=cn.det ‘John is reading.’

litsxxw(-​ti)=s read-​3=pn.det

Johni John

In intransitive Dependent clauses a pronominal subject is represented by a Series II suffix on the verb and not a Series III pronoun, which cannot co-​occur with or replace a Series II suffix. Based on the fact that the Series II functions as 3rd person agreement with a full DP, as in (7), following Hunt (1993) I claim that pro occupies this argument position and agrees with the Series II suffix: (8)

a. yukw=hl imperf=cn.det ‘I’m reading.’

litsxxw-​’yi  proi read-​1sg.II

b.

*yukw=hl litsxxw-​ti ’nii’yi

c.

*yukw=hl litsxxw ’nii’y

5  The 3rd person agreement suffix -​t often disappears in normal speech. Some transcriptions do not include it all, and there is speculation that it has disappeared altogether in contemporary speech. I use brackets to indicate this variability. See Tarpent (1987: 207) for a detailed discussion of the optional 3rd person agreement in Nisga’a, which also applies to Gitksan (Hunt 1993: 115–118).

 

1014   Tyler Peterson As such, DPs and Series III pronouns have the same distribution in Independent intransitive clauses. However, Series II suffixes can co-​occur as agreement with a DP in a Dependent clause but not with a Series III pronoun. These patterns are summarized in Table 41.3. Table 41.3 The distribution of Series III and II in intransitives clauses Independent Dependent

Dep Dep

V

S

V V

DP III

V-​IIi V-​IIi

DPi proi/​*IIIi

Turning to transitive sentences, in an Independent clause a subject agrees with the Series II suffixes, while the object lacks agreement: (9) Nisǥa’a (Tarpent 1987: 235) hlimoo-​i(-​ti)=s Maryi=t help-​tr-​3=pn.det Mary=pn.det ‘Mary helped Peter.’

Peter Peter

Analyzing Series II as subject agreement (and not object agreement) in Independent clauses is supported by the fact that a pronoun object is represented by a Series III pronoun as in (10), and a pronominal subject is represented by a Series II suffix, as in (11): (10) Gitksan hlimoo-​i-​ti=s Maryi ’nii’y help-​tr-​3=pn.det Mary=pn.det 1sg.III ‘Mary helped me.’                  (Tarpent 1987: 234) (11)

a. hlimoo-​i-​’y help-​tr-​1sg.II ‘I helped Mary.’ b. hlimoo-​i-​’y help-​tr-​1sg.II ‘I helped you.’

Mary Mary ’niin 2sg.III

Given that a full DP subject triggers Series II agreement, as (10) shows, and the Series III independent pronoun can occur in the object argument position as full DPs do, as (11)

 

ALIGNMENT across Tsimshianic    1015 shows, we might expect that the Series III pronoun can occur in the subject argument position and trigger Series II subject agreement. This turns out not to be the case. In these constructions a pronominal subject can only be represented by a Series II morpheme, and a Series III independent pronoun cannot occur in this position, as (12a) and (12b) show. Following Hunt (1993) I extend the pro analysis above to these cases, and claim that null pro occupies this position, which agrees with the Series II, as in (12c): (12)

a. *hlimoo-​i-​’yi help-​tr-​1sg.II ‘I helped you.’ b. *hlimoo-​i-​ti help-​tr-​3sg.II ‘I helped you.’ c. hlimoo-​i-​’yi help-​tr-​1sg.II ‘I helped you.’

’nii’yi 1sg.III

’niin 2sg.III

’nii’yi 1sg.III proi

’niin 2sg.III

’niin 2sg.III

These patterns exemplify standard pro-​drop behavior:  the person and number features of the Series II agreement are rich enough to compel the pro-​drop observed in (12c).6 As with Dependent intransitives, Dependent transitives are also introduced by a Dependent marker (again using yukw prog for ease of exposition). However, in Dependent transitives, the function of Series II agreement now shifts to representing the object, while another phi-​indexing paradigm, the Series I morphemes given Table 41.4, assume the role of representing the subject. The Series I morphemes surface preverbally and typically (but not always) encliticize to the Dependent marker, characterizing a typical second position clitic. What this Table 41.4 IT Series I enclitics 1 2 3

sg

pl

=ni =mi =t

tip (mi)...sim =t

Sources: Rigsby (1986); Tarpent (1987)

6  The one possible difference with a typical pro-​drop language such as Italian, where a pronoun can be optionally used for emphasis, is that pro-​drop is obligatory in IT (see Hunt 1993 for a detailed analysis of pro in Gitksan). However, Tarpent (1987) reports cases in Nisga’a where a Series III pronoun can replace pro while triggering 3rd person Series II agreement.

 

1016   Tyler Peterson means is that in there is both subject and object agreement in Dependent transitive clauses, as shown in example (13): (13) yukw=ti tsap-​tj=s prog=3.I cook-​3sg.II=pn.det ‘Sheila’s doing up the fish.’

Sheilai=hl Sheila=cn.det

honj fish

This example is not particularly revealing, as the 3rd person singular in both Series I and II are the same morpheme =t/​-​t. However, replacing a singular object for the plural pronoun object in (14) reveals that the Series II on the verb stem does indeed represent the object: (14) yukw=ti stil-​diitj=s prog=3.I accompany-​3pl.II=pn.det ‘Sheila’s going with them.’

Sheilai   proj Sheila

The sentences in (15) using pronouns in both subject and object positions clearly exemplify the Series I and II paradigms: (15)

a. yukw=ni hlimoo-​n prog=1sg.I help-​2sg.II ‘I’m helping you.’ b. yukw=mi hlimoo-​’y prog=2sg.I help-​1sg.II ‘You’re helping me.’

The transitive clause types and the Series agreement patterns are summarized in Table 41.5. Table 41.5 Template of distribution of Series I, II, and III in transitive clauses

Independent Dependent

Dep=Ii

V

A

O

V-​IIi

DPi/​proi/​*IIIi

DP/​*pro/​III

V-​IIj

DPi/​proi/​*IIIi

DPj/​proj/​*IIIj

By integrating these observations with the pronoun and Series II agreement patterns in intransitive clauses, the following generalizations emerge: Series I only ever mark an A argument, characterizing an ergative distribution; Series III can only ever mark a S or O argument, therefore absolutive in distribution. Series II is pivotal in the sense that it

 

ALIGNMENT across Tsimshianic    1017 Table 41.6 Split ergativity in IT

Series I Series II Series III

Independent

Dependent

—​ A S/​O

A S/​O —​

can represent an A, a S or an O, depending on the clause type: it represents an ergative argument in Independent clauses while absolutive in Dependent clauses. The overall distribution of Series I, II, and III is summarized in Table 41.6. A cursory evaluation of the mapping of the three phi-​indexing Series to semantic roles, as conditioned by clause type, confirms that Rigsby in his initial study was prudent in not labelling the different Series as ‘ergative’ or ‘absolutive,’ instead giving them the descriptively neutral labels I, II, and III. This table also leads to the main generalization that all sentences in Gitksan and Nisga’a are ergative, with the Series II markers serving as a kind of lexical pivot across the Independent/​Dependent clause types (with the Series II pivot itself showing an ergative alignment across clause types). The IT Series III independent pronouns could be analyzed as being morphologically complex, consisting of the Series II suffixes suffixing to the ‘base’ ’ni(i)-​, along the lines of (16) (see Livingston 1989 and Peterson 2006b for such an approach): (16)

a. gya’a-​n=s see-​2sg.II=pn.det ‘You saw Alvin.’

Alvin Alvin

b. t’ugwantxw ’ni-​n fall.down      ’ni-​2sg.II ‘You fell down.’ From a purely morphological standpoint this analysis seems plausible, as all of the Series II morphemes could combine with ’ni(i)-​ to form a Series III pronoun (cf. Table 41.1). One of the consequences of this analysis is that there are only two sets of phi-​ indexing morphemes plus the now-​derived independent pronouns: I, II, and ’ni(i)-​II. From here, an independent analysis of the clausal morphosyntax could explain why a derived independent pronoun as in (16b) appears in argument position, as the full DP does in (16a). Déchaine and Wiltschko (2002) claim that this is the right analysis for a similar set of facts in Halkomelem and Shuswap (Salish) pronouns. Under this line of reasoning ’ni(i)-​could be a kind of pronominal determiner, thus allowing the Series II morphemes to occur in argument position as a DP. This would have the added advantage of explaining the lack of a lexical determiner on the non-​derived Series III pronouns (cf. (6)).

 

1018   Tyler Peterson Analyzing the Series III pronouns as morphologically complex could also reveal what looks like a nominative alignment in the Series II agreement, as both the A in (16a) and the S in (16b) are represented by Series II in the same clause type (i.e. Independent), thus yielding a more familiar ergative/​nominative split conditioned by clause type (Dependent and Independent). However, replacing the lexical DP with a derived Series III pronominal object would neutralize the nominative alignment of the Series II suffixes, as A, S, and O would all now be represented with the Series II suffixes. Using Independent clauses as an example, the 2sg.II suffix -​n represents S in (16b) and A and O in (17): (17) tim hlimoo-​i-​’y ’nii-​n fut help-​tr-​1sg.SII det-​2sg.SII ‘I’ll help you.’ Appealing to an analysis of ’ni(i)-​ as a determiner (or even a morphological case marker of some kind) does not rescue this analysis; in fact, the data shows ’ni(i)-​could only be an absolutive case marker, as it marks a Series II S in (16b) and O in (17). This would only serve to confirm an ergative/​absolutive orientation.

41.3.2 Coast Tsimshianic 41.3.2.1 Agreement and Pronouns in CT Aside from some phonological and morphosyntactic differences, CT agreement and pronouns are cognates with those of IT. The close similarity of forms can be most easily observed in the Series I  and II paradigms (Dunn 1979a; Mulder 1994), given in Tables 41.7 and 41.8. Table 41.7 CT Series I (cf. Table 41.4)

1 2 3

sg

pl

n m t

dp m-​...-​sm t

Table 41.8 CT Series II (cf. Table 41.2)

1 2 3

sg

pl

-​u /​ -​i -​n -​t

-​m -​sm -​dit

Given these similarities, IT and CT diverge in the form and function of the Series III. In IT the Series III are pronouns and independent words; however, the CT cognates are bound suffixes on the verb stem. I label this in CT as Series IIIa in Table 41.9, using the ‘a’ in IIIa to reflect this difference with IT. Another point of divergence between IT and CT

 

ALIGNMENT across Tsimshianic    1019 is found in an additional pronominal paradigm found only in CT. I call these Series IIIb, given in Table 41.10.7 Table 41.9 CT Series IIIa (cf. Table 41.1)

1 2 3

sg

pl

-​’nu -​n -​t

-​’nm -​’nsm -​t

Table 41.10 CT Series IIIb

1 2 3

sg

pl

’nüüyu ’nüün ’niit

’nüüm ’nüüsm ’niit

What is striking is how the CT Series IIIb pronouns also resemble the IT Series III pronouns (hence the ‘b’ label): both are independent words, and both appear to be composed of a ‘base’ proto-prefix (IT ’nii-​, CT ’nüü-​) plus their respective Series II morphemes. In the absence of the relevant primary data in CT, I will remain neutral toward this analysis in CT, but I will examine the syntactic distribution and semantic roles of the Series IIIb in more detail in what follows.

41.3.2.2 Dependent Clauses in CT It is convenient to begin by looking first at Dependent clauses in CT, as these are very similar to those in IT with respect to the agreement and pronouns. Intransitive Dependent clauses in CT trigger Series II agreement with S exactly as in IT. Example (18) is a Dependent clause triggered by the imperfective morpheme yagwa (IT yukw), and the 3rd person plural Series II suffix -​dit agrees with the plural S: (18)

yagwa bus-​laxs-​diti [dp Allen di=s Tom]i imperf chop-​wood-​3pl.II assoc Allen and=pn.det Tom ‘Allen and Tom are wood-​ chopping.’              (Mulder 1994: 51)

7  A point about terminology must be made here, as I depart from the glosses used the CT literature (i.e. Dunn 1979a; Mulder 1994). In the CT literature the phi-​indexing paradigms are called dependent pronouns, which I relabel in the following way: the Series I in this chapter are called the subjective in the CT literature, the Series II are called the objective and the Series III are called the definite objective (labels that originated with Boas 1911 and extended by Dunn 1979a). In this chapter and in other work (Peterson 2006a) I advocate the adoption of Rigsby’s IT Series numbering for CT. There are a number of reasons for doing this. First, the labels subjective, objective and definite objective are ambiguous and imply an analysis as, for example, the subjective as having something to do with the grammatical role of subjects, and the objective with objects, and the definite objective with definiteness (and objects). These analyses have never been clearly supported in the CT literature. Additionally, we saw in the previous section how the phi-​indexing paradigms can shift in their orientation in representing A, S, and O, thus making the traditional CT labels potentially confusing. On a more practical note, the differing labels between CT and IT obscure the observation that the paradigms between CT and IT are nearly identical in form. As such, the Rigsby labels I adapt to CT aid in the direct comparison of IT and CT.

 

1020   Tyler Peterson In (19) the dependent markers ɫa dm trigger a Dependent clause that has a pronominal S indexed by Series II agreement with pro in the argument position: (19) ɫa dm baa-​yui proi about fut run-​1sg.II ‘I am going to run soon.’

(Mulder 1994: 85)

Also as in IT, in Dependent transitive clauses Series I agrees with the A, and Series II with the O, whether as agreement with full DPs or as pronouns via co-​indexation with pro, as examples (20)-​(23) show: (20) yagwa-​ti t’uus-​tj=it prog=3.I push=3.II-​pn.det ‘John is pushing Mary.’

Dzoni=at John=pn.det

(21) yagwa-​ni dzab-​(tj)=a prog-​1sg.I make-​3sg.II-​cn.det ‘I am making a basket.’

proi

(22) ada wil-​ti way-​uj haasi-​it and then-​3.I find-​1sg.II dog-​dem ‘And then (this) dog found me.’ (23) yagwa dpi babuud-​nj prog 1pl.I wait-​2sg.II ‘We are waiting for you.’

Melij Mary

duuɫkj basket

(Dunn 1979a: 67)

(Mulder 1994: 86)

proj (Boas 1911: 384)

proi proj (Mulder 1994: 79)

These data confirm a generalization across Tsimshianic: in Dependent clauses Series I  always represents A, Series II represents S or O, and the Series IIIa/​b pronouns do not occur. At this point it is useful to look at the Series IIIb independent pronouns before turning to the Independent clauses. Mulder (1994) notes that the Series IIIb independent pronouns are not common in normal speech; rather, they are used for ‘emphatic contrast’ (p. 64). Focus is a typical grammatical strategy used for this purpose, whereby a DP is displaced to the front of a sentence. As mentioned in the previous section, in the Tsimshianic languages displacement of any constituent to a preverbal position triggers Dependent clause morphology, along with the expected Series I agreement with A and Series II agreement with S or O. When a DP is focused, as in (24a), Series II agreement with the focused subject occurs. It is in this preverbal focus position that we find the Series IIIb independent pronouns, as in (24b) —​along with 3rd person Series II agreement on the verb:

 

ALIGNMENT across Tsimshianic    1021 (24) a. awtai   uks-​ haytg-​iti gi-​sga lax porcupine  toward.water-​stand-​3sg.II dem-​cn.prep loc maɫiitg-​ m   kyoox green-​attrib  grass ‘It was the porcupine who stood at the edge on the green grass.’ (Mulder 1994: 135) b. ’nüüyui nah algyag-​ati-​a Sm’algyax 1sg.IIIb past speak-​3sg.II-​prep Sm’algyax ‘I was the one who was speaking Sm’algyax.’         (Mulder 1994: 65) This suggests that Series IIIb pronouns in CT are DPs, and that Series I and II 3rd person agreement occurs with a DP, a Series IIIb pronoun, or pro in argument position. In CT (and IT) A, S, or O can be focused (see Davis and Brown 2011 for a detailed discussion of the extraction facts in Gitksan). If the Series IIIb pronouns can occur in the same position as a full DP and trigger agreement the same way, we predict that they can be focused regardless of what semantic role they represent. This turns out to be the case; (24b) is a focused S, and in (25) the A is focused using a Series IIIb pronoun: (25)

’nüüyui dm-​ti in baa-​(tj)-​’n bootj 1sg.IIIb fut-​3.I top run-​3sg.II-​trans boat ‘I’m the one who will run the boat.’

(Mulder 1994: 70)

With the focused pronoun object in (26), the Series IIIb pronoun does not occur sentence initially, but it is still in the preverbal field, along with the predicted Series I agreement enclitics on the dependent discourse marker ada: (26) ada-​ti ’niidj=a nah niidz-​(tj)=a and-​3.I 1sg.IIIb=cn.det past see-​3sg.II=cn.det ‘And I’m the one the bear saw.’

oli bear (Dunn 1979a: 343)

I return to the distribution of the Series IIIb pronouns in the next section, where additional evidence is provided in support of analyzing the Series IIIb as DPs.

41.3.2.3 Independent Clauses and the Person Hierarchy in CT At first blush, CT Independent clauses look very similar to IT Independent clauses: as with IT, Series IIIa in CT represent intransitive pronominal S in Independent clauses, either with a typical intransitive verb (27) or a predicative adjective (28). The main difference is morphological: the CT Series IIIb are suffixes on the verb: (27)

t’aa-​’nu da sit-​1sg.IIIb prep ‘I sit at the table.’

awaa dem

ha’lihaɫels table

(Mulder 1994: 79)

 

1022   Tyler Peterson (28) ’wiileeks-​a’nu ada amap’aas-​a’nu big-​1sg.IIIb and beautiful-​1sg.IIIb ‘I am big and beautiful.’                  (Mulder 1994: 57) In the transitive clauses in (29)–​(30) Series II agrees with a DP Agent or pro, and there is no agreement with O, exactly as in IT: (29) dm ts’laayk-​ti=a hana’ai amt’aa da gwii fut visit-​3sg.II-​cn.det woman loon prep dem ‘The woman will visit the loon there.’ (30) ǥaliimks-​idi=a proi throw-​3sg.II-​cn.det ‘He threw dirt in her face.’

yuup dirt

da prep

ts’al-​t face-​3.poss

But this is where the similarity with IT ends. Unlike the IT languages, which are entirely ergative in both clause types, the CT languages have a number of splits in Independent clauses conditioned by a person hierarchy (Silverstein 1976: 113; see also Müller and Thomas, Chapter 12, in this volume): (31) Person hierarchy in CT 1 ≥ 2 > 3 > DP The basic effect of a typical person hierarchy such as (31) is that the further down the person is on the hierarchy, the more likely that argument will be marked with either ergative/​absolutive case or agreement. Conversely, the higher up on the hierarchy the more like the person is marked with nominative/​accusative agreement. While argument and person hierarchies are a central conditioning factor cross-​linguistically in split ergativity, Mulder (1994) suggests that the resulting splits in CT are less common: in Independent transitive clauses with pronouns as arguments, the application of the hierarchy in (31) to the CT Series morphemes yields additional splits that are neither ergative-​absolutive nor nominative-​accusative. These are summarized in Table 41.11.

Table 41.11 Split ergativity in CT in sentences with pronoun arguments Independent Series I Series II Series IIIa

Dependent

O>A

A1 > O2

A1/​2 > O3

A —​ S/​O

A O S

—​ A/​O S

A S/​O —​

 

ALIGNMENT across Tsimshianic    1023 In comparing the CT groupings in Independent clauses with the IT we find the lack of Series II representing A and Series III representing S/​O, as found in IT (cf. Table 41.6). Nonetheless, the O > A  grouping is an ergative one. What Mulder (1994: 56) calls the contrastive alignment emerges from the A1 > O2 grouping, when a 1st person subject acts on a 2nd person object. The A1/​2 > O3 grouping is transitive alignment, as the person hierarchy forces a higher ranking first or 2nd person subject, and lower ranking object to both be represented with Series II. When O outranks A on the person hierarchy A is represented with Series I and the O with Series IIIa. This is another point of divergence from IT: the Series I morphemes in both CT and IT are preverbal, but this creates what superficially looks like a Dependent clause in CT, only without a Dependent marker. The 1st person O outranks the 2nd person A in (32a) and the 3rd person A in (32b): (32)

a. m waay-​i’nu 2sg.I find-​1sg.IIIa ‘You found me.’ b. t ap’aǥ-​a’nu 3.I remember-​1sg.IIIa ‘He remembered me.’

The next question is whether CT Series IIIa are arguments or agreement morphemes. In (33) the pronoun in object position and the O > A hierarchy forces the pronoun object to be realized as Series IIIa: (33)

ti waay-​i’nu haasi-​it 3.I find-​1sg.IIIa dog-​dem ‘The dog found me.’                        (Boas 1911: 385)

Recall that in IT the Series III are independent pronouns that occupy an argument position. However, the CT facts are different in two respects: first, unlike IT, the CT Series IIIa are verbal suffixes occupying the same position as the Series II agreement morphology.8 Second, a problematic result arises if the CT Series IIIa are treated as pronouns occupying an argument position, along the lines of IT Series III. In the Gitksan (IT) sentence in (34a) the pronoun object is realized as a Series III independent pronoun ’nii’y. If the CT Series IIIa suffixes were indeed pronouns the order of arguments would be reversed, as (34b) shows: (34) a. Gitksan gya’a-​ti=hl see-​3sg.II=cn.det ‘The bear saw me.’ 8 

[smaxi]A bear

[’nii’y]O 1sg.III

Although the data is not conclusive, the Series IIIb appear to occupy argument positions in the same way that the Series III pronouns do in IT. This suggests that the CT Series IIIa are agreement morphemes and not pronouns, and that the CT Series IIIb instead fulfill this function.

 

1024   Tyler Peterson b. Sm’algyax na-​ti ’niidz[-​a’nu]O past-​3.I see-​1sg.IIIa ‘The bear saw me.’ ≠ ‘I saw the bear.’

[oli]A bear

(Dunn 1979a: 63)

Treating the CT Series IIIa morphemes as agreement—​and not pronouns as their IT Series III counterparts—​makes them amenable to a pro analysis: pro occupies the object (or subject) argument position which triggers Series IIIa agreement on the verb. (35) a. mi ap’aǥ-​a’nuj [proi]A [proj]O b. na-​ti ’niidz-​a’nuj [oli]A [proj]O

(cf. (32b)) (cf. (34b))

An examination of the distribution of Series I and IIIa in examples (32a–​33) yields an ergative orientation (AI:S/​OIII) in Independent clauses. At this point, the only notable difference between IT and CT is that the O > A hierarchy forces the Series I agreement with the lower ranked agent (where this is Series II in IT without the argument hierarchy). The other difference is largely morphosyntactic in nature: unlike IT, in CT the Series IIIa are bona fide agreement. The divergence between IT is CT is more noticeable when A outranks O. For example, in (36) the 1st person A outranks the 2nd person singular O in (36) and 2nd person plural O in (37). In cases such as these A is represented by Series I and O by Series II, again resulting in what superficially looks like a Dependent clause (the preverbal future particle tim/​dm does not trigger a Dependent clause in any of the Tsimshianic languages): (36) n-​dm man-​gad-​n 1sg.I-​fut up.through-​take-​2sg.II ‘I will take you up.’                  (Mulder 1994: 51) (37) n-​dm ɫümoon-​sm 1sg.I-​fut help-​2pl.II ‘I’ll help you (pl).’          (Mulder 1994: 57; compare with IT in (17)) Examples (36) and (37) taken together with the intransitive pronominal S in (38), which are always Series IIIa in Independent clauses, reveals the contrastive orientation: each semantic role is represented by a different Series (AI:SII:OIIIa): (38) dm al tgyi ksgooga-​’nu fut emph down be.first-​1sg.IIIa ‘I’ll go down first.’                  (Mulder 1994: 51)

 

ALIGNMENT across Tsimshianic    1025 When a 1st or 2nd person subject A outranks a 3rd person object O (A1/​2 > O3), as in (39a) and (39b), both represented by Series II. The result is a sentence that has two Series II morphemes, both suffixed to the verb stem: (39)

a. ap’ax-​d-​u remember-​3sg.II-​1sg.II ‘I remember him.’                  (Dunn 1979a: 225) b. ‘dm waal-​u-​t,’ daya=ga awta fut do-​1sg.II-​3sg.II say=cn.det porcupine ‘I will do it,’ said the porcupine.           (Mulder 1994: 58)

Because both A and O are marked with Series II, this yields Mulder’s transitive orientation (A/​OII:SIIIa). However, given the person hierarchy there should never be any ambiguity in which argument the stacked Series II morphemes represent: the 3rd person agreement will always be present and always represents the pronoun object, regardless of the whether the subject agent is 1st or 2nd person. This explains the variation in the order of agreement between (39a) and (39b). In constructions like these we find the Series IIIb independent pronouns again; example (40) shows how a Series IIIb pronoun co-​refers with Series II on the verb: (40) niidz-​u-​ti ’nüüni see-​1sg.II-​3sg.II 2sg.IIIb ‘I see you.’          (Mulder 1994: 71 [Boas 1911: 386]; Lit. ‘I see it, you.’) Mulder (1994) notes that in constructions such as (40) the Series II is not agreement, but rather the expression of a pronoun. This suggests that Series IIIb in these cases is a kind of right dislocation (as the literal translation suggests). However, another example shows Series II co-​referring with a full DP: (41)

nah ɫa ’niidz-​di-​u ’yuutai dim-​ti in baa-​n past just see-​3sg.II-​1sg.II man fut-​3.I rel run-​caus ‘I just saw the man who will run the boat.’           (Mulder 1994: 143)

The Series II object co-​refers with the head of the relative clause ’yuuta (man). If the right dislocation analysis is correct the literal translation of (41) is predicted to be ‘I just saw him, the man who will run the boat.’ The relevant primary data is lacking that would show whether this involves syntactic agreement or simply semantic co-​reference. Either way, it provides evidence for a DP analysis of Series IIIb pronouns in CT. The distribution of Series I, II, and IIIb in transitive CT sentences is summarized in the templates given in Table 41.12 (compare with Table 41.5 in IT).

 

1026   Tyler Peterson Table 41.12 Template of distribution of Series I, II, and IIIb in transitive clauses V

A

O

Ii Ii

V-​IIi V-​IIi-​IIj V-​IIj V-​IIIbj

DPi/​proi proi proi proi/​DPi

DP/​*pro proj/​DPj(/​IIIbj) proj proj

Dep=Ii

V-​IIj

DPi/​proi

DPj/​proj

Independent A1/​2 > O3 A1 > O2 O>A Dependent

It is important to emphasize that the CT languages do not have a rare or exotic property, given that these alternative alignments are not commonly found in the world’s languages. Malchukov (2010) reports a similar situation in a number of Iranian languages, where it is called the double-​oblique pattern. In Vafsi, both the A and O are marked by the oblique case in the past tense (Malchukov 2010: 184; see also Haig, Chapter 20, in this volume): (42) luas-​i kærg-​e=s fox-​obl.sg chicken-​obl.sg=3sg ‘The fox carried off (the) chicken.

bææ-​værdæ pfv-​take.pst

Malchukov claims that this pattern can be made sense of from a diachronic perspective. The double-​oblique case construction in (42) is the end point of grammaticalization path that has its origins in the multifunctional dative-​genitive case in these languages. While applying a diachronic analysis to the CT facts would likely shed light on the emergence of the transitive and contrastive alignments, it is beyond the scope of this chapter. Although Mulder’s analysis is essentially accurate, I suggest there is a deeper generalization that captures these groupings. The alignments in Table 41.12 can be described as expansions of ergativity. They are essentially ergative: in each of the possible alignments A and S are never grouped in the same clause type, meaning both clause types are ergative; and they are ‘expansions’ as the other possible semantic to grammatical role mappings are realized using the three Series morphemes, while still preserving an ergative alignment. Under this view, the labels ‘transitive’ and ‘contrastive’ just characterize subtypes of ergativity. As such, we can sustain the claim that the Tsimshianic languages are purely morphologically ergative.9 Equipped with an analysis of the agreement and pronoun morphology and how they are aligned in IT and CT we are now in an ideal position to examine the other area of Tsimshianic that apparently shows sensitivities to the mapping between semantic and grammatical roles: the connectives. 9 

See also Khan (Chapter 36, this volume) for a discussion of the diversity of types and degrees of ergativity across the Neo-​Aramaic dialects.

 

ALIGNMENT across Tsimshianic    1027

41.4  Tsimshianic Connectives A prominent feature of Tsimshianic is what are called the connectives in the Tsimshianic literature. Connectives have determiner-​like properties in that they precede an NP and encode a specific kind of semantic information about that NP. In IT the common noun connective =hl (cn.det) in (43a) encodes the status of horse as a common noun. In (43b) the proper noun connective =t (pn.det) encodes the proper noun status of the individual Alvin. A  curious feature of the connectives is the syntactic constituency-​word mismatch between the NP and its connective: IT connectives typically (but not always) encliticize to the immediately preceding constituent in the sentence—​regardless of that constituent’s status in the syntax of the clause, as the bracketing of the examples in (43) shows: (43) IT connectives (examples from Gitksan) a. kuxw[=hl kyuwatan] run.away=cn.det horse ‘The horse ran away.’ b. kuxw[=t run.away=pn.det ‘Alvin ran away.’

Alvin] horse

c. hlimoo-​i(-​t)[=s Tom][=t help-​tr-​3sg.II=pn.det Tom=pn.det ‘Tom helped Mary.’

Mary] Mary]

The proper noun determiner has a morphosyntactically conditioned allomorph =s, which occurs in Independent transitives as in (43c), and in Dependent transitives marking a proper noun subject, as in (44) (Hunt 1994): (44) hla yukw dim kuxw-​(ti)=s incept prog fut run.away-​3sg.II=pn.det ‘Alvin is about to run away.’

Alvini Alvin

Peterson (2004, 2006a, 2006b) analyzes the connectives as semantically weak determiners. Under this analysis the connectives =t/​s and =hl simply represent the argument status of the NPs the precede—​they do not have any typical determiner semantics (i.e. definiteness or uniqueness), but only a common vs. proper noun distinction. Because they are determiners, and not case morphology or agreement, we expect the connectives not to encode any mappings between grammatical and semantic roles. However, Tarpent (1987: 474) argues against =t/​s allomorphy, citing cases where =s co-​occurs with another connective, the associative tip:

 

1028   Tyler Peterson (45) giikw-​i=s dip Lisa ǥan=t Henry=hl wilp buy-​ctrl=pn.det assoc Lisa coord=pn.det Henry=cn.det house ‘Lisa and Henry bought a house.’             (Forbes 2013: 36) Forbes (2013) provides a fine-​grained theoretical analysis of the internal structure of the Gitksan DP that captures the co-​occurrence of =s and tip in sentences such as (45). Based on its semantic properties (proper noun status and associativity), Forbes calls dip a ‘true’ determiner. Because of this, tip is merged in a different syntactic position than =t and =hl, thus they are able to co-​occur. This analysis is compatible with treating =t and =s as allomorphs. A connective-​as-​determiner analysis is challenged by the CT languages, which have a considerably more complex paradigm of connectives. In addition to encoding the common/​proper noun distinction, enclitic CT connectives are sensitive the semantic role of the argument NP they mark. The Sm’algyax reduced connectives are given in Table 41.13, and the Sgüüxs in Table 41.14 (Dunn 1979a; Mulder 1994).10 What determines the arrangements in Tables 41.13 and 41.14 is the transitivity of the clause coupled with the Independent/​Dependent type of the clause. Taken together, what emerges is all of the logically possible alignments of semantic and grammatical roles, as summarized in Table 41.15 (adapted from Mulder 1994: 97).

Table 41.13 Sm’algyax connectives

Table 41.14 Sgüüxs connectives

Proper noun Common noun connective connective A S O

=as/​=dit =as =at

=a/​=da =a =a

Proper noun Common noun connective connective A S O

=as/​=dit =as =it

=i/​=di =i =i

Table 41.15 Alignments of the CT Connectives (both Sm’algyax and Sgüüxs)

Independent clauses Dependent clauses

10 

Proper noun

Common noun

A/​S:O (nominative) A:S:O (contrastive)

A/​S/​O (neutral) A:S/​O (ergative)

Mulder’s (1994) study of the Boas (1911) texts claims that there are additional Sm’algyax connectives that encode a present/​absent distinction (1994: 33). However, there are many missing connectives in this expanded paradigm, and these additional meanings have not been conclusively determined. As such, these additional connectives are not discussed here. Rather, Tables 41.13 and 41.14 are what Mulder calls the reduced connectives.

 

ALIGNMENT across Tsimshianic    1029 A discussion of the CT connective system is relevant in two ways: first, it is claimed in the CT literature that the CT connectives reveal a nominative alignment (Dunn 1979a,b; Mulder 1994). The person hierarchy is used to support this claim, as a proper noun (a human individual) would be represented by nominative/​accusative morphology, while the lower ranked common noun the ergative/​absolutive. Although plausible, this would be surprising from another perspective, since determiners cross-​ linguistically do not encode the alignments of grammatical and semantic roles. I show that a reconstruction of agreement morphology and determiners in CT, supported by a comparative analysis with IT determiners, reveals that the multiplicity of alignments in Table 41.15 result from the fusion of an invariant determiner with an independently attested agreement marker. As such, like their IT counterparts, CT connectives are not aligned in any way. To begin with, the neutral alignment of the common noun connectives, analyzed as determiners, in Independent clauses comes for free: (46) Sm’algyax: Neutral alignment a. nah t’uus(-​ti)[=a y’uuti]A[=a hanak’]O past push-​3sg.II=cn.det man=cn.det woman ‘The man pushed the woman.’                  (Dunn 1979a: 63) b. banm=xstoox[=a ɫguwoomɫk]S pretend=sleep=cn.det child ‘The child is pretending to sleep.’           (Stebbins 2003: 397) Peterson (2004, 2006a) claims that the connectives in CT are in fact morphologically complex: their synchronic form is the result of the fusion of the 3rd person Series II agreement and a common or proper noun determiner. The linear order of these fused elements involves the 3rd person Series II agreement and a following determiner in IT, as schematized in (47):11 (47) CT: Verb=[agr.det] IT: Verb-​[agr]=[det] Under this analysis the paradigm of CT connectives is drastically simplified, and all of the alignments in Table 41.15 disappear. The inventory of Tsimshianic connectives-​as-​ determiners is given in Table 41.16. Using (48) as an example, the ergatively aligned common noun determiner in Dependent clauses is reanalyzed as non-​aligned as follows: the common noun determiner =a marks the common nouns duus ‘cat,’ hoon ‘fish’ and sts’ool ‘beaver,’ each of which represents a different semantic role (just as with the IT common nouns in (46)). 11 

See also Dunn (1979b) and Mulder (1994: 43) who discuss the viability of this approach.

 

1030   Tyler Peterson Table 41.16 Tsimshianic determiners (Peterson 2006)

IT CT

Common noun cn.det

Proper noun pn.det

Plural assoc.det

=hl =a/​=i

=t (or =s)

tip

The original common noun connective =da which is claimed to only mark A  in Dependent clauses is isolable as 3rd person Series II object agreement (-​t) plus the common noun determiner =a marking the subject, via (47). This fission analysis is applied in the second lines of the examples in (48): (48) Sm’algyax: Ergative reduced to non-​aligned a. yagwa-​t huum=da duus=a hoon yagwa-​ti huum-​tj[=a duusi]A[=a hoonj]O prog-​3sg.I swim-​3sg.II=cn.det cat=cn.det fish ‘The cat is sniffing the fish.’               (Mulder 1994: 32) b. yagwa hadiks(-​ti)[=a sts’ooli]S da ts’m in t’aaks prog swim-​3sg.II=cn.det beaver prep in pond ‘A beaver is swimming in the pond.’             (Dunn 1979a: 63) The same results of the fission analysis obtain in Sgüüxs in (49), where the common noun determiner =i is a cognate of the Sm’algyax =a. (49) Sgüüxs: Ergative reduced to non-​aligned a. yágwa-​t níis=di óli=i hoon yágwa-​ti níis-​tj[=i ólii]A[=i hoonj]O prog-​3.I glare.at-​3sg.II=cn.det bear=cn.det fish ‘The bear glared at the fish.’              (Dunn 1979b: 133) b. dzagha dá’uhl(-​ti)[=i y’axwi]S t’ei lu on ’asdi nak across go-​3sg.II=cn.det man across on both sides ‘A man went across the inlet.’               (Dunn 1990: 2) Voicing of intervocalic obstruents is a robust phonological process across Tsimshianic, hence the [d]‌allomorph of the underlying 3rd person /​-​t/​agreement in (48a) (Rigsby and Ingram 1990; Brown 2010). This analysis explains part of the original alignments in Table 41.15: the original A-​marking connectives (dit, da, di) in CT all contain Series II object agreement (/​-​t/​) fused with a subject-​marking enclitic determiner (=t, =a, =i). What emerges is an exact correspondence with IT.

 

ALIGNMENT across Tsimshianic    1031 Davis and Forbes (2015) provide a detailed analysis of the morphophonological rules that condition that =s allomorphy that occurs with the proper noun determiners in IT, which also applies in CT. The isolable proper noun determiner =t, which is realized as the =s allomorph with transitive subjects, now reduces the nominative alignment to neutral, coupled with epenthetic [a]‌or [i], which breaks up the underlying /​-​t=t/​ and C=t in sequences in general (see Tarpent 1987: 852, Mulder 1994: 24, and Davis and Forbes 2015 for details): (50) Sm’algyax: Nominative reduced to non-​aligned a. niic(-​t)=as Nadine=at Isabelle niic(-​ti)[=t Nadinei]A[=t Isabelle]O see-​3sg.II=pn.det Nadine=pn.det Isabelle ‘Nadine saw Isabelle.’                (Stebbins 2003: 19) b. haytg=as simon da na-​dzooǥ=a aks haytg[=t simon]S da na-​dzooǥ=a na-​dzoog=aks stand=pn.det Simon prep poss-​edge=cn.det water ‘Simon stood by the edge of the water.’            (Mulder 1994: 79) Finally, we can apply fission to the original proper and common noun connectives in dependent clauses that give rise to the contrastive and ergative alignments, showing via agreement reconstruction that these are non-​a ligned, as in (51): (51)

Sm’algyax: Contrastive reduced to non-​aligned a. ɫa-​t k’yilum=dit dzodz=a baaysik das dzon ɫa-​ti k’yilum-​tj=[t dzodzi]A[=a baaysikj]O das dzon past-​3.I give-​3sg.II=pn.det G.-​cn.det bicycle prep J. ‘George gave a bicycle to John.’              (Mulder 1994: 40) b. yagwa-​t t’uus-​dit yagwa-​ti t’uus-​tj=[t prog=3.I push=3.II-​pn.det ‘John is pushing Mary.’ c. yagwa yawxg(-​t)=as yagwa yawxg(-​ti)[=t prog eat-​pn.det ‘Amy is eating.’

ami amii]S Ami

Dzon=at Dzoni]A[=t John=pn.det

Meli Melij]O Mary

(Dunn 1979a: 67)

(Mulder 1994: 68)

In sum, previous analyses of the CT connectives claim that there is a part of the CT grammar that is nominatively aligned, in addition to having contrastive and neutral

 

1032   Tyler Peterson alignments. Along with dissolving the alignments in Table  41.15, this analysis shows there is no alignments of any kind in the CT connective system, and by extension, no nominative alignment in the morphology of Tsimshianic. What we are left with is a fairly standard paradigm of determiners in CT that, despite being different lexemes, correspond exactly to those in IT.12

41.5 Conclusion Dixon’s generalization regarding the lack of a purely morphologically ergative language is challenged by the Tsimshianic languages: the alignments of agreement and pronoun distribution in both IT and CT are purely ergative (with the exception of the CT Series IIIb, which appear to be non-​aligned). The IT languages do have a split, one conditioned by clause type. Nonetheless, both sides of the split have an ergative alignment; even the pivot, the Series II agreement morphemes across Independent and Dependent clauses, is ergative. I also claim that the CT languages are purely morphologically ergative, in the sense that what lies on the other side of the split—​the contrastive and transitive alignments—​are simply expansions of how A, S, and O are represented in the agreement patterns. Essentially, A and S never overlap, meaning CT has an ergative core. This analysis also shows that the CT languages are not exotic in any way: these unusual alignments follow logically from the available agreement options in CT, as conditioned by the person hierarchy.13 This analysis does not shed light on the long standing issue in the Tsimshianic literature of how to define the clause types, as it does not provide an explanation as to why the Independent/​Dependent clause types constitute the main split in Tsimshianic. This conforms to the cross-​linguistic generalization that embedded environments are typically ergative in a split ergative language. Mulder (1994) defends a mood analysis of the Independent/​Dependent clause types in CT, and provides the most detailed account of what role tense, aspect and the person hierarchy plays in the clause types. We saw the effects of the last of these in the agreement/​pronoun patterns. Nonetheless, a mood-​based analysis seems untenable in IT. Evidence from displacement and other non-​mood meanings suggest the two clause types are the result of purely syntactic factors. 12 

Supporting evidence for this analysis is found in some sentences in Boas (1911: 362–​363), where he transcribes the CT common noun determiner =a as =ɫ, the same lateral fricative [ɬ] as the IT common noun determiner =hl. 13  See Kikusawa (Chapter 23, this volume) for a similar comparative/​historical approach involving the merger of pronominal sets and how this affects the ‘de-​evolution’ of ergative alignments in Austronesian.

 

ALIGNMENT across Tsimshianic    1033 Dunn (1979b) first suggested the possibility of segmenting that Sm’algyax connectives, but only to argue against it, as “no great increase in efficiency or simplicity can be gained by considering the connectives a series of subsyllabic suffixes. The segmented analysis requires more morphs with greater ambiguity (measured as morphs per homonym set) than does the unsegmented analysis” (1979: 136). What Dunn seems to be concerned with is the difficulty a Sm’algyax learner would face with the homonymy of the proper noun connective and 3rd person agreement, and with the added complication of =s allomorphy. Davis and Forbes (2015) express a similar concern from another perspective:  the morphophonological rules (of which I only briefly summarized above) needed to derive the right surface forms of the connectives introduce their own unique complexities for the learner. Further research is needed to explore this balance. Nonetheless, what is not clear is the payoff in terms of learnability: the unsegmented connectives afford a smaller lexicon, but at the expense of creating a sui generis class of morphemes with highly specialized meanings. Cross-​linguistically we do find fusions of determiners with other adjacent lexical categories in the syntax, such as the preposition-​determiner contractions in the Romance languages. However, these fusions do not encode semantic roles as the unanalyzed connectives do. Additionally, the unsegmented connectives not only obscure regular, predictable patterns of agreement and argument marking that are independently attested CT, but also the correspondences that hold between the CT and IT agreement and determiners. Finally, within the context of a study on the ergative properties of Tsimshianic it is worth mentioning that there has been some investigations into syntactic ergativity. Mulder (1994) claims that Sm’algyax is syntactically ergative, adducing tests from extraction, imperatives, coordination, subordination, and relativization that target A differently than S or O in Sm’algyax (see, for example, Aissen, Chapter 30, in this volume). Hunt (1993: 41) suggests that the same results do not obtain in Gitksan. I also leave it to future research to cast a contemporary light on these, and how it connects to the morphological ergativity described in this chapter.

Acknowledgments Special thanks to my Gitksan consultants Doreen Jensen, Fern Weget, Clara Weget, Gwen Simms, Barbara Sennott (Harris), and Louise Wilson. Thanks also to Will Oxford, Jason Brown, Henry Davis, and Bruce Rigsby for their helpful comments and corrections, and the research that lead to this chapter. This research was made possible with grants from the Endangered Language Fund, Jacobs Research Fund (Whatcom Museum Society), and The Endangered Languages Documentation Program (SOAS), awarded to the author. All errors are my own. Examples not cited are from fieldwork. The Tsimshian practical orthography is used: ḵ = [q]‌; hl and ɫ = [ɬ]; g = [G]; x = [χ]; a = [ə]; ü = [ɯ]. See Rigsby (1986); Tarpent (1987); Stebbins (2003); Brown (2010) for additional background.

 

1034   Tyler Peterson

Abbreviations 1, first person; 2, second person; 3, third person; attrib, attributive; dem, demonstrative; caus, causative; cn.det, common noun determiner; cn.prep, connective preposition; con, connective; coord, coordination; ctrl, control; incept, inceptive; loc, locative; obl, oblique; pfv, perfective; pl, plural; pn.det, proper noun determiner; poss, possessive; prep, preposition; prog, progressive; pst, past; rel, relative marker; sg, singular; tr, transitive; trans, transitive.

 

Chapter 42

What be i ng a syn tactically e rg at i v e l anguage me a ns for Katu kina-​K a na ma ri Francesc Queixalós

42.1 Introduction The following is an attempt at describing a fragment of the morphosyntax of a language that is remarkably suited for raising pivotal issues on grammatical relations. The language, Katukina-​Kanamari (hereafter KatKan), spoken between the Purus and the Javari, southern tributaries of the middle Amazon, is probably the last remnant of the small family Katukina.1 My purpose is twofold. First, to provide with some empirical facts the database on which any framework aiming at a general account of grammar’s structure and/​ or function checks its own validity. Second, to precisely define ‘grammatical relation’ in order to describe this particular language and to further account for an infrequent type of ergative grammar. For these ends, I will employ the usual tools of typology and descriptive linguistics, with an eye on empirical findings achieved by more theoretically oriented approaches. After profiling the major typological features of KatKan morphosyntax, I present the internal structure of phrases, a crucial prerequisite for establishing argument alignments. Some conclusions will then be drawn regarding grammatical relations and formal vs. functional motivations for some argument structure changing devices.

1  The compound name refers to two dialects, formerly considered as two distinct languages. Adelaar (2000) argues for a genetic relationship with Harakambut, an isolate spoken in Peruvian Amazonia.

 

1036   Francesc Queixalós

42.2  Typological Outline Limited morphology, strong constituency, and neatly asymmetric form of transitive verb arguments are the major distinctive features of KatKan. Parts of speech include verbs, nouns, adverbs, postpositions, and particles. Subclasses are: mono-​and di-​valent nouns and verbs, with an additional partition of monovalent verbs into unergatives and unaccusatives.2 Phrases, including predicate phrases, are headed by all parts of speech except particles. Non-​verbal predicates do not require a copula, but an existential copula can optionally occur. A derived voice reduces the valence of divalent verbs. Other argument structure changing mechanisms include applicative, causative, and noun incorporation.

42.3 Phrases In this section I will present the major constituents of the KatKan basic independent clause, first reviewing their syntactic positions, then describing their internal structure. The bulk of semantic and pragmatic information is distributed over the following syntagmatic units in the clause: predicate, argument(s),3 adjunct(s), as well as tense, aspect, modality, information-​structure, and discourse-​connecting particles. (1) panihanadjunct naparticle Poroyaargument todyipredicate boparticle! yesterday focus Poroya ComeBack exclamation ‘It is yesterday that Poroya came back!’ The basic word order is predicate initial, contrary to the example (1). Unless dislocated as in (1), adjuncts occur after the core (i.e. predicate + argument(s)). Particles are either final, second position, or free. The latter prove useful for testing constituency: in (2), ok/​* show the positions respectively available/​unavailable for hosting the connecting particle niama, ‘then.’4 2  Since the language has neither three-​place nor adpositional verbs, the mono-​/​di-​valent distinction for verbs is tantamount to intransitive/​transitive. This terminological choice allows me to highlight the parallel between argument structures of verbs and nouns. 3  A terminological proviso is in order. In my usage, ‘argument’ is exclusively taken as what is sometimes labelled ‘syntactic argument,’ that is, the linguistic expression (phonological zero included) of a core participant, the latter being a value that saturates a semantic role slot required by a given predicate. Consistently, ‘internal/​external’ merely denote the arguments’ occurrence location. Much of the following may get biased if ‘argument’ is understood as related to some intermediate level of structure between semantic (“thematic”) roles and linguistic expression in the vein of Manning (1996) and others (see also section 42.6). 4  This particle has a range of clause-​linking functions in discourse and intra-​sentential domains.

 

Katukina-Kanamari: a syntactically ergative language    1037 (2)

ok

wa * hinuk5 * nahoho ok kotyia woman group call otter ‘Then the women called the otter again.’

ok

tuda again

ok

Phrase heads distribute over two main categories depending on whether their argument structure comprises an internal argument. As for constituency, a remarkable isomorphism pervades phrases containing an internal argument. In order to visually capture this far-​reaching generalization (see below), the number of argument places allotted by a given head will hereafter be expressed through left-​exponents: e.g. Iverb, IIverb. Monovalent heads yield rather simple constituents containing Inouns (kotyia in (2)) and Iverbs (todyi in (1)), as well as pronouns, (idi:k in (4)) and lexical adverbs (panihan in (1)). Heads that require an internal argument include IInouns (naobatyawa in (3)) and IIverbs (nahoho in (2)), as well as postpositions (nakatu in (3)).6 (3)

waok-​dyi7 Nodia naobatyawa Hanani arrive-​cntrp Nodia wife Hanani ‘Nodia’s wife arrived here together with Hanani.’

nakatu ​CommInstr

Instances of verbal, nominal, adverbial, and postpositional predicates are (3), (4), (5), (6), respectively.8 (4)

tukuna idi:k HumanBeing you ‘You are a human being.’

(5)

kodo kamodya UpThere MonkeySp. ‘The monkey sp. is up there.’

(6) ita iki karipan fire locative SweetPotato ‘The sweet potato is in the fire.’ NPs also occupy argument positions, external to the predicate phrase as kotyia in (2), internal as wahinuk in (2), Nodia in (3) and Hanani in (3). Adverbial and postpositional phrases, for their part, also occupy adjunct positions, panihan in (1) and Hanani nakatu in (3). 5 /​ u/​is a high back unrounded vowel; /​:/​equals long vowel. Grammatical morphemes appear in their

basic allomorph shape. 6  Mono-​/​di-​valent for nouns is the morphosyntactic correlate of the alienable/​inalienable semantic distinction. 7  Dialectal variant: -​di. 8  Lexical adverbs and postpositions can be subsumed in a single category (heads of adverbial phrases), making the parallelism with mono-​/d ​ i-​valent nouns and verbs complete.

 

1038   Francesc Queixalós A special mention must be made to a NP displaying the following two properties: (1) it is case marked with the only morpheme in the language entitled to be called, prima facie, a case marker: -​na, ‘allative’; (2) its distribution is restricted to that of adverbial adjuncts, with no access to predicative function. (7) yo-​dahu wankurun hak-​na 1sing-​carry pot house-​allative ‘I brought the pot to the house.’ Now, the reader has probably noticed that the three IIheads contained in examples (2) and (3), the noun naobatyawa ‘wife,’ verb nahoho, ‘call,’ and postposition nakatu, ‘with,’ all begin with a syllable na. When preceded by their internal argument NP, IIheads obligatorily begin with na. In effect, besides the allative on adverbial adjuncts, the language has a single case marker na for NPs as internal arguments of IIheads. See both na in: (8) Kopa-​na= nuhuk poako Kopa-​MkCase= give paddle ‘Kopa gave the paddle to Dyano.’

Dyano-​na Dyano-​allative

Plausibly the argument case marker has its etymon in the allative suffix. (This hypothesis has far-​reaching consequences regarding the origin of argument alignment in KatKan, see section 42.8 and Queixalós 2013: 57ff.). Now, (8) is phonologically /​Kopa # nanuhuk # poako # Dyanona /​. I assume the phonological positioning of the first na to be the result of a diachronic process of procliticization to the head undergone by the case suffix grammatically attached to the immediately preceding internal argument. We are left with two synchronically homophonous case markers, one semantic (“inherent”), the allative plain suffix, and one grammatical (“structural”), the IIhead proclitic for internal argument. So as to simultaneously visualize structural na’s grammatical attachment to the left and phonological attachment to the right, I represent this morpheme as -​na= in examples. Moreover, and in order to highlight its unitary function of marking dependence to a head, I uniformly gloss it as MkCase. It must be added here that the postposition argument endures differential undermarking when non-​animate, see example (6). I turn now to the comparison between basic IIclauses and Iclauses in terms of the morphosyntactic properties of arguments. As for a preliminary distinction between the arguments of IIverbs, I will rely on their prototypical semantic role correlates, agent and patient.

42.4 Alignment Nominal case morphology is modest. In verbal IIclauses, -​na= marks the argument expressing the agent participant. The other argument lacks overt case marking. The

 

Katukina-Kanamari: a syntactically ergative language    1039 argument expressing the agent necessarily precedes the verb and, canonically, the argument expressing the patient follows it. In Iclauses, the single argument has the same properties for marking and order as the argument expressing the patient in IIclauses.    (9) Ino-​na= dyuman tahi Ino-​MkCase= pour water ‘Ino poured the water.’ (10) datikan pi:na sink hook ‘The hook sank.’ The two preceding examples show that case marking establishes an absolutive and an ergative category. Verb morphology confirms the existence of these categories, as can be seen in the asymmetry between the minimal pronominalization of each argument. The absolutive NP yields a zero (pro), whereas the ergative NP leaves behind a personal prefix on the verb. In other words, the internal argument must be overt, either lexically or pronominally. (The person prefix belongs to a paradigm of three persons and two numbers, with no gender–​class or inclusive/​exclusive distinction.) (11) a1-​dyuman Ø2 3sing-​pour ‘She1 poured it2.’ (12) datikan sink ‘It sank.’

Ø

(The mandatory realization of the internal argument also obtains in postpositional phrases and IINPs, appealing to the same morphological device; see section 42.3.) Order between the ergative argument and its verb is fixed due to their respective constituency properties. For the same reason, the absolutive NP can be separated from the verb either through fronting, (13)–​(14), or by an intervening particle, (15)–​(16). (13) [tahi] [ [Hanani-​na=] dyuman] water Hanani-​MkCase= pour ‘Hanani poured the water.’ (14) [pi:na] [datikan] hook sink ‘The hook sank.’

 

1040   Francesc Queixalós (15)

[ [Hanani-​na=] (*niama) dyuman] Hanani-​MkCase= pour ‘Then Hanani poured the water.’

(16) [datikan] (OKniama) sink then ‘The hook sank.’

(OKniama) then

[tahi] water

[pi:na] hook

A demonstrative may appear as, or within, an absolutive NP, respectively (17)–​(18) and (19)–​(20). Neither is possible for an ergative NP.9 (17) wa:pa-​na= ti [itiyan] dog-​MkCase= kill ProximalDemonstrative ‘The dog killed this one.’ (18) ki:tan [itiyan] sleep ProximalDemonstrative ‘This one slept.’ (19) yo-​tikok [ityian 1sing-​know ProximalDemonstrative ‘I know this white man.’

dyara] WhiteMan

(20) ki:tan [itiyan10 wa:pa] sleep ProximalDemonstrative dog ‘This dog slept.’ Coordinated NPs can only be absolutive. No formal means other than sequence and prosody are at work in coordinating NPs. (21) opatyin-​na= wu a-​wa nyama child-​MkCase= want 3sing-​grn mother ‘The child loves her mother and her sister.’11

a-​ponhanya 3sing-​sister

(22) tyuku Nodia Owik die Nodia Owik ‘Nodia and Owik died.’ 9 

See Dixon (1972: 218) for a comparable restriction in Dyirbal. Itiyan/​ityian are dialectal variants. 11  GRN: generic relational noun. ‘Mother’ is a Inoun. INouns related to an overt ‘possessor’ unfold a complex structure whereby the head noun is preceded by a IINP itself consisting of the generic relational noun -​wa as head and its internal argument (Queixalós forthcoming). ‘Sister’ is a IInoun. 10 

 

Katukina-Kanamari: a syntactically ergative language    1041 Contrastive focus is accomplished on absolutive phrases by a following particle kana (or na, as in (1)) and, in a vast majority of its occurrences, by placing the focused constituent in initial position. It does not apply directly to ergative phrases. (23)

pi:na kana Aro-​na= hook focus Aro-​MkCase= ‘It’s a hook that Aro brought.’

dahu carry

(24) waro kana ki:tan-​nin parrot focus sleep-​durative ‘It’s the parrot that is sleeping.’ A questioned constituent is fronted and replaced by an invariable interrogative pronoun. Among verb arguments, this only involves absolutive phrases. Depending on the dialect, an interrogation particle appears clause final or after the questioned constituent.12 (25)

hanin tan no-​toman InterrPrn here 2sing-​shoot ‘What did you shoot here?’

(26) hanian tu InterrPrn interrogation ‘Who is arriving?’

yu? interrogation

waok-​dyi-​nin? arrive-​centrp-​durative

Relativizing a noun simply consists of introducing the relative clause with a reference conferring form nyan, possibly a truncated version of anyan, third person free pronoun.13 Relatives, as other subordinate clauses, mark their dependence through the -​nin suffix (assumed to descend diachronically from the same etymon than the durative -​nin appearing in the previous example; see section 42.8 and Queixalós 2013: 57ff.), and are reputed to be non-​finite since they lack tense–​aspect–​mode particles. The relativized noun, which can only be an absolutive inside the relative clause, preferably follows the deictic nyan but may be left in situ, (27)b. In this latter case, no ambiguity arises as to the identity of the relativized noun in IIclauses thanks to the impossibility to relativize the ergative. (27) a yo-​hi:k nyan [tukuna Makuana-​na= dahu-​dyi-​nin] 1sing-​see/​look deictic HumanBeing Makuana-​MkCase= carry-​centrp-​dep ‘I know the man that Makuana brought here.’ 12  Hanin/​hanian and yu/​tu are also dialectal variants. The latter is homophonous with the negation mark. 13  Itself a formerly demonstrative form. A few relatives in the data are introduced by anyan.

 

1042   Francesc Queixalós b yo-​hi:k nyan [Makuana-​na= dahu-​dyi-​nin tukuna] 1sing-​see/​look deictic Makuana-​MkCase= carry-​centrp-​dep HumanBeing ‘idem’

(28) tyo-​tikok nyan [tukuna hoki-​nin tyo-​katu] 1plural-​know deictic HumanBeing talk-​dep 1plural-​CommInstr ‘We know the man who talked with us.’ (In the other dialect, relativization is accomplished without the deictic nyan but with obligatory fronting of the relativized noun: (29) yo-​tiok [dyara yo-​toman-​nin] 1sing-​know NonIndian 1sing-​shoot-​dep ‘I knew the white man that I shot.’) So far, we have seen three limitations on argument extraction that are typically imposed, cross-​linguistically, by ergative syntaxes (e.g. several Mayan languages, Grinevald & Peake 2012). Coreference phenomena depict a situation where the existence of a clear pivot in a given construction is in an inverse ratio to the exposure of the latter to pragmatic factors such as situational context and background information on the participant referents, as well as semantic ones related to world knowledge. At the intra-​clausal level, indeterminacy prevails as to the control of the “possessive” third person prefixes in argument NPs: either the absolutive, (30), or the ergative, (31), can be the antecedent. In principle, disjoint reference triggers the insertion of a free pronoun as in (32) and (33); nonetheless, (34) is excerpted from a spontaneous text. (30) [ [a1-​obatyawa-​na=] hoho] [Nodia1] 3sing-​wife-​MkCase= call Nodia ‘Nodia1’s wife called him1 (more lit.: His1 wife called Nodia1).’ (31) [ [Waro1-​na=] kiunyuk] [a1-​tyo] Waro-​MkCase= comb 3sing-​daughter ‘Waro1 combed her1 daughter.’ (32) [ [ [anyan1-​na=] obatyawa-​na=] hoho] 3sing-​MkCase= wife-​MkCase= call ‘His1 wife called Nodia2.’

[Nodia2] Nodia

(33) [ [Waro1-​na=] kiunyuk] [ [anyan2-​na=] tyo] Waro-​MkCase= comb 3sing-​MkCase= daughter ‘Waro1 combed his/​her2 daughter.’

 

Katukina-Kanamari: a syntactically ergative language    1043 (34) [ [pi:da1-​na=] mato:kdakan] jaguar-​MkCase= lap ‘The jaguar1 lapped his2 blood.’

[a2-​mimi] 3sing-blood

As (30) and (31) show, neither the order precedence nor the constituency hierarchy (c-​ command) are relevant to the control of this kind of anaphora.14 The third person prefix in postpositional adjuncts is straightforwardly controlled in an ergative fashion; that is, it looks for its antecedent in the absolutive argument. Beyond anaphoric prefixes, lexical adverbs with manner or locative meaning are also ergatively oriented. (35)

Dyomi1-​na= tohi:k Aro2 Dyomi-​MkCase= see Aro ‘Dyomi1 saw Aro2 in his2 house.’

(36) Konta1-​na= tohi:k pi:da2 Kontan-​MkCase= see jaguar ‘Kontan1 saw the jaguar2 up there2.’15

a2-​wa 3sing-​ngr

hak house

naki inessive

kododi2 UpThere

At the inter-​clausal level, auxiliaries deserve special mention. Inter-​clausal dependence is mainly marked with the -​nin suffix, already met in relatives. Two lexical verbs, IIwu, ‘want,’ (21), and Ibak, ‘be good/​well/​beautiful,’ (37), are respectively recycled as purpose and intensive auxiliaries, (39)–​(42). (37)

bak-tu korion BeGood-NEGATION vine ‘The vine is not good (i.e. is useless).’

As auxiliaries and independently of either their inherent valence or that of the subordinate verb, wu and bak lend a clause structure, [a…]a in (38), where the formally matrix verb (i.e. the auxiliary) heads a IIphrase, [b…]b, provided with its external, [c…]c (here fronted) and internal, [d…]d, arguments. (38)

[a [c tukuna]c [b [d hoki-​nin=]d wu-​nin]b]a itan niama adyi HumanBeing     talk-​dep= want-​dep here then come ‘When people want to talk, they come here (lit.: people wanting to talk come here).’

14  See Queixalós (2013: 72) for the idea that such indeterminacy in coreference control may be characteristic of an early stage in the diachronic evolution of ergativity. 15  In elicitation session, with the help of gestures, I tried to get the informant understand (35) as taking place at Dyomi’s place and (36) as mentionning Kontan’s location, but both readings were rejected. For the former, the informant insisted: “No, at Aro’s.” For the latter, he appealed to something equivalent to ‘Kontan was sitting up there, (and) he saw the jaguar.’

 

1044   Francesc Queixalós The internal argument, containing the subordinate lexical verb, marks its dependence to the matrix through the -​nin suffix, which behaves syntagmatically like the case marker -​na= as we saw: grammatically hosted by the constituent at its left but phonologically bound to the phrase head at its right. In the example, the dependence marker -​nin occurs twice, first as a proclitic subordinating the lexical verb ‘talk’ to the auxiliary ‘want,’ then as a suffix subordinating the whole phrase ‘people wanting to talk’ to the clause matrix verb ‘come.’16 In terms of the mapping of semantic roles onto arguments, IIwu appears to generate a VP pattern comparable to that to be seen at a later point in the chapter (section 42.5) in accusatively aligned clauses: the “wanted” manner of existing17 is projected onto the internal argument, ‘my wife cooking (the meat)’ in (39),18 and ‘my bathing’ in (40). (39) [ [yo-​obatyawa-​na= toda-​nin=]IN wu]    1sing-​wife-​MkCase= cook-​dep= want    ‘My wife wants to cook the wild meat.’ (40) [ [kodyi-​nin=]IN wu]    bathe-​dep= want    ‘I want to bathe.’

barahaiEX WildMeat

aduEX 1sing

Though less transparently, Ibak brings about a parallel pattern: the “good” (i.e. intensified) manner of existing is projected on the internal argument: ‘they calling (the manatee)’ in (41), and ‘(the canoe) being full’ in (42). (Thinking of bak as an unaccusative verb, it makes some sense to view its single argument, denoting the participant which ‘is good,’ actually a manner of existing, as an internal one.)19 (41) niama [ [ma-​oko-​nin=]IN bak] then   3plural-​call-​dep= BeGood ‘Then they keenly called the manatee.’ (42) [ [dyo:-​nin=]IN bak]    BeFull-​dep= BeGood    ‘The canoe is pretty full.’

tyoponaEX manatee

podakEX canoe

The external argument of both the subordinate and the matrix clauses (absolutive and nominative, respectively) exhibits crucial properties. First, in the subordinate clause, 16 

Besides marking dependence and durative aspect, -​nin appears in several more or less lexically frozen nominalizations, see section 42.8. 17  A catch-​all phrase for whatever semantic relation a predicate bears to its participant(s): action, event, process, state, property, inclusion, etc. 18  VP and its internal argument bracketed. And: IN/​EX: internal/​external. 19  Thanks to Andrés Pablo Salanova for pointing out this aspect to me. I also owe him many valuable comments on an earlier draft of this chapter.

 

Katukina-Kanamari: a syntactically ergative language    1045 the only overt argument is the internal-​ergative, lexically in (39), pronominally in (41); the external-​absolutive, on its part, is void phonologically. Second, as long as wu’s dependent clause is monovalent, the external-​nominative argument of the matrix clause is a semantic participant of the matrix verb, e.g. the “wanter” of (40). If wu’s dependent clause is divalent, the external-​nominative argument of the matrix clause, ‘wild meat’ in (39), is no longer a semantic participant of wu, ‘want.’ Neither are ‘manatee’ in (41), and ‘canoe’ in (42) semantic participants of matrix bak, ‘be good.’ In spite of wu appearing to be a good candidate for a control verb generating a PRO in the dependent clause, the need for a unified account for auxiliary constructions is clear from the foregoing and from the co-​occurrence of both auxiliaries, (43). Moreover, it is worth noting that wu further gets recycled for coding imminent aspect, (44), a fact that somehow reflects the lack of semantic link between the auxiliary and the participant expressed by its external argument. I will therefore assume that the external-​absolutive argument of the lexical subordinate verb in auxiliary constructions uniformly raises to the external-​nominative position in the matrix clause. (43) [a[b [ca1-​oma-​nin=]cIN bak-​nin=]bIN      3sing-​buy-​dep= BeGood-​dep=     ‘He1wants to really buy it2.’

wu]a Ø2EX want

(44) [ [wara-​nin=]IN wu] ØEX    GiveBirth-​dep= want    ‘She was about to give birth.’ Another case of raising is observed with tikok, ‘know,’ when used as a complement-​ taking verb. Again, the external-​absolutive in the completive surfaces as the external-​ nominative in the matrix, independently of whether the involved participant is a semantic role of the latter, and independently of the alignment in the IIcompletive: see Iverb in (45) and IIverb in (46). Note that in (46) mok is no participant of the matrix verb ‘know.’ As in auxiliary constructions, the matrix VP is built on the accusative pattern, since its internal argument is the ‘known’ manner of existing. Now, the lack of procliticization of the dependence marker -​nin (phonetically [waikpanin # tikok # adu]) is evidence that we are dealing here with something different from an auxiliary construction. (45) [a [b [cwaikpa-​nin]c [Ø1] ]b      song-​dep      ‘I know how to sing.’

tikok]a know

(46) [a [b [cyo1-​toman-​nin]c [Ø2] ]b tikok]a      1sing-​shoot-​dep know      ‘I know how to shoot the tapir.’

adu1 1sing mok2 tapir

 

1046   Francesc Queixalós With verbs other than tikok as ‘know how to,’ the argument selectional requirements of the subordinate verb appear to be independent from those of the matrix verb, therefore yielding non-​raising structures. This can be observed in the following complement clause, where not only no argument is coindexed with an argument of the main verb, but all the subordinate verb arguments are realized in situ. (47) Nodia1-​na= tohi:k [ [Warohan2-​na= ikubaraman-​nin] [wiri3] ]… Nodia-​MkCase= see Warohan-​MkCase= AimAt-​dep WildPig …mokawa katu  gun     CommInstr ‘Nodia saw Warohan aiming at the wild pig with a gun.’ Inducing the speaker to utter (47) with Warohan aiming at Nodia results in having the latter referent overtly resumed in the subordinate clause, (48). Likewise, the mutual independence in argument selectional restrictions between both verbs makes the difference between wu, ‘want,’ as an auxiliary verb (see above) and as a full verb, (49). (48) Nodia1-​na= tohi:k [ [Warohan2-​na= ikubaraman-​nin] [Nodia1] ] Nodia-​MkCase= see Warohan-​MkCase= AimAt-​dep Nodia ‘Nodia1 saw Warohan2 aiming at him1.’ (49) Nodia1-​na= wu-tu [ [opatyin2-​na= biwi:k-​nin] [kapayo3] ] Nodia-​MkCase= want-NEGATION    child-​MkCase= suck-​dep papaya ‘Nodia1 does not want the child2 to eat the papaya3.’ As for subordinate Iclauses, all the data show the sole argument as overtly standing between the main and the subordinate verbs, which could be indicative of an absolutive argument of the matrix verb, as assumed in (50). An alternative analysis might be fronting of the absolutive argument inside the subordinate clause, as in (51), but: (1) as discussed, no data is available containing the argument in its canonical post-​predicate position; (2) their highly presupposed character generally makes subordinates immune to pragmatic pressures (Van Valin 2013);20 and (3) I must acknowledge that besides relativization no methodic scrutiny of fronting in subordinates has been undertaken. (50) Owik1-​na= ba:bu Tikon2 Owik-​MkCase= order/​send Tikon ‘Owik sent Tikon to hunt.’ (51)

20 

?

Owik-​na= ba:bu Owik-​MkCase= order/​send

[ [dyan-​nin] [Ø2] ]     hunt-​dep

[ [Tikon]   Tikon

[dyan-​nin] ] hunt-​dep

Absolutive argument fronting in one dialect’s relative clauses is not motivated by pragmatic factors (see above).

 

Katukina-Kanamari: a syntactically ergative language    1047 Regarding subordinate IIclauses in the ergative pattern, control structures are far from being as straightforward as for subordinate Iclauses. The general constraint on obligatory realization of the internal argument discards an ergative PRO since the argument representing the agent in the dependent clause will of necessity be either a case marked NP or a person verb prefix, as shown in the two following examples. (52)

a1-​obatyawa2-​na= nobuk21 [ [Dyano1-​na= tuku-​nin] 3sing-​wife-​MkCase= send/​order   Dyano-​MkCase= chop-​dep ‘Dyano's wife sent him (lit.: his1 wife2 sent Dyano1)22 chop the log3.’

(53)

ma1-​man-​na niama [ [a2-​man-​nin] 3plural-​do/​make-​centrf then 3sing-​do/​make-​dep ‘Then they sent23 him to go get a wild pig.’

[oman3] ] log

[wiri3] ] WildPig

Turning to example (54), which contains the purpose subordinator niama, the issue is then the precise properties of the zero pronoun in such a subordinate IIclause, an aspect that is still at an early stage of understanding: the argument position in the example is controlled, but the proscription of a NP in it (i.e. it being necessarily controlled) has to be empirically validated yet. (54)

Hi:wuk1-​na= hoho Mayon2 a2-​ama-​hoki Ø1 niama Hi:wuk-​MkCase call Mayon 3sing-​applicative-​talk purpose ‘Hi:wuk1 called Mayon2 to have the latter2 talk to him1.’

Comrie (1985) and Falk (2006: 135ff.) point out that in control constructions a semantic cline tends to endow the participant represented by the controlled argument with agentive characteristics. This entails that, in a syntactically ergative language like KatKan with a mandatory overt ergative argument, the lack of a genuine PRO in a subordinated IIclause is expected. I will however leave open the issue since (1) absolutive controlled PROs, though scarce cross-​linguistically, are attested elsewhere (see Dyirbal and other languages in section 42.6); (2) a supplement of KatKan data (with prototypical control verbs for the niama clause) would certainly afford a more subtle analysis. Adverbial subordinate clauses are introduced by postpositions (except temporal clauses, see (38)). As in auxiliary constructions, the dependence suffix -​nin cliticizes to the phrase head at its right, here a postposition. Now the ergatively oriented coreference pivot, coindexing two absolutives, is only preferential.

21  A dialectal variant of ba:bu. Note that a-​is an argument within the NP and that control of its reference by the internal argument of the subordinate clause is another blatant instance of the neutrality of constituent hierarchy on coreference control. 22  Even more literally: “his wife sent him (for) Dyano chopping the log.” 1 1 1 23  The verb man is polysemous: ‘do, make, get, say, send, order.’ More semantically concrete is buhuk, ‘make, manufacture.’

 

1048   Francesc Queixalós (55)

ityaro1-​na= hak don2 [noporan2-​nin=] woman-​MkCase= spear fish gasp-​dep= ‘The woman1 speared the fish2 while (it2 was) gasping.’

ton locative

(56) a1-​makaudyaran Ø2 [dyahian2-​nin=] ama 3sing-​StrideOver StandUp-​dep= goal 'He1 strode over her2 to (have her2) stand up.’ Coordination between clauses is the less restricted domain within the sentence. It is thus reasonable to expect that even in languages with a strong tendency to omit, under coreference with a given argument in the first clause, the expression of a given argument in the second clause, functional factors will often tend to prevail: as for pragmatics, situational or interactional priorities tuning reference; as for semantics, world knowledge precluding unintended readings. KatKan is no exception. Coordination is paratactic in terms of formal devices strengthening the link between consecutive independent clauses. Prosody, mainly global intonation, seems to be the basic clue for distinguishing coordinated clauses from just successive utterances in a discourse stretch. The connecting particle niama can also play a role in tightening together otherwise paratactic clauses. Statistically, and in speaker's judgements, coreference in coordinated clauses is moderately biased to ergatively oriented pivots, as in (57). An accusatively oriented pivot in (58) is accepted, but the consulted speaker finds more natural the reading whereby the people go away, in accordance with the preceding example and in spite of the myth, where Tamakori is the one that actually goes away (see (70)). Clauses between [a…]a. (57) [a [b [cKontan1-​na=]c tudiok]b [Dyan2] ]a [atona Ø2 niama]a    Kontan-​MkCase= scold Dyan GoAway then    ‘Kontan1 scolded Dyan2 and then she2 went away.’ (58) [a [b [cTamakori1-​na=]c buhuk]b [tukuna2] ]a [a [tona Ø2/1 niama]a ]    Tamakori-​MkCase= make HumanBeing   GoAway then    ‘Tamakori1 created the people2 and then they2 / he1 went away.’ All told, depending on the phenomenon under scrutiny, coreference regulations can be said as ranging from inexistent to ergatively colored to obligatorily ergative. Intraclausally, neither precedence in linear order nor constituency hierarchy appear to play any role in establishing a coreference controller for third person prefixes on core NPs. Conversely, the external absolutive NP controls the person prefix of NPs within postpositional phrases, as well as the orientation of lexical adverbs. The starkest ergatively based phenomena are found in raising structures such as verb-​auxiliary sequences and completives of some verbs: no internal-​ergative argument is ever involved. In adverbial subordination and in coordination, controllers rather select the external-​absolutive argument respectively contained in the matrix or in the first clause. Finally, control

 

Katukina-Kanamari: a syntactically ergative language    1049 constructions proper have not been uncovered to date involving the ergative pattern in the subordinate clause. Oriented nominalizations are the last domain I will mention as inducing an absolutive category. The unique and IIverb-patient participants are directly rendered by the deictic nyan (seen above with a different distribution) following the otherwise unaltered verb. The IIverb retains the person prefix expressing its agent participant. (59)

[yo-​tikok nyan] pok 1sing-​know deictic canoe ‘I know this canoe (lit.: this canoe is my known one).’

(60) [adyi nyan]-​na= nuhuk tawa come deictic-​MkCase= give manioc ‘The one who came gave the manioc (lit.: the coming one gave…).’ (This is quite different a matter from the amply documented absolutive cline of non-​ oriented nominalizations (Alexiadou 2001)  in and outside Indo-​European, since the cline in question is about argument retrieval, independently of basic alignment,24 whereas we are dealing here with the denotatum of the nominalization itself.) By subsuming in a single and simple formal device the unique and the IIverb patient participants, nominalization shows that, lexically, the IIverb is patient-​oriented.

42.5  Split Transitivity Transitive splits are congenital to ergativity (although not all need be related to ergativity; e.g. some differential “subject” marking phenomena). First, they appear cross-​ linguistically as being most pervasively linked to ergativity.25 Second, synchronic conditions under which clauses depart form the basic ergative pattern not only unveil important properties of the ergative pattern, either functional (e.g. patient-​oriented tense-​aspect) or formal (e.g. more or less noun-​like verbs) but correlatedly provide precious hints as to the diachronic conditions that pave the way for the spreading of ergative make-​ups into basic independent clauses (e.g. semantically or pragmatically unexpected

24  In Sikuani, as an instance of a nominative-​accusative language (Queixalós 2012), the IIverb argument expressing the patient appears in nominalizations as a genitive, as does the sole argument of a Iverb. A number of cross-​linguistically documented phenomena display an absolutive-​ergative alignment independently of the basic alignment in the language. Queixalós (2013: 11) subsumes them under the label ubiquitous ergativity. 25  According to Ramirez (2003), Yanomami would be splitless in both morphology and syntax. Reference-​tracking mechanisms seem to be discourse-​driven, with a neat preference for absolutive pivots.

 

1050   Francesc Queixalós agent, patient focusing, agent downgrading, deverbalizing devices, complex-​verb constituents, oblique marking). In KatKan an unindividuated patient participant triggers an accusatively aligned IIclause.26 Here again we see the basic predicate-​ initial order and the asymmetry between both arguments. But now: (1) the internal argument expresses the patient participant; (2) this accusative argument, obligatorily overt and lexical, bears no explicit case marker; (3) the verb lacks person morphology, as do Iverbs; (4) the external argument expresses the agent; (5) this nominative argument has, as such, the same formal properties as the sole argument of Iverbs. (For an argumentation as to the differences between the accusative argument and an incorporated noun, see section 42.8). In the following, examples a show accusatively aligned clauses, to be contrasted with ergatively aligned clauses in b: (61) a pi:da ohiya Hi:wuk jaguar fear Hi:wuk ‘Hi:wuk fears jaguars.’ b panihan Hi:wuk-​na= ohiya yesterday Hi:wuk-​MkCase= fear ‘Yesterday, Hi:wuk feared a/​the jaguar.’

pi:da jaguar

(62) a tahi dyuman Hanani water pour Hanani ‘Hanani poured water.’ b Ino-​na= dyuman Ino-​MkCase= pour ‘Ino poured the water.’

tahi water

Comparing clause constituency: (63) Accusative [ [tahi]NP  dyuman]VP [Hanani]NP water   pour      Hanani Ergative [ [Hanani-​na=]NP dyuman]VP Hanani pour

26 

[tahi]NP water

Dialects, and speakers intra-​dialectally, diverge as to the strength of this semantic condition. “Unindividuated”: either a generic notion (a collection of individuals taken as an internally unstructured set), or a mass notion. This condition on accusative clauses does not entail that patients in the ergative pattern have to be individuated, specific, referential, definite, and so on. They tend to (and plausible diachronic reasons for this can be adduced), but no strict condition prevails.

 

Katukina-Kanamari: a syntactically ergative language    1051 Inspection of other syntactic properties of arguments in the accusative clause has not, comparatively, been carried out in a systematic fashion, and several gaps subsist. However, a few hints to asymmetries are available. The nominative NP can be elided, (64), fronted, (65), focused, (66), and raised in auxiliary constructions, (67). The accusative is immune to all four, for instance (68). (64) [ [tukuna] makoniok] HumanBeing advise ‘He1 advised the humans.’

Ø1

(65) pi:da [ [ityaro] botyana] jaguar woman track ‘The jaguar tracks the women.’ (66) mapiri na [ [takara] duni] anaconda focus hen catch ‘It is the anaconda that catches hens.’ (67) [ [ [wiri] pu-​nin=] wu-tu] adu WildPig eat-​dep= want-NEGATION 1sing ‘I don’t want to eat wild pig.’ (68) *oba na biwi:k ityaro tobacco focus suck woman ‘(intended:) It’s tobacco that the woman smokes.’27 Nominative coreference pivots are at work in coordination and in subordination between matrix and adverbial clauses:  see coordinated clauses in (70) as divalent + monovalent, in (71) as divalent + divalent, and in (72) hierarchized clauses as IImatrix + divalent adverbial. No pivot combining the accusative argument and any other argument—including short-distance anaphora between intraclausal co-​arguments, (69)— has been detected so far (recall the unindividuated condition attached to the patient participant; clauses between [a...]a). (69) [ [anya1] otohi:k] a2/​*1-​okpu woman LookFor 3sing-​son ‘His2 son looks for some woman1.’

27  When proposed this sequence, which would contain a focus particle between the accusative NP and the verb, the speaker did not even think of focus, and judged the idea of tobacco being able to smoke the woman as nonsense. That is to say, thanks to the homophony between this focus marker and the case suffix, he analyzed the sequence as following the ergative pattern: [ [oba-​na=] biwi:k] [ityaro].

 

1052   Francesc Queixalós (70) [a [ [tukuna1] buhuk] Tamakori2]a [a atona Ø2 niama]a HumanBeing make Tamakori GoAway then ‘Tamakori2 created the people, then he2 went away.’ (71) [a[ [bamak1] wandoki] Dyoraidi2]a, [a[ [nokonana3] wandoki] Ø2 tyin]a FishSp. cook Dyoraidi FishSp. cook also ‘Dyoraidi2 cooked fish sp.1, and she2 also cooked fish sp.3.’ (72) [a[ [mokawa1] wu] adu2]a [a[ [wiri3] toman2-​nin=] ama]a gun want 1sing WildPig shoot-​dep= goal ‘I2 want a gun1 in order to shoot2 wild pigs3.’ We saw that control constructions with a subordinate IIclause in the ergative pattern are less than clear. There I suggested that this might follow from (1) the semantic propension to have the controlled argument express a somehow agentive participant, and (2) the formal properties of that argument in the ergative clause. If such a semantic condition does play a role, then the accusative clause meets it and should allow a controlled PRO. (73) seems to be in keeping with this expectation. (73) Bada1-​na= ba:bu Topiana2 [ [wiri toman-​na-​nin] [Ø2] ] Bada-​MkCase= order/​send Topiana WildPig shoot-​centrf-​dep ‘Bada sent Topiana shoot wild pigs over there.’ As seen while reviewing nominalizations of verbs in basic, ergatively aligned, clauses (section 42.4), a nominalization oriented toward the patient of the IIverb is built along the lines of that oriented toward the sole participant of a Iverb (i.e. nyan following the otherwise unaltered verb). In accusative clauses, it is the nominalization oriented toward the agent, (74), which is identical (verb unaltered) to that of the unique argument of Iverbs, (60), thus lending a nominative category. (74) waok-​dyi [dyurakon tokman nyan] arrive-​centrp LatexTree cut deictic ‘The rubber tapper arrived here.’ In sum, asymmetries between both arguments of the accusative clause point in the direction of a nominative argument clearly prominent on syntactic grounds: constituency, elision, fronting, extraction, raising, coreference requirements (including PRO), and nominalization.

42.6  Subject and Object The syntactic prominence mentioned in 42.5 regarding the nominative argument in accusative clauses is ample evidence for subjecthood. Now, the fact is, this evidence is

 

Katukina-Kanamari: a syntactically ergative language    1053 very much the same than that adduced in section 42.4 for the absolutive argument in ergative clauses. And, of course, the same that is displayed by the sole argument of Iverbs. I therefore consider the absolutive argument of IIverbs as being the subject of its clause as well. We are, thus, subsuming under the notion of subject as the set of arguments privileged for accessibility to syntactic rules (accessibility that creates asymmetries between IIclause arguments), several categories that conform a somewhat heterogenous class when defined in case terms: (1) the argument expressing the agent in accusative clauses or “nominative”; (2) the argument expressing the patient in ergative clauses or “absolutive”; to what we can add: the sole argument of Iverbs, “nominative” when compared to the argument expressing the agent in accusative clauses, but “absolutive” when compared to that expressing the patient in ergative clauses.28 As for the non-​subject argument of the IIverb (unmarked “accusative” and marked “ergative” in their respective clause type), I will dub it the object of the clause in spite of its syntactic properties be less homogeneous. Let us now consider a factor of heterogeneity between IIclause subjects: PRO, detected for the accusative clause subject, but problematic for the ergative clause subject. While uncommon, clear absolutive PROs are not unattested cross-​linguistically. Dyirbal might well be an instance of a syntactically ergative language with an absolutive PRO.29 Assuming that the “purpose” marker in the following example, excerpted from Dixon (1994: 169), entails non-​finiteness: (75)

yabu1-​Ø ŋuma2-​ŋgu giga-​n Ø1 gubi3-​ŋgu mawa-​li mother-​abs father-​erg send-​NonFuture PRO doctor-​erg examine-​purpose ‘Father sent mother to be examined by the doctor.’

A cogent account of this cross-​linguistic paucity is that the functional dimension of PRO (namely: agentivity) is more central to the issue than commonly acknowledged. In this respect it is worth recalling Comrie’s (1985) observation that, in English would-​ be counter-​examples to PRO as agentively colored, the speaker is, by using the control structure, somehow endowing the participant with a non-​null amount of agentivity. Be it as it may, PRO is perhaps to be seen as not so much entitled to any particularly universal status of hallmark for subjecthood. Due to its functional agentive bias, it simply fits well accusative syntaxes.30 All in all, however, it remains that in many languages, the two co-​arguments of the IIclause show a clear imbalance regarding their respective syntactic properties. Moreover, these asymmetries in their turn point to a notion of hierarchy: (1) one

28 

And, additionally, the sole argument of non-​verbal predicates. Gone unnoticed, to my knowledge, except for Aldridge (2008c). Bittner & Hale (1996b), for instance, spot a rather far-​fetched ergative-​agent PRO in an example from Dixon (1994: 134) (see Queixalós 2007 for comments). Several Austronesian languages feature an absolutive PRO (Toba-​Batak, Manning & Sag 1999; South-​Tama, Trick 2006; Nehan, Glennon 2014: 102). 30  Because, along with other aspects of descriptive and theoretical paraphernalia, it has been worked out within the realm of accusativity. See Van Valin (2013) for a similar dependent-​marking bias on PRO. 29 

 

1054   Francesc Queixalós argument has convergent prominence in constituency, access to syntactic operations, and coreference regulations (the two latter appearing to a certain extent as by-​products of the former);31 (2) such prominence makes this argument similar to the sole argument of the Iclause. In our case at hand, there are in KatKan good grounds for thinking that the arguments labelled absolutive and nominative in the foregoing presentation are subjects in their respective clauses. I do not want to claim, though, that an ergative morphosyntax is but a mirror image of an accusative one. It is not, pace Martinet (1979) and Marantz (1984). From the very “inversion” (Manning 1996) in the respective mappings of semantic roles onto formal arguments emerge a number of contrasts between ergative and accusative syntaxes (e.g. splits) that raise challenges to several approaches to morphosyntax. For the time being, it is sufficient to say that we have robust evidence for assuming that IIclauses in this language have a subject and an object as the syntactic correlates, respectively, of the absolutive/​nominative on one side, and the ergative/​accusative on the other side. KatKan has grammatical relations in the sense of what, by the current accounts, Lakhota (Van Valin 1985), Cayuga (Mithun 1991a), Kannada and Manipuri (Bhat 1991), and other so-​called active languages have not.32 A corollary of confining subjecthood to the strict syntactic level is that, in dealing with ergativity, we need no (1)  special conceptual apparatus discarding the notion of subject (counter to Dryer 1986, but well in the spirit of Marantz 1984); (2) reduction of “subject” to the product of mere semantically driven phenomena (counter to Dixon's 1994 use of purpose clauses, reflexives and imperatives, and the consequent transfer of other would-​be subject properties to “pivot”); (3) splitting of subjecthood between a level of “argument structure” and a level of “grammatical structure,” a subspecies of (2) (counter to Manning 1996 and Aldridge 2008c);33 and, last but not least, (4) absorption of grammatical relations in a level defined in terms of case morphology or constituency/​government. Let me briefly comment on (4). We have in KatKan three types of complex phrases built on the following IIheads: noun, verb, and postposition (section 42.3). A fourth type of complex phrase is brought about by the IIverb in accusative clauses (section 42.5). Now, the internal argument of all four phrase types features a certain amount of heterogeneity in terms of coding properties. With a verb head in the ergative clause and a noun head in the NP, it takes the -​na= marker. With a postposition head it endures differential undermarking (see section 42.3 and example (77)). With a verb head in the accusative clause, it takes no overt case marker at all. Hence, morphological case and constituency 31  As in the constituency-​based definition of grammatical relations in the government and binding vintage of generative grammar (Chomsky 1981). 32  Mainly, should I add, those displaying rich morphology, since the critical factor superseding the need for this kind of syntactic hierarchy, and consequently for the level of structure where subject and object belong, might well be the amount of referent information carried by morphology at different points of the clause/​sentence, rather than active alignment per se (Mithun 1991a makes this point for Selayarese, not an active language). 33  From the latter: “the ergative NP is the argument structure subject [control, binding, imperative], and the absolutive NP is the grammatical structure subject [extraction, wide scope]” (brackets mine).

 

Katukina-Kanamari: a syntactically ergative language    1055 location are distinct notions and the generalizations attained by either level of structure, although they may intersect, are not superposable. Heterogeneity is also detectable at the level of syntactic behavior between the internal argument of IInouns, postpositions, and IIverbs in ergative clauses. Contrary to the ergative argument, the genitive and the argument of postposition do not need to be extracted to undergo pronominalization by a demonstrative, (76)–​(77). The genitive can be interrogated in situ, (78), whereas the argument of postposition obligatorily “pied-​ pipes” its head (i.e. the whole constituent moves) (79). None of these is available to the verb internal argument. The same applies to focalization, (80). (76) [ [itiyan-​na=] tyo] tona ProxDem-​MkCase= daughter GoAway ‘This one’s daughter went away.’ (77) yo-​toman wiri tya bo [ [itiyan] katu]! 1sing-​shoot WildPig prospective exclamation ProxDem CommInstr ‘I am going to shoot a wild pig with this!’ (78) [ [hanian-​na=] okpu] InterrPrn-​MkCase= offspring ‘Whose son is he?’

tu? interrogation

(79) [ [hanian-​na=] katu] tu adyi? InterrPrn-​MkCase= CommInstr interrogation come ‘With whom did he come?’ (80) [ [Dyano-​na=] ama] kana Dyano-​MkCase= benefactive focus ‘It’s for Dyano that Aro brought the hook.’

Aro-​na= dahu Aro-​MkCase= bring

pi:na hook

We will see in the next section how the IIverb internal argument accesses intransitivizing voice so as to have its participant realized by an expression that allows full accessibility. On the contrary, the internal argument of IInouns and postpositions is banned from any involvement in voice. Hence, constituency and grammatical relations are also distinct notions leading to distinct generalizations. Much the same can be said when correlating identically case-​marked NPs (-​na=) and their contrasting syntactic behavior, as also shown by the preceding examples. The notion of grammatical relation supplies a meaningful account for such mismatches between the constituency and case marking properties of arguments and their syntactic behaviors: among internal arguments, only the verb argument is an object; that is, a co-​argument in the clause domain, competing with the other co-​argument for certain formal properties.34 34 

The notion of “case” competition is in Bittner & Hale (1996b).

 

1056   Francesc Queixalós

42.7 Voice The specificity of voice phenomena as argument structure changing mechanism lies in the existence of the kind of syntactic hierarchy briefly outlined in the preceding paragraphs. In the basic clause, the verb entertains a privileged relation to a given argument, which, for this reason, we commonly identify as the subject. By way of mechanisms that more often than not verb morphology mirrors, this privileged relation is altered. Typically, some other semantic role35 takes on the subject relation. In sum, voice changes encroach with the basic subject-​verb relation.36 The more crucial formal correlates select the subject in the basic clause, the more the motivation for voice change will be formal in nature. This is the case of KatKan antipassive. Four devices converge in shifting from an ergative clause structure, (81), to an antipassive one:  (1)  the VP lacks an object; (2)  an invariable morpheme wa-​ preempts the person prefix slot that in basic verb morphology hosts the pronominalized agent argument; (3) the agent participant is promoted to subject; (4) the patient participant is either omitted, (82), or realized in a postpositionally marked adjunct position, (83).37 (81) [ [takara-​na=] biwi:k] hen-​MkCase= suck ‘The hen ate the papaya.’ (82) [wa-​pu] niama antip-​eat then ‘Then we ate.’38

[kapayo] papaya

[adi:k] 1plural

(83) [wa-​wu] [dyara] [tukuna anya-​na= antip-​want NonIndian HumanBeing woman-​MkCase= ‘White men are fond of Indian women.’

katu] CommInstr

Besides a slight propension to have a more or less unindividuated patient participant, the motivation for voice is formal: allow the argument expressing the agent to encompass the properties allotted to the subject. Crucially, those concerning accessibility to the syntactic prerogatives listed in section 42.4. For the sake of brevity, I will illustrate this with a small sample (cf. Queixalós 2010 for additional examples). 35 

Or no semantic role at all, cf. impersonal voices. Active languages (see footnote 32) are thus not expected to have voice, e.g. Lakhota (Van Valin 1985). A purely functional view of voice allows Mithun (2006) to identify in Central Pomo a “passive” in a verbal form blocking any mention to the agent participant. 37  In the other dialect, it remains unmarked. 38  This and the following examples are excerpted from a narrative where the speaker longs for old days’ style of life. 36 

 

Katukina-Kanamari: a syntactically ergative language    1057 (84) Focus Dapoma-tu kana wa-​man Dapoma-NEGATION Focus antip-​do/​make ‘It is not Dapoma who did it (speaking of a murder).’ (85)

Raising in auxiliary constructions bara      ayotu-​nin [  [ wa-​pu1-​nin=] bak] adi:k1 bo! WildMeat BePlentiful-​dep antip-​eat-​dep= BeGood 1plural exclamation ‘When wild meat is plentiful, we eat a lot!’

Concerning reference-​tracking, we have (1) at the complex sentence level, temporal subordinate Iclause subject ⇄ antipassive main clause subject in (86); and, parallelly, (2) over a stretch of discourse, Iclause subject ⇄ antipassive subsequent clause subject in (87). (86) nuk1 waokdyiwaokdyan-​nin wa-​ pu      Ø1 group ArrivePlural-​dep antip-​eat ‘As the mob1 arrived (home) they1 ate.’ (87) anyan1 niama tona tyo!... 3sing then GoAway exclamation ‘She1 then flew away (from jaguar)!...’ ...pakuruko niama wa-​da-​dokman Ø1 OilLamp then antip-​BeforeGoing-​light ‘... She1 then lit the oil lamp and went away.’ (88) Agent-​oriented nominalization wa-tokman nyan tukuna antip-cut deictic HumanBeing ‘The Indian is a rubber tapper (lit.: a cutter, see (74)).’ Nominalization is therefore another formal device showing that the lexical semantics of the IIverb is straightforwardly oriented toward the patient participant. In synthesis, voice phenomena reinforce the previous conclusions that (1) the basic, ergative clause is starkly hierarchized as to the properties of arguments in a number of syntactic domains, (2) its absolutive argument, expressing the patient, stands on top of this hierarchy, entitled as such to the status of subject. Voice is thus, as a direct consequence of (1) and (2), mainly motivated by the need to bestow this same properties to the formal correlate of the agent participant. Given that picture, and summarizing in a nutshell section 42.6, the ergative clause mappings between semantic roles, syntactic arguments, and grammatical relations may be set out as follows: (89) patient absolutive agent ergative

subject object

 

1058   Francesc Queixalós The accusative pattern is symmetrical to the ergative one in various aspects, but since it represents no basic active voice it has no intransitivizing voice alternation. The plausible motivations for a passive voice are complied by different and independent devices. As for formal motivations, a syntactically high patient argument or a low agent argument need not be derived: they are in the essence of the ergative pattern. As for functional motivations, the ergative clause provides a dereferentialized agent through the third plural person prefix: (90) ma-​dahu tyowipikon tyo! 3plural-​carry CollarBeads exclamation ‘They/​someone carried away the collar beads!’

42.8  Non-​Voice Alternations I will review here constructions involving causatives, applicatives, and noun incorporation in order to show that the inversion in KatKan’s way of projecting semantic roles onto formal entities, as compared to an accusative grammar, is consistently preserved in these areas of morphosyntax.39 Noun incorporation requires a more detailed account since, as announced in section 42.5, I will address here the basis on which we can distinguish an incorporated noun from an accusative argument. In causativization a participant not included in the semantic structure of the verb enters the scene as the initiator of the manner of existing described in the clause. It is realized as an ergative argument, internal to the VP. IVerbs take a suffix -​tiki giving way to a monoclausal construction. The verb nobuk/​ba:bu, ‘send, order,’ yields an indirect causation bi-​clausal sequence40 where the basic lexical Iverb/​IIverb becomes dependent of the cause verb. The verb man, ‘do, make, say, etc.’ behaves either as -​tiki, that is, a suffix, or as nobuk/​ba:bu, auxiliary. In order to save space, I will focus on -​tiki causation (instances of two-​clause causation appear in (52)–​(53); their properties do not impinge on the issue at stake). The causee retains the absolutive position. (91) horon barahai burn WildMeat ‘The meat burned.’ (92) [a-​obatyawa-​na= horon-​tiki] 3sing-​wife-​MkCase= burn-​causative ‘His wife burnt the meat.’ 39 

barahai WildMeat

As in many other strongly ergative languages (e.g. Dyirbal and Macuxi, Dixon 1994: 138), the reflexive derives from mere unaccusativization. Reciprocal in KatKan is a subspecies of reflexive. 40  With (yet unaccounted) exceptions.

 

Katukina-Kanamari: a syntactically ergative language    1059 Applicative preverbs allow a semantically peripheral participant to surface as the absolutive argument of an ergative clause. Some preverbs are clearly related to postpositions, compare (93)/​(80). IVerbs turn divalent but IIverbs preserve their original valence since no trivalent construction, neither primary nor derived, exists in the language. The participant occupying the absolutive position in the basic IIclause surfaces as a postpositional adjunct. (93)

yo-​ama-​wandoki 1sing-Applicative-​cook ‘I’ll cook the fish for you.’

idi:k 2sing

don katu fish CommInstr

wa future

Noun incorporation pertains to the redistributive type:  the verb keeps its valence unchanged but semantic participants are reallocated in argument/​non-​argument positions. Functionally, it accomplishes the task of promoting to verb argument realization (absolutive) a semantically or pragmatically salient participant that in the non-​derived construction occurs as the internal argument of a non-​verb head (i.e. IInoun/​postposition). Only IIverbs in ergative clauses together with unaccusative verbs incorporate. And only the lexical head of an absolutive NP is eligible for incorporation. Moreover, the inherent valence of the incorporated noun determines the form of the incorporation: (1) direct for IInouns, (94), or (2) mediated by the applicative for Inouns, (95) (compare (93)). (94) a tuktu yo-​baki shiver 1sing-​hand ‘My hand shivers.’ b baki-​tuktu adu hand-​shiver 1sing ‘idem (lit.: I hand-​shiver)’ c a-​hi:k yo-​wadik 3sing-​see/​look 1sing-​name ‘He knows my name.’ d a-​wadik-​hi:k adu 3sing-​name-​see/​look 1sing ‘idem (lit.: he name-​knows me)’ (95)

Dyomi-​na= ama-​amatyuru-​man-​na Dyomi-​MkCase= Applicative-​FishSp.-​do/​make-​centrf ‘Dyomi went getting some fish sp. for me.’

adu 1sing

Both types of incorporation are subsumed in a single syntactic mechanism whereby a head (IInoun in direct incorporation and postposition in applicatively mediated) joins

II

 

1060   Francesc Queixalós the verb in a new verb complex to which it brings its own internal argument as the absolutive argument. The original participant hosted in absolutive position gets its expression incorporated by itself (IInoun), or together with the postposition (Inoun). As an effect, valence remains unaltered. A handful of frozen direct incorporations of Inouns, that is, non-​mediated by the applicative, involve an unindividuated nominal notion and some semantically light II verb, e.g. don-​man, fish-​do >‘to fish,’ bara-​man, WildMeat-​do >‘to hunt.’ Excursus. It is rather usual for incorporees to denote unindividuated notions and become syntactically inert.41 Of course, this sounds much like what was said earlier on the accusative argument, and thus raises the issue of the foregoing account for the accusative clause (section 42.5) being recast as one for noun incorporation. Starting from the ergative clause: unindividuated patients do not surface as absolutive arguments but as incorporated nouns; hence, the vacuous absolutive argument position may host the agent participant and the verb loses its internal argument position, becoming thus monovalent; the absolutive agent accesses the syntactic properties of the sole argument of Iverbs. We would, thus, have here the productive counterpart of the direct Inoun incorporation mentioned in the previous paragraph. In my opinion, the accusative clause hypothesis has some advantages over the direct productive Inoun incorporation hypothesis. I will resume one previous example before setting up the discussion. (96) [ [ pi:da] ohiya] [Hi:wuk]   jaguar fear Hi:wuk   ‘Hi:wuk fears jaguars.’ Admittedly, (96) is no prototypical transitive clause in the sense of Hopper & Thompson (1980) scalar account of transitivity based on multifactor semantic and pragmatic dimensions. Which favors at first sight the direct productive Inoun incorporation hypothesis. However, we would expect some amount of homogeneity between the different types of Inoun incorporation, as we detected homogeneity among the uncontroversial IIhead incorporations. This is clearly not the case. First, the semantics of the incorporee in the direct productive Inoun incorporation (my accusative argument) is in no way classificatory as it is in the frozen type; neither are the verbs involved in the putative productive type of incorporation akin to semantically light verbs (see examples (95) and following). The direct productive Inoun incorporation hypothesis yields three highly heterogenous sorts of Inoun incorporation. Second, another factor of heterogeneity brought about by the direct productive I noun incorporation hypothesis lies in one functional aspect of IIhead incorporations. In these two incorporation types, both participants (the incorporee and the promoted)

41 

Possibly not in Mohawk (Baker 1996: 307ff.) or Chuktchee (Spencer 1995), among others.

 

Katukina-Kanamari: a syntactically ergative language    1061 entertain some kind of mutual semantic link independently from the manner of existing described by the verb (roughly, actual or potential “possessee”/​“possessor”). In what I label accusative clause, the link is a strict product of the manner of existing whereby any transitive verb connects its two participants. Third, semantic animacy restrictions on incorporated nouns apply even in languages reputed to feature syntactically active incorporees (see n.  41). The direct productive I noun incorporation hypothesis would generate a number of KatKan counter-​examples such as (64) and (65), renumbered here. The accusative clause hypotheses only imposes an unindividuated internal argument. (97) [ [tukuna] makoniok] HumanBeing advise ‘He1 advised the humans.’

Ø1

(98) pi:da [ [ityaro] botyana] jaguar woman track ‘The jaguar tracks the women.’ So far I have, on semantic considerations, sustained my preference for the accusative clause hypothesis. There are also formal grounds for doing so. (1) Resuming homogeneity concerns, on the one hand the putative productive incorporation of (96) is starkly unlike the applicatively mediated incorporation as we saw in this chapter. On the other hand, the frozen incorporations cannot be seen as mere lexicalizations of the productive type, since the latter would lend an intransitive clause whereas the former does preserve transitivity, as in (99). (99) Dyomi-​na= don-​man-​na Dyomi-​MkCase= fish-​do/​make-​centrf ‘Dyomi fished a fish sp. over there.’

amatyuru FishSp.

(2) In section 42.4 we reviewed auxiliary constructions, for instance (40) (renumbered): (100) [ [kodyi-​nin=] wu] bathe-​dep= want ‘I want to bathe.’

adu 1sing

The accusative clause supplies an independently motivated pattern of constituency for such constructions: compare VPs (bracketed) in (100) and (97). Interestingly enough, the parallel between both internal arguments (unindividuated noun in (97), lexical verb in (100)) suggests that the kind of dependence uncovered by the nin= suffix in auxiliary constructions should prevail in the accusative clause. Which points to the argumenthood of the preverbal noun in (97). The direct productive Inoun incorporation hypothesis introduces unnecessary formal prolixity.

 

1062   Francesc Queixalós (3) Because of its promoting effect, intransitivizing noun incorporation is sometimes endowed with antipassive function, as in the Maya family (Grinevald & Peake 2012). For sure, the direct productive Inoun incorporation hypothesis for KatKan lends itself rather naturally to an antipassive reading. Notwithstanding, promoting the agent participant is already the basic task of wa-​verb antipassive (section 42.7). This somehow weakens the direct productive Inoun incorporation hypothesis. However, it also opens the issue of the function allotted to the accusative clause. Obviously, such a pattern affords the speaker with a means to express unindividuated patients. But in my opinion, the essence of the accusative clause lies somewhere else.42 We are led to the last justification (which I also take to be the more important in spite of its speculative character) for the existence of the accusative clause in this language. The lack of comparative evidence makes it difficult to figure out what may have been the patterns of alignement in the language’s past grammatical architecture. Nevertheless, several clues point to a nominalized verb origin for the ergative clause. Identical genitive and ergative nominal cases; identical person prefixes on nouns and verbs; identical constituency between IINPs and IIVPs; and: a verbal suffix -​nin= occurring in some lexicalized patient-​oriented nominalizations (in one dialect; a bit more productively in the other),43 as well as in dependent verbs (see above, particularly adverbial subordinate clauses introduced by postpositions such as in (55) and (56)) and, finally, as finite verb durative aspect, e.g. (24). Somewhere in the history of the language ‘Hi:wuk calls Mayon’ ends in a nominal predication Mayon (is) Hi:wuk’s called one, in which the argument position filled by the patient participant captures the syntactic privileges attached to subjects.44 The accusative clause in the foregoing discussion would thus be a relic of the basic clause in pre-​ergative times, preserving constituency and the syntactic prominence of the argument expressing the agent, but left with mere unindividuated objects because unindividuated notions are poor topics and hence poor candidates to

42 

Hence the antipassive cline to capturing unindividuated patients. Such a functional overlap may reveal a zone of diachronic instability. 43 Kanamari: kirian-​nin, ‘the thunder,’ from kirian, ‘to thunder.’ Katukina: donman-​na tyo-​tiok-​nin ToFish-​centrf 1plural-​know-​nominaliser ‘the one we know went fishing over there (lit.: our known one went...)’ The nyan nominalizations (section 42.4) are from Kanamari. Katukina has created totally different forms. The fact that two so closely related variants have each generated its own nominalization mechanism independently of the other (1) suggests that the advent of ergativity is recent in diachronic time; (2) confirms the nominalized etymon for today’s basic finite verb. 44  By default: the pragmatic motivation for such a move would rather be agent backgrounding than patient focusing. Queixalós (2013) is an attempt (in an admittedly speculative fashion) at getting, out of the purported conundrum of ergativity, an intelligible object of thought in functional terms. Two universal but antinomic pressures would be simultaneously at work in the framing of IIclause grammar: the cognitive prominence of the agent participant, and the communicational recurrent need to downgrade it. A marked/​unstable grammatical architecture arises when (and where, for there certainly are social conditions interfering) the latter factor eventually takes the upper hand.

 

Katukina-Kanamari: a syntactically ergative language    1063 grammaticalization as subjects in the supervening pattern. Would this scenario prove appropriate, the idea that the accusative clause is no synchronically derived construction would gain some strength. End of excursus. Back to non-​voice alternations, taking the mappings in (89) as a correct account of facts, we get, prima facie, a rather startling picture when portraying the argument structure changes by means of a direct correlation between hosting positions and grammatical relations. To wit: (1) In applicatives, the promoted/​ new participant (beneficiary, etc.) occurs as subject. (2) In causatives, the new participant (causer) occurs as object. (3) In incorporation, a subject, including that of a IIverb, has the expression of its participant incorporated. On the contrary, the projection of hosting positions directly onto semantic roles takes on a different complexion, certainly more sound, and encompasses cross-​linguistically the ergative as well as the accusative pattern: (1) In applicatives, the promoted/​new participant (beneficiary, etc.) occurs in a non-​ agent position. (2) In causatives, the new participant (causer) occurs in an agent position. (3) In incorporation, the expression of a non-​agent is incorporated. I submit that quite apparently different syntaxes (critically, ergative and accusative ones, see Queixalós 2014) are at once in requiring that generalizations as to argument structure variations like the ones under scrutiny (applicative, causative, incorporation) be cast in semantic role terms (no matter how prototypical or abstract, e.g. “agent/​non-​ agent”) instead of formal properties of expressions (absolutive, object, etc.).45

42.9  Final Remarks In describing the ergative morphosyntax of this language, my aim is to suggest that among the levels at which languages structure what Seiler (1989) labelled “participation” (in my phrasing, formal and functional correlates of entities involved in the manners of existing described by predicates), grammatical relations belong to a level of their own. Notions of subject and object need to be defined so as to capture generalizations that 45  This seems to be a plausible way out of Baker’s (1988) ban on the actual existence of an ergative syntax, based on the following (here simplified) line of reasoning: (1) incorporation singles out the internal argument; (2) in an ergative syntax the argument expressing the agent must be internal; and (3) no incorporation of the agent expression is cross-​linguistically attested.

 

1064   Francesc Queixalós other levels of structure do not.46 Basically, properties beyond semantic role and topicality, but also beyond coding and constituency. As for the latter, voice alternation, for instance, shows that object in KatKan is something different from internal. The appropriate domain for articulating such generalizations is the differential accessibility to those properties that syntax imposes on arguments (cf. Anderson 1976; and Keenan’s 1976 “behavior and control,” among others). Once this requirement is fully acknowledged, we will observe languages where grammatical relations play no role, languages with more or less diffuse/​mixed assignation of such properties to arguments, and languages endowed with a clearly distinct, rather homogenous, level of grammatical relations. KatKan and, say, English, are instances of languages possessing grammatical relations. In so far as these considerations go, the mapping of a functional level (agent, topic) on a formal level is, grossly, not relevant for identifying subjects and objects. For this reason, and again in Marantz’s (1984) spirit if not letter, an argument expressing the agent may perfectly be an object in the basic transitive clause of a given language. That said, behind the synchronic mechanics stand the diachronic scenarios that depict the paths through which things got to be as they are. History is no explanation per se, but it is an unavoidable part of it. This is a cardinal matter regarding ergativity with all its splits and other at first sight tortuous inconsistencies in the formal treatment synchronic grammar gives to participants. But this is also where the functional levels make full sense. Participants (or semantic roles) are, mutually, in no equal footing regarding their cognitive import and, beyond, their projection on formal categories. Neither are the referents manipulated by speakers. These asymmetries (Foley & Van Valin 1984; Shibatani 1991; Dixon 1994; Givón 2001a) exert competing (even antinomic) pressures on the formal make up of grammar. Their outcome is rendered more intelligible by diachrony just because ergativity in a particular language is but a synchronic heterogeneous slice in this continuous flow of interconnected adjustments.

Abbreviations abs, absolutive; antip, antipassive; centrf, centrifugal motion; centrp, centripetal motion; CommInstr, commitative–​instrumental; dep, dependence; erg, ergative; grn, generic relational noun; InterrPrn, interrogative pronoun; MkCase, marked case; np, noun phrase; ProxDem, proximal demonstrative; sing, singular; vp, verb phrase

46 

An extension of the neutralization condition to which Van Valin & Lapolla (1997: 274) submit the relationship between semantic roles and grammatical relations.

 

Chapter 43

Ergativit y i n J ê l anguag e s Andrés PABLO Salanova

The Jê language family is one of the major language families of indigenous languages in lowland South America. It consists of approximately ten living languages spoken in a vast area in the eastern half of Brazil, from near the mouth of the Amazon to the extreme south of the country.1 The language family, whose depth was calculated at 54mc by Swadesh (cited in Kaufman 1990), is quite diverse in many respects, but there is remarkable structural similarity between all languages of the family. Ergativity in particular is a relatively homogeneous phenomenon in Jê languages. The basic facts of Jê ergativity were first described qua ergativity by Urban (1985) in a language of the southern branch, and by dos Santos (1997) in a language of the northern branch. In the central branch ergativity has not been explicitly addressed, but sufficient information can be gathered from the recent description found in Estevam (2011). In

1 

The languages (followed in parentheses by their ISO 639-​3 codes) are Mẽbengokre (txu), Apinayé (api), Timbira (gvp, ram, xra, xre, xri), Kĩsêdjê (suy) and Panará (kre) in the northern branch, Xavante (xav) and Xerente (xer) in the central branch, and Kaingang (kgp, zkp) and Xokleng (xok) in the southern branch. Kaingang itself is a dialect continuum approaching various distinct languages, whereas the pairs consisting of Mẽbengokre and Apinayé, on the one hand, and Xavante and Xerente on the other might in fact each be a single language with two mutually intelligible dialects. Timbira is normally taken to be a dialect continuum, but proximity between variants is greater than normally assumed, and no intelligibility barriers seem to exist. Related to Kĩsêdjê is the nearly extinct and almost completely undocumented Tapayúna, whose proximity to Kĩsêdjê is not clear. Beyond Jê proper, several languages or small families in south-​central and eastern Brazil and eastern Bolivia are linked to Jê to form a Macro-​Jê stock. These include Maxakali (mbl), Karajá (kpj), Jabutian (jbt, ark), Bororoan (bor, otu, umo), Rikbaktsa (rkb), and Chiquitano (cax), for which the link is plausible, and more tentatively Fulniô (fun), the moribund Krenak (kqq), Guató (gta) and Ofayé (opy), and the extinct Karirian (kzw), Purian (xxr, prr), Kamakã (vkm) and Jeikó (no code), of which all but the first are known solely through short wordlists. For classification, see Campbell (2012).

 

1066   Andrés PABLO Salanova this chapter we do not introduce any new facts, but rather attempt a new synthesis of what is known. The facts can be summarized as follows: 1. The verb in all Jê languages has two basic stems. In much of the literature, these are called the short and long forms, but other names exist. The latter is nominal-​ adjectival, while the former is finite verbal. 2. The long forms introduce ergative traits in the clause. Typically, verbs in the long form inflect for absolutive rather than accusative person (as short form verbs do), using a series of markers that is only slightly different from accusative, and ergative free pronouns are used with the long form, rather than nominative free pronouns. Ergative has some characteristics of an inherent case, and has the particularity that it can be dropped and occur superfluously, when a subject is already indicated by another type of pronoun or person marker. 3. The northern Jê language Panará, in addition to ergativity in noun phrases that broadly resembles that found elsewhere in the family, exhibits an ergative-​ accusative split in person clitics based on a realis-​irrealis opposition that is independent from the long versus short opposition in verbs. This adds an extra layer of complexity not found in other Jê languages. 4. In virtually all languages of the family, only clauses with long-​form verbs can be embedded. Embedded clauses are straightforwardly ergative. 5. Various constructions exist where a long-​form verb is used in an independent clause. All such constructions have ergative traits, sometimes mixed with accusative traits. 6. Independent clauses with short-​form verbs are straightforwardly nominative-​accusative. 7. Intransitive predicates fall into two classes: one truly verbal, contrasting long and short forms and following the split pattern described earlier, and one that is invariably nominal, always inflecting for its subject with absolutive person markers. Transitive predicates with noncanonical subjects (i.e. dative) are also invariably nominal, and mark their object in the absolutive rather than in the accusative. 8. Absolutive markers are morphologically identical to the genitive markers used to mark possessors of inalienably possessed nouns. In some sense, then, Jê languages could be seen to pattern like many better-​described languages where ergativity is associated with nominalizations (cf. Alexiadou, Chapter 15, in this volume). However, like in other language families of the Americas and elsewhere, “nominalizations” have such a wide array of uses that ergativity itself is pervasive in all types of clauses. The chapter is organized around the descriptive points as enumerated. Section 43.1 describes the basic morphosyntax of Jê languages, and introduces ergativity. Section 43.2 describes ergativity in embedded clauses. Section 43.3 describes straightforwardly ergative independent clauses. Section 43.4 describes independent clauses that are straightforwardly accusative. Section 43.5 describes independent clauses with mixed ergative-​accusative traits. Section 43.6 gives a full sketch of ergativity phenomena in one northern Jê language, showing how all of these traits come together. Section 43.7 concludes by raising some general points of interest for a theory of ergativity.

 

Ergativity in Jê languages    1067

43.1  The Morphosyntax of Jê Languages Jê languages are head final, mostly head-​marking, and essentially analytic. Some deviations from strict head finality might be seen in a few languages (e.g. Panará, Xokleng) but only in finite independent clauses. Some synthesis can likewise be found in finite independent clauses in a cluster of temporal, aspectual, and modal particles in the left periphery of the clause, which also often include person markers for a high participant, but lexical heads are not normally marked for more than one of their dependents in any language, with the possible exception of Panará, where clitics for participants and TAM particles are described as leaning to the right onto the lexical verb, and Xavante, where transitive verbs with second person subjects agree with both subject and object with distinct morphemes, according to Estevam (2011). The position of direct objects is typically immediately before the verb, while oblique participants (whether selected or adjoined) and verbal modifiers appear after the subject and before the object. Fronting of the object is common, and is accompanied by the presence of a person index for the object on the verb, which is absent when the object is contiguous to the verb. Other constituents of the clause may also be fronted. The postverbal position may be occupied by verbal modifiers, some of which govern the long form of the lexical verb and consequently affect the alignment of the clause. In Arregui et al. (2014) we describe the postverbal modifiers of Mẽbengokre as the proper auxiliaries of the language, while we take the various modifiers that appear in the left periphery of the clause to be adverbial particles. We believe that this is essentially correct for the whole language family. The following examples, taken from each of the three branches of the family, illustrate the basic structure of independent clauses: (1)

Xavante (central Jê; Estevam 2011, p. 445) Taha na warazu ma da-​ãma aj’uté za’ra re. that with white_​person pfv 3-​prv defeat pl dpr ‘It’s with that (trick) that the white person defeated the (Xavante) people.’

(2)

Panará (northern Jê; Dourado 2001, p. 115) ĩkjẽ ti= rã= sari nãkã I 3sg.erg= 1sg.abs= bite.perf snake ‘I was bit by the snake.’

(3)

Mẽbengokre (northern Jê; author’s data) Kàx o nẽ ba arỳm tep nhirênh o=nhỹ knife with nfut 1nom already fish cut obl=sit.sg.v ‘It’s with a knife that I’m (sitting) cutting up the fish.’

hẽ erg

 

1068   Andrés PABLO Salanova (4)

Xokleng (southern Jê; Urban 1985) konhgàg tẽ wũ ti penũ man det 3nom he shoot ‘The man shot him.’

mũ act

In the preceding examples, the first phrase to occur in the clause is a highlighted element which may or may not be the subject: it is an oblique in (1) and (3), the direct object in (2), and the subject in (4). However, a strict SOV order occurs after this highlighted initial position, and in (2) and (4) one can even see how the highlighted constituent is taken up again in situ by a pronominal clitic. Other than in the Panará example, where the ergative subject noun phrase appears postverbally (possibly another mode of highlighting), all that one sees after the verb in these examples are aspectual or other modifiers: a plural and depreciative scoping over the object in Xavante, an “active aspect” in Xokleng, and a progressive involving the postural verb nhỹ (‘sit’) in Mẽbengokre. Jê languages typically have two main series of person markers, one nominative and one used elsewhere (herein called absolutive, following our nomenclature for Mẽbengokre). Absolutive person markers are used on their own as absolutive or accusative verbal prefixes (though accusative is distinct from absolutive in the third person in most northern and central languages), as prefixal possessors for inalienably possessed nouns, and as prefixal complements of postpositions, including the postposition used to mark ergative, when analyzable. Nominative person markers are free pronouns, and are referred to as nominative pronouns herein. The following tables show the two main series in Mẽbengokre, Xavante and Xokleng, and the three series of Panará. The exact status of the nominative series is not always clearly articulated in the available descriptions.2

2  The equivalents of what I call nominative pronouns in Mẽbengokre are called “person markers” in Xokleng by Urban (1985). In Xavante, the series of nominatives is highly impoverished, and is called by Estevam (2011) “person indices” in conjunction with a set of postverbal modifiers used for indefinite and non-​referential subjects. The nominative series of Xokleng and Xavante may cooccur with “emphatic pronouns” standing for the subject, which are formed with the absolutive person markers and an emphatic marker. It might be possible to analyze nominative pronouns as inflected auxiliaries, but other than a quick reference in our own work, we do not know of any concrete proposal in this direction, though see in Wiesemann (1986a) the discussion of Xerente nominative markers, which are fused with modal elements. In the tables, when plurals are not given it is because they are indicated by separate particles which do not fuse with the person markers. The form given for third person pronouns is the least marked, used for non-​deictic third person reference. Xavante is unique in the family in opposing honorific and unmarked forms in the second person. Honorific forms are excluded in this abbreviated presentation. Xokleng 3sg.m requires tã before wũ. It might therefore be more correct to say that 3sg.m.nom is tãwũ.

 

Ergativity in Jê languages    1069 (5)

Xokleng (from Urban 1985) Absolutive Nominative 1sg ẽg nũ 1pl ãg nã 2 a mã 3sg.m ti wũ 3sg.f di wũ 3pl og wũ

(6) Mẽbengokre Absolutive 1sg i-​ 12sg ba-​ 2sg a-​ 3sg ku-​/∅ ​

Nominative ba gu ga ∅

(7)

Xavante (from Estevam 2011) Absolutive Nominative 1sg ĩĩ-​ wa 1pl wa-​ wa 2 ai-​ te 3 ti-​/∅ ​ te

(8)

Panará (B. Bardagil-​Mas, p.c. 12/​2014) Absolutive Nominative 1sg ra= ∅ 2sg a= ti 3sg ∅ ti

Ergative rê ka ti

Third person absolutive forms involve some rather complicated nonconcatenative exponency that is examined comparatively in Salanova (2011a). There is no overt case morphology for noun phrases in direct cases. Absolutive NPs are normally in complementary distribution with third person absolutive pronominal prefixes. Only northern and central Jê languages seem to distinguish accusative from absolutive in the third person pronominal prefixes. Accusative prefixes (third person ku-​in Mẽbengokre, and ti-​in Xavante, given in (6) and (7)) are used with a subset of transitive verbs in clause types that have nominative-​accusative alignment, and are also governed by some postpositions. In Xavante, furthermore, a handful of intransitive verbs lexically assign accusative to their subjects. Absolutive prefixes (third person ∅ in Mẽbengokre and Xavante, given in (6) and (7)) are used elsewhere.

 

1070   Andrés PABLO Salanova Overt case morphology is also normally absent in noun phrases in the nominative, though the position of nominative noun phrases is clearly distinct from absolutive/​accusative, as can be ascertained by the relative position of TAM markers and other modifiers of the proposition. Markers that are sometimes interpreted as nominative enclitics in analyses of individual languages are actually emphatic markers, TAM particles that are obligatory in post-​subject position, or a third person “person marker” or nominative pronoun that repeats a subject that is highlighted through fronting. Oblique cases, which include ergative, are marked with postpositions that govern absolutive or accusative (in northern Jê languages) person markers. In the case of the ergative, the dative and a specific type of locative, all of which are used to express subjects in certain types of clauses, there is some degree of fusion with the absolutive person in a few languages, notably Apinayé (cf. Oliveira 2005) and Xerente, without becoming opaque. Person hierarchy effects are virtually absent in Jê languages, contrary to what happens in many language families in the South American lowlands. Thus, the semantics of the noun phrase participants does not affect alignment in the family (pace dos Santos 1997). Other than in Panará, the main trait governing alignment in the clause is whether the main predicate is nominal or verbal. Some predicate types (e.g. most stative predicates) are generally nominal in Jê languages, and have their own invariable ergative-​like alignment. This is described in detail in section 43.6, though only for Mẽbengokre. True verbs, with which we are concerned here, have both a finite verbal form (the short form), and a nominal-​adjectival form (the long form). Exponency of this opposition is irregular, even though it is mostly realized by the suffixation of a single consonant.3 In our glosses we do not segment these exponents, even when they are in principle segmentable, and simply indicate nominal-​adjectival forms as .n, and finite verbal forms as .v. Nominal-​adjectival forms of verbs are related to ergative alignment in a not completely straightforward way, while finite verbal forms are strictly associated to nominative-​ accusative alignment. The association will become clear as we develop our description in what follows. Also relevant to the topic of ergativity, though only tangentially, is the opposition between singular and plural forms of verbs. This has irregular exponency, being manifested through reduplication of the root, suppletion of the root or stem, or changes in the prefixes that indicate valency. Typically, plural verbs “agree” with a plural absolutive argument, though they can also serve to describe events with repeated action, even when all participants are singular. In complex constructions involving a lexical verb and an intransitive postural verb functioning as an auxiliary (see the discussion of the progressive in section 43.6), number marking on each of the verbs might agree with a different participant. 3  In fact, the direction of the derivation is not clear. While nominal-​adjectival forms are normally larger than the finite verbal forms, and would appear to have received a consonantal suffix, the extra consonant is not predictable, suggesting rather that finite verbal forms are derived through apocope.

 

Ergativity in Jê languages    1071 I round out the description of Jê morphosyntax with a short discussion of valence-​changing morphology, even though this is of relatively little import for the discussion of ergativity. Verb stems in Jê languages often consist of a root plus a prefix indicating transitivity, but these prefixes are not productive morphology in any of the contemporary Jê languages. No synthetic valency-​increasing mechanisms exist. Two classes of intransitivizing prefixes exist in a few Jê languages, antipassive, and anticausative. These are true intransitivizers rather than, e.g. dummy objects, as they make the subjects of the verbs to which they attach to be absolutive in clauses that have ergative-​absolutive alignment. Prefixal reflexive and reciprocal inflection, on the other hand, though apparently occupying the same morphological slot, do not intransitivize the verb to which they attach, and should be properly thought of as part of the paradigm of person prefixes. No other important derivational processes exist.

43.2  Ergativity in Embedded Clauses To our knowledge, plain ergativity is universal in embedded clauses with nominal-​ adjectival forms of verbs in Jê languages. The following are examples from the three branches of the family (subordinate clauses are in brackets):    (9) Xokleng (­Urban 1985, exs. 59, 60) a. [ti tawi kũ] mã ti weg tẽ he arrive.sg.n when 2nom he see.v imp “When he arrives, you are going to see him.” b. [ẽ tõ ujol tãñ kolkũ] tã tawig tẽ coref erg tapir kill.n after he arrive.sg.v imp “After he kills the tapir, he is going to arrive.” (10)

Xavante (from Estevam 2011, pp. 337, 311) a. [te-​te sazu ñerẽ] te za a-​wajbu 3-​erg 3.burn.n though 3nom fut 3coll-​gather ‘Though they burn them, they will gather them.’ b. [aj-​wara wamhã], ... 2-​travel.n when ‘when you travel, ...’

(11)

mẽbengokre a. ba [kute tep krẽn] pumũ 1nom 3erg fish eat.n see.v ‘I saw him eating fish.’

 

1072   Andrés PABLO Salanova b. [i-​nhõt mỳrỳri] nẽ im 1-​sleep.n while nfut 1dat ‘He spoke to me while I slept.’

kabẽn 3.speak

That subordinate clauses are systematically ergative in Jê languages follows from the fact that subordination in the family normally embeds nominal forms. However, one does find in several Jê languages sequences of clauses which involve some sort of semantic subordination, as in the following Mẽbengokre example. These maintain a nominative-​ accusative alignment. (12) ba a-​mã ku-​ngã ga 1nom 2-​dat 3acc-​give.v 2nom ‘I’ll give it to you so that you can see it.’

omũ 3.see.v

In the cases we have examined more carefully, such constructions are still manifestly paratactic or coordinated in synchronic analyses of the languages. We do not exclude the possibility that some Jê languages might have developed subordination of nominative-​ accusative finite clauses out of such paratactic constructions.

43.3  Straight Ergativity in Independent Clauses As we will see, independent clauses with finite verb forms have nominative-​accusative alignment. Before we go into such cases, however, we examine some cases where a nominal-​adjectival form of verbs can be used in independent clauses, with full ergative alignment, as in the following examples (note the “redundant” or emphatic use of nominative in Xavante): (13) Timbira (from Alves 2004) a. kahãj te ih-​pỳn woman erg 3-​carry.n ‘The woman carried him.’ b. a-​wrỳk 2-​descend.n ‘You got down.’ (14) Xokleng (from Urban 1985) a. ti tẽg wã he go.n stat ‘He went.’

 

Ergativity in Jê languages    1073 b. ti tõ ti penũ wã he erg he shoot stat ‘He shot him.’ (15)

Xavante (from Estevam 2011, pp. 448, 305) a. duréjhã si’õno wa wa-​te   ñamri za’ra before basket 1nom 1-​erg 3.weave pl ‘In ancient times we used to weave baskets.’ b. ĩĩ-​wara õ di za 1-​travel neg expl prosp “I’m not going to travel.”

(16)

Panará (from Dourado 2001) a. ĩprĩ ti= sari nãka hẽ boy 3sg.erg 3sg.bite.n snake erg ‘The snake bit the boy.’ b. kamera jy how= ria= kõri ĩkô how you real with 2pl   drink.n water with ‘You drank water.’ (lit., ‘with water’)

In examples (14) and (15) we see that a stative particle or another postverbal modifier follows the nominal-​adjectival form of verb. In the remaining examples, the nominal-​ adjectival form of the verb is outright “insubordinated”:4 the nominal-​adjectival form of the verb, normally reserved for subordinate contexts, appears in an independent clause without any modification. Among the postverbal modifiers that require a nominal form of the verb and that trigger pure ergativity are: Xokleng The stative aspect markers wa and ko, and possibly others; cf. Urban (1985), p. 173: “the system of verbal suffixes is extremely complex and need not be described in detail here.” Kaingang For Wiesemann (1972), the nominal form is simply a variant of the verbal form that is required whenever cooccurring with any aspect marker. Wiesemann (1986a) states more specifically that it is used with negation tũ, with the adverbial e ‘much’, and with narrative (mũ), descriptive (nĩ) and procedural (tĩ) aspect particles.

4  The term insubordination was coined by Evans (2007) to describe situations in which a verb form or clause type that normally occurs in subordinate contexts is used as an independent clause with a particular marked value (normally related to illocutionary force or information structure, but also other things).

 

1074   Andrés PABLO Salanova Xavante/​Xerente Negation (õ), intensifiers, diminutives, and several other postverbal modifiers. Mẽbengokre/​Apinayé/​Kĩsêdjê Negation (kêt), three types of prospective aspect (’ỳr, kadjy, mã), certain manner modifiers (mex ‘good’, pyro ‘ready’) and aspectual modifiers (rã’ã ‘still’). In the cases where a nominal-​adjectival form of a verb is used in an independent clause without some sort of postverbal operator, this results in a particular marked value for aspect, with meanings ranging from generic and habitual, through resultative and existential perfect, to recent past, when these meanings have been described. This is how it has been described in those languages of the family where it occurs: Xavante Aorist, described by Estevam (p.c., 3/​2015) as describing an event “not linked to utterance time.” Timbira Immediate past (Popjes and Popjes 1986; Alves 2004). Mẽbengokre Resultatives for verbs that involve a change of state; existential perfects; habituals or generics for verbs that denote plural activities (Salanova 2007). Apinayé Insubordinated nominal forms are claimed to be vanishingly rare except for the verbs kuma ‘to hear’ and omũ ‘to see’, which become mar ‘to know by having heard’ and omũnh ‘to know by having seen’ (Oliveira 2005) Panará Perfect (Dourado 2001). This form appears to be more frequent in independent clauses in Panará than in other languages of the family. The existence of such “insubordinated” nominal forms has prompted authors in the functionalist tradition to consider that nominal-​adjectival forms of verbs have been reanalyzed as finite forms, taking on particular aspectual meanings (Gildea 2008). While this is plausible in principle, there is no reason to believe that it is happening in contemporary Jê languages: nominal clauses abound in these languages, always with an existential meaning. A clause headed by a nominal-​adjectival form of a verb could still be a nominal clause synchronically, as long as it is possible to interpret it as an existential clause. This analysis cannot be developed here for reasons of space, but see Salanova (2007).

43.4  Nominative-​Accusative Independent Clauses Nominative-​ accusative is the unmarked alignment in independent clauses in all branches of the family, though in Panará ergativity in independent clauses is more prevalent than in the remaining languages. Note that other than in a few northern and

 

Ergativity in Jê languages    1075 central Jê languages (Mẽbengokre, Apinayé, Kĩsêdjê, Xavante), accusative prefixes are not different from absolutive. (17)

Mẽbengokre a. ba arỳm ku-​ma 1nom already 3acc-​hear.v ‘I heard it already.’ b. ba arỳm mã tẽ 1nom already away go.v ‘I’m already going.’

(18)

Xokleng a. konhngàngi tẽ wũi ti penũ mũ man def 3nom he shoot.v act ‘The man shot him.’ b. tã wũ he 3nom ‘He went.’

(19)

tẽ go.v

mũ act

Xavante (Estevam 2011: 183) a. Alexandre u wa Alexandre to 1nom ‘I’m going to Alexandre’s.’ b. wa za 1nom fut ‘I will grab it.’

mo go.v

ti-​ö 3acc-​grab.v

Verbs in nominative-​accusative clauses are generally in their finite verbal form, which means that they are not subordinated to any other predicate. Given this, it might seem surprising to see that they are followed by what looks like an auxiliary in Xokleng, i.e. the active marker mõ. In fact, there is a whole paradigm of “active” aspectual markers in Xokleng (and in Kaingang) that is parallel to the progressive paradigm of Mẽbengokre (i.e. various verbs of posture and of movement, holding a selectional relation with the subject), which we describe in what follows and treat as clearly subordinating a nominal-​ adjectival form of the lexical verb. The proper analysis of the construction in Xokleng is not evident to us at this stage. It is possible that, rather than subordination of the lexical verb, what one sees here is a type of verb serialization involving two finite verbs. Serialization is found elsewhere in the family, but it is usually either lexically idiosyncratic or employed with causal or

 

1076   Andrés PABLO Salanova consequential meanings (see the construction in (12)), rather than to indicate aspectual value, as in this case.

43.5  Mixed Ergativity in Independent Clauses A mixed ergative pattern in progressive and other constructions has been described in Northern Jê languages (cf. Salanova 2008a; Gildea and de Castro Alves 2010), but something analogous may already be found in Urban’s (1985) description of number agreement in Xokleng, which we omit here for reasons of space. The general idea is that ergative-​ absolutive alignment and nominative-​accusative alignment coexist in the same clause, with ergative alignment appearing in the lower (i.e. more deeply embedded) part of the clause, and accusative alignment in the higher part of the clause. This sometimes results in arguments being indexed by more than one element. We exemplify this with Mẽbengokre: (20) bai ii-​djàpêx 1nom 1-​work.n ‘I’m working.’

o=nhỹ with=sit.v

(21) bai aj-​mar o=nhỹ 1nom 2-​hear.n with=sit.v ‘I’m listening to you.’ As can be seen in these examples, the lexical verb (djàpêx, mar) appears in a nominal-​ adjectival form, and as a consequence gets absolutive inflection. In the higher part of the clause, however, one gets a nominative pronoun. When the lexical verb is intransitive, this nominative pronoun duplicates the argument already indexed by the absolutive prefix. When the lexical verb is transitive, the nominative pronoun replaces the ergative pronoun. This is indicated with the subscript indices. Ergative and accusative case never appear in the progressive construction. Intuitively, one can link this particular combination of accusative and ergative alignment to the fact that, in addition to the lexical verb, there is an additional verbal word, the progressive “auxiliary,” which is more lexical than would seem at first sight. The lexical verb, in a nominal-​adjectival form, would be responsible for the partially ergative alignment, while the finite auxiliary would be responsible for the partially accusative alignment. In section 43.6 we develop just such an analysis for the Mẽbengokre progressive construction. See also Baker’s and Coon and Preminger’s Chapters 31 and 10, respectively, in this volume. However, the mixed pattern also arises in constructions where there is no finite auxiliary. One such case is negation in several northern Jê languages. We illustrate this with some Timbira examples taken from Alves (2004) , where it occurs only in the future tense:

 

Ergativity in Jê languages    1077 (22) wa ha i-​wryk 1nom irr 1-​go down.n ‘I will not go down.’ (23)

wa ha ih-​pyr 1nom irr 3-​grab.n ‘I will not grab it.’

nare neg

na neg

(24) wa ha wakà pyr 1nom irr knife grab.n ‘I will not grab the knife.’

na neg

Admittedly, there is evidence that negation is also more lexical than might appear at first sight in Jê languages, and that what we have in these examples are again complex clauses. However, such cases force us to relinquish the link between nominative case and finiteness, or to redefine finiteness independently of the n/​v opposition. Such matters are still the subject of current research.

43.6  A Sketch of Mẽbengokre Ergativity This section is an abbreviated version of Salanova (2008b). In Mẽbengokre, there exist clauses that display accusative alignment, and others that display ergative alignment. The contrast between the two types of clauses may be seen in the following examples. In embedded contexts, one only finds clauses with ergative alignment (25a, b). In main clauses, the two types coexist, and the contrast between the two types of alignment is associated with the aspectual difference mentioned in section 43.3, and described in more detail in what follows. (25)

a i-​rwỳk 1-​go.down.n ‘I’ve gone down’

a.’

ba ruw 1nom go.down.v ‘I’ll go down’

b ije byr 1erg 3.grab.n ‘I’ve grabbed’

b.’

ba ku-​by 1nom 3acc-​grab.v ‘I will grab’

The analysis that we provide here essentially consists in characterizing the forms that head ergative clauses as being nominal, while the remaining constructions are headed by fully finite verbs.

 

1078   Andrés PABLO Salanova Before we go ahead with the analysis of these constructions, we take up again certain descriptive points about Mẽbengokre that were made in Reis Silva and Salanova (2000) and Reis Silva (2001), in particular regarding word classes in Mẽbengokre.

43.6.1 Predicates and Word Classes in Mẽbengokre Practically all open-​class items may appear as main predicates in Mẽbengokre clauses. We provide a few examples here, all of which are underived lexemes, covering the whole morphosyntactic spectrum of independent predicates: (26) ba ku-​dji 1nom 3acc-​put.v ‘I put it vertically.’

(31) i-​kra 1-​child ‘I have children.’

(27) ba onhuw 1nom 3.shoot.v ‘I shoot it with an arrow.’

(32) i-​mã ành 1-​dat 3.sweet ‘I find it tasty.’

(28) ba kàxiràx o ite 1nom nail with drive ‘I drove the nail [into something].’

(33) i-​nhõ kà 1-​poss canoe ‘I have a canoe.’

(29) ba bôx 1nom arrive.v ‘I arrive.’

(34) i-​bê tep 1-​loc fish ‘I’m a fish.’

(30) i-​ngryk 1-​angry ‘I’m angry.’ We will classify predicates mainly according to the way in which their arguments are marked. In the examples given in what follows, one can identify the following types:

1. 2. 3. 4. 5. 6.

Predicates that take a nominative subject and an accusative object: (26). Predicates that take a nominative subject and an absolutive object: (27). Predicates that take a nominative subject and an oblique object: (28). Predicates that take a nominative subject: (29). Predicates that take an absolutive subject: (30) and (31). Predicates that take an oblique subject (dative, locative, or possessive), and sometimes also an absolutive object: (32), (33), and (34).

We might need to make further divisions in this classification on the basis of other criteria; for example, we might need to distinguish between (30) or (31), or (32), or (33), on

 

Ergativity in Jê languages    1079 the basis of their semantics or some formal property. Likewise, the importance of the distinction between (26) and (27), treated differently by different authors (Reis Silva and Salanova 2000; Ferreira 2003; Oliveira 2005), could be called into question. The first important point to raise about this classification is that predicates of Types 1, 2, 3, and 4 take on two different forms, and the opposition between these two forms is correlated with the difference in argument alignment that we described in (25). This may be synopsized as follows: (35)

Type 1 Type 2 Type 3 Type 4 Type 5 Type 6 alignment A erg–​abs erg–​abs erg–​obl abs abs obl–​(abs) alignment B nom–​acc nom–​abs nom–​obl nom —​ —​

We take the fact that Types 1–​4 have two different forms which correlate with two types of alignment to be the main criterion to distinguish the following two major lexical categories in Mẽbengokre: verbs (Types 1–​4) and nouns (the rest). This distinction is reinforced by a series of other morphological criteria: (1) verbs are often morphologically complex, consisting of a “class” prefix and a root; (2) many verbs have suppletive forms for number; (3) verbs may take on valence-​changing morphology. None of these additional criteria will be essential for what follows.

43.6.1.1 Verbs Dependent marking with verbs in the A alignment is identical to dependent marking with nominal predicates. This is self-​evident for intransitive verbs (Type 4), which are parallel to intransitive nominal predicates (Type 5). For transitive verbs (Types 1, 2, 3), the parallel with transitive nominal predicates (Type 6) lies in that ergative shares traits with certain oblique cases such as dative (in (32)) or locative (in (34)). It is primarily because of this parallel that we call the verb form that goes with alignment A the nominal form of a verb, represented as .n in the glosses. This form is opposed to the verbal form of verbs, associated with alignment B, and glossed as .v. Note that in this analysis there are no non-​finite verb forms in Mẽbengokre distinct from the nominal form. In some previous work we called the nominal form the non-​finite form of verbs. That designation has the advantage of highlighting the parallel between these forms and participial and gerundive forms found in better-​known languages, which are not unlike them in meaning and use. But this terminology adds redundancy: just as there is no opposition between nouns and adjectives in Mẽbengokre, there seems to be no opposition between nominal forms of verbs and non-​finite (i.e. adjectival) forms of verbs in the language.

43.6.1.2 Nouns The core function of nouns is to serve as heads of referring expressions. In Mẽbengokre, only nouns (including nominal forms of verbs) can occupy this function. There are no referring expressions that contain finite verbs. Morphologically, there are two types of

 

1080   Andrés PABLO Salanova noun in Mẽbengokre: relational and non-​relational. Relational nouns always require an argument to be expressed. This is generally a possessor of an inalienably possessed noun, the whole of the part expressed by the head, and so on. It comes marked in the absolutive case. In nominal forms of verbs this argument corresponds to the sole argument of intransitives, or to the object of transitives, as in (36c,d): (36) a. i-​pa 1-​arm ‘my arm’ b. i-​prõ 1-​wife ‘my wife’ c. i-​tẽm 1-​go.n ‘my going’ d. i-​pumũnh 1-​see.n ‘my being seen’ Nouns that are not relational do not take this absolutive argument, and can appear with an oblique possessor or other oblique arguments.

43.6.2 Nominal Forms of Verbs Nominal forms of verbs are used in lexical “nominalizations,” internally headed relative clauses, and nominal eventive clauses. We will not deal with the first here, as they are seldom accompanied by arguments, thus preventing any sort of clear alignment pattern to arise. The following examples illustrate the other two uses: (37) a. [kubẽ kute àktire krõr] nẽ jã barbarian 3erg Gavião pacify.n nfut this ‘This is the white man (barbarian) that made peace with the Gavião.’ b. i-​mã [aje amũ =jã=’ã kubẽkà jadjàr]=jã ngã 1-​dat 2erg yesterday clothes wear.pl.n=this give.v ‘Give me those clothes you were wearing yesterday.’ (38) a. ba [kute tep krẽn] 1nom 3erg fish eat.n ‘I saw him eating fish.’

pumũ see.v

 

Ergativity in Jê languages    1081 b. ba [kute tep o=àbir] mỳrỳri 1nom 3erg fish obl=3.pull_​up.n while ‘I talked to him while he was fishing.’

kum 3dat

i-​kabẽn 1-​speak

As can be seen, the meaning of the sentences in (37) is one of the participants in the eventuality described by the verb, while in those in (38) the reference is to the event itself. That is, (38a) should be understood as meaning that I saw the event of him eating fish. The eventive meaning is easier to identify in complements of verbs of direct perception, but is present also in a number of other constructions involving nominal forms of verbs, as we will see. Internally headed relative clauses and embedded eventive clauses are superficially identical and share the same basic structure. A fuller discussion of Mẽbengokre internally headed relative clauses and eventive clauses may be found in Salanova (2011b, 2013).

43.6.3 Structure and Meaning of Nominalizations Nominal eventive clauses and internally headed relative clauses share several characteristics: (a) they have ergative alignment; (b) they have certain passive traits; (c) they lack several positions that exist in the left periphery of main clauses. Like with relational nouns that are not related to verbs, the absolutive argument in nominal forms of verbs is always obligatory, and always represented by a noun or pronominal index directly to the left of the predicate. This is a fact that will be important for the analysis of progressive clauses that follows. Ergative, on the other hand, is optional. If omitted, the subject is interpreted generically (it is for this reason that we claim that nominal clauses have certain passive traits), or, if the clause is embedded, as obligatorily controlled by a higher subject. In contrast to this, the nominative subject of (finite) verbal forms of verbs, is always anaphoric to a topical referent if it is left unexpressed. This is shown in the contrast between (39) and (40): (39)

a. ba rádio kate 1nom radio break.v ‘I broke the radio.’ b. ∅ rádio kate 3nom radio break.v ‘He/​she/​it broke the radio.’ (not ‘someone broke the radio’)

(40) a. ije rádio ka’êk 1erg radio break.n ‘my breaking the radio’

 

1082   Andrés PABLO Salanova b. rádio ka’êk radio break.n ‘[someone] breaking the radio’, ‘the radio breaking’ This contrast might suggest that ergative is an inherent case that comes with the thematic role of agent or causer (for a discussion of ergative as inherent case, see Sheehan, Chapter 3, this volume). This is further encouraged by the fact that the ergative marker seems related (though not identical) to a postposition that introduces (propositional) causes: (41) bri pyma=je muw frog fear=p cry.v ‘He cried for fear of the frog.’ The case for inherent ergative case is not as clear cut, however. On the one hand, there are transitive perception verbs with experiencer external arguments, and these are marked ergative just like agent external arguments. On the other, ergative case cannot be freely added to any intransitive nominal predicate to make it transitive. In fact, the following is ungrammatical:5 (42)

* ije muw 1erg cry.v Intended: ‘I made him cry.’

I conclude that ergative is not inherent in Mẽbengokre. Rather, the ergative could be considered analogous to the by phrase of English passive clauses, which, though optional, transmits the thematic role associated with the external argument of the active form of the verb. A final point to make about the constructions in question is that they lack certain left-​peripheral positions that independent clauses have. If one compares (43) with (44), one sees that the first cannot have the tense-​marking particle nẽ ‘nfut,’ nor a focalized noun phrase. (43) (*kukryt) (*nẽ) tapir (foc) nfut ‘my killing tapir’

(*ije) 1erg

arỳm already

(44) kukryt nẽ ba arỳm tapir (foc) nfut 1nom already ‘It’s tapir that I killed.’ 5 

ije kukryt bĩn 1erg tapir kill.sg.n ku-​bĩ 3acc-​kill.sg.v

However, see discussion about the pleonastic ergative below.

 

Ergativity in Jê languages    1083 It is also apparent that while a nominative pronoun appears outside aspectual particles, ergative appears inside. We extract from these facts the conclusion that embedded nominal constructions have a smaller structure than independent clauses.

43.6.4 Uses of the Nominal Form of Verbs I claim that whenever a nominal form of a verb is used, it is either “insubordinated” in the sense introduced earlier, or subordinated to the following element. In other words, all propositions involving a nominal form, if not insubordinated, are complex constructions in one way or another. It is our position that there is no grammaticalization of nominal forms (or combinations of a nominal form and another element) into finite forms in Mẽbengokre. An alternative approach to nominal forms in independent contexts can be found in Gildea (2008) and Alves (2010). The following examples cover the spectrum of independent clause types involving nominal forms of verbs. Examples (45) and (47) are fully ergative, while (46) has a mix of ergative-​absolutive and nominative-​accusative elements, which can be ascribed to the presence of a finite auxiliary in addition to the nominal lexical verb. (45)

ije tep krẽn 1erg fish eat.sg.n ‘I haven’t eaten fish.’

(46) ba tep krẽn 1nom fish eat.sg.n ‘I’m eating fish.’

kêt neg o=nhỹ obl=sit.sg.v

(47) ije tep kur 1erg fish eat.pl.n ‘I (generally) eat fish.’ Example (45) represents the form taken by all negative clauses, and by clauses modified by certain postverbal aspectual and manner elements. Example (46) represents the form of all progressive clauses. Example (47), a case of insubordination, is used to express generic or resultative meanings. We examine the structure of each type of construction in a separate subsection.

43.6.4.1 Negative Clauses The negative particle kêt, used for clausal negation, is also a negative existential predicate used with underived nouns:

 

1084   Andrés PABLO Salanova (48) a. tep kêt fish neg ‘There is no fish.’ b. me õ kuwy kêt=ri pl 3.poss fire neg=when ‘[This is about] when people had no fire.’ c. i-​kêt=kam 1-​neg=in ‘[This is] during the time I didn’t exist.’ In fact, if we take the negative existential use to be basic, we can throw new light on negated clauses. We contend that these require nominal forms of verbs precisely because the whole clause is an argument (and hence nominal) to the negative predicate. (49) a. ije pi’ôk byr kêt 1erg paper grab.n neg ‘I didn’t grab the paper.’ b. bà=kam i-​mõr kêt forest=in 1-​go.pl.n neg ‘I didn’t go into the forest.’ In negative clauses in Mẽbengokre, the semantic composition is ‘there is no X-​ing (by Y).’

43.6.4.2 Progressive Clauses Mẽbengokre progressive clauses were initially described by Reis Silva (1996). We reproduce that analysis here. Progressive constructions are partially ergative-​absolutive, and partially nominative-​ accusative. While the person marker on the verb is always absolutive, there is a further nominative pronoun which either stands for the second argument or duplicates the reference of the absolutive if the latter indexes an intransitive subject: (50) a. ba tep   krẽn o=nhỹ 1nom fish   eat.sg.n obl=sit.sg.v ‘I’m eating fish.’ b. ba i-​tor 1nom 1-​dance.n ‘I’m dancing.’

o=dja obl=stand.sg.v

 

Ergativity in Jê languages    1085 The morphosyntax of this construction, described in more detail in Arregui et al. (2014), indicates that the auxiliary, a postural verb that corresponds to the shape that the subject takes during the action, is the main verb in the clause. This is not unlike what we said about negation. There are two differences, however. One is that the main predicate is a finite verb in this case, meaning that the alignment of its arguments will be nominative-​accusative. The second difference is that there is a semantic relation between the postural verb and the nominative subject, making the latter into the postural verb’s logical subject. The absolutive NP, on the other hand, is an argument of the embedded nominal lexical verb. These elements are suggestive of a control construction. Control in Mẽbengokre has a quirk compared to its English equivalent, which is that only embedded ergative is deleted under control. Absolutive remains, whether it represents a subject or an object. This is the source of the double marking found in progressive constructions with intransitive verbs: (51)

a. Ergative is “equi-deleted” in the control construction obligatory coreference and “equi-deletion” ba [ ije tep kr˜en ] o=nhy˜ 1nom 1erg fish eat obl=sit.sg.v logical subjects b. Absolutive is not “equi-deleted”, as it is morphologically required by the verb obligatory coreference without “equi-deletion” ba [ i-tor ] o=dja 1nom 1-dance-N obl =stand.sg.V logical subjects

As we anticipated, we believe this to be a morphological quirk of absolutive, which is an obligatory prefix of nominal forms of verbs.

43.6.4.3 Generic Constructions as Existential Clauses If we are correct about the characterization of a certain form of the verb as nominal-​ adjectival, a sentence such as (47) is counterintuitively headed by a noun. I claim that this presents no difficulties in the context of Mẽbengokre nominal clauses, and that no grammaticalization of the nominal-​adjectival form into a finite verbal form needs to be appealed to. In Mẽbengokre, any noun phrase is liable to being interpreted as a complete proposition. I gave examples of this in (31) and (33). From the translations one may gather that nominal clauses are interpreted as existential constructions.

 

1086   Andrés PABLO Salanova What I propose for insubordinated nominal-​adjectival forms of verbs is that they are interpreted, like underived nouns heading independent clauses, as heading existential constructions. There are many subtle points to make in relation to their precise aspectual meaning which we cannot make here due to limitations of space. For discussion, see Salanova (2007). It is however telling that among their meanings one finds the existential perfect, as well as the habitual, both of which involve claiming existence of one or multiple eventualities without any particular anchoring with respect to topic time. Expressed through a paraphrase, the sense of a clause with an insubordinated nominal-​adjectival verb form is akin to “there was (a) V-​ing of NPobj by NPsubj.”

43.6.5 Summary The Mẽbengokre split is essentially between normally subordinated nominal constructions, which are straightforwardly ergative-​absolutive, and independent finite verbal constructions, which are straightforwardly nominative-​accusative. The complexity of the alignment system of the language comes not from subtleties of the case system itself, but rather from the ubiquity of nominal constructions, which are required with negation, with various types of postverbal auxiliary, and so on, in addition to regular syntactic subordination to a lexical predicate. This is, with only slight deviations, an accurate description of what happens in other languages of the family as well. The extent to which nominal-​adjectival forms of verbs are used in independent clauses and the possible elements that may combine with them varies from one language to the next, but the essential traits, in particular the relationship between nominal-​adjectival forms and ergative alignment, and the exclusivity of ergative alignment in subordinate clauses, are the same.

43.7 Conclusions The general picture that Jê languages offer is one where ergativity is associated to nominal-​ adjectival forms of the verb, strengthening the link between nominalizations and ergativity explored by Alexiadou (2001, Chapter 15, this volume); see also Aldridge’s chapter in this volume (Chapter 21). An important difference between Jê languages and the better-​known languages examined in Alexiadou’s original study is the extent to which nominal forms of predicates are used in the former. In addition to being required in all contexts of subordination, they are used when clauses are modified by a number of elements which normally include negation and manner predicates, in addition to most constructions with aspectual auxiliaries. Before we conclude, it is important to dwell briefly on the morphological identity between genitive (for inalienable possessors) and absolutive in all Jê languages.

 

Ergativity in Jê languages    1087 This identity, common in various languages of the eastern Amazon region (cf. Gildea 2008 for other examples, as well as Queixalós in this volume, Chapter 42, for a central Amazonian example), is easily accounted for if one accepts the nominal character of long form verbs. Note however that this pattern is not the only type of identity found between a genitive case and a core clausal case. In another common pattern, found in Mayan and Inuit-​Aleut languages (cf. in this volume Aissen, Chapter 30, and Compton, Chapter 34), genitive is identical to ergative case rather than to absolutive. For these languages, nominalization has also been claimed to play a role in the ergative alignment found at the clausal level. The structures of nominal(ized) clauses in the two types of languages are necessarily quite distinct. At this stage of our knowledge, we can only put these two types side by side in a simple typology.6 There are other traits for which ergativity in Jê languages is remarkable: Panará offers one example of a predominantly SVO language that has robust ergativity (on the link between constituent order and ergativity, see Taraldsen, Chapter 14, this volume); Jê languages provide a variety of examples of “redundant marking” of participants, often overlapping ergative and accusative alignment, going well beyond to the “pragmatic ergativity” of Tibeto-Burman languages described by DeLancey (2011); finally, Jê languages show how some superficially striking patterns (such as an ergative split conditioned by negation or by manner modifiers) arise from the interplay of syntactic constraints rather than from functional pressures.

Acknowledgments I would like to thank my main Mẽbengokre consultants, Ikrô Kayapó and Bepkamrêk Kayapó, for the information that allowed me to develop a description of Mẽbengokre ergativity. I would like to thank Adriana Machado Estevam for discussing her description of Xavante, and Bernat Bardagil for information regarding person markers in Panará.

Abbreviations The glosses used in this chapter are as follows: 1, 12, 2, 3—pronominal markers of first person, including first person, second, and third person, respectively; erg, dat, acc, nom, poss, loc, obl—case markers or postpositions corresponding to ergative, dative, accusative, nominative, possessive, locative and oblique; the absolutive, which corresponds also to the genitive case, is left unmarked in this chapter; sg, pauc, pl—singular, paucal, plural; n, v—nominal and verbal form of the verb; neg—negation; anticaus—anticausative; antipass—antipassive; dim—diminutive; det—determiner; fut, nfut—future, non-​future; foc—focalized phrase. 6 

Note that the ergativity found in the noun phrases of better-​known languages (i.e. those discussed in Alexiadou 2001) patterns more like Jê ergativity than like Inuit-​Aleut or Mayan ergativity, even if there is no morphological identity between the genitive and one of the core cases: in those languages, which include English, the genitive case has an absolutive distribution, whereas the second argument in noun phrases is introduced by means of a by phrase.

 

1088   Andrés PABLO Salanova Morphological categories that can’t be segmented are separated by . in the gloss; separations between morphemes are indicated by -​, and between a clitic and its base by =. We’ve chosen to render all examples orthographically, converting from the phonemic transcriptions found in some sources. Our glosses have been homogenized and do not correspond exactly to glosses given in the sources.

 

Chapter 44

Interacti on of ergativit y a nd inf ormation st ru c t u re in Jaminjung ( Au st ra l ia ) Eva Schultze-Berndt

44.1 Introduction This chapter discusses the manifestations of ergativity in Jaminjung, a language of the small Mirndi family, spoken in northern Australia. Ergativity in Jaminjung is of interest in several respects. First, at the morphological level, Jaminjung is what has been described as an optional ergative language, since in many lexical and grammatical environments, the Ergative1 marker may or may not be present on agent NPs. This state of affairs thus differs from a split ergative system where the distribution of case marking is predictable and consistent on the basis of one or more grammatical factors such as the position of the agent referent on the animacy hierarchy, or the temporal or aspectual properties of the clause. It will be argued that the Jaminjung pattern can be described as a system of differential agent marking which is sensitive to multiple factors without strict constraints. One important factor is information structure, in that the presence of ergative marking strongly correlates with (but is not restricted to) focal status of the agent referent. However, factors that have been shown to trigger splits in split ergative systems—​person, animacy, verb class, and tense/​aspect—​also play a role. Second, as an alternative to the Ergative case, the Ablative case (i.e. the case form that is also used to flag the source location in motion expressions) is also found to mark agents, albeit much less frequently. When attempting to unravel the factors behind differential agent marking, one therefore needs to consider not only two but three 1  I will use capital letters to refer to a language-​specific case, and lower case to refer to the cross-​ linguistic category.

 

1090   Eva Schultze-Berndt possibilities: Ergative marking (default), Ablative marking, and zero marking. It will be argued in section 44.4 that information structure is an important factor responsible for Ergative or Ablative marking of agents. Third, Jaminjung exhibits a—​cross-​linguistically recurrent—​identity of ergative and instrumental case. Considered together, these three phenomena raise questions about the nature of case assignment and grammatical relations, since the use of the two agent cases is determined neither purely lexically nor purely structurally. It will be argued that a construction-​ based framework combined with the notion of violable constraints is well suited for the analysis of ergativity in Jaminjung. In other words, speakers have at their disposal a range of argument structure constructions (for our purposes, ergative-​marked NP, ablative-​ marked NP, and zero-​marked NP). Since ergative marking in Jaminjung is restricted to bivalent predicates (in a sense to be refined in sections 44.2.1 and 44.2.3), the predicate licenses agent marking, but is not, on its own, sufficient for assigning case. Rather, the choice of construction in a particular context is also constrained by tense and aspect of the clause, person and animacy of the agent referent, and its discourse status. In this chapter, the language name Jaminjung is used as a shorthand for a dialect cluster comprising of two named varieties, Jaminjung and Ngaliwurru (ISO-​639: DJD). These are the only remaining members of the Jaminjungan (also Yirram or Western Mirndi) group of the Mirndi family, a discontinuous language family discussed by Chadwick (1984, 1997) and Harvey (2008). The traditional country of the Jaminjung and Ngaliwurru people is situated immediately north and south of the Victoria River in the Northern Territory, between the present-​day settlements of Victoria River Crossing and Timber Creek. Today speakers of both varieties—​alongside members of other language groups—​live in the traditional Ngaliwurru and Nungali country around Timber Creek as well as in Wadeye to the north and in the larger settlements of Katherine to the east and Kununurra to the west. Because of the scattered nature of the present-​day population, it is difficult to accurately estimate the number of speakers, but it is probably fewer than 100. A large-​scale language shift is underway to Kriol (also called Northern Territory Kriol and Roper River Kriol; ISO-​639: ROP), an English-​lexified Creole language that emerged as a result of the colonial situation and now functions as a lingua franca between Aboriginal people throughout a large area of northern-​ central Australia (Sandefur 1984; Harris 1986; Schultze-​Berndt et al. 2013; Meakins 2014a). Jaminjung and Ngaliwurru have been described in two unpublished grammars (Cleverly 1968; Bolt et al. 1971), in a number of papers by Hoddinott and Kofod ( 1976a, b, c), three unpublished PhD theses (Schultze-​Berndt 2000; Simard 2010; Hoffmann 2011)  and a number of published papers and conference presentations by the same authors (including Schultze-​Berndt 2001, 2006a, 2007, 2010; Schultze-​Berndt and Simard 2012; Simard 2013b, 2013a, 2015). The discussion in this chapter is based on annotated corpora compiled by the author between 1993 and 2012. Corpus examples are followed by speaker’s initials and a unique reference number which identifies them in the archived corpus (DoBeS Endangered Languages Archive; http://​dobes.mpi.nl/​ research); this will be made accessible on request. Glosses follow the Leipzig Glossing conventions. Underline marks Kriol words. The chapter is structured as follows: section 44.2 provides a brief overview of the main relevant grammatical characteristics of Jaminjung, and adduces arguments for an analysis

 

Ergativity and information structure in Jaminjung    1091 of the Ergative case as corresponding to the semantic role “effector”, covering agents, natural forces, and instruments. Section 44.3 addresses the question of whether or not Jaminjung displays syntactic ergativity (the answer will be negative). The factors responsible for differential agent marking will be discussed in section 44.4. Section 44.5 is a summary with a brief discussion of the wider implications of the language-​specific findings.

44.2  Ergativity and Argument Structure In terms of the morphosyntactic encoding of arguments and adjuncts Jaminjung is, in many ways, rather typical for languages of the Australian linguistic area (for an overview, see Gaby 2008b), and in particular for the northern languages which do not belong to the large Pama-​Nyungan language family. Core arguments are indexed on the verb, and argument roles are also indicated by a rich case system which in addition to the ergative/​instrumental includes dative, purposive, motivational, origin (encoding place of origin, source, or cause), and comitative case markers, as well as the spatial cases locative, allative, and ablative. Unlike many Pama-​Nyungan split-​ergative languages of Australia, Jaminjung has no differential object marking, i.e. there is no overt object marking (accusative case) for nominals of any category, or in any clause type. Transitive objects and intransitive subjects are always unmarked (a term used here interchangeably with Absolutive case). The form of the Ergative case marker is =ni, often denasalized to =di following a stop. The Ergative marker has been reported to be formally identical to the locative case (Hoddinott and Kofod 1976b); however, in my data, the Locative case marker, in the Ngaliwurru variety only, is =ni postvocalically, but =gi elsewhere (in Jaminjung it is =gi ~ =gi). The diachronic source of the Ergative marker is very likely a third person pronoun or demonstrative, a point that will be further discussed in section 44.5. Case marking in Jaminjung is phrasal in nature, as already pointed out by Hoddinott and Kofod (1976b: 398). This is illustrated in (1) to (3) for the Ergative, but holds for the other case markers as well. Case markers can follow each word in a complex noun phrase, or only one of the words (they are not restricted to the edge of the phrase). The case markers are therefore analysed as clitics for the purposes of this chapter. The factors conditioning the variation in case position are not well understood at present and will not be further discussed here. (1)

[ngiya=biya buluwuj=di]TOP digirrij ganiny-​nginama!2 PROX=SEQ egg=ERG die 3SG>3SG-​step/​weigh.PRS ‘these (bags of) eggs are killing you (by weighing down on you when you attempt to carry them)!’ (IP; ES08_​A04_​06.145)

2  See section 44.4.3 for a definition of the information structure categories employed in the analysis.

 

1092   Eva Schultze-Berndt (2) “…”

gani-​yu=nu 3SG>3SG-​say/​do.PST=3SG.OBL ‘ “…” THIS man said to him’

[yinju=ni PROX=ERG

gurang]FOC older.man (ERa; ES08_​A20_​03.119)

(3) [janyju .. mangarra .. gagawurli]TOP gan-​ijja-​ny .. DEM plant.food long.yam 3SG>3SG-​poke-​PST [ngarrgina=ni  jungurniny=ni]FOC 1SG.POSS=ERG husband=ERG ‘that long yam food, it was dug it up by my husband’ (ERa; ES12_​A02_​01.002-​03) While the phrasal nature of case marking is evidence for an (at least partially) configurational structure, Jaminjung is a “free word order” language in the sense that constituent order within NPs and complex predicates is variable to some extent and clause-​level constituent order is more closely tied to information structure than to grammatical relations (see section 44.4.3; Simard 2010; Schultze-​Berndt and Simard 2012). A frequent strategy is the distribution of information semantically belonging to a single clause across several intonation units (Simard 2013a). Discourse-​given participants need not be represented by overt NPs; a clause minimally consists of a (simple or complex) verb. As a consequence, the frequency of overt NPs, including agents, is relatively low. A text count conducted on a small corpus of texts revealed that only approximately 25% of all transitive predicates were accompanied by an overt agent NP (Schultze-​Berndt 2006b). This will have to be kept in mind when considering the function and significance of case marking on agents. The remainder of this section provides relevant information on the morphological and syntactic correlates of verbal valency (44.2.1), the interrelation of aspect and argument structure (44.2.2), and the range of functions of the Ergative marker (44.2.3), resulting in a definition of agents employed in the following discussion.

44.2.1 Morphological and Syntactic Correlates of Argument Structure Before embarking on the discussion of argument structure and argument marking in Jaminjung, it will be necessary to clarify the application of the notion of “verb” for this language. One of the distinguishing properties of Jaminjung is the existence of two parts of speech which are restricted to predicative function. Inflecting Verbs (IVs) are a class of stems taking obligatory prefixes for person/​number and a number of modal categories as well as tense/​aspect suffixes (tense and aspect can also be marked by complete or partial stem suppletion). IVs form a closed class with only about 30 members (with some dialectal and individual variation). Only IVs can serve as the main predicate of finite, independent clauses.

 

Ergativity and information structure in Jaminjung    1093 The second class is an open class of stems which cannot take any of the verbal inflections and for this reason is labelled Uninflecting Verbs (UVs), following McGregor (2002). UVs can function as the main predicate in some types of subordinate clause (see section 44.3). Their most frequent use is in complex predicates in combination with Inflecting Verbs. Jaminjung complex verbs meet the widely accepted definition of complex predicates as monoclausal constructions where two or more predicative constituents jointly contribute to the argument structure of the clause, share at least one argument, and share values for tense, aspect, modality, and polarity (see e.g. Butt 1997: 108; 2010). In terms of both their lexical semantics and their argument structure they form nuclear junctures in the sense of Role and Reference Grammar (Foley and Van Valin 1984; Van Valin and LaPolla 1997), or “merger” constructions in the terminology of Baker and Harvey (2010). In other words, the complex predicate behaves like a single verb in that it takes a single set of arguments, and neither of the components can introduce an additional participant that would duplicate a grammatical function already contributed by the other, as would be the case e.g. for a serial verb construction involving two bivalent predicates each taking different objects. Two complex predicates involving the same UV but differing in their overall argument structure are illustrated in (4). (4)

a. bag ga-​jga-​ny=ni break 3SG-​go-​PST=DS ‘the jar broke’

bottle bottle

(IP; ES97_​A03_​01.090, Frog Story)

b. bag-​bag gana-​nangga=biyang mama=ni RDP-​break 3SG>3SG-​chop.IPFV=SEQ mother’s.brother=ERG ‘my uncle used to break them (quartz stones) up’ (EH, ES08_​A04_​02.234) By virtue of being semantically generic, the closed-​class IVs take on the nature of overt event classifiers. For example, the intransitive IV –​ijga ‘go’ in (4) regularly occurs in expressions of motion, but also in expressions of change of state, and thus specifies the general type of event that the complex predicate encodes. The argument structure and semantic composition of Jaminjung complex predicates is discussed in more detail elsewhere (Schultze-​Berndt 2000, 2007, 2010, 2012, 2015). Morphologically, inflecting verbs (IVs) fall into two non-​ overlapping classes. Intransitive IVs only inflect for the person/​number of a single argument (S), while transitive IVs index two arguments; these are the Actor (A)  and Undergoer (P)  for monotransitive predicates and (usually) the Actor and Recipient for ditransitive predicates. Reflexive–​reciprocal forms can be formed from most morphologically transitive IVs; the resulting forms are morphologically and syntactically intransitive, i.e. they only index a single participant (with the intransitive person paradigm). As a default, the S argument of intransitive verbs (including reflexive/​reciprocal verbs) is encoded by a zero-​marked (absolutive) NP (if present), as is the P argument of transitive verbs, and both the Recipient and Theme arguments of ditransitive

 

1094   Eva Schultze-Berndt verbs (Schultze-​Berndt 2010). The A  argument of transitive and ditransitive verbs, on the other hand, has the three case-​marking possibilities mentioned above (zero, ergative, or ablative). There is no class of transitive verbs that is singled out for “special” subject marking (as found e.g. for experiencers in many languages). Intransitive subjects are never ergative-​marked, unlike in a number of other “optional ergative” languages. One consequence, relevant in the context of ergativity, of the complex predicate structure just discussed is that Jaminjung has no ambitransitive predicates corresponding e.g. to English to break or to open. IVs as simple predicates are either strictly monovalent, bivalent, or trivalent. In complex predicates, transitivity alternations with the same UV always involve a change of IV, as illustrated in (4) (Schultze-​ Berndt 2015). This means that the presence of ergative case marking is unlikely to be related to the need to distinguish a transitive from an intransitive reading of the same predicate. While, as a general rule, the number of core arguments and the case frame used with a particular (simple or complex) predicate corresponds to the morphological transitivity of the IV, there are a number of deviations from this general rule. These involve specific secondary meanings of IVs in complex predicates, which are at odds with their morphological class. For example, the morphologically transitive IV -​ma ‘hit’ can be used to form complex verbs of emerging and appearance which are syntactically intransitive, allowing only a single, absolutive NP, as illustrated in (5). Two other cases of mismatch between morphological and syntactic argument structure will be discussed in sections 44.2.2 and 44.2.3; for details see Schultze-​Berndt (2015). (5) jarlig majani bul child(ABS) maybe emerge ‘maybe the child appears on her’

gani-​ma-​m=nu 3SG>3SG-​hit.PRS=3SG.OBL (JM; ES12_​A04_​01.193)

Apart from the derivational reflexive/​ reciprocal formation and the possibility of combining the same UV with different IVs, Jaminjung has no valency changing devices. Neither does it have voice alternations such as a passive or antipassive constructions (though see the remarks on the ‘pseudo-​progressive construction’ in section 44.2.2).

44.2.2 Aspect and Argument Structure As indicated above, tense and aspect are marked on the Inflecting Verb (IV) in Jaminjung. The categories that are distinguished—​variably marked by suffixes or stem suppletion—​are Present Tense (PRS; unspecified for aspect), Past Perfective (PST) and Past Imperfective (IPFV). Modal categories are marked by prefixation; in addition to Imperative mood, Jaminjung has an untensed Hypothetical modal category (HYP), mainly used in contexts of negative and undesirable possibility, and a possibility

 

Ergativity and information structure in Jaminjung    1095 (Potential) modal (POT) which in untensed clauses has future and desiderative functions and which is also compatible with the Past Imperfective in a past possibility reading. The correlation between inflectional tense/​aspect and the presence or absence of ergative marking will be considered in section 44.4.2. In addition to the inflectional imperfective, there is an imperfective construction (termed ‘Pseudo-​Progressive’ by Schultze-​Berndt 2012) which involves iterative marking on an Uninflecting Verb and the choice of one of the two morphologically intransitive Inflecting Verbs -​yu ‘be’ or -​ijga ‘go’. This construction has some relevance for the discussion of ergativity since it is incompatible with ergative marking even in the case of bivalent predicates. If overt, both core participants are in the (unmarked) absolutive case, as shown in (6). (6) en janyungbari burlug-​mayan ga-​yu gugu and another(ABS) drink-​ITER 3SG-​be.PRS water(ABS) ‘and the other one (animal) is drinking water’ (DMc; ES96_​ A13_​01.020) A double absolutive frame in progressive or otherwise imperfective constructions is a cross-​linguistically recurrent phenomenon; it has been described, for example, by Laka (2006a) for Basque and by Forker (2012) for Nakh–​Daghestanian languages. The absence of agent marking in this case is conditioned by the use of an intransitive predicate (originally with locative semantics), as is cross-​linguistically common in the encoding of a progressive structure (e.g. Bybee et al. 1994: 129–​137; Laka 2006a; Coon 2013a). If one considers mainly the intransitive properties of the construction, it has affinities with antipassive constructions, which are often imperfective in nature (Cooreman 1994: 57–​58; Polinsky 2013). The pseudo-​progressive overlaps with a canonical antipassive in featuring an absolutive Agent, while failing to meet the criterion of an oblique or obligatorily demoted object. If one considers mainly the transitive nature of the construction—​i.e. the fact that it allows for an absolutive object in the case of bivalent predicates—​it represents a clear ergative split based on aspectual properties, as opposed to the more fluid factors influencing the presence or absence of ergative marking to be discussed in section 44.4.

44.2.3 Actors, Agents, and Instruments: The Notion of Effector The Jaminjung case form which is the default agent marker and is therefore labelled Ergative case here is found with arguments in a range of semantic roles. It can mark volitional human agents as in (3), non-​volitional animate actors as in (7), inanimate forces as in (8), and other inanimate causes of an event as in (1). There is no evidence that differential agent marking in Jaminjung is conditioned by conscious choice or volitionality of the agent, as can be the case e.g. in Hindi/​Urdu (Mohanan 1994a: 72ff.; Butt 1997: 122; de Hoop and Narasimhan 2005, 2009).

 

1096   Eva Schultze-Berndt (7) jarlig=biya bardawurru gani-​ma-​ya \ .. [gumurrinyji=ni=malang]FOC child=SEQ many 3SG>3SG-​have-​PRS emu=ERG=GIVEN ‘it has many children, the emu (does)’ (DM, MH96_​A19_​01.049–​50) (8) [wilarung=ni ganurr-​ina-​m]FOC lightning=ERG 3SG>3PL-​chop-​PRS ‘the lightning strikes them, “daaaa” ’

“da:::” (sound)

(IP, ES08_​A16_​02.012)

In all the examples so far, the ergative-​marked NP is coreferential with a pronominal index on the inflecting verb in the first, A(ctor) slot (see section 44.2.1). The Ergative case is however also used to mark instruments as in (9) and (11), a functional overlap common in Australian languages, but also cross-​linguistically. As expected, the verb in this cases indexes person and number of the agent, not the instrument, and the agent can be represented by a second ergative-​marked noun phrase as shown in (9). (9) thanthu=ni=biyang warn gani-​bili wirlga=ni DEM=ERG=SEQ entangle 3SG:3SG-​POT:get/​handle foot=ERG/​INST ‘that one (the owl) is going to grab him (the boy) with its claws’ (IP, ES97_​A03_​01.170) In most traditional accounts, ergative and instrumental are analysed as homophonous markers of two distinct grammatical roles (e.g. Blake 2001 [1994]:  49–​50). Laughren (1992) accounts for ergative marking on instruments in Warlpiri by analysing instrumental NPs as secondary predicates controlled by the agent, since ergative marking of instruments is restricted to transitive predicates. However, the distinction between “instruments” and “agents” is not clear-​cut. While Ergative-​marking of instruments is likewise restricted to transitive predicates in Jaminjung, the Ergative case is also found with the inanimate agent co-​occurring with a morphologically intransitive verb, -​irna ‘burn, be affected by heat’, illustrated in (10). (10) jarlig wuju ga-​rna child small 3SG-​burn.PST ‘the little child got burnt by the fire’

guyug=di fire=ERG

(JM, fieldnotes 1993)

I argue here that the Jaminjung Ergative case has a single overall function, that of marking the role of “effector” (Van Valin and Wilkins 1996; Bickel 2000: 599f.) or “initiator” (Legate 2012a: 183). This is taken to be a language-​specific semantic role (in the sense of a semantic role directly reflected in morphosyntax) involving a participant that is causally involved in an event which impinges on an undergoer (hence the restriction of ergative marking to two-​participant scenarios, as opposed to simply “agentive” scenarios). The basic assumption underlying this analysis is that arguments

 

Ergativity and information structure in Jaminjung    1097 form constructions in their own right and consequently express constructional meanings which have to be compatible with, but are not determined either by the meaning of the predicate (“lexical case”), or by the overall structure of the clause (“structural case”) (Goldberg 1995; Barðdal 2011). This assumption is in turn based on the observation that the functions of case markers are not equivalent cross-​linguistically even if they share some core uses (e.g. the occurrence on typical agents of typical transitive clauses). The analysis of the Jaminjung Ergative case as encoding the role of effector subsumes all functions discussed above, including that of the heat source with a ‘burn’ predicate, the sentient participant with experiencer predicates, and the controlling participant with transitive predicates of possession. The Effector role differs from the role of the Actor participant which is indexed by the first (A) prefix of transitive predicates, which is much more restrictive in that it is only compatible with participants which are the ultimate cause (first cause) of an event (DeLancey 1991; Van Valin and LaPolla 1997: 146). This analysis is supported by the fact that the ergative-​marked heat source with intransitive IV -​irna ‘burn’ is always inanimate (fire, ashes, or the sun). With animate agents (or the sun)—​ultimate causes of a fire—​the transitive ‘burn’ verb -​irriga ‘cook, burn’ is required, as in (11). (11)

(yurlgi) guyug=di burru-​rriga=nu, jag=gu leech fire=ERG/​INST 3PL>3SG-​cook.PST=3SG.OBL go.down=DAT ‘they burnt (the leech) with fire for her, so that it would fall off ’  (IP, ES97_​A03_​02.179)

In sum, the morphological Ergative case in Jaminjung is the marker of an agent role in a broad sense (“effector”, “instigator”), subsuming instruments and natural forces. The discussion of ergativity in the following sections will be restricted to agents in the narrow sense (i.e. the agentive arguments of transitive predicates), excluding instruments. Agents in Jaminjung will be defined as arguments that (i) are indexed by a person/​number prefix on the verb and (ii) are potentially ergative-​marked. This definition excludes the single argument in cases like (5), which meets criterion (i) but not (ii). Criterion (ii) also excludes the—​obligatorily absolutive—​ f irst argument of bivalent predicates formed with intransitive inflecting verbs in the “pseudo-​progressive” imperfective construction (44.2.2). Instruments are excluded by criterion (i) since they are not indexed on the verb. However, the definition includes Ergative-​marked phrases encoding a heat source with the morphologically intransitive verb -​irna ‘burn’, as in (10). On agents thus defined, the Ergative is the default case in terms of its frequency, but is not obligatory; moreover, the Ablative case can also mark the agent role, but does not fully overlap in meaning with the Ergative in that it is restricted to animate agents (see section 44.4.4).

 

1098   Eva Schultze-Berndt

44.3  The Issue of Syntactic Ergativity A question not addressed so far is whether Jaminjung exhibits not only a morphological ergative–​absolutive case marking pattern, but also syntactic ergativity, i.e. whether any syntactic pattern singles out either the ergative (A) or the absolutive (S/​P) function. Standard diagnostics for syntactic ergativity are argument omission in clausal coordination, relativization, control relations between main and subordinate clause, and switch-​reference constructions. The coordination test does not work for Jaminjung because, as discussed in section 44.2, all discourse-​given arguments can be omitted. Relativization is not a good diagnostic either because Jaminjung only has a generalized subordinate clause corresponding to Hale’s (1976a) adjoined relative clause which, in addition to reference-​identifying (“relative clause”) functions, also has temporal and circumstantial uses and thus do not require a head (see also Nordlinger 2006). Only control turns out to be an unambiguous diagnostic for Jaminjung, and to reveal nominative–​accusative patterning. This is because direct objects (in absolutive case, and indexed by the Object prefix of transitive verbs) and some types of Dative-​marked objects control the unexpressed subject of a type of simultaneous non-​finite clause marked with the clitic =bina ~ =mina, formally identical to the Allative case. Examples (12) and (13) provide illustration. A parallel construction is found in other languages in the same linguistic area, e.g. Warlpiri (Simpson and Bresnan 1983; Legate 2012a: 184) and Wardaman (Merlan 1994: 287). (12) [gurrija=bina]CL-​NFIN waj yirrurr-​unga-​ny \ digging=ALL leave.behind 1PL.EXCL>3PL-​move.away.from-​PST ‘we had left themi (while theyi were) digging’ (DP, ES96_​A09_​03.156) (13) “gujarding=ni  ngarrgina=ni    gan-​anjama-​ny mother=ERG  1SG.POSS=ERG  3SG>3SG-​bring-​PST yinju=biya=nggu \       wajgany” \  gan-​unggu-​m=nu \ PROX=SEQ=2SG.OBL  honey     3SG>3SG-​say/​do-​PRS=3SG.OBL mugurn=bina \ lie/​sleep=ALL ‘ “my mother brought this sugarbag for you” she said to heri, (while shei was) lying down’ (VP; ES99_​V01_​06a.105–​7) In contrast, simultaneous non-​finite clauses where the unexpressed subject corresponds to either the S (14) or the A argument (15) of the matrix clause can take the iterative marker =mayan if the subject is understood to be agentive. (14)

Cave Spring=bina yurr-​ijga-​ny=mulu \ =ALL 1PL.INCL-​go-​PST=COLL ‘all of us went to Cave Spring, gathering ashes’

[bilij garrb=mayan] \ ashes gather=ITER (IP, ES08_​A04_​01.007–​8)

 

Ergativity and information structure in Jaminjung    1099 (15)

yirr-​angu yag, [murruny=mayan] 1PL.INCL>3SG-​get/​handle.PST fish in.heap=ITER ‘we (were making a camp and) caught fish, heaping it up’ (IP, ES08_​A04_​06.094)

Control phenomena thus single out the object of transitive clauses as opposed to the subject of transitive or intransitive clauses—​a nominative/​accusative pattern. A nominative/​accusative pattern is also displayed by an additional switch-​reference construction which applies to a sequence of finite clauses. This is marked by an enclitic on the inflecting verb, which signals non-​identity between the S or A of the clause hosting the clitic and that of a preceding clause, as illustrated in (16) to (18). (16)

buwu ga-​rdba-​ny gugu=bina, guyug=biya  (S ≠ S) dive 3SG-​fall-​PST water=ALL fire=SEQ ngarnarnabma ga-​rna=ni gurunyung=gi glowing 3SG-​burn.PST=DS head=LOC (of a mythical man who had tied up a firestick in his hair when crossing a river) ‘he dived into the water, (but) the fire kept burning on his head’ (IP, ES01_​A03_​07.006-​7)

(17)

ngab gan-​ma, naru nga-​rdba-​ny=ni  (A ≠ S but O = S) miss 3SG>1SG-​hit.PST dodge 1SG-​fall-​PST=DS ‘he missed me (when trying to throw something at me), (because) I dodged’ (EH, ES03_​A02_​01.375)

(18)

ERa: gub come.off

gani-​bili, ngali?  3SG>3SG-​POT:get/​handle TAG

(A ≠ A)

JM: dugulg gani-​minda-​ny=ni swallow 3SG>3SG-​eat-​PST=DS (watching a video of a woman trying to pull a hook out of a fish she has caught): ERa: ‘she wants to pull it (the hook) out, right?’—​JM: ‘it (the fish) swallowed it’ It should be noted that the different subject marker is not obligatory in such contexts, and moreover appears to have a wider application of indicating lack of topic continuity, rather than necessarily flagging different subject status in the grammatical sense. An example is (19), from a description of a number of photos showing the speaker’s family engaged in building a shade. The two sentences—​which are separated by a distinct pause—​have in fact the same subject but represent descriptions of two different scenes. (19)

gujarding-​guluwa gurdij ga-​yu Jidipimen \ mother-​KIN2 stand 3SG-​be.PRS CDEP:man ‘(here look), your mother is standing up, the CDEP person’ (… 3sec pause, new photo selected)

 

1100   Eva Schultze-Berndt ngiya=biya gujarding-​guluwa tharda ga-​yu=ni bayi-​bayirr PROX=SEQ mother-​KIN2 face.away 3SG-​be.PRS=DS RDP-​supported ‘here is your mother facing away, being all over the top (of the roof)’ (IP; ES97_​A03_​10.032–​33) Intriguingly, the different subject marker actually shares a form with the ergative marker, and both plausibly originate in a third person singular pronoun or deictic, an observation that we will return to in section 44.5. In sum, there is no evidence for an absolutive (or ergative) pivot in any complex construction, and therefore Jaminjung cannot be considered a syntactically ergative language. In fact, two complex constructions—​a control construction involving allative-​ marked non-​finite subordinate clause, and a switch-​reference construction applying to a sequence of finite clauses—​display nominative/​accusative alignment.

44.4  Factors Influencing Morphological Ergativity The phenomenon known as “fluid A  marking” or “optional ergative marking” has received considerable attention in the recent linguistic literature, as has the wider research area of which it forms a part, differential argument marking (see section 44.4.1 for further discussion and references). As already indicated in section 44.1, Jaminjung has a three-​way differential agent marking system, with the options of ergative marking (the most frequent option, found on 72% of the agents in a preliminary dataset), ablative marking (the least frequent option, found on less than 7% of agents), and zero marking (22%). This section provides an overview of the—​multiple—​factors triggering the presence or absence of case marking on Agents in Jaminjung. It will be argued that differential agent marking in Jaminjung is sensitive to information structure, in addition to various factors that have long been recognized as triggering split agent marking: verb semantics, tense and aspect, person, and animacy. The discussion of optional ergativity in Jaminjung presented here is based on a preliminary discourse study of a total number of 32 texts (amounting to approximately 6,600 intonation units) representing a range of discourse genres including mythological and personal narratives, procedural texts, retellings of picture books, other descriptions of visual stimuli, and conversations. These were manually scanned for overt Agent expressions (as defined in section 44.2.3), and the relevant examples coded for information structure category of the agent phrase (argument focus, broad focus, and topic), person, animacy in the narrow sense, their formal expression (pronoun, pronominal demonstrative, noun phrase with lexical head, and, for objects, also quotation and zero), constituent

 

Ergativity and information structure in Jaminjung    1101 order, tense/​aspect/​modality (using the inflectional categories of the verb as described in section 44.2.2), semantic class of verb, and speaker. The number of clauses with Agent phrases in this set of texts is relatively low (282, amounting to approximately one agent noun phrase in 20 intonation units) which is not unexpected given cross-​linguistic observations about the frequency of overt agent noun phrases in any language allowing for argument omission. An expanded study, based on a larger dataset and detailed discussion of the factors identified for Agent marking, is presented in Schultze-​Berndt and Meakins (in prep.). The following discussion reports on the most robust findings. In order to contextualize the discussion, I will briefly review the literature on optional ergativity and the factors identified in studies of other languages (section 44.4.1). Section 44.4.2 deals with the semantic factors which can been identified as responsible for the presence vs. absence of agent marking (subsuming the Ergative and Ablative variants). Section 44.4.3 provides a brief definition of the information structure categories employed and discusses their influence on overt agent marking. Section 44.4.4 examines in more detail the conditions under which the (infrequent) Ablative marking of agents is found.

44.4.1 Variable Ergative Marking Cross-​Linguistically Fluid A marking (“optional” ergative marking) is attested in many ergative languages of Australia, New Guinea, and the Himalayan region, but also outside these regions, e.g. in languages of the Chibchan family (Quesada 1999) and in the Kawapanan language Shiwilu, spoken in the Peruvian Amazon (Valenzuela 2011). Factors that have been described as influencing the presence vs. absence of ergative case in these languages include those that had long been identified as underlying clear-​cut split systems. One of these is verb semantics and/​or degree of affectedness of the object, in that some verb classes require or favour ergative marking, e.g. the verb ‘know’ in the Tamangic language Manang Gurung (Bond et al. 2013). A second factor is tense and aspect. For example, obligatory ergativity in past perfective contexts, combined with optional ergative marking in non-​past and imperfective contexts, has been reported for languages of the Himalayan region including Nepali (Li 2007) and spoken Lhasa Tibetan (Tournadre 1995; Denwood 1999: 195–​198); again this distribution is in line with that described for tense/​aspect-​based split ergative systems. A third factor is the position of the agent argument on the referential (empathy or “animacy”) hierarchy (local pattern), or the relative position of agent and undergoer on the hierarchy (global pattern). Both are considered relevant for Kuuk Thayorre by Gaby (2008a). In many fluid A languages, ergative marking is in fact obligatory for inanimate As such as natural forces, e.g. Gooniyandi (McGregor 1992), Umpithamu (Verstraete 2010), Gurindji Kriol (Meakins 2011: 225), Nepali (Li 2007; Bickel 2010: 407), Fore (Scott 1986) and Qiang (LaPolla and Huang 2008: 80) (see also McGregor 2010: 1617). For referents other than inanimates, animacy in the narrow sense rarely appears to be relevant for differential A marking of any sort (Fauconnier and Verstraete 2014: 14–​16). However, the distinction between speech act participants and third persons—​a frequent factor in

 

1102   Eva Schultze-Berndt split ergative systems especially in the Australian context—​also plays a role in the distribution of ergative marking in “fluid A” languages, e.g. Japhug Rgyalrong (Jacques 2010); this also applies to Jaminjung. In addition, a growing number of studies point to the importance of factors related to discourse organization and speaker and hearer’s common ground in accounting for variable ergative marking. This parallels a similar recognition of information structure as a factor responsible for differential object marking (Nikolaeva 1999, 2001; Iemmolo 2010; Dalrymple and Nikolaeva 2011). Due to its dependence on the discourse context, variable agent marking of this kind cannot result in a clear-​cut split based on traditional grammatical or semantic factors in the same way as e.g. verb class and animacy; in other words, these factors will not be found in systems traditionally classified as “split”. Recent overviews and further references can be found in McGregor (2010), Chelliah and Hyslop (2011), DeLancey (2011), Malchukov (2008a) and Fauconnier and Verstraete (2014). The discourse–​pragmatic function of ergative marking has been described in different terms by different authors. Notions that are often invoked to account for the presence of ergative marking are “focus/​new information” (e.g. Tournadre 1991; Tournadre 1995; Denwood 1999: 197; Malchukov 2008a; Chelliah 2009; Hyslop 2010: 13–​17; Suter 2010; Verstraete 2010; Fauconnier and Verstraete 2014), contrast (e.g. Tournadre 1991; Chelliah 2009; Jacques 2010), topic/​actor switch (e.g. Lidz 2011; Bond et al. 2013), or expectedness of the agent (McGregor 1992, 1998; Gaby 2008a). Some authors employ a notion of prominence or (contrastive) emphasis, cross-​cutting focus, and topic categories (Tournadre 1991, 1995; LaPolla and Huang 2008; Meakins 2009: 78; Meakins and O'Shannessy 2010; 2011: 228–​236). The notion of “prominence”, and a similar notion of “(argument) strength” is given an even wider interpretation—​variably encompassing a high rank on the animacy hierarchy and discourse topicality as well as perfectivity of the clause and volitionality of the agent referent—​in some recent works addressing differential argument marking from an optimality-​theoretical perspective (Legendre et al. 1993: 684–​688; Aissen 1999b; de Hoop and Malchukov 2007). As de Hoop and de Swart (2009: 14) point out, employing such a broad notion of prominence leads to the somewhat unsatisfactory conclusion that in some languages it is highly prominent agents that are case-​marked and in others, agents low in prominence. Languages exhibiting discourse–​pragmatic uses of agent markers also vary in whether or not these are extended to the single arguments of intransitive verbs, thereby raising the question of the appropriateness of an “ergative” analysis of the relevant construction (Tournadre 1991: 100–​102; Li 2007; Gaby 2008a; Meakins 2009; Gaby 2010; Hyslop 2010; Suter 2010; Meakins 2015). Of particular interest are reported cases of “ergative” markers with uses which extend even further from that of case marker of transitive agents, and which are therefore analysed as general markers of contrastiveness or focus. Examples are Meithei (Chelliah 2009)  and Jingulu, a language remotely related to Jaminjung and featuring an ergative /​focus marker which is cognate with the Jaminjung ergative marker (Pensalfini 1999). As indicated in the previous discussion (section 44.2.1), in Jaminjung ergative marking is restricted to transitive predicates.

 

Ergativity and information structure in Jaminjung    1103 The fourth factor just discussed, discourse–​ pragmatic function and information structure, is shown to interact with some of the other factors in many studies of “optional” ergativity. To give just a few examples, in Kurtöp, some verb classes require ergative marking of agents but for others, ergative marking signals contrastive focus (Hyslop 2010). Jacques (2010: 131) claims that for Japhug Rgyalrong, 1st and 2nd person pronouns rarely receive ergative marking, but if they do it is in contrastive function. In Qiang, ergative marking is obligatory with inanimate agents, but is only present in an “emphatic” function with other categories on the animacy hierarchy (LaPolla and Huang 2008: 76–​88). The conditions favouring the presence vs absence of case-​marking on agents in Jaminjung are of a similar multifactorial nature, as will become clear in the next three subsections.

44.4.2 Semantic Factors Responsible for Presence vs. Absence of Agent Marking The factors to be considered in this section are verb class, tense and aspect, and person and animacy of the agent referent. These are the semantic factors identified in 44.4.1 as having parallels in systems with obligatory split ergativity. This will be followed in 44.4.3 by an investigation of the role of information structure in favouring the presence or absence of ergative marking. For expository reasons, in this and the following section, Ergative and Ablative marking will be subsumed under the category of “(case-​)marked”. Ablative marking will be considered separately in section 44.4.4. Considering only the semantics of the (simple or complex) predicate, the transitive verb of possession (–​muwa ~ –​ma ‘have’) and expressions of speech in conjunction with a quotation behave strikingly different from all other predicate classes in Jaminjung. Whereas the frequency of case-​marking (as opposed to zero marking) lies between 80 and 90% for all other predicate classes, with transitive possession and direct speech expressions only just over 50% of agents are case-​marked. The most plausible explanation for this distribution is “effectiveness” in the sense of Tsunoda (1981b) and DeLancey (1981), i.e. degree of affectedness of the undergoer. This is low in the case of “uncontrolled” possession, i.e. inalienable (part–​whole) relationships, as in (20) (incidentally one of the rare cases where inanimate As can be unmarked; see further below), or the case of animals having eggs or young, which accounts for many of the zero-​marked examples in the dataset. (20) [gardawalng]TOP gana-​ma-​ya egg 3SG>3SG-​have-​PRS ‘the egg has little things inside for kids’

wuju-​wuju mali jalig-​gina RDP-​small thing child-​POSS (describing kinder surprise eggs)

In contrast, whenever –​muwa ~ –​ma ‘have’ is used with an animate object in a sense of ‘look after, keep’, as in (21), it always occurs with a case-​marked agent in the dataset.

 

1104   Eva Schultze-Berndt Note however that the contrast is not clear-​cut—​in the sense that one could assume two distinct senses of ‘have’ with two distinct case-​frames—​because even part–​whole relationships can be expressed with case-​marked agents/​possessors, e.g. under conditions of focus (see 44.4.3). (21) [galwarrang rait] TOP .. gana-​ma-​ya [^jayiny girl right 3SG>3SG-​have-​PRS mother’s.mother yinju=ni]FOC PROX=ERG ‘the girl, all right, she is looked after by her grandmother’ (VP, ES99_​V01_​06a.095) As already mentioned, a high ratio of unmarked agents is found in clauses involving the generic verb -​ju(nggu) ‘say/​do’, with reported speech filling the role of undergoer, as illustrated in (2). Formally, this verb has the main transitive properties—​the ability to occur with ergative marking, and the use of the transitive person prefix paradigm. It is however a defective transitive verb in that it fails to form a reflexive/​reciprocal (see section 44.2.1). Moreover, the quotation is not a typical object, and the clause containing the agent phrase may be more appropriately described as a “framing” construction than as a main clause with embedded reported speech (McGregor 1994; Rumsey 2010). Prosodically, the speech framing constructions can have “parenthetical” status as reflected in an overall lower pitch, narrower pitch range, and lower intensity. In addition to the overall effects of particular predicates, it is quite possible that the presence vs. absence of agent marking can indicate a difference in affectedness of the object even with the same predicate, as has been argued for Gooniyandi by McGregor (1992, 1998). This possibility is illustrated in (22) and (23), which feature the same inflecting verb of motion –​unga ‘leave’, but encode events which differ in impact on the object left behind: a pet frog leaving its ‘owner’ in (22) and a mother leaving her child—​even if temporarily, to go hunting—​in (23). Note however that the two sentences also differ in information structure. (22) [malara=biya]TOP dibard ganuny-​ngunga-​m, ba-​ngawu frog=SEQ jump 3SG>3DU-​leave-​PRS IMP-​see ‘the frog now is leaving the two jumping away, look’ (Frog Story; IP, ES97_​A03_​01.047) (23) majani [gujarding=ni waj maybe mother=ERG leave.behind ‘maybe the mother leaves her (a child)’

gan-​unga-​m]FOC 3SG>3SG-​leave-​PRS (VP, ES99_​V01_​06a.078)

Such more fine-​grained effects have been left out of consideration in the overall analysis. However, the saliently distinct behaviour of the two verbs –​ju(nggu) ‘say/​do’ and

 

Ergativity and information structure in Jaminjung    1105 –​muwa ~ –​ma ‘have’, compared with other transitive verbs, will be kept in mind when investigating the remaining factors responsible for ergative marking. The next potential factor to be discussed is tense, aspect, and modality. Here we find a contrast between past perfective clauses and all other inflectional tense/​aspect/​modal categories (see section 44.2.2): if one excludes clauses with quotations as objects—​for the reasons discussed above—​case-​marking is nearly always present in past perfective clauses (94% in the preliminary study) but only present in around 75% of clauses for all other categories. (Note however that periphrastic imperfective constructions categorically have unmarked agents, as discussed in section 44.2.2.) This is of course precisely what is predicted on the basis of what is known about split ergative systems—​if there is a split on the basis of tense or aspect, past and perfective categories will have ergative-​marked agents, and non-​ past and imperfective categories will have zero-​marked agents. In the case of Jaminjung, the (relatively weak) tendency mirrors the split found in other languages. Modality, on the other hand—​described as a factor in fluid ergativity for languages such as Meithei (Chelliah 2009: 389)—​does not play a role in Jaminjung: modal clauses with future time reference are not any less likely to have ergative-​marked Agents than non-​modal clauses. Finally, split ergativity is frequently associated with position on the so-​called animacy (or empathy) hierarchy, although animacy in the narrow sense may not be the most important factor. Rather, splits are particularly likely to occur between 1st/​2nd person and all other persons (DeLancey 1981: 645; Mallinson and Blake 1981: 86; Aissen 1999b: 679; Song 2001: 170), or pronouns vs. lexical NPs (DeLancey 1981: 645; Comrie 2013a). Moreover, as indicated in section 44.4.1, in many languages with optional ergativity, ergative marking has been described as obligatory for inanimate agents. Jaminjung shows a very clear tendency for speech act participants (1st/​2nd person pronouns) to appear as unmarked agents, mirroring the pattern in consistently split ergative languages: in the preliminary study, only 17% of speech act participant pronouns were marked with either of the two agent cases. For all other human referents, on the other hand, agent marking is the clear default, found in 81% of all cases. For non-​human animates, the ratio increases to 87% of marked agents. Inanimates are almost invariably ergative marked, but this is not an exceptionless generalization, as illustrated by example (20). Proper names or kinship terms, just like other human categories, allow for either presence or absence of case marking; ergative marking with a proper name is illustrated in (24). (24) [^Ben=ni=biyang]FOC du gan-​ijja-​ny=burri gurunyung proper_​name=ERG=SEQ shoot 3SG>3SG-​poke-​PST=3PL head ‘BEN then shot it (an echidna) in the head for them’ (IP, ES08_​A04_​06.468) Thus, two clear trends emerge for the interaction of animacy with agent marking: near-​obligatory marking on inanimates, and zero-​marking as a default for 1st and

 

1106   Eva Schultze-Berndt 2nd person pronouns, both in line with cross-​linguistic expectations. It is moreover instructive to consider the exceptions to the generalization: all six instances of case-​ marked personal pronouns are NPs in narrow (argument) focus. This leads us to the next factor, information structure.

44.4.3 The Role of Information Structure in the Case-​Marking of Agents The literature reviewed in section 44.4.1 points to a role of information structure—​and possibly, additional factors of discourse organization—​in determining the presence or absence of case marking on agents in Fluid Ergative systems. This section discusses evidence for some influence of information structure on agent marking in Jaminjung. In order to base the investigation on reasonably clearly defined notions, the information structure category of overt agent phrases was determined on the basis of contextual criteria—​to be briefly introduced below—​in combination with prosodic criteria and constituent order. This builds on the detailed, quantitative prosodic analysis by Simard (2010), which establishes the prosodic correlates of topic and focus phrases in Jaminjung. In line with convergent views in recent research, the analysis assumes a definition of focus as the expression that fills a variable in an open proposition, the latter corresponding to the background part of the clause (which can be conceptualized as an implicit, or sometimes explicit, question under discussion). In other words, a focused item evokes alternatives that are relevant for its interpretation, and the focused item is the candidate that is selected from these alternatives to fill the variable in the question (e.g. Rooth 1992; Dik 1997: 328; Roberts 1998 [1996]; Krifka 2006: 18). Two subtypes of focus were distinguished based on scope: narrow focus or argument focus (the focused constituent is a noun phrase), and broad focus (the focused part of a clause can be identified with the entire comment on an explicit or implicit topic, or with the entire assertion). Due to the lack of clear-​cut correlates, information focus and contrastive focus were not distinguished for the purposes of this study. Focused items in Jaminjung are characterized by a pronounced falling contour (marked by ^ in the examples), the onset of which is aligned with the first prosodic word in the scope of focus belonging to a lexical category (Simard 2010: 221, 296). Prosodic phrasing over and above the correlates of noun phrases (see Simard 2010: 146–​165) was not specifically investigated, but at least impressionistically, there is no difference in prosodic phrasing between focused elements involving or not involving case marking of agents. The text fragment in (25) illustrates three instances of broad focus, of which the first includes an (ergative-​marked) Agent phrase. (25) [^dibard=biya jump=SEQ

gani-​b-​arrga 3SG>3SG-​POT-​approach

mugmug=ni]FOC owl=ERG

 

Ergativity and information structure in Jaminjung    1107 [^wirrij ga-​yu=nu]FOC [ba-​^ngawu mung!]FOC angry 3SG-​be.PRS=3SG.OBL IMP-​see look.at ‘he is going to be jumped at by the owl, it is angry at him, look!’ (IP, ES97_​A03_​01.230) The staged dialogue in (26) illustrates agent phrases—​again, all ergative-​marked—​in narrow (argument) focus; it also shows that constituents in argument focus are variable in their position. Constituent order appears to be related to accessibility of a discourse referent rather than to the information structure category or a feature like exhaustiveness or contrastiveness. (26) “[nanggarni]FOC gan-​uga=rrgu ngarrgina dubuluj?” who:ERG 3SG>3SG-​take.PST=1SG.OBL 1SG:POSS bag imin     sei    from dijan,  dijan 3SG:PST say from     this ‘ “who took my bag from me?” she said, this one (did, to) that one’ “aa majani [jarlig= ^burlu=ni]FOC .. ah maybe child=COLL=ERG ‘ “ah, maybe the children took it” ’

burr-​uga” 3PL>3SG-​take.PST

(…) “yawayi, yawayi, burr-​uga yes yes 3PL>3SG-​take.PST (…) ‘ “yes, yes, the children took it” ’

[jarlig=burlu=ni”]FOC child=COLL=ERG (JM, ES08_​A20_​03.080-​86)

Agent phrases occurring as afterthoughts, i.e. separated from the remainder of the clause by a strong prosodic break, were subsumed under the category of narrow focus (for an example see (7)). Afterthoughts serve to assert additional information which semantically belongs to the preceding unit. Afterthoughts in Jaminjung are distinguished from right-​dislocated topics (see below) by the presence of a pitch reset and a focal contour (Simard 2014); this distinction is recurrent cross-​linguistically (Aijmer 1989; Ziv 1994; Averintseva-​Klisch 2008). The analysis of topics is based on a generally accepted definition of a (sentence) topic as an overt constituent denoting “what the sentence is about”; the topic specifies an address (or “filing card”) in the common ground between speaker and addressee to which the information in the remainder of the utterance is to be added (Reinhart 1981; Lambrecht 1994; Jacobs 2001; Krifka 2006; Maslova and Bernini 2006). For the purposes of this overview, no further distinctions e.g. between contrastive and shifted topics are applied. As is cross-​linguistically common, topical constituents occur in clause-​initial position in Jaminjung. They are identifiable as topics because they precede the constituent associated with a focal pitch contour at its left edge, as illustrated for an agent noun

 

1108   Eva Schultze-Berndt phrase in (27) and (22), and for an object noun phrase in (28). The prosodic contour of the topic constituent itself can be variable; while it often ends in a rise (indicating continuation), this is not necessarily the case. (27) [yirri=biya]TOP-​L [^munuwi yirr-​arra-​m]FOC 1PL.EXCL=SEQ native.bee 1PL.EXCL>3SG-​put-​PRS ‘as for us, we call them munuwi’ (as opposed to speakers of other languages who have other terms for the native bee) (IP, ES97_​A03_​01.188) (28) [gunurr]TOP-​L ^[jarlig=ni bag child=ERG ‘the bag, a child is carrying it’

gan-​antha]FOC 3SG>3SG-​take.PRS

(IP, ES08_​A20_​01.093)

Topical constituents can also be positioned at the right edge of a clause in Jaminjung; these signal reactivation of a topic rather than shift to a new topic (Simard 2014), again in line with cross-​linguistic observations (Givón 1983a; Ziv 1994; Lambrecht 2001). Right-​edge topics are distinguished from afterthoughts (see above) by a flat, non-​focal contour. In the Frog Story retelling from which (29) is taken, the dog has been a protagonist from the beginning of the narrative, and has been frequently mentioned. This example also shows that a right-​edge topic (here, the agent) can co-​occur with a left-​edge topic (here, the object). (29) [janju=malang mulanggirrng]TOP-​L [jag na gan-​angu,]FOC DEM=GIVEN dangerous go.down now 3SG>3SG-​get/​handle.PST [wirib=ni=malang]TOP-​R dog=ERG=GIVEN ‘those dangerous ones (i.e. bees), he got them down, the dog’ (CP, ES96_​A18_​02.069) Importantly, information structure as defined here is distinct from referential status in discourse, as captured by categories such as givenness, accessibility to the hearer, specificity, or definiteness. There are of course well-​known correlations e.g. between topicality and high accessibility, and discourse-​new referents and focal status, but these are not absolute. Referential status was not investigated, partly because Jaminjung lacks clear formal correlates such as definite articles. The exclusion of definiteness also seems justified by the result of a recent study reporting no influence of definiteness on differential agent marking (as opposed to differential object marking) cross-​linguistically (Fauconnier and Verstraete 2014). The investigation of the interaction of information structure categories and agent marking in Jaminjung reveals clear trends but—​again—​no clear-​cut association, in other words, ergative marking is variable with all information structure categories. Excluding from consideration clauses with quotations as objects, which favour zero-​marking (section 44.4.2), and moreover in most cases feature agent phrases in the background rather than as topic or focus, agents are almost exclusively case-​marked under both broad and narrow focus (in well over 90% of cases in the preliminary study). Most of

 

Ergativity and information structure in Jaminjung    1109 the exceptions are speech act participant pronouns—​in other words, the strong tendency for focal agents to be marked can be overridden by the factor person (see section 44.5 for further discussion). Prosodically integrated agent NPs in argument focus and afterthoughts do not differ in their likelihood of attracting case marking. Agent phrases as topics are much less likely to be case-​marked, though marking is still the default (approximately two thirds of all topical agents receive case marking). The preliminary study suggests some effect of prosodic detachment, with prosodically (left-​ or right-​) detached topics more likely to be zero-​marked. In addition, the semantic factors discussed in section 44.4.2 play a major role in determining the presence or absence of case marking on topical agents (for details see Schultze-​Berndt and Meakins in prep.); for example, in the preliminary study, no first or second person pronoun in topic position receives ergative marking. In sum, in addition to the semantic factors of verb class, tense/​aspect, person, and animacy, a basic hierarchy of information structure categories (Focus > Topic) allows one to predict the likelihood of overt agent marking. Focal agents are almost invariably marked; this constraint is overridden mainly by pronominal agents. As a default, topical agents are still case-​marked, but in addition to prosodic detachment, the semantic factors favouring zero-​marking are more likely to exert their force. Potential explanations for these findings will be discussed in more detail in section 44.5. In the following section, we will briefly investigate the factors underlying ablative-​rather than ergative-​ marking of agents in Jaminjung.

44.4.4 Focal Ablative Marking of Agents In this section, we will return to the issue of a further differentiation within the “case-​ marked agent” category, achieved by the use of two contrasting case markers. The low overall proportion (below 7%) of Ablative as agent marker strongly suggests that it is constrained in its distribution compared with the Ergative. The first such constraint is a restriction of Ablative agent marking to animate entities,3 while, as shown in section 44.2.3, ergative-​marking is regularly found on inanimate agents. The second striking finding is that the majority (68% in the preliminary study) of Ablative-​marked agents in the dataset fall into the category of argument focus, whereas only 44% of all ergative-​marked agents were instances of argument focus (although due to the higher total frequency of ergative marking, the majority of agents in argument focus are still ergative-​marked). An example is (30). (30) “ba-​manggu Imp-​hit

[^nami=ngunyi]”FOC 2SG=ABL

3  In its spatial function, naturally, the Ablative case occurs frequently on inanimates.

 

1110   Eva Schultze-Berndt ^burru-​yu=ngunggu ^jarlig=ni=gun, ngih? 3PL>3SG-​say/​do.PST=2SG.OBL child=ERG=CONTR TAG ‘ “kill it YOURSELF!” they said to you, the children, didn't they?!’ (account of echidna hunting, the children being squeamish about handling the spiky creature) (EH, ES08_​A04_​06.467) This difference in distribution suggests that Ablative marking is favoured by contexts where the identity of the agent is noteworthy or under discussion. This in fact also holds for the few cases where the Ablative-​marked agent falls under a different information structure category. All three examples of topical Ablative-​marked agents in the preliminary study involve an explicit contrast between two topics, as in (31). (31) [waitbala=ngunyi]TOP Blue Bone burr-​arra-​m white.person=ABL Blue Bone 3PL>3SG-​put-​PRS ‘(what do we call that fish species?), the white people call it Blue Bone’ (EH, ES08_​A04_​06.087) Given the low number of Ablative-​marked agents in the preliminary dataset (and in the corpus data more generally), and the lack of any historical documentation of the rise of this construction, these observations do not allow for firm generalizations. The striking correlation between Ablative-​marking and argument focus, however, suggests that this infrequent construction is mainly used to indicate unexpected or contrastive agents, in a way that is very similar to the focal ergative marker in Warrwa (McGregor 2006; McGregor and Verstraete 2010: 1614–​1615). Moreover, while Ablative and Ergative both serve as agent markers, they do not correspond to the same semantic role. The Ergative encodes the semantic role of effector, encompassing any participant which serves to bring about an effect, including natural forces, other inanimates, and instrument (section 44.2.3), while the Ablative requires that the agent is animate and volitional.

44.5  Discussion and Implications The investigation of “optional” ergativity in Jaminjung presented here reveals yet another instance of “soft constraints mirroring hard constraints” (Bresnan et al. 2001)—​or possibly the reverse. In other words, the same factors that are also potentially relevant for consistently split ergative systems play a role in fluid or variable systems of agent marking (see also McGregor 2010: 1616): lower animacy of the agent, past perfective tense/​aspect, and a higher degree of effectiveness/​impingement of the event on an undergoer. In addition, the findings show a strong tendency for focal agents to be marked, which intriguingly manifests itself in the existence of an infrequent second, “focal” ergative marker (taking the form of the Ablative case) which mainly occurs in the condition of argument (narrow) focus.

 

Ergativity and information structure in Jaminjung    1111 These findings reflect a competition between two overarching constraints which has been invoked in recent applications of optimality theory to differential case marking: semantic Faithfulness, favouring overt marking of the agent role (corresponding e.g. to Aissen’s 1999b *∅, Zeevat and Jäger’s 2002  “Generation”, Malchukov’s 2008a “indexing”, and de Hoop and Malchukov’s 2007 “Identify”), and Economy, favouring the absence of any marking (corresponding e.g. to Aissen’s 1999b *STRUC and Zeevat and Jäger’s 2002 and Malchukov’s 2008a “Economy”). In some of these works, the interaction of the two constraints has been modelled by a notion of “argument strength” which directly correlates with the marking pattern, but in turn is determined by a—​not clearly specified—​language-​specific interaction of factors (de Hoop 1999: 103; de Hoop and Narasimhan 2009:  72–​76). The problem with this notion of argument strength is that it allows for the mapping but does not in itself provide an explanation for the cross-​linguistic patterns found in the data, or encompass predictions of the nature and direction of the factors underlying it. More interesting generalizations can be arrived at by examining the actual factors which seem to recurrently influence differential case marking—​in this case case-​marking on agents. Some of the factors also applying to Fluid Ergative marking in Jaminjung, but also to split and Fluid systems elsewhere, relate to the completedness of the action and affectedness of an object—​both favouring the presence of ergative marking (section 44.4.2). Others correspond to the factors of person and animacy, originally proposed by Silverstein (1986) and supported by neurolinguistic evidence (see Bornkessel-​Schlesewsky and Schlesewsky 2015 for a review). The discourse expectation (cf. Zeevat and Jäger’s (2002) “bias”) for speech act participants to be agents makes ergative-​marking redundant, i.e. Economy will prevail; conversely, the expectation for inanimates to be non-​agents will result in near-​categorical marking of inanimates in the agent role (section 44.4.2). The role of information structure is similarly motivated. An expectation for agents to be topics will result in a near-​categorical agent-​marking in focus position (section 44.4.3): given the definition of focus assumed here, the role of a given entity as an agent is, by definition, under discussion when the corresponding constituent is in the scope of focus. This favours the actual marking of the agent role, a constraint which can however be overridden by semantic factors, in particular person. It does not follow, however, that the identity of the agent can be predicted when the corresponding constituent is in topic position. Since in languages like Jaminjung, the tendency is for any argument to be elided if discourse-​given, the default for an agent in the role of discourse topic will be omission. If an agent phrase functions as an overt sentence topic, it is therefore usually because it contrasts with another potential agent, is a shifted topic, or has not been mentioned in the immediately preceding discourse. In this view, the relatively high rate of ergative marking on topics is still consistent with the overarching motivation of unexpectedness favouring overt marking of the agent role. Under the account just presented, ergative (or ablative) marking does not serve to mark focus, any more than it serves to mark a low rank in the animacy hierarchy. These conditions are merely those where the entity in the agent role is less expected to fill this role, and hence more likely to be marked (because of the intrinsic association of focus

 

1112   Eva Schultze-Berndt with an open variable in the first case, and the heuristics of associating agents with higher animates and in particular with speech act participants in the second case). Rather, synchronically, Ergative marking is closely tied to the (language-​specific) semantic role of effector (see section 44.2.3). This language-​specific semantics comes into play over and above the general tendencies discussed above governing the likelihood of agent marking in general, and thus constrains their applicability. Thus, while in some languages only animate, controlling, or volitional agents may receive ergative marking (or indeed appear in first argument position), in Jaminjung the Ergative marker is compatible with all participants which have the semantic role of “effector”, i.e. which play a causal role in bringing about an event impinging on an undergoer (which includes instruments); volitionality or control is entirely irrelevant. A distinct semantic representation is required for the second agent marker (formally identical to the Ablative), since it is only found with animate, controlling agents which are the ultimate cause of an event. The factors of verb semantics, tense/​aspect, animacy, and information structure merely make the marking of the role more relevant, or conversely make omission of the markers, under considerations of economy, more permissible. There is however some intriguing evidence that the diachronic origin of the Ergative marker =ni is more directly related to marking expectedness in discourse than it is to marking the semantic role of agent. The origin of the Ergative marker in Jaminjung/​ Ngaliwurru is, in all likelihood, a third person pronoun/​demonstrative; compelling comparative evidence for this claim is gender agreement of ergative case in the related languages Nungali and Jingulu (Chadwick 1976; Pensalfini 1999; McGregor 2008). Synchronically, in Jaminjung/​Ngaliwurru, the pronominal ni is retained in 3SG>3SG verbal prefix gani-​and is a plausible origin for the verbal enclitic =ni marking switch-​ reference across finite clauses (see section 44.3 for discussion and examples). Plausibly, the discourse use of a third singular pronoun to disambiguate reference in the case of a switched or new subject/​agent could be the origin (via distinct pathways of grammaticalization) for both the switch-​reference construction and the Ergative marker (the latter by an association of switched or new subjects in apposition with the pronoun with the role of transitive agent, a scenario discussed in more detail by McGregor (2008: 311–​ 316)). For a number of other Australian languages, too, there are claims that markers of a special discourse status have been reanalysed as ergative markers, or vice versa (Jingulu, Pensalfini 1999; Kuuk Thaayorre, Gaby 2010). In sum, the findings for Jaminjung/​Ngaliwurru can be accounted for well by the assumption of a language-​specific semantics of the Ergative marker (or more specifically, the two agent markers in this language), in conjunction with universal but “soft” constraints on the presence vs. absence of marking, which include information structure. They raise issues, on the other hand, for a strict separation of lexical and structural case, since the lexical semantics of the predicate can license Ergative case, but not determine its presence. The factors that do favour or disfavour its presence, moreover, are not related to structure in the strict syntactic sense. Even if information structure was incorporated into the structural assignment of case, the factors of position on the animacy hierarchy and tense/​aspect still need to be accounted for.

 

Ergativity and information structure in Jaminjung    1113 Finally, our findings give rise to a cautionary note about the data underlying claims about argument marking systems—​specifically the way in which elicited data may miss discourse organizational factors responsible for case marking. This is a point also poignantly made by DeLancey (2011: 13) as well as by Simpson (2012), who points to information structure as a possible (but, given the current state of documentation, difficult to verify) factor in the diachronic development of the “unusual” Arrernte system which has ergative marking on 1st person pronouns, but not on other pronouns. It is an intriguing thought that the “discovery” of variable argument marking and the factors—​including discourse–​pragmatic factors—​underlying it correlates with the availability of corpora of sound-​linked, unedited speech from various genres, and the affinity of authors with corpus-​based analyses, as opposed to grammatical descriptions based on elicited utterances and heavily edited texts. In the case of Jaminjung, had the available data not contained instances of personal pronouns in argument focus, Jaminjung might well have given the impression of a language with a strict ergative split between pronouns and other types of nouns, since it turns out that pronouns in an Agent role are never marked outside the argument focus context (this is true not just for the restricted dataset in the preliminary study discussed here, but for every single one of the handful of cases of ergative-​marked pronouns found in a relatively large corpus). More generally, the study demonstrates the importance of the availability of both prosodic information and discourse context, as ensured by sound-​linked corpora.

Abbreviations The glosses used in this chapter are as follows: 1, 2, 3 1st, 2nd, 3rd person; ABL, Ablative; ABS, Absolutive; ALL, Allative; COLL, Collective; CONTR, Contrastive clitic; DAT, Dative; DEM, Demonstrative (distance-neutral); DS, Different subject (switch reference marking clitic); DU, Dual; ERG, Ergative; EXCL, Exclusive 1st person; FOC, Focus; GIVEN, Marker of discourse-given status; IMP, Imperative; INCL, Inclusive 1st person; INST, Instrumental; IPFV, Past Imperfective; ITER, Iterative; IV, Inflecting verb; KIN2, Kinship-term specific 2nd person possessive pronoun (“your kin”); LOC, Locative; OBL, Oblique pronominal clitic; PL, Plural; POSS, Possessive; POT, Potential/Future; PROX, Proximal demonstrative; PRS, Present tense; PST, Past (perfective or underspecified); RDP, Reduplication; SEQ, Sequential clitic (“now,” “and then”); SG, Singular; TAG, Question tag particle; TOP, Topic; TR, Transitivity marker (Kriol); UV, Uninflecting verb.

 

Chapter 45

Alignment a nd orientat i on i n Kart vel ia n (Sou th Cau c asia n) Kevin Tuite

45.1 Introduction Kartvelian (also known as South Caucasian) is one of the three endemic language families of the Caucasus (Daniel & Lander 2011), along with Abkhaz-​Adyghean (West Caucasian) and Nakh–​Dagestanian (East Caucasian). Despite numerous attempts to link some or all of these languages to larger families or phyla, no convincing demonstration has been made that the three endemic families are related to each other, nor to languages groups elsewhere (Schulze 1997; Nichols 1997, 2003: 208; Tuite 2008). Georgian, the national language of the Georgian republic, with over four million speakers and fifteen centuries of use as a written medium, is the best-​known member of the Kartvelian family. The other members are the closely related Mingrelian and Laz languages—​ grouped together as the Zan branch of the family—​and the outlier Svan.

45.2  The History of the Study of Kartvelian Alignment If the Georgian literary language and the modern vernaculars “present the longest history among the languages of the ergative type” (Boeder 1979: 435), Georgian has also been the object and the medium of a grammatical tradition going back centuries

 

Alignment and orientation in Kartvelian    1115 (Tsagareli 1873; Schuchardt 1895, 1896a, 1896b, 1897; Chikobava 1965; Iluridze 2006). I thus believe it will be of interest to survey the attempts, by native and foreign grammarians, to name and describe the case ending glossed as “ERG” in the present chapter,1 and, more broadly, to squeeze the oddly shaped peg of Georgian morphosyntax into the neatly symmetric holes provided by the grammatical theories of their times. To give an impression of the complexity of the facts confronting grammarians and linguists, here are examples from the Old Georgian Bible translations, illustrating some aspects of case assignment with transitive and intransitive verbs. In these two renderings of John 5:21, the first from the late-​ninth-​century Adish Gospels, the second from the mid-​eleventh-​century translation by Giorgi Mtac’mideli, the same transitive verb root appears in the future/​optative and present tenses: (1)

vitarca-​igi mama-​man aɣ-​a-​dg-​in-​n-​e-​s mk’wdar-​n-​i as-​DET father-​ERG up-​ApV-​stand-​CAUS-​Pl.N-​OPT-​V3sg dead-​PL-​NOM ‘as the father (ERG) raises the dead (NOM)’

(2)

vitarca mama-​j aɣ-​a-​dg-​in-​eb-​s as father-​NOM up-​ApV-​ stand-​CAUS -​SM-​V3sg ‘as the father (NOM) raises the dead (DAT)’

mk’wdar-​ta dead-​OBL.PL

The first verb contains the aorist stem (-​dgin-​), whereas the second is built from the present stem (-​dgin-​eb-​). Transitive verbs assign ERG case to their subjects in the aorist series of tenses, whereas they assign NOM case in the present/​imperfect series. As with Hindi–​Urdu and several Iranian languages, the case assignment properties of the verb are linked to aspect and/​or tense (see in this volume the chapters by Haig, Nash, and Butt (20, 8, and 33, respectively)). Compare the above pattern to those associated with intransitive verbs: (3)

mk’wdar-​n-​i aɣ-​dg-​e-​n uxrc’el-​n-​i [aorist stem] dead-​PL-​NOM up-​stand-​OPT-​V3pl incorruptible-​PL-​NOM ‘The dead (NOM) shall rise incorruptible.’ (I Corinth 15:52)

(4)

mk’wdar-​n-​i aɣ-​dg-​eb-​i-​an dead-​PL-​NOM up-​stand-​SM-​PASS-​V3pl ‘The dead (NOM) rise.’ (Mt 11:5)

[present stem]

(5) da i-​kux-​a upal-​man qm-​ita did-​ita [aorist stem] and SbV-​thunder-​V3sg.PST lord-​ERG voice-​INS great-​INS ‘And the Lord (ERG) thundered with a mighty voice.’ (I Kings (I Samuel) 7:10 ms B, 1743) 1  In keeping with the theme of this volume, I will apply the label “ergative” (ERG) to the Kartvelian case endings which mark, minimally, the subject of transitive verbs in the aorist series, even when their distribution has little to do with ergativity in the strict sense (Harris 2006).

 

1116   Kevin Tuite (6) da kux-​d-​a upal-​i ... qm-​ita k’rčxialeb-​isa-​jta [present stem] and thunder-​IMP-​V3sg.PST lord-​NOM voice-​INS howl-​GEN-​INS ‘And the Lord (NOM) thundered with a howling voice.’ (I Kings (I Samuel) 7:10 ms O, 978)

The Old Georgian verb meaning “stand up, rise” assigns NOM case to its single argument in all tenses. The verb kux-​ “thunder”—​ordinarily employed in impersonal weather expressions—​manifests the same shift in case assignment as the transitive verb in examples (1)–​(2). The three types of verbs illustrated above—​ERG-​assigning transitives, ERG-​assigning intransitives, and NOM-​only intransitives—​exist in Old Georgian and the modern dialects, as well as Svan and Laz (the special case of Mingrelian will be examined in 45.6.3). Further complicating matters, in a third series of tenses, including the present perfect, pluperfect, and perfect subjunctive, the agreement morphology, and case-​assigning properties of ERG-​assigning transitives and intransitives undergo what is commonly called “inversion”. The argument corresponding to the subject is assigned DAT case and controls object agreement (Set M) in the verb, whereas the direct object, if there is one, is assigned NOM case and controls subject agreement (Set V affixes). (7) brdzola-​j da dzleva-​j aɣ-​Ø-​u-​dg-​in-​eb-​ie-​s struggle-​NOM and effort-​NOM up-​M3-​ObV-​stand-​CAUS-​SM-​PERF-​V3sg ioane-​ s    mc’valebel-​ ta  zeda John-​DAT  heretic-​OBL.PL  on ‘John (DAT) has raised up struggle and effort (NOM) against the heretics.’ [Eprem Mcire, Cxovrebaj C’m. Ioane Damask’elisa, 11th c.] (8) arɣa x-​e-​q’ivn-​o-​s katam-​sa vidremde not-​EMPH M3-​ObV-​crow-​OPT-​V3sg rooster-​DAT until uvar=m-​ q’-​ o     me deny=M1sg-​do-​OPT  me ‘The cock (DAT) will not have crowed before you deny me.’          [Jn 13:38, Khanmeti]

Leaving out complicating factors to be presented later, the alignment situation in the Old Georgian language was as shown in Table 45.1. Table 45.1 Agreement and case marking in Old Georgian Tense/​aspect/​ mode series

Transitive verbs “A”

“O”

“B”

Intransitive I “S”

“B”

M+DAT V+NOM M+DAT

Intransitive II “S”

“B”

V+NOM

M+DAT

I. Present series

V+NOM M+DAT

II. Aorist series

V+ERG

M+NOM M+DAT V+ERG

M+DAT

V+NOM

M+DAT

III. Perfect series

M+DAT

V+NOM

—​—​

V+NOM

M+DAT

—​—​

M+DAT

 

Alignment and orientation in Kartvelian    1117 Three types of grammatical roles will be distinguished in the presentation to follow: 1. The semantic macroroles of A (“agent”, “subject” of transitive verb), O (“object” or “patient”), S (“subject” of intransitive verb), and B (“beneficiary”, encompassing various sorts of actants realized as indirect objects). These are in part inspired by Dixon’s “universal syntactic-​semantic primitives” (1994: 6), albeit without the accompanying claims of universality. Their primary purpose in this context is to keep track of actants across shifts in case assignment and agreement. 2. The Kartvelian verb has two sets of person agreement affixes. The clausal argument cross-​referenced by Set V agreement markers is the morphological subject (cf. Boeder 2004: 25). This argument is assigned either NOM or ERG case (as shown in Table 45.1). The argument (or arguments) linked to Set M agreement is the morphological object. There is a further distinction to be made between the indirect object, marked consistently by Set M agreement and DAT case; and the direct object (transitive verbs only), associated with shifting case assignment (DAT and NOM). The morphological subject is marked by the darker tint background in Table 45.1. 3. The role of syntactic subject is not as prominent in Kartvelian as it is in many West European languages. The most reliable tests for syntactic subjecthood in the Kartvelian languages are the binding of reflexive and reciprocal pronouns; and the argument denoting the addressee of imperatives (Harris 1981:  23–​27; Tuite 1998: 29–​31). The syntactic subject tends to appear toward the beginning of the sentence, although word order is quite free. I will employ the term direct syntax to characterize constructions where the syntactic subject coincides with the morphological subject; and indirect syntax when the morphological object has the attributes of syntactic subjecthood, as in examples (7) and (8).

45.2.1 Nominative–​Accusative Model Imposed on Georgian The earliest grammatical descriptions of Georgian (Maggio 1643, Shanshovani 1737, among others) employed the same categories as in Greek and Latin grammar (Iluridze 2006). The Georgian ERG suffix, if mentioned at all, was classified as a variant of the NOM.2 Since direct objects could also be assigned NOM case (in the aorist and perfect series), the suffix /​-​i/​, which is in fact an allomorph of the NOM, was listed as the accusative. The identification of the Georgian ERG as a type of nominative was to have a long history, persisting up to the grammars of Janashvili (1906: 15) and Zorell (1930: 92–​93), and resurfacing, with very different arguments, in Nebieridze (1987). 2  In the first printed grammar of Georgian, produced by the Italian missionary Francisco-​Maria Maggio (1643), the data themselves are distorted to fit a nominative–​accusative paradigm. The 3rd-​ person ERG pronoun iman was generalized as the only subject form, and applied across the board to all verb types in all tenses (e.g. the ludicrously ungrammatical iman aris ’s/​he-​ERG is’!).

 

1118   Kevin Tuite The first grammarian to describe the ERG as a case distinct from the NOM was Catholicos Anton I ((1767) 1885), who gave it the designation of “narrative” case (motxrobiti). This curious name reflects the high frequency of the ERG case in narratives of past events, where the aorist tense predominates. It was adopted by Ioane Bagrationi ((1828) 1936), and has continued to appear in the writings of Georgian linguists and language pedagogues up to the present day (Iluridze 2006: 45).3 Neither Anton I, nor his successors up to the turn of the twentieth century (Zhordania 1889; Xundadze 1901), took note of the distinctive patterning of the ERG case with verb class.

45.2.2 Ergativity Identified in the Kartvelian Languages The next significant breakthrough was due to the wide-​ranging interests of Hugo Schuchardt, who investigated the languages of the Caucasus after having already studied Basque grammar. Schuchardt extended the passive analysis of the ergative construction—​which had been applied previously to languages such as Basque and Inuktitut—​to the Kartvelian ERG, so that a Svan sentence of the form ‘man-​ERG horse-​NOM tied-​AOR’ was glossed “von dem Mann wurde das Pferd angebunden” (1895: 36). The interpretation of sentences with an ERG subject as fundamentally passive was adopted by Marr in the Soviet Union (Marr 1910, 1925; Marr & Brière 1931);4 and Western linguists such as Dirr (1928: 58). One challenge faced by proponents of the ergative analysis is the existence of intransitive verbs, as in (5), having the same case-​assignment properties as transitives. Linguists who ascribe an ergative–​absolutive construction to Georgian either (i)  treat active intransitive verbs as underlyingly or formerly transitive (Schuchardt 1895: 40, 70–​7 1; Rudenko 1940: 170–​174; Hewitt 1995; Nash 1995, Chapter 8, this volume); and/​or (ii) assume that the aorist–​series paradigms of active intransitive verbs, along with their morphosyntactic properties, were borrowed from transitive conjugations (Shanidze 1953: 363–​364, 486–​489; Vogt 1971: 134).

45.2.3 Kartvelian Languages as Active–​Inactive By the mid-​twentieth century, most foreign specialists in the field of Georgian or Kartvelian linguistics employed the term “ergative” to designate Anton I’s narrative case, including those who rejected Schuchardt’s passive analysis and the stadialist presuppositions of Marr and his disciples. The alignment of the language was described as what later would be called “split-​ergative”, with ergative–​absolutive alignment in the 3  As defined in the Georgian grammar of Ioane Kartvelishvili (c.1809), the narrative case indicates “of whom or of what the narrative is told (moitxrobis), or [by whom or by what] some act is caused” (cited by Iluridze 2006: 52). 4  On Marr and linguistic stadialism, see Sériot (2005, 2008); Tuite (2008, 2011).

 

Alignment and orientation in Kartvelian    1119 aorist series, and nominative–​accusative alignment in the present/​imperfective series (Deeters 1930; Tschenkéli 1958; Vogt 1971; Silverstein 1976; Dixon 1979). One early voice of dissent belonged to Aronson (1970), who questioned whether Georgian was ergative at all, pointing to the subject-​like prominence of the argument assigned ERG case (rather than the one assigned NOM case), the active–​passive voice distinction, and the existence of ‘middle verbs’ (= active intransitives) with ERG subjects. Shortly afterwards, and independently of each other, G. A. Klimov (1976, 1977) and Alice Harris (1981: 235–​246) came to the conclusion that the alignment of Georgian, Laz, and Svan in the aorist tense series would be most accurately described as active–​inactive rather than ergative–​absolutive. Although Klimov repudiated Marrist linguistic evolutionism, his theory of “contensive typology” was similarly informed by a unidirectional concept of morphosyntactic change. According to Klimov, a language’s dominant alignment (active–​inactive (i.e. split-​intransitive), ergative–​absolutive, nominative–​accusative) is grounded in fundamental structural features of its lexicon, such as the classification of verbal and nominal roots. At any given time, the grammar and lexicon of a language can bear traces of its past stages, alongside indications that it is undergoing shift toward a structure further along in Klimov’s hierarchy. The Kartvelian languages struck him at first as “polystadial”, having features characteristic of an early active stage, a more recent ergative stage, as well as a pronounced trend toward nominative typology (1973: 238). Klimov subsequently modified his view, attributing the “pre-​nominative” traits of the Kartvelian languages to an ancestor of the active type, from which they began acquiring nominative–​language characteristics without ever passing through an ergative phase (1976; 1977: 217–​230; 1979). About the same time as Klimov, Harris classified Georgian case alignment in the aorist series as active–​inactive (Harris 1981:  235–​246), although she rejected Klimov’s link between this type of alignment and a specific typological profile, and his evolutionary scheme (Harris 1985). In her view (1981, 1982, 1985, 1990), the case-​assignment properties of inactive and active intransitive verbs arise from their underlying grammatical relations, as these are understood in the Relational Grammar framework: Verbs such as aɣdgeb-​ ‘stand up’ have an initial direct object but no subject (“unaccusative” verbs), whereas active intransitives have an initial subject but no direct object (“unergative” verbs).5 As demonstrated in an exhaustive study by Holisky (1981), the semantic distinction between Georgian “unergative” (active intransitive) and “unaccusative” (inactive intransitive) verbs is not strictly along the lines of agentivity or stativity. Many intransitives that rarely if ever take animate arguments assign ERG case (c’q’al-​ma i-​duɣ-​a ‘water-​ERG boiled’; al-​ma i-​gizgiz-​a ‘flame-​ERG crackled’); whereas many inactive intransitive verbs take agentive subjects (st’umar-​i a-​mɣer-​d-​a ‘guest-​NOM 5 

Harris’ publications on Georgian syntax have also spawned a veritable cottage industry of reformulations of her unaccusative/​unergative analysis of Georgian alignment in various formal linguistic frameworks (Anderson 1984; Bittner & Hale 1996a; McGinnis 1997; Butt & King 2003; Blevins 2005; among others). Very few of these works contain Kartvelian data that has not already been discussed by Harris herself, so I will have no more to say about them here.

 

1120   Kevin Tuite began-​to-​sing’; gogo bič’-​s e-​saubr-​a ‘girl-​NOM conversed-​with boy-​DAT’). The principal distinction between the Kartvelian unergative class and the remaining intransitives is aspectual: Most active intransitive verbs are atelic, that is, they denote activities without reference to a beginning or end point.6 It should be pointed out, however, that the Georgian unergative class includes some statives (Holisky 1981: 152–​169), and also telicized verbs derived by the addition of preverbs (Tuite 1996). On the other hand, some agentive inactive intransitives appear with ERG subjects in informal usage or in non-​ literary Georgian dialects (Boeder 1979; Harris 1981: 270–​273). Representations of Georgian and Kartvelian morphosyntactic alignment could be said to have followed Klimov’s hierarchy in reverse, from nominative to ergative to active. In the following overview of Kartvelian grammar, I will seek to demonstrate that none of these models is entirely accurate.

45.3  Nominal Morphology: Declension By the mid-​nineteenth century, the case names had been established which are still used by Georgian linguists and schoolteachers. The Old Georgian declensions of common nouns and proper names is shown in Table 45.2 (Vogt 1947). Table 45.2 Case and number desinences in Old Georgian Case names

Common nouns Abbr

Short

Long

nominative (saxelobiti)

NOM

Ø

-​i

narrative (motxrobiti); ergative

ERG

-​man

dative (micemiti)

DAT

-​s

genitive (natesaobiti)

GEN

-​is

active (mokmedebiti); instrumental

INS

-​it

transformative (vitarebiti); adverbial

ADV

vocative (c’odebiti)

VOC

Names

Plural N/​T

EB

Ø

-​n-​i

-​eb-​i

Ø

-​t-​a

-e​ b-m ​ an

-​s-​a

-​s

-​t-​a

-​eb-​sa

-​is-​a

-​is

-​t-​a

-​eb-​isa

-​it-​a

-​it

—​—​

-​eb-​ita

-​(a)d

-​(a)d

—​—​

-​eb-​ad

-​o

Ø

-​n-​o

-​eb-​o

Old Georgian had two plural declensions: one formed by insertion of the suffix /​-​eb-​/​ before the case endings; and a fusional system opposing the rectus plural suffix /​-​n-​/​ to the oblique /​-​t-​/.​ The long endings are derived by the addition of the vowels /​-​i/​in the NOM, and /​-​a/​in the oblique cases, to the short case endings (Boeder 1995: 157–​159). According to some, the long endings are the vestige of postposed demonstrative 6 

Comparativists have sought to integrate the Kartvelian facts into more general theories of split-​ intransitivity; see among others Merlan (1985); Van Valin (1990); Lazard (1994); Arkadiev (2008a).

 

Alignment and orientation in Kartvelian    1121 elements, which had functions similar to those of definite articles (Klimov 1962; Harris 1988; Tuite 2004). Traces of the earlier distinction are especially evident in the distribution of the short NOM, equivalent to the bare stem, which appears in predicate nominals (tkwen xq’avt igi kwab-​Ø avazak’ta [Lk 19:46] ‘you made it a den of thieves’); naming constructions (romelsa hrkwian betlem-​Ø [Lk 2:4] ‘which they call Bethlehem’); and noun stems with generic reference incorporated into compound verbs (nu k’ac=k’lav [Mk 10:19] ‘thou shalt not kill’, literally: ‘thou shalt not person=kill’). The derivation of some components of Kartvelian case morphology from ancient demonstratives has been invoked to account for the rich allomorphy of Svan declension7 (Svan being the only Kartvelian language with true declension classes, as in Indo-​European or East Caucasian languages), and the surprising diversity of ERG case markers in the four Kartvelian languages. The case and number suffixes of the modern Kartvelian languages are shown in Table 45.3. Table 45.3 Case and number desinences in the modern Kartvelian languages Case

PKrt

Georgian

Mingrelian

Laz

Svan

sing.

plural

sing. plural

sing.

plural

singular

plural

NOM *-​i, Ø

-​i, Ø

-​eb-​i

-​i

-​ep-​i

Ø

-​epe

*-​i, Ø

-​ǣl, -​ær

ERG

—​—​

-​m(a) -​eb-​ma -​k

-​en-​k

-​k

-​epe-​k -​em, -​d

DAT

*-​s

-​s

-​eb-​s

-​s

-​en-​s

-​s

-​epe-​s

GEN

*-​iś

-​is

-​eb-​is

-​iš

-​ep-​iš

-​ši

-e​ pe-š​ i -​iš, -​eš

INS

?*-​it

-​it

-​eb-​it

-​it

-​ep-​it

-​te(n) -​epe-​te -​šw, -​wš

ADV

*-​d

-​ad

-​eb-​ad

-​o(t) -e​ p-o​ (t) —​—​

—​—​

-​d

-​ǣl-​d, -​ær-​d

VOC

—​—​

-​o

-​eb-​o

—​—​ —​—​

—​—​

—​—​

—​—​

—​—​

-​ǣl-​d, -​ær-​d

-​s, -​w, -​am -​ǣl-​s, -​ær-​s -​ǣl-​eš, -​(a)r-​eš -​ǣl-​šw, -​ar-​šw

In Old Georgian, and some conservative modern dialects, proper names have no endings in NOM and ERG contexts. Some linguists consider the declension properties of proper names to have once been characteristic of all types of nouns (Chikobava 1948; Boeder 1979), a hypothesis with implications for the reconstruction of Proto-​Kartvelian.   (9) asap šv-​a iosapat’ Asaph:“ERG” bear-​V3sg.PST Jehosaphat:“NOM” ‘Asaph begat Jehosaphat.’                     [Matt 1:8] (10)

7 

avtandil ked-​n-​i maɣal-​n-​i gada-​i-​ar-​a A:“ERG” range-​PL-​NOM high-​PL-​NOM across-​SbV-​go-​V3sg.PST ‘Avtandil crossed high mountain ranges.’ [(Glola Rachan dialect) Dzidziguri 1937:90]

The allomorphy of the Svan ERG and some other oblique cases is to an extent correlated with lexical class (pronouns vs. common nouns); cf. Arkadiev (2012).

 

1122   Kevin Tuite Since Maggio (1643: 68–​69), the Georgian 1st-​and 2nd-​person pronouns, as well as those of Svan, Laz and Mingrelian, are attributed paradigms marked by both syncretism and suppletion. A single form of each pronoun appears in the core grammatical contexts, where ERG, NOM or DAT case would be assigned. In the 1st singular for all Kartvelian languages, and the 2nd sing. for all but Georgian, an etymologically unrelated stem is used to form the GEN and other oblique cases (Klimov 1962: 105–​106). This odd configuration raises the question whether the rectus (ERG–​NOM–​DAT) pronominal stems originated as indeclinable particles, which can appear in precisely those contexts where personal affixes in the verb indicate the grammatical status of 1st-​and 2nd-​person arguments. The GEN, INS, and ADV cases are built from a distinct adjectival stem (Table 45.4). Table 45.4 1st and 2nd person singular pronouns in Kartvelian Svan

Georgian

Mingrelian

Laz

particle (NOM/​ERG/​DAT)

mi

me

ma

ma

adjective stem (other cases)

mišgw-​

čem-​

čkim-​

čkim-​

particle

si

šen

si

si

adjective stem

isgw-​

šen-​

skan-​

skan-​

1st singular

2nd singular

The Kartvelian languages have two (distal, proximal) or three (distal, 2nd-​person linked, proximal) sets of 3rd-​person pronouns (Mart’irosovi 1964). These also have suppletive paradigms, with a distinct rectus (NOM-​case) stem that probably originated as an indeclinable deictic particle (Table 45.5). Table 45.5 Partial paradigms of proximal pronoun (“this”) in Kartvelian (singular only) Georgian

Svan

Mingrelian

Laz

NOM

es(e)

ala

atena

aja

ERG

ama-​n

am-​nēm

ate-​k

amu-k​

DAT

ama-​s

am-​ən

ate-​s

amu-​s

GEN

am-​is

am-​iša

ate-​ši

amu-š​ i

45.4  Verb Morphology 45.4.1 Verb Types The Kartvelian verb has a more or less agglutinative structure, although segmental transparency can be obscured by vowel syncope, ablaut, and other morphophonological

 

Alignment and orientation in Kartvelian    1123 processes (especially in Svan). The architecture of the verb can be represented as layers of morphemes built outward from the root (Table 45.6). Table 45.6 The internal structure of the Kartvelian verb I. root

ROOT

II. stem formants

causative/​ passive, verbal plurality, series marker

III. verb class, tense/​mood

version

IV. person and number V. clitics and preverbs

imperfect, tense/​ mood

V/​M

person/​ number

preverbs, clitics

clitics

The verb classifications employed in recent grammars and dictionaries of Georgian and its sister languages are based to varying degrees on the system developed by Shanidze (1953:  433–​502). For Shanidze, the primary division among verb types was according to active, passive or medial “voice” (gvari), which can be reduced to a distinction between those verbs that can assign ERG case, and those that cannot. I will retain the label “active” for the former; rather than designate the latter as “passive”—​a label which can only be accurately applied to a subset of these verbs—​I will call them the “inactive” class (Table 45.7): Table 45.7 Kartvelian verb classes Verb class

Examples from Georgian

I. ERG-​assigning (‘active verbs’) a. Transitive

qn-​av-​ ‘plow’, č’am-​ ‘eat’, a-​nt-​eb-​ ‘light’

b. Medioactive

t’ir-​‘weep’, čkar-​ob-​‘rush’, livliv-​eb-​‘sparkle (light on water)’

II. Non-​ERG-​assigning (‘inactive verbs’) a. ‘Dynamic passives’ a. i. Root intransitive

dn-​‘melt’, k’vd-​‘die’, drk’-​‘bend’

a. ii. Suffixal intransitive

grdzel-​d-​ ‘become long’, mc’ip-​d-​ ‘ripen’

a. iii. Prefixal intransitive

i-​c’v-​‘is burning’, e-​mal-​v-​‘hide oneself from’

b. Stative passive

c’er-​i-​‘is written’, tes-​i-​‘is sown’

c. Mediopassive

q’var-​‘love’, gav-​‘resemble’, sxed-​‘(many) sit’

 

1124   Kevin Tuite

45.4.2 Person/​Number Affixes All Kartvelian languages have two sets of person/​number affixes. Following Shanidze (1953: 198–199), I will designate them as ‘Set V’ and ‘Set M’, after the 1st-​person prefixes. The prefixes indicate person and role ([morphological| subject or object; Boeder 2002: 91). The suffixes indicate number—​a distinction originally limited to the Set V markers—​and, in the 3rd-​person, tense, mood, and verb class (Table 45.8). Table 45.8 Person and number affixes in the Kartvelian verb Old Georgian

Georgian

Laz–​Mingrelian

Svan

Set V (“subject”) 1sg

v-​

-​Ø

v-​

-​Ø

v/​b-​

-​Ø, r, k

1pl

v-​

-​t

v-​

-​t

v/​b-​

-​t, rt

xw-​

-​Ø

excl

xw-​

-​d

incl

l-​

-​d

2sg

x/​h/​Ø-​

-​Ø

Ø-​

-​Ø

Ø-​

-​Ø, r, k

x/​ Ø-​

-​Ø

2pl

x/​h/​Ø-​

-​t

Ø-​

-​t

Ø-​

-​t, rt

x/​ Ø-​

-​d

3sg

-​s,a/​o,n,ed

-​s, -​a/​o

-​s, u, n

-​Ø, -​s

3pl

-​an,en,es,ed

-​an,en,es,nen

-​an,es,nan

-​x

Set M (“object”) 1sg

m-​

m-​

m-​

1excl

m-​

gv-​

m-​

-t​ ,an,es,nan

n-​

1incl

gw-​

gv-​

m-​

-t​ ,an,es,nan

gw-​

2

g-​

g-​

k/​g/​r-​

2pl

g-​

g-​

3

x/​h/​Ø-​

Ø/​s/​h-​

3pl

x/​h/​Ø-​

Ø/​s/​h-​

-​t

k/​g/​r-​

m-​

ǰ-​ -t​ ,an,es,nan

Ø-​ -​t

Ø-​

ǰ-​

-​x

x/​ Ø-​ -t​ ,an,es,nan

x/​ Ø-​

-​x

Georgian, Laz and Mingrelian, but not Svan, have paired sets of V3sg and V3pl suffixes. With the exception of conjunctive /​-​s/​, Svan shows no obvious cognates to any of the Georgian–​Zan V3 suffixes.8 Synchronically, the 3rd-​person singular affix for Svan non-​ conjunctive verbs can be represented as a zero morpheme. Although the 1st-​and 2nd-​person have both Set V and Set M prefixes, they compete for a single slot in the Kartvelian verb. Should a verb have a 1st-​person (morphological) 8 

Kaldani (1978) surmised that a lost suffix *-​a could account for the stem vocalism in the V3sg of some Svan aorists, but such a suffix is nowhere attested in surface forms.

 

Alignment and orientation in Kartvelian    1125 Table 45.9 Cognate V3 suffixes in Old Georgian and Zan tense/​mood/​aspect

Present/​permansive

Conjunctive

Past indicative

Iterative/​atelic

3sg

3pl

3sg

3pl

3sg

3pl

3sg

3pl

Georgian

-​s

-​en/​-​an

-​s

-​n

-​a

-​es

-​n/​-​ed

-​ed

Laz-​Mingrelian

-​s

-​an

-​s

-​n

-​u

-​es

-​n

-​nan

Proto-​Georgian-​Zan

*-​s

*-​en

*-​s

*-​n

*-​a

*-​es

*-​n

?*-​ed

subject and a 2nd-​person (morphological) object, or vice-​versa, only the Set M prefix appears.9 The person of the subject can be determined from the suffixes only. (11)

(me) (šen) (I/​me) (you) ‘I saw you’

g-​nax-​e-​Ø M2-​see-​AOR-​V1/​2

//​ (šen) (me) //​ (you) (I/​me) //​ ‘you saw me’

m-​nax-​e-​Ø M1-​see-​AOR-​V1/​2

45.4.3 Verbal Plurality The stems of certain classes of Svan verbs can be marked with one of the pluralizers /​ -​ǣl-​/​, /​-ə​ ̄r-​/​or /​-i​ ēl-​/​. These suffixes indicate (i) repeated action, (ii) absolutive plurality, that is, plurality of the direct object or intransitive subject (Sharadzenidze 1954; Topuria 1967: 231–233; cf. Dressler 1968) (12)

Svan verbal nouns with pluralizers (Upper Bal dialect) [Gujejiani & Palmaitis 1985] li-​jēlw-​e ‘to sweep sthg’ li-​jēlw-​ǣl-​i ‘to sweep sthg many times’ li-​c’b-​īn-​e ‘to dangle’ li-​c’ēb-​ǣl ‘to dangle many things, to dangle many times’ li-​šxb-​i ‘to sew one thing’ li-​šxb-​iēl-​i ‘to sew many things’ li-​pt’-​e ‘to bore, drill’ li-​pt’-​ə̄r-​e ‘to drill many holes’

Old Georgian also had a suffix which shared some of the functions of the Svan verbal pluralizer, and which occupied a similar position close to the verb root (Harris 1985: 209–​230). The suffix /​-​(e)n-​/​indicated the presence of a plural direct object or intransitive subject, but only if the verb was a transitive or prefixal intransitive verb, and only in tenses based on the aorist stem (including the pluperfect). Most prefixal intransitives are based on transitive verb stems, and have the meaning of passives, as in (14).10 9 

For two distinct interpretations of this phenomenon, see Harris (1981: 30–​31), Boeder (2002). Prefixal intransitive morphology can also signal the backgrounding or deletion of the patient, rather than the agent. Antipassives of this kind—​called “deponents” by Shanidze (1953: section 366)—​ occur in Old Georgian and all modern Kartvelian languages (Tuite 2002; cf. Lazard 1994: 255). 10 

 

1126   Kevin Tuite It would appear, then, that Old Georgian /​-​(e)n/​was limited to marking the plurality of NOM arguments of verbs that are transitive at some level of their derivation, although there are indications that it might once have functioned as a verbal pluralizer comparable to its Svan counterpart (Harris 1985: 222–​228; Tuite 1998: 68–​72, 91–​97) (13) sit’q’va-​n-​i čem-​n-​i ara da-​marx-​n-​a word-​PL-​NOM my-​PL-​NOM not Pb-​preserve-​Pl.N-​PST.V3sg ‘He did not keep my words.’               [I Samuel 15,11] (14) or-​n-​i-​ve igi da-​i-​marx-​n-​i-​an two-​PL-​NOM-​same the:NOM Pb-​PASS-​preserve-​Pl.N-​PRM-​V3pl ‘Both will be preserved.’                  [Matthew 9:18]

45.5 Verb Classes, Case Assignment, and Subjecthood The relationship between Kartvelian verbs and their clausal arguments can be summed up in the following principles: 1. Verb stems can be divided into two major classes: the “active” class (which assigns ERG case in the aorist series, and undergoes inversion in the perfect series); and the “inactive” class, which does not. 2. Only one argument per verb can control Set V agreement; whereas two or more can potentially control Set M. When verbs have a DO and one or more IOs, these arguments “compete” for the prefixal agreement slot. A 1st-​or 2nd-​person object is ordinarily preferred over a 3rd-​person object. A 3rd-​person NOM-​case DO of a transitive verb in the aorist series has no chance of Set M agreement. 3. An Set V or M agreement marker may be lexically specified but nonreferential. 4. In principle, any core argument can have the attributes of syntactic subjecthood. Constructions with indirect syntax, in which the morphological indirect or direct object functions as syntactic subject, are not uncommon in Kartvelian. Verbs associated with indirect syntax due to their meaning can be grouped together as indirect verbs. Unlike the inversion transformation undergone by active verbs in the perfect tense series, the lexically conditioned indirect syntax associated with indirect verbs appears in the present and aorist series as well. The verbs in (15) and (16) are both transitive, but that in (16) is linked to indirect syntax. (15)

es gogo-​eb-​i ertmanet-​s Ø-​a-​cin-​eb-​en M3-​ApV-​laugh-​SM-​V3pl this:NOM girl-​PL-​NOM each.other-​DAT ‘These girls (NOM) make each other (DAT) laugh.’

 

Alignment and orientation in Kartvelian    1127 (16)

am gogo-​eb-​s ertmanet-​i Ø-​a-​int’eres-​eb-​t this:OBL girl-​PL-​DAT each.other-​NOM M3-​ApV-​interest-​SM-​M3pl ‘These girls (DAT) are interested in each other (NOM).’ (lit. ‘each other make these girls interested’)

Every possibility generated by the four principles listed above is realized by at least a small group of verbs in each Kartvelian language. Active and inactive verb types are shown in Table 45.10 Table 45.10 Georgian verb types grouped by argument structure Active class:

Type

Ac1a. Vs-​Mo

active transitive

Ac1b. Vo-​Ms

Examples

(i) DO only in Set M slot

v-​h-​k’lav ‘I kill DO’

(ii) IO only in Set M slot

v-​u-​k’ben ‘I bite IO’

(iii) DO & IO compete for Set M

v-​a-​cnob ‘I introduce DO to IO’

indirect transitive (i) IO as syntactic subject

m-​i-​panckaleb-​s ‘it makes my (heart) flutter’

(ii) DO as syntactic subject

m-​a-​int’ereseb-​s ‘it interests me’

Ac2. Vs

active intransitive

v-​q’ep ‘I bark’

Ac3. Vs-​MØ

active w/​nonreferential Set M (i) nonreferential DO

v-​a-​k’ueb ‘I fart audibly’

(ii) nonreferential IO

v-​u-​k’rav ‘I play (instrument)’

Ac4. VØ-​Ms

agentless transitive

m-​axveleb-​s ‘I cough’

Ac5. VØ

active impersonal

c’vim-​s ‘it rains’

Inactive class: In1a. Vs-​Mo

relative intransitive

v-​e-​maleb-​i ‘I hide from IO’

In1b. Vo-​Ms

indirect relative intransitive

m-​i-​q’var-​s ‘I love MS’

In2. Vs

absolute intransitive

v-​ar ‘I am’

In3. Vs-​MØ

inactive w/​nonreferential Set M

aɣ-​v-​e-​srulebi ‘I am finished, I die’

In4. VØ-​Ms

inactive w/​nonreferential Set V

m-​ɣvidzav-​s ‘I am awake’

In5. VØ

inactive impersonal

bnel-​a ‘it is dark’

(subscript s= syntactic subject, o = syntactic object; Ø = nonreferential agreement marker; MS, DO, IO = morphological subject, direct object, indirect object)

Among the verb types shown are two which have only a single surface argument in the DAT case, marked by Set M agreement. The inactive DAT-​only group (In4) includes verbs denoting reactions to psycho-​physiological states (m-​šia ‘I am

 

1128   Kevin Tuite hungry’; m-​civa ‘I am cold’); and a large, open, class of verbs indicating the desire for some thing or activity, e.g. m-​e-​mɣereba ‘I feel like singing’, m-​e-​šok’oladeba ‘I feel like eating chocolate’ (Shanidze 1953: 299–​301). The three dozen or so agentless transitives (Ac4) denote observable, usually involuntary, responses to internal physiological conditions, such as shivering, sneezing, etc. (Shanidze 1953: 195–​196; Tuite 2009): (17) ertxel kimi-​is lekcia-​ze da-​m-​a-​mtknar-​a once chemistry-​gen lecture-​at Pb-​M1sg-​ApV-​yawn-​PST.V3sg ‘Once I yawned at a chemistry lecture.’ (chat group Tbilisi forum) (18) xazarula-​s siciv-​isa-​gan še-​Ø-​a-​žržol-​a X.-​dat cold-​gen-​from Pb-​M3-​ApV-​shudder-​PST.V3sg ‘Xazarula shuddered from the cold.’ (Nodar Dumbadze Xazarula) The number of inactive and active DAT-​only verbs in Georgian and the other Kartvelian languages is by no means negligible. As a consequence, three types of intransitive verbs, rather than two, can be set alongside the active transitives. The resulting alignments for Georgian and Svan are shown in Tables 45.11 and 45.12, for verb agreement and case assignment (the special cases of Laz and Mingrelian will be discussed later). Alignment A (verb agreement in Series I and II; case marking in Series I): split-​ intransitive, according to direct vs. indirect syntax (the DAT-​only verbs = O-​coding intransitives). With respect to Ardeşen Laz, Kutscher (2008, 2009) characterizes the basis of this opposition as [+/​-​control] (“Kontrollfähigkeit”). Alignment B (verb agreement and case marking in Series III): split-​intransitive, with lexical aspect as the principal factor, although with complications. Most A-​coding intransitives are non-​telic (atelic active intransitives and stative experiencer verbs). The set of O-​coding intransitives, however, includes both change-​of-​state and stative verbs, including many with agentive subjects. Alignment C (case marking in Series II): split-​intransitive, with three sets of intransitives according to case marking:  A-​coding (active intransitives), O-​coding (inactive intransitives with direct syntax), and B[eneficiary]-​ coding (DAT-​ only verbs). Table 45.11 Alignment of person markers (all three persons) Type (Tab 45.10)

Ac1, Ac3

Ac2, (Ac 5)

In1/​2/​3/​(5)

Ac4, In4

A

O

S-​active

S-​inactive

S-​dative

Series I (present)

V

M

V

V

M

Series II (aorist)

V

M

V

V

M

Series III (perfect)

M

V

M

V

M

 

Alignment and orientation in Kartvelian    1129 Table 45.12 Alignment of case assignment (3rd-​person) Type (Tab 45.10)

Ac1, Ac3

Ac2, (Ac 5)

In1/​2/​3/​(5)

Ac4, In4

A

O

S-​active

S-​inactive

S-​dative

Series I (present)

NOM

DAT

NOM

NOM

DAT

Series II (aorist)

ERG

NOM

ERG

NOM

DAT

Series III (perfect)

DAT

NOM

DAT

NOM

DAT

Left out of the picture so far are the various types of indirect verbs with two or three arguments. These number in the hundreds in all four Kartvelian languages. Although operating from different theoretical standpoints, Harris (1981:  128)  and Dixon (1994: 121) equate the DAT subjects of these verbs with regular transitive subjects (Dixon’s A, Harris’s “initial 1”). Bringing bi-​and tri-​valent indirect verbs into the alignment picture—​whether as variants of transitives or as construction types in their own right—​does not significantly change the already complex situation described in this section.

45.6  Verb Categories and Case Assignment in Svan, Laz, and Mingrelian 45.6.1 Svan Although Svan is the outlier in the Kartvelian family, its morphosyntax is very similar to that of Georgian. Active verbs, including a significant number of intransitives, have distinct case-​assignment patterns in the present, aorist and perfect tense series. (19a)

eǰjær næj tæš-​s gw-​a-​hwd-​i-​x they:NOM us cheese-​DAT M1incl-​ApV-​give-​PRS-​V3pl ‘They (NOM) are giving us the cheese (DAT) [present].’

(19b) eǰjær-​d næj tæš la-​gw-​ēm-​x they-​ERG us cheese:NOM Pb-​M1incl-​give.AOR-​V3pl ‘They (ERG) gave us the cheese (NOM) [aorist].’ (19c)

eǰjær-​s nišgwej-​d tæš loxwhodax {la-​x-​o-​hod-​a-​x} they-​DAT our-​ADV cheese:NOM Pb-​M3-​ObV-​give-​PERF-​M3pl ‘They (DAT) have given us the cheese (NOM) [present perfect].’

 

1130   Kevin Tuite Also as in Georgian, DAT-​subject constructions are very frequent in Svan. These include inversion constructions as in (19c), as well as active and inactive indirect verbs: (20) sk’odi ǰævr gvi-​s x-​o-​c’xvav-​da-​x       æl deep worry:NOM heart-​DAT M3-​ObV-​torment-​IMP-​M3pl  this č’q’int’-​i     dede-​s       i     mama-​s [Lent’ex dialect; elicited] boy-​GEN  mother-​DAT  and  father-​DAT ‘The boy’s mother and father (DAT) were tormented by intense anxiety.’ (lit. “Deep worry (NOM) pained the heart for the boy’s mother and father (DAT)”;

45.6.2 Laz The distinctive characteristic of Laz morphosyntax is the near-​total elimination of case-​ pattern shift according to series. A case assignment pattern similar to that employed in series II in Georgian has been extended to series I and most of the evidential tenses (Harris 1985: 297). In each of these series, ERG case (in -​k, as in Mingrelian) is assigned to the subjects of active verbs, and direct objects are marked with NOM case:11 (21) xurma-​Ø i-​mxor-​s ǰoɣoi-​k [K’iziria 1982:90] persimmon-​NOM SbV-​eat-​V3sg dog-​ERG ‘The dog is eating a persimmon.’ (Georgian: xurma-​s Ø-​č’am-​s dzaɣl-​i [persimmon-​DAT eats dog-​NOM]) (22)

k’oči-​k do-​q’vil-​ere-​n ɣeǰ-​i [Harris 1985:298] man-​ERG Pb-​kill-​PERF-​V3sg pig-​NOM ‘The man has killed a pig.’ (Georgian: k’ac-​s da=Ø-​u-​k’lav-​s ɣor-​i [man-​DAT has-​killed pig-​NOM])

The use of the ERG case with intransitive verbs is not precisely correlated with verb class. A few verbs of the active intransitive type only take NOM subjects (e.g. ts’k’ai xaxal-​am-​s ‘water-​NOM boils’), and some others appear with subjects in either the ERG or NOM, even in the same text (Lacroix 2009: 581–​2; 599–​604), e.g.: (23) ar k’oči-​k mtugi-​ši patišai-​ša one man-​ERG mouse-​GEN king-​ALL ‘One man went to the mouse king.’ (24)

11 

i-​gzal-​u [K’iziria 1982: 85] SbV-​go-​PST.V3sg

mskibu-​ša i-​gzal-​u xčini-​Ø mill-​ALL SbV-​go-​PST.V3sg old.woman-​NOM ‘The old woman went to the mill.’

[K’iziria 1982: 85]

Evidence that the DOs of transitive verbs in series I were once assigned DAT case, as in Georgian and Svan, has been found in Laz folk poetry (Chikobava 1936: 181–​182).

 

Alignment and orientation in Kartvelian    1131 ERG case can also be optionally assigned to the subjects of agentive inactive verbs, as in some modern Georgian dialects (Klimov 1976: 153; K’iziria 1982: 84–​85), e.g.: (25)

bere-​k razi d-​iv-​u child-​ERG agreed Pb-​become-​PST.V3sg ‘The child agreed (‘became in agreement’).’

(Lacroix 2009: 602)

The only exception to this pattern of case assignment is the perfect series, in which active verbs undergo inversion (Harris 1985:  297–​298). As in Mingrelian, a few agentive inactive verbs also undergo inversion in the perfect series (Lacroix 2009: 364–​372). In some Laz subdialects the case system appears to be losing its diacritical function. In Vic’e, series I Class A verbs can assign “NOM” (i.e. unmarked) case to both morphological subject and DO: but’k’uǰi-​Ø topri-​Ø i-​kum-​s [bee-​“NOM” honey-​“NOM” SbV-​make-​V3sg] “the bee is making honey” (Marr 1910: 77–​78; Chikobava 1936: 181). In the subdialect of Ardeşen, there are no longer distinct NOM, ERG and DAT cases, the root form of the nominal being used in core grammatical contexts (Harris 1985: 385–​359; Kutscher 2009).

45.6.3 Mingrelian The Mingrelian language, although closely related to Laz, differs from it sharply with respect to case assignment. The ERG is only assigned by verbs in the aorist series, as in Georgian and Svan, but has been extended to nearly all subjects, regardless of the transitivity or class of the verb: (26) k’oč-​k do-​ɣur-​u man-​ERG Pb-​die-​PST.V3sg ‘The man-​ERG died.’ (Georgian: k’ac-​i mo-​k’vd-​a [man-​NOM die:IIp:V3sg]) Briefly put, any constituent that is assigned NOM case in series I is assigned ERG case in series II, including the morphological subjects of indirect verbs, and predicate nominals (Uridia 1960; Harris 1985: 372–​373; K’art’ozia et al. 2010: 329–​330): (27)

vezir-​iš naragadu-​k kə-​g-​Ø-​a-​šin-​u te boši-​s Pb-​Pb-​M3-​ObV-​recall-​PST.V3sg this  boy-​DAT vezir-​GEN said-​ERG ‘The boy (DAT) remembered the vezir’s words (ERG).’ [Xubua 1937: 229]

(28)

tina-​k ok’o i-​ʔu-​a-​s tiš komonǰ-​k [Xubua 1937: 35] he-​ERG must SbV-​become-​OPT-​V3sg her husband-​ERG ‘He (ERG) must become her husband (ERG).’

 

1132   Kevin Tuite In Mingrelian, agentive inactive verbs can undergo inversion in the perfect series, assigning DAT case to their subjects (K’art’ozia et al. 2010: 328–​329): (29) ge-​Ø-​u-​dgin-​u dzɣabi-​s Pb-​M3-​ObV-​stand-​PST.V3sg girl-​DAT ‘The girl has stood up.’ (Georgian: amdgara gogo-​Ø [stood-​up girl-​NOM])

45.7  Number Agreement and Morphosyntactic Orientation By “morphosyntactic orientation” I refer to the distribution of morphological and syntactic privileges within the clause. In many languages, a single category of clausal arguments accumulates the lion’s share of such privileges as agreement with the verb, initial position in the clause, marking with the nominative or absolutive case, etc., thereby emerging as the syntactic subject. As was mentioned in the introduction to section 45.2, the role of syntactic subject is less prominent in Kartvelian than in more familiar European languages. Since the Kartvelian verb has two sets of agreement markers, and word order is relatively free, the distribution of privileges is not so asymmetric. Zero anaphora is possible (and frequent) in morphological subject, direct object, and indirect object positions; though the likelihood increases when the argument in question functions as syntactic subject (Tuite 1998: 43–​46). In his analysis of subjecthood in Georgian, Aronson (1970: 294) noted that number agreement, especially with 3rd-​ person arguments, is a useful diagnostic of shifts in the morphosyntactic orientation of the Kartvelian languages and dialects. Type A:  split-​ergative (morphological subject) orientation. In Old Georgian, Set V agreement included marking of number in all three persons, whereas the Set M prefixes only indicated person.12 In addition, the suffix /​-​en-​/,​ mentioned above, agreed with plural NOM arguments in the aorist series and the pluperfect. In other words, NPs coded by either the unmarked (NOM) case, or the unmarked (Set V) agreement markers, had greater number-​agreement privileges than other arguments (in particular, arguments assigned DAT case). This was also the case until recently in the conservative Georgian dialects of the northeast highlands and the village Glola in Upper Rach’a (Tuite 1998). Consider the following two sentences from the Old Georgian gospels. Both have verbs associated with indirect syntax: an indirect intransitive verb in the first example, and a transitive verb in the pluperfect, with inversion, in the second. The verbs agree in number with the 3pl NOM-​case morphological subjects, not with their syntactic subjects: 12 

The prefix /​gv-​/​, which in modern Georgian signals 1st-​person plural, originally marked 1st-​person inclusive in contrast to the 1st-​exclusive prefix /​m-​/.​

 

Alignment and orientation in Kartvelian    1133 (30)

g-​i-​q’war-​d-​en tkwen M2-​ObV-​love-​IMP-​V3pl youpl:DAT ‘Youpl should love those who love youpl.’

(31)

(mat) da-​x-​e-​drik’-​n-​es p’ir-​n-​i mat-​n-​i they:DAT PV-​M3-​ObV-​bend-​PLNOM-​ V3pl face-​PL-​NOM their-​PL-​NOM kweq’an-​ad earth-​ADV ‘They had turned their faces toward the ground.’ [Lk 24:5]

moq’ware-​n-​i lover-​PL-​NOM

tkwen-​n-​i yourpl:PL-​NOM [Lk 6:32]

Type B: nominative (syntactic subject) orientation. The grammars of languages such as English, German and Latin are oriented toward a single clausal argument, which alone can control verb agreement. In Svan, Laz, standard Georgian, and most of its lowland dialects, there is no uniform marking of syntactic subjects: they can be assigned NOM, ERG or DAT case, and the 3rd-​person agreement markers in the verb vary according to tense, mood, and verb class. On the other hand, syntactic subjects control number agreement in all three persons, even in indirect-​syntax constructions, where the subject is assigned DAT case. To show how the morphosyntactic orientation has shifted since the Old Georgian period, here are the same two gospel passages translated into Modern Standard Georgian: (32)

g-​i-​q’var-​d-​e-​t tkven M2-​ObV-​love-​IMP-​PL youpl:DAT ‘Youpl should love those who love youpl.’

(33)

(mat) da-​Ø-​e-​drik’-​a-​t p’ir-​eb-​i mic’-​is mimart they:DAT Pb-​M3-​ObV-​bend-​V3sg-​PL face-​PL-​NOM earth-​GEN toward ‘They had turned their faces toward the ground.’

tkveni yourpl

moq’vare-​eb-​i lover-​PL-​NOM

In the southwest Georgian dialects of Guria and Ach’aria, the ERG case marker is increasingly taking on the function of coding subjecthood. The ERG case is frequently assigned by inactive verbs in Gurian, including some clearly nonagentive verbs (xenc’ipe-​m mok’t’a ‘the king-​ERG died’ (Zhghent’i 1936: 69)). In a Lower Ach’arian subdialect spoken near the Turkish border, verbs of either class can assign ERG case to their subjects in the present as well as the aorist series (K’iziria 1974: 78–​79; Boeder 1979: 445; Harris 1985: 376–​380). In these dialects, one observes both the extension of ERG case marking to the present series, as in Laz, and its extension to inactive intransitive subjects, as in Mingrelian. (34) sakonel-​ma ar mo-​di-​s [K’iziria 1974:79] cattle-​ERG not Pb-​come-​V3sg ‘The cattle are not coming.’ (Standard Geo. sakonel-​i ar mo-​di-​s [cattle-​NOM not comes])

 

1134   Kevin Tuite Type C: discourse-​prominence orientation. Some Georgian dialects from the southeast (Ingiloan, Fereidanian, K’axetian) and northwest (Imeretian, Lechxumian, Rach’an) represent a very different type of morphosyntactic orientation. These dialects differ from the first group in that the agreement privileges distinguishing NOM from DAT NPs, and Set V from Set M, are no longer found. They differ from the second group in that syntactic subjects have few if any morphosyntactic privileges. Any core argument can in principle control number agreement (NA) in all three persons in any series. The primary feature characterizing Type C is that animate and presupposed arguments have greater agreement privileges than inanimate or newly introduced arguments. In Type C dialects, those NPs controlling number agreement almost always denote topical, animate referents. This privilege is independent of case, agreement set and syntactic role. In (35), the clitic /​-​q’e/​signals the plurality of the 3rd-​person indirect object. Note that the first verb does not agree in number with its object, while the last two verbs do. This reflects the degree of presupposability of the argument in question, which is introduced as new information before the first verb, and is briefly maintained as discourse topic. In (36), the first verb does not agree in number with its plural subject (the suffix is 3sg), whereas the following verbs do. (35) danarčom bič’-​eb-​s da dad-​eb-​sb Ø-​u-​k’eteb-​en remaining boy-​PL-​DAT and bridesmaid-​PL-​DAT M3-​ObV-​make-​V3pl plav-​s,    čey-​s;     Øb      Ø-​a-​č’mev-​en-​q’e,      da pilaf-​DAT  tea-​DAT  Ø:3pl:DAT  M3-​ApV-​feed-​V3pl-​M3pl  and Øb    Ø-​a-​levineb-​en-​q’e Ø:3pl:DAT  M3-​ApV-​drink-​V3pl-​M3pl ‘For the remaining boys and bridesmaids they prepare pilaf and tea; they feed them and give them something to drink.’ [Ingiloan, GTK:244] (36) dila-​ze gare-​dan pxak’a-​pxuk’-​it mo-​vid-​a morning-​at outside-​from scratching-​INS Pb-​come-​PST.V3sg tagv-​eb-​im       da   Øm         ga-​a-​ps-​en        xvimir-​i mouse-​PL-​NOM  and  Ø:3pl:ERG  Pb-​ApV-​fill-​V3pl  hopper-​NOM pul-​it. money-​INS ‘Mice came with a scratching sound and Ø filled the hopper with money.’ [Lower Imeretian, GTK:474]

 

Alignment and orientation in Kartvelian    1135

45.8  Acquisition of Georgian Case Assignment Despite the appearance of bewildering complexity that they present to foreigners, the basic case-​assignment rules of Georgian are mastered by children before their third birthdays (Imedadze & Tuite 1992: 90–93). In particular, there is no evidence that the children attempt to employ the ERG marker with all subjects in the aorist series, nor that they restrict its use to transitive subjects. Split-​intransitive patterning emerges with relatively few errors. The one type of persistent case-​marking “mistake” is one that is also attested in the speech of adults in many non-​standard dialects: the overextension of ERG marking to the subjects of agentive inactive verbs, as in the following sentence produced by a 34-​month-​old girl: (37)

k’at’a-​m xe-​ze a-​vid-​a cat-​ERG tree-​on up-​go-​PST.V3sg ‘The cat climbed the tree.’

(K’axadze 1969)

45.9  Proto-​K artvelian Morphosyntax Like Yenisean, as reconstructed by Vajda (2007), the Kartvelian ancestral language appears to have had a more straightforward morphosyntactic alignment than its daughters.

45.9.1 Proto-​Kartvelian Case Marking Unlike the other primary case endings, the various forms of the ERG suffix in Georgian, Laz-​Mingrelian and Svan do not appear to go back to a common ancestor. Klimov (1960; 1962: 139–​143) derives Mingrelian-​Laz /​-​k/​, Georgian /​-​ma-​/​and the Svan oblique formant /​-​m-​/​ from the demonstrative roots /​*-​g-​/​ and /​*-​m-​/​, attested in pronominals and locative adverbials. Another allomorph of the Svan ERG is identical with the Kartvelian ADV ending in /​-d ​ /​. Chikobava (1948: 114–​5) and Boeder (1979: 457) question whether the ancestral language had a distinct ERG case marker at all. They hypothesize that the declension of proper names in Old Georgian is a vestige of the case marking of all nominals in Proto-​Kartvelian. As Boeder (1979) notes, the West Caucasian language

 

1136   Kevin Tuite Circassian provides an instructive parallel: The formal distinction between ergative and absolutive cases is marked on definite or specific nouns only; there is no such marking on non-​specific or generic nouns, names and 1st and 2nd person pronouns (Colarusso 1992: 51–​52; Kumakhov & Vamling 2009; Testelec & Arkadiev 2014). (39) Besleney Kabardian (Testelec & Arkadiev 2014) čʼele-​ʁesa apxʷede pjəsme jə-​txə-​ne-​q’əm letter 3sg.erg-​write-​fut-​neg boy-​well.behaved such ‘A well-​behaved boy (“ERG”) wouldn’t write such a letter (“NOM”).’ (i.e. this letter is such that no decent boy would write it) In other languages as well, the optional marking of ergative case on nouns is pragmatically motivated (Rumsey, San Roque, & Schieffelin 2013). Du Bois (1987b) has shown, on the basis of discourses recorded in languages of various alignments, that new topics and narrative protagonists are introduced preferentially in intransitive subject and transitive object positions, whereas transitive subjects tend to represent old or given information (typically conveyed by pronouns or zero anaphora). The Kartvelian ERG suffixes, which emerged independently in each branch of the family from demonstrative stems, could have originally signalled the exceptional appearance of a full NP in transitive subject position.

45.9.2 Proto-​Kartvelian Verb Classes The two major verb classes I labelled “active” and “inactive” are distinguished by stem morphology as well as case-​assignment properties, and therefore can be attributed to the ancestral language. Furthermore, the four principles governing verb-​argument relations (mentioned in section 45.5) are shared by all members of the family, and are likely to go back to Proto-​Kartvelian. The system of three tense/​aspect series, with the accompanying shift in case assignment and agreement for active verbs, emerged relatively late. The Series I (present) stem, which for most verbs is marked by a suffix (“present/​future stem formant”, or “series marker”), originated as a type of antipassive, associated with durative aspect (Chikobava 1948; Harris 1985). Being formally intransitive, the Series I forms assigned NOM case to their subjects, and DAT to the demoted former direct object. The Series III forms were stative passives with resultative meaning, reinterpreted as transitive perfects (m-​i-​c’er-​ie-​s ‘it is written for/​by me’ > ‘I have written it’). The class of active intransitives seems also to be of recent origin. Evidence from Old Georgian and Laz indicates that in Proto-​Kartvelian many, perhaps all, of these verbs either assigned NOM case to their subjects, or did not appear in the aorist tense series. The verb /​q’iv-​/​‘crow’, which assigns ERG case in Modern Georgian, is attested in Old Georgian with a NOM subject in the aorist (katam-​i q’iv-​a [rooster-​NOM crow-​PST. V3sg] ‘the cock crowed’, Matt 26: 74; Harris 1985: 347).

 

Alignment and orientation in Kartvelian    1137

45.9.3  Proto-​Kartvelian Morphosyntax The integration of the formerly antipassive Series I and the resultative passive Series III into the paradigms of active verbs, along with the emergence of a new class of ERG-​ assigning intransitives, is largely responsible for the mix of split-​intransitive alignment types in modern Georgian and Svan. The relation-​marking system of Proto-​Kartvelian would have been for the most part of the head-​marking type, with incipient dependent marking—​in the form of optional, pragmatically conditioned demonstrative suffixes—​ for common nouns. The resulting morphosyntactic system gives the impression of being split between nominative–​ accusative agreement and ergative–​ absolutive case marking, but that is without taking the various groups of indirect verbs into consideration, including those with a single surface argument assigned DAT case and marked with Set M prefixes. There are no grounds for assuming that indirect verbs are of recent origin in Kartvelian. Therefore, I believe it justified to postulate split-​intransitive alignment for the head-​marking component of Proto-​Kartvelian morphosyntax, alongside an emergent ergative–​absolutive case system, and verbal-​plurality marking linked to absolutive arguments (Table 45.13). Table 45.13 Proto-​Kartvelian relation marking Active (transitive)

Inactive (intransitive)

DAT-​only

A

O

Sa

So

1 /​2 p.

V

M

V

M

3rd p.

V

—​—​

V

M

case (optional)

(ERG)

(NOM)

(NOM)

DAT

verbal plurality

—​—​

agreement

agreement

—​—​

role st

nd

The principal sentence types in Proto-​Kartvelian, with respect to case marking and agreement, would have been as follows (plural arguments are shown, to illustrate the patterning of number agreement and verbal plurality marking): Transitive (continued by active verbs in Series II tenses) 1st & 2nd person: you we/​us M-​verb-​Pl.N-​V1/​2-​PL 3rd person: X-​(ERG) Y-​(NOM) verb-​Pl.N-​V3pl Bivalent intransitive (includes ancestors of Series I and III tenses of active verbs) 1st & 2nd person: you we/​us M-​verb-​Pl.N-​V1/​2-​PL 3rd person: X-​(NOM) Y-​DAT M-​verb-​Pl.N-​V3pl

 

1138   Kevin Tuite Monovalent intransitive, NOM-​only 1st & 2nd person: you verb-​Pl.N-​V1/​2-​PL 3rd person: X-​(NOM) verb-​Pl.N-​V3pl Monovalent intransitive, DAT-​only 1st & 2nd person: we/​us M-​verb-​[VØ] 3rd person: Y-​DAT M-​verb-​[VØ]

Acknowledgments I wish to express special thanks to my esteemed colleague Winfried Boeder, who on this, as on countless other occasions, has unstintingly supplied me with detailed comments and bibliographical suggestions. Thanks also go to Lisa Travis and an anonymous reader for their helpful advice on how to make this text at least slightly more reader-​friendly. I take sole responsibility for all shortcomings.

Abbreviations ALL, allative; AOR, aorist; ApV, applicative version (transitive, superessive); CAUS, causative; DET, determiner; EMPH, emphatic; IMP, imperfect; ObV, objective version; OPT, optative; PASS, passive; PERF, perfect; Pl.N, plural nominative; PRM, permansive; PST, past; SbV, subjective version; SM, series marker.

 

References

All links accessed October/​November 2016, unless otherwise indicated. Abadie, Peggy. 1974. ‘Nepali as an ergative language.’ Linguistics of the Tibeto-​Burman area 1: 156–​177. Abney, Steven. 1987. ‘The English noun structure in its sentential aspect.’ PhD dissertation, MIT. Abramovitz, Rafael. 2015. ‘Another look at the Chukchi spurious antipassive.’ MS, University of Chicago. Ackerman, Farrell & John Moore. 2001. Proto-​properties and grammatical encoding: A correspondence theory of argument selection. Stanford, CA: CSLI. Adams, Karen L. & Alexis Manaster-​Ramer. 1988. ‘Some questions of topic/​focus choice in Tagalog.’ Oceanic Linguistics 27: 79–​101. Adelaar, K. Alexander. 1992. Proto Malayic: The reconstruction of its phonology and parts of its lexicon and morphology. 119. Sydney: Dept. of Linguistics, Research School of Pacific Studies, Australian National University. Adelaar, Alexander (ed.). 2013. Voice variation in Austronesian languages of Indonesia, vol. 54 of NUSA. Tokyo University of Foreign Studies. Adelaar, Alexander & Nikolaus Himmelmann (eds.). 2005. The Austronesian languages of Asia and Madagascar. London: Routledge. Adelaar, Wilhem. 2000. ‘Propuesta de un nuevo vínculo genético entre dos grupos lingüísticos indígenas de la Amazonía occidental. Harakmbut y Katukina,’ in L. Miranda Esquerré (ed.), Actas del I  Congreso de Lenguas Indígenas de Sudamérica, 2, Lima, Universidad Ricardo Palma, 219–​236. Agha, Asif. 1990. ‘Lexical structure and grammatical categories in Lhasa Tibetan.’ PhD dissertation, University of Chicago. Agha, Asif. 1993. Structural form and utterance context in Lhasa Tibetan. Peter Lang: New York. Aguirre, Carmen. 2003. ‘Early verb development in one Spanish-​speaking child,’ in D. Bittner, W. U. Dressler, & M. Kilani-​Schoch (eds.), Development of verb inflection in first language acquisition: A cross-​linguistic perspective. Berlin: Mouton de Gruyter, 1–​27. Aguirre, Carmen. 2006. ‘What do overregularizations tell us about morphological knowledge?’ Círculo de lingüística aplicada a la comunicación, Facultad de Ciencias de la Información, 26: 3–​11. Ahlsén, Elisabeth. 2006. Introduction to neurolinguistics. Amsterdam & Philadelphia, PA: Benjamins. Ahmed, Tafseer. 2006. ‘Spatial, temporal and structural usages of Urdu ko,’ in Miriam Butt & Tracy Holloway King (eds.), Proceedings of the LFG06 conference. Stanford, CA: CSLI, 1–​13. Ahmed, Tafseer. 2010. ‘The unaccusativity/​unergativity distinction in Urdu.’ Journal of South Asian Linguistics 3: 3–​22.

 

1140   References Aijmer, Karin. 1989. ‘Themes and tails: The discourse functions of dislocated elements.’ Nordic Journal of Linguistics 12(2): 137–​154. Aikhenvald, Alexandra Y. 2009. ‘Syntactic ergativity in Paumarí,’ in S. G. Obeng (ed.), Topics in descriptive and African linguistics: Essays in honor of distinguished professor Paul Newman. Munich: Lincom Europa, 11–​127. Aissen, Judith. 1999a. ‘Agent focus and inverse in Tzotzil.’ Language 75: 451–​485. Aissen, Judith. 1999b. ‘Markedness and subject choice in optimality theory.’ Natural Language and Linguistic Theory 17: 673–​7 11. Aissen, Judith. 2003a. ‘Differential coding, partial blocking, and bidirectional OT,’ in Pawel Nowak and Corey Yoquelet (eds.), BLS 29. Berkeley, CA: Berkeley Linguistics Society, 1–16. Aissen, Judith. 2003b. ‘Differential object marking: Iconicity vs. economy.’ Natural Language and Linguistic Theory 21: 435–​483. Aissen, Judith. 2011. ‘On the syntax of agent focus in K’ichee,’ in Kirill Shklovsky, Pedro Mateo Pedro, & Jessica Coon (eds.), Proceedings of formal approaches to Mayan linguistics I (MIT Working Papers in Linguistics, Vol. 63). Cambridge, MA: MIT, 1–​16. Akhtar, Raja Nasim. 1999. ‘Aspectual complex predicates in Punjabi.’ PhD dissertation, University of Essex. Alberdi, Xabier. 2003. ‘The transitivity of borrowed verbs in Basque: An outline,’ in Bernard Oyharçabal (ed.), Inquiries into the lexicon–​syntax relations in Basque. Supplements of Anuario de Filología Vasca Julio de Urquijo 46: 23–​56. Albizu, Pablo. 2002. ‘Basque verbal morphology:  Redefining cases,’ in X. Artiagoitia, P. Goenaga, & J.A. Lakarra (eds.), Erramu Boneta: Festschrift for Rudolf P.G. de Rijk. Bilbao: UPV-​EHU, 1–​19. Albizu, Pablo & Luis Eguren. 2000. ‘An optimality theoretic account for “ergative displacement” in Basque,’ in W.U. Dressler, O.E. Pfeiffer, M.A. Pöchtrager, & J.R. Rennison (eds.), Morphological Analysis in Comparison. Amsterdam & Philadelphia, PA: Benjamins. Albright, Adam. 2002. ‘The identification of bases in morphological paradigms.’ PhD dissertation, UCLA. Aldai, Gontzal. 2008. ‘From ergative case marking to semantic case marking: The case of historical Basque,’ in Mark Donohue & Soren Wichmann (eds.), The typology of semantic alignment. Oxford: Oxford University Press, 197–​218. Aldai, Gontzal. 2009. ‘Is Basque morphologically ergative?’ Studies in Language 33–​4: 783–​831. Aldridge, Edith. 2002. ‘Wh-​movement in Seediq and Tagalog,’ in Andrea Rackowski & Norvin Richards (eds.), Proceedings of AFLA 8: The 8th meeting of the Austronesian Formal Linguistics Association (MIT Working Papers in Linguistics, Vol. 44). Cambridge, MA: MIT Press, 1–​28. Aldridge, Edith. 2004. ‘Ergativity and word order in Austronesian Languages.’ PhD dissertation, Cornell University. Aldridge, Edith. 2006. ‘Absolutive case in Tagalog,’ in Jackie Bunting, Sapna Desai, Robert Peachey, Chris Straughn, & Zuzana Tomkova (eds.), Proceedings from the 42nd Annual Meeting of the Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 1–​15. Aldridge, Edith. 2008a. ‘Generative approaches to ergativity.’ Language and Linguistics Compass 2(5): 966–​995. Aldridge, Edith. 2008b. ‘Phase-​based account of extraction in Indonesian.’ Lingua 118: 1440–​1469. Aldridge, Edith. 2008c. ‘Minimalist analysis of ergativity.’ Sophia Linguistica 55: 123–​142. Aldridge, Edith. 2009. ‘Minimalist questions for the nominalist analysis of Tagalog syntax.’ Theoretical Linguistics 35(1): 51–​62.

 

References   1141 Aldridge, Edith. 2010. ‘Directionality in word order change in Austronesian languages,’ in Anne Breitbarth, Christopher Lucas, Sheila Watts, & David Willis (eds.), Continuity and change in grammar. Amsterdam & Philadelphia, PA: Benjamins, 169–​180. Aldridge, Edith. 2011. ‘Antipassive in Austronesian alignment change,’ in Dianne Jonas, John Whitman, & Andrew Garrett (eds.), Grammatical change: Origins, nature, outcomes. Oxford: Oxford University Press, 331–​345. Aldridge, Edith. 2012a. ‘Event existentials in Tagalog,’ in Lauren Eby Clemens Greg Scontras & Maria Polinsky (eds.), Proceedings of the 18th Annual Meeting of the Austronesian Formal Linguistics Society (AFLA 18). Online publication hosted by the University of Western Ontario, 16–​30. Aldridge, Edith. 2012b. ‘Antipassive and ergativity in Tagalog.’ Lingua 122: 192–​203. Aldridge, Edith. 2013a. ‘Nominalization source of ergativity in Tagalog.’ Paper presented at National Tsinghua University, Graduate Institute of Linguistics. Aldridge, Edith. 2013b. ‘Origin of ergative variation in austronesian languages.’ Paper presented at Diachronic Generative Syntax (DiGS), Ottawa, 2013. Aldridge, Edith. [2015.] ‘A minimalist approach to the emergence of ergativity in Austronesian languages.’ Linguistics Vanguard 1(1): 313–326.Edith.2016‘Ergativity from subjunctive in Austronesian languages.’Language and Linguistics Alexiadou, Artemis. 1997. Adverb placement: A case study in antisymmetric syntax. Amsterdam & Philadelphia, PA: Benjamins. Alexiadou, Artemis. 1999. ‘Remarks on the syntax of process nominals: An ergative pattern in nominative–​accusative languages.’ NELS 29: 1–​15. Alexiadou, Artemis. 2001. Functional structure in nominals:  Nominalization and ergativity. Amsterdam & Philadelphia, PA: Benjamins. Alexiadou, Artemis. 2004. ‘Inflection class, gender and DP internal structure,’ in G. Müller et al. (eds.), Explorations in nominal inflection. Berlin: Mouton de Gruyter, 21–​50. Alexiadou, Artemis. 2008. ‘Tense in the nominal domain: Implications for grammar architecture.’ Linguistic Variation Yearbook 8: 33–​65. Alexiadou, Artemis & Elena Anagnostopoulou. 2001. ‘The subject in situ generalization, and the role of case in driving computations.’ Linguistic Inquiry 32: 193–​231. Alexiadou, Artemis & Elena Anagnostopoulou. 2006. ‘From hierarchies to features,’ in Cedric Boeckx (ed.), Agreement systems. Amsterdam & Philadelphia, PA: Benjamins, 41–​62. Alexiadou, Artemis, Elena Anagnostopoulou, & Christina Sevdali. 2014. ‘Opaque and transparent datives, and how they behave in passives.’ Journal of Comparative Germanic Linguistics 17: 1–​34. Alexiadou, Artemis, Elena Anagnostopoulou, & Florian Schäfer. 2006. ‘The properties of anticausatives crosslinguistically,’ in M. Frascarelli (ed.), Phases of interpretation. Berlin: Mouton de Gruyter. Alexiadou, Artemis, Elena Anagnostopoulou, & Florian Schäfer. 2009. ‘PP-​licensing in nominalizations.’ Proceedings of NELS 38: 38–​52. Alexiadou, Artemis, Gianina Iordachioaia, Mariangeles Cano, Fabienne Martin, & Florian Schäfer. 2013. ‘The realization of external arguments in nominalizations.’ Journal of Comparative Germanic Linguistics 16: 73–​95. Alexiadou, Artemis, Gianina Iordachioaia, & Florian Schäfer. 2011. ‘Syntactic realization of plural in Romance and Germanic nominalizations,’ in K. Arregi et al. (eds.), Romance linguistics 2008:  Interactions in Romance. Amsterdam & Philadelphia, PA:  Benjamins, 107–​124.

 

1142   References Alexiadou, Artemis, Gianina Iordachioaia, & Elena Soare. 2010. ‘Number/​aspect interactions in the syntax of nominalizations: A Distributed Morphology approach.’ Journal of Linguistics 46: 537–​574. Alexiadou, Artemis, Terje Lohndal, Tor Afarli, & Maren Berg Grimstad. 2015. ‘Language mixing: A distributed morphology approach.’ Proceedings of NELS 45: 25–38. Alexiadou, Artemis & Florian Schäfer. 2006. ‘Instrument subjects are agents or causers,’ in Donald Baumer, David Montero, & Michael Scanlon (eds.), Proceedings of the 25th West Coast conference on formal linguistics. Somerville, MA: Cascadilla Proceedings Project, 40–​48. Allen, Cynthia L. 1995. ‘On doing as you please,’ in Andreas H. Jucker (ed.), Historical pragmatics:  Pragmatic developments in the history of English. Amsterdam & Philadelphia, PA: Benjamins, 275–​308. Allen, Shanley E.M 1996. Aspects of argument structure acquisition in Inuktitut. Amsterdam & Philadelphia, PA: Benjamins. Allen, Shanley E.M. 2000. ‘A discourse–​pragmatic explanation for argument representation in child Inuktitut.’ Linguistics 38: 483–​521. Allen, Shanley E.M. 2013. ‘The acquisition of ergativity in Inuktitut,’ in E.L. Bavin & S. Stolle (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 71–​105. Allen, Shanley E.M. & Heike Schröder. 2003. ‘Preferred argument structure in early Inuktitut spontaneous speech data,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 301–​338. Allen, W. Sidney. 1960. ‘Notes on the Rājasthānī verb.’ Indian Linguistics 21: 4–​13. Allen, W. Sidney. 1964. ‘Transitivity and possession.’ Language 40: 337–​343. Alsdorf, Ludwig. 1936. ‘Vasudevahiṃḍī, a specimen of archaic Jaina Maharastri.’ Bulletin of the School of Oriental Studies 8(2–​3): 319–​333. Alves, Flávia de Castro. 2004. ‘O Timbira falado pelos Canela Apãnjekrá.’ PhD dissertation, Universidade Estadual de Campinas. Alves, Flávia de Castro. 2010. ‘Evolution of alignment in Timbira.’ International Journal of American Linguistics 76: 439–​475 Amberber, Mengistu. 2009. ‘Differentical case marking of arguments in Amharic,’ in Andrej L. Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford:  Oxford University Press, 339–​355. Amritavalli, Raghavachari. 1979. ‘The representation of transitivity in the lexicon.’ Linguistic Analysis 5(1): 71–​92. Anagnostopoulou, Elena. 2005. ‘Strong and weak person restrictions: A feature checking analysis,’ in Lorie Heggie & Francisco Ordonez (eds.), Clitic and affix combinations: Theoretical perspectives (Linguistics Today 74). Amsterdam & Philadelphia, PA: Benjamins, 199–​235. Anagnostopoulou, Elena. 2006. ‘Clitic doubling,’ in Martin Everaert & Henk van Riemsdijk (eds.), The Blackwell companion to syntax, vol. 1. Oxford: Blackwell, 519–​581. Anand, Pranav & Andrew Nevins. 2006. ‘The locus of ergative assignment:  Evidence from scope,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity:  Emerging issues: Studies in natural language and linguistic theory, vol. 65. Dordrecht: Springer, 3–​25. Andersen, Paul Kent. 1986. ‘Dieta-​partizipialkonstruktion bei a soka:  Passiv oder ergativ?’ Zeitschrift für vergleichende Sprachforschung 99: 75–​95. Andersen, Torben. 1988. ‘Ergativity in Päri, a Nilotic OVS language.’ Lingua 75: 289–​324. Andersen, Torben. 1991. ‘Subject and topic in Dinka.’ Studies in Language 15: 265–​294.

 

References   1143 Andersen, Torben. 1992. ‘Morphological stratification in Dinka: On the alternations of voice quality, vowel length and tone in the morphology of transitive verbal roots in a monosyllabic language.’ Studies in African Linguistics 23: 1–​63. Andersen, Torben. 2000. ‘Anywa and Päri, II:  A  morphosyntactic comparison.’ Afrika und Übersee 83: 65–​87. Andersen, Torben. 2002. ‘Case inflection and nominal head marking in Dinka.’ Journal of African Languages and Linguistics 23: 1–​30. Anderson, Greogory & Randall Eggert. 2001. ‘A typology of verb agreement in Burushaski.’ Linguistics in the Tibeto-​Berman Area 24: 235–​254. Anderson, Neil & Martha Wade. 1988. ‘Ergativity and Control in Folopa.’ Language and Linguistics in Melanesia 19: 1–​16. Anderson, Stephen R. 1971. On the linguistic status of the performative/​constative distinction. Indiana: Indiana University Linguistics Club. Anderson, Stephen R. 1976. ‘On the notion of subject in ergative languages,’ in Charles N. Li & Sandra Thompson (eds.), Subject and topic. New York: Academic Press, 1–​23. Anderson, Stephen R. 1977. ‘On mechanisms by which languages become ergative,’ in Charles N. Li (ed.), Mechanisms of syntactic change. Austin: University of Texas Press, 317–​363. Anderson, Stephen R. 1984. ‘On representations in morphology: Case, agreement and inversion in Georgian.’ Natural Language and Linguistic Theory 2(2): 157–​218. Anderson, Stephen R. 1988. ‘Morphological change,’ in Frederick Newmeyer (ed.), Linguistics: The Cambridge survey, vol. 1. Cambridge: Cambridge University Press, 324–​362. Anderson, Victoria & Yuko Otsuka. 2006. ‘The phonetics and phonology of “definitive accent” in Tongan.’ Oceanic Linguistics 45: 33–​53. Andrews, Avery D. 1990. ‘Case structures and control in modern Icelandic,’ in Joan Maling & Annie Zaenen (eds.), Syntax and semantics 24:  Modern Icelandic syntax. San Diego, CA: Academic Press, 187–​234. Andvik, Erik E. 1999. ‘Tshangla grammar.’ PhD dissertation, University of Oregon. Annamalai, E. & P.P. Giridhar. 1991. ‘Multiple case markers and case roles in Angami: A raising solution.’ Bulletin of the Deccan College Research Institute 51/​52, Professor S.M. Katre, Felicitation Volume (1991–​1992), 221–​228. Anton I, Catholicos. 1885. Kartuli ɣrammatika [based on 1767 version]. Tbilisi:  Ekvtime Xeladzis stamba. Aoshima, Sachiko, Colin Phillips, & Amy Weinberg. 2004. ‘Processing filler-​gap dependencies in a head-​final language.’ Journal of Memory and Language 51: 23–​54. Ariel, Mira. 1990. Accessing noun–​phrase antecedents. London: Routledge. Ariel, Mira. 2000. ‘The development of person agreement markers: From pronouns to higher accessibility markers,’ in Michael Barlow & Suzanne Kemmer (eds.), Usage-​based models of language. Palo Alto: Center for the Study of Language and Information. Ariel, Mira. 2001. ‘Accessibility theory:  An overview,’ in Ted Sanders, Joost Schilperoord, & Wilbert Spooren (eds.), Text representation:  Linguistic and psycholinguistic aspects. Amsterdam & Philadelphia, PA: Benjamins, 29–​87. Ariel, Mira, Elitzur Dattner, John W. Du Bois, & Tal Linzen (2015). ‘Pronominal datives: The royal road to argument status.’ Studies in Language 39(2): 257–​321. Arka, I. Wayan. 2004. Balinese morphosyntax: A lexical–​functional approach. Canberra: Pacific Linguistics. Arka, I. Wayan & Christopher D. Manning. 1998. Voice and grammatical relations in Indonesian: A new perspective. Stanford, CA: CSLI.

 

1144   References Arka, I. Wayan & Christopher D. Manning. 2008. ‘Voice and grammatical relations in Indonesian:  A  new perspective,’ in Peter K. Austin & Simon Musgrave (eds.), Voice and grammatical relation in Austronesian languages. Stanford, CA: CSLI, 45–​69. Arka, I. Wayan & Malcolm Ross (eds.). 2005. The many faces of Austronesian voice systems: Some new empirical studies. Number 571 in Pacific Linguistics. Canberra: Research School of Pacific and Asian Studies, ANU. Arkadiev, Peter. 2008a. ‘Differential argument marking in two-​term case systems and its implications for a general theory of case marking,’ in H. de Hoop & P. de Swart (eds.), Differential subject marking. Dordrecht: Springer, 151–​173. Arkadiev, Peter. 2008b. ‘Thematic roles, event structure, and argument encoding in semantically aligned languages,’ in Mark Donohue & Søren Wichmann (eds.), The typology of semantic alignment. Oxford: Oxford University Press, 101–​117. Arkadiev, Peter. 2012. ‘On the meaning of allomorphy: The case of the ergative.’ Paper presented at the 15th International Morphology Meeting, Vienna, February 9–​12. Arkadiev, Peter & Alexander Letuchiy. 2008. ‘Derivacii antipassivnoj zony v adygejskom jazyke,’ in Vladimir Plungian & Sergey Tatevosov (eds.), Issledovanija po otglagol’noj derivacii. Moscow: Jazyki slavjanskix kul’tur, 77–​102. Arnold, Jennifer E. 2003. ‘Multiple constraints on reference form: Null, pronominal, and full reference in Mapudungun,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 225–​245. Aronoff, Mark. 1980. ‘Contextuals.’ Language 56: 744–​758. Aronoff, Mark. 2013. ‘Varieties of morphological defaults and exceptions.’ ReVel 7: 84–​97. Aronson, Howard. 1970. ‘Towards a semantic analysis of case and subject in Georgian.’ Lingua 25: 291–​301. Arregi, Karlos & Andrew Nevins. 2008. ‘Agreement and clitic restrictions in Basque,’ in Susann Fischer Roberta D’Alessandro & Gunnar Hrafn Hrafnbjargarson (eds.), Agreement restrictions. Berlin: Mouton de Gruyter, 49–​86. Arregi, Karlos & Andrew Nevins. 2012. Morphotactics: Basque auxiliaries and the structure of spellout. Studies in Natural Language and Linguistic Theory, vol. 86. Dordrecht: Springer. Arregui, Ana, Mariá Luisa Rivero, & Andrés Pablo Salanova. 2014. ‘Cross-​linguistic variation in imperfectivity.’ Natural Language and Linguistic Theory 32: 307–​362 Artawa, Ketut. 1994. ‘Ergativity in Balinese syntax.’ PhD dissertation, La Trobe University. Arteatx, Iñigo. 2007. ‘Euskarazko oharmen aditzen osagarrietako perpaus jokatugabeak.’ Uztaro 63: 31–​63. Arteatx, Iñigo. 2012. ‘Perception verb complements in Basque,’ in U. Etxeberria, R. Etxepare, & M. Uribe-​Etxebarria (eds.), Noun phrases and nominalizations in Basque. Amsterdam & Philadelphia, PA: Benjamins, 397–​436. Artiagoitia, Xabier. 2001. ‘Seemingly ergative and ergatively seeming,’ in J. Herschenson, E. Mallén, & K. Zagona (eds.), Features and interfaces in Romance: Essays in honor of Heles Contreras. Amsterdam & Philadelphia, PA: Benjamins, 1–​22. Artiagoitia, Xabier. 2003. ‘Complementation,’ in J.I. Hualde & J. Ortiz de Urbina (eds.), A grammar of Basque. Berlin: Mouton de Gruyter, 634–​709. Artshuler, Daniel. 2013. ‘There is no neutral aspect,’ in T. Snider (ed.), Proceedings of SALT 23, 40–62. Ashby, William J. & Paola Bentivoglio. 1993. ‘Preferred argument structure in spoken French and Spanish.’ Language Variation and Change 5: 61–​76.

 

References   1145 Ashby, William J. & Paola Bentivoglio. 2003. ‘Preferred argument structure across time and space:  A  comparative diachronic Analysis of French and Spanish,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 61–​80. Assmann, Anke, Doreen Georgi, Fabian Heck, Gereon Müller, & Philipp Weisser. 2015. ‘Ergatives move too early: On an instance of opacity in syntax.’ Syntax 18(4): 343–387.] Aurnague, Michel. 2001. ‘Entités et relations dans les descriptions spatiales: L’espace et son expression en basque et en français. Habilitation thesis.’ Université de Toulouse-​Le Mirail. Austin, Jennifer. 2007. ‘Grammatical interference and the acquisition of ergative case in bilingual children learning Basque and Spanish.’ Bilingualism:  Language and Cognition 10(3): 315–​331. Austin, Jennifer. 2009. ‘Delay, interference and bilingual development:  The acquisition of verbal morphology in children learning Basque and Spanish.’ International Journal of Bilingualism 13(4): 447–​479. Austin, Jennifer. 2010. ‘Rich inflection and the production of root infinitives in child language.’ Morphology 20(1), 41–​69. Austin, Jennifer. 2012. ‘The case–​agreement hierarchy in acquisition: Evidence from children learning Basque.’ Lingua 122: 289–​302. Austin, Jennifer. 2013. ‘Ergativity in child Basque,’ in Edith L. Bavin & Sabine Stoll (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 35–​69. Austin, Peter. 1981a. ‘Switch-​reference in Australia.’ Language 57: 309–​334. Austin, Peter. 1981b. A grammar of Diyari, South Australia. Cambridge: Cambridge University Press. Austin, Peter. 1982. ‘Transitivity and cognate objects in Australian languages,’ in Paul J. Hopper & Sandra A. Thompson (eds.), Studies in transitivity (Syntax and Semantics 15). New York: Academic Press, 37–​47. Austin, Peter. 2013. A grammar of Diyari (version 2.5). Cambridge:  Cambridge University Press. Austin, Peter, Barry Blake, & Margaret Florey. 2001. ‘Explorations in valency in Austronesian languages.’ La Trobe Papers in Linguistics 11. Bundoora: La Trobe University. Authier, Gilles. 2009. Grammaire Kryz:  Langue caucasique d’Azerbaidjan dialecte d’Alik. Leuven: Peeters. Authier, Gilles. 2011. ‘From adlocative case to debitive mood without desubordination,’ in Gilles Authier & Timur Maisak (eds.), Tense, aspect, modality and finiteness in East Caucasian languages. Bochum: Brockmeyer, 67–​94 Authier, Gilles & Katharina Haude. 2012. Ergativity, valency and voice. Berlin:  Walter de Gruyter. Averintseva-​Klisch, Maria. 2008. ‘To the right of the clause: Right dislocation vs. afterthought,’ in C. Fabricius-​Hansen & W. Ramm (eds.), ‘Subordination’ versus ‘Coordination’ in sentence and text:  From a cross-​linguistic perspective. Amsterdam & Philadelphia, PA:  Benjamins, 217–​239. Axenov, Serge. 2006. The Balochi language of Turkmenistan:  A  corpus-​based grammatical description. Uppsala: Acta Upsaliensis. Babaliyeva, Ayten. 2013. ‘Études sur la morphosyntaxe du tabasaran littéraire.’ PhD dissertation, École pratique des hautes études Paris. Bach, Kent & Robert Harnish. 1979. Linguistic communication and speech acts. Cambridge, MA: MIT Press.

 

1146   References Badecker, Willliam & Frantisek Kuminiak. 2007. ‘Morphology, agreement and working memory retrieval in sentence production: Evidence from gender and case in Slovak.’ Journal of Memory and Language 56(1), 65–​85. Baerman, Matthew. 2009. ‘Case syncretism,’ in Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 219–​231. Baerman, Matthew & Dunstan Brown. 2013. ‘Case syncretism,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. [available online at http://​wals.info/​chapter/​28, accessed 2014–​12–​11] Baerman, Matthew, Dunstan Brown, & Greville G. Corbett. 2005. The syntax–​morphology interface: A study of syncretism. Cambridge: Cambridge University Press. Bagrationi (Bat’onishvili), Ioane. 1936. K’almasoba. T’pilisi: Saxelgami. Baker, Brett, & Mark Harvey. 2010. ‘Complex predicate formation,’ in Mengistu Amberber, Brett Baker, & Mark Harvey (eds.), Complex predicates: Cross-​linguistic perspectives on event structure. Cambridge: Cambridge University Press, 13–​47. Baker, Mark C. 1985. ‘The mirror principle and morphosyntactic explanation.’ Linguistic Inquiry 16: 373–​415. Baker, Mark C. 1988. Incorporation:  A  theory of grammatical function changing. Chicago: University of Chicago Press. Baker, Mark C. 1996. The polysynthesis parameter. Oxford: Oxford University Press. Baker, Mark C. 1997. ‘Thematic roles and syntactic structure,’ in Lilianne Haegeman (ed.), Elements of grammar. Dordrecht: Kluwer, 73–​137. Baker, Mark C. 2001. The atoms of language. New York: Basic Books. Baker, Mark C. 2003. Lexical categories:  Verbs, nouns and adjectives. Cambridge: Cambridge University Press. Baker, Mark C. 2008. The syntax of agreement and concord. Cambridge: Cambridge University Press. Baker, Mark C. 2014a. ‘On dependent ergative case (in Shipibo) and its derivation by phase.’ Linguistic Inquiry 45: 341–​380. Baker, Mark C. 2014b. ‘Pseudo noun incorporation as covert noun incorporation: Linearization and crosslinguistic variation.’ Language and Linguistics 15: 5–​46. Baker, Mark C. 2015. Case: Its principles and its parameters. Cambridge: Cambridge University Press. Baker, Mark C. & Nadezhda Vinokurova. 2010. ‘Two modalities of case assignment in Sakha.’ Natural Language and Linguistic Theory 28: 593–​642. Baker, Mark C, Kyle Johnson, & Ian Roberts. 1989. ‘Passive arguments raised’. Linguistic Inquiry 20: 219–​251. Baker, Mark C, Roberto Aranovich, & Lucía Golluscio. 2005. ‘Two types of syntactic noun incorporation:  Noun incorporation in mapudungun and its typological implications.’ Language 81: 138–​177. Bakker, Peter & Yaron Matras. (eds.). 2013. Contact languages: A comprehensive guide. Language Contact and Bilingualism 6. Boston/​Berlin: de Gruyter Mouton. Ball, Douglas. 2007. ‘On ergativity and accusativity in Proto-​Polynesian and Proto-​Central Pacific.’ Oceanic Linguistics 46(1): 128–​153. Ball, Douglas. 2009. ‘Clause structure and argument realization in Tongan.’ PhD dissertation, Stanford University.

 

References   1147 Ballard, Lee. 2011. ‘Phonology, grammar, morphophonemics,’ in Ameda, Chimcas, Gonzalo A. Tigo, Vicente B. Mesa, & Lee Ballard (eds.), Ibaloy. Baguio: Diteng and the Cordillera Studies Center, University of the Philippines, 772–​891. Bar-​Asher Siegal, Elitzur Avraham. 2007. ‘The origin and the typology of the pattern Qtil Li in Syriac and Babylonian Aramaic,’ in A. Maman, S.E. Fassberg, & Y. Breuer (eds.), Shaʻare Lashon:  Studies in Hebrew, Aramaic, and Jewish languages in honor of Moshe Bar-​Asher. Jerusalem: Bialik Institute, 360–​392. [in Hebrew] Bar-​Asher Siegal, Elitzur Avraham. 2011. ‘On the passiveness of one pattern in Jewish Babylonian Aramaic:  A  linguistic and philological discussion.’ Journal of Semitic Studies 56: 111–​143. Bar-​el, Leora, Henry Davis, & Lisa Matthewson. 2005. ‘On non-​culminating accomplishments.’ Proceedings of NELS 35: 79–​92. Barker, Chris. 1998. ‘Episodic -​ee in English: A thematic role constraint on new word formation.’ Language 74: 695–​727. Baranzehi, Adem. 2003. ‘The Sarawani dialect of Balochi and Persian influence on it,’ in Carina Jahani & Agnes Korn (eds.), The Baloch and their neighbours. Ethnic and linguistic contact in Balochistan in historical and modern times. Wiesbaden: Reichert, 75–​111 Barotto Alessandra. 2014a ‘Split ergativity in the NENA dialects,’ in G. Khan & L. Napiorkowska (eds.), Neo-​Aramaic and its linguistic context. Piscataway, NJ: Gorgias Press, 232–​249 Barotto, Alessandra. 2014b. ‘Typology of case alignments in NENA Dialects.’ Ricognizioni [online] 1(1). Barðdal, Jóhanna. 2002. ‘ “Oblique subjects” in Icelandic and German.’ Working Papers in Scandinavian Syntax 70: 61–​99. Barðdal, Jóhanna. 2011. ‘Lexical vs. structural case: A false dichotomy.’ Morphology 21(3–​4): 619–​654. Barðdal, Jóhanna. 2013. ‘Construction-​ based historical–​ comparative reconstruction,’ in Thomas Hoffmann & Graeme Trousdale (eds.), The Oxford handbook of construction grammar. New York: Oxford University Press, 438–​457. Barðdal, Jóhanna. 2015. ‘Syntax and syntactic reconstruction,’ in Claire Bowern & Bethwyn Evans (eds.), The Routledge handbook of historical linguistics. Abingdon, UK: Routledge, 343–​373. Barreña, Andoni. 1995. Gramatikaren jabekuntzagarapena eta haur euskaldunak. Bilbao: Servicio Editorial de la Universidad del País Vasco. Bashir, Elena. 1999. ‘The Urdu and Hindi ergative postposition ne: Its changing role in the grammar,’ in Rajendra Singh (ed.), The yearbook of South Asian Languages and linguistics. New Delhi: Sage, 11–​36. Bashir, Elena. 2009. ‘Wakhi,’ in G. Windfuhr (ed.), The Iranian Languages. London: Routledge Curzon, 825–​862. Basilico, David. 2003. ‘The topic of small clauses.’ Linguistic Inquiry 34: 1–​35. Basilico, David. 2004. ‘Antipassive, anticausative and reflexive: Phases and reflexive morphology.’ Paper presented at the 35th Meeting of the Northeastern Linguistic Society. NELS 35. University of Connecticut, October. Basilico, David. 2008. ‘The syntactic representation of perfectivity.’ Lingua 118: 1716–​1739. Basilico, David. 2012. ‘The antipassive and its relation to scalar structure,’ in María Cristina Cuervo & Yves Roberge (eds.), The end of argument structure. Bingley:  Emerald Group, 75–​104. Basri, Hasan & Daniel Finer. 1987. ‘The definiteness of trace.’ Linguistic Inquiry 18: 141–​147.

 

1148   References Basri, Hasan. 1999. ‘Phonological and syntactic reflections of the morphological structure of Selayarese.’ PhD dissertation, State University of New York at Stony Brook. Bath, Darbhe Narayana Shankara. 1991. Grammatical relations: The evidence against their necessity and universality. London & New York: Routledge. Bauer, Brigitte. 2000. Archaic syntax in Indo-​European: The spread and transitivity in Latin and French. Berlin: Mouton de Gruyter. Bauer, Winifred. 1983. ‘On the category “direct object” in Maori.’ Paper presented at the Linguistic Society of New Zealand Conference, Auckland. Bauer, Winifred. 1997. The Reed reference grammar of Māori. Auckland, NZ: Reed. Bavant, Marc. 2014. Resultatif, diathese et possession en basque, vieux perse et elamite. Utrecht: Netherlands Graduate School of Linguistics. Bavin Edith L. 1992. ‘The acquisition of Warlpiri,’ in D.I. Slobin (ed.), The crosslinguistic study of language acquisition, vol. 3. Hillsdale, NJ: Lawrence Erlbaum, 309–​372. Bavin Edith L. 2000. ‘Ellipsis in Warlpiri children’s narratives:  An analysis of frog stories.’ Linguistics 38: 569–​589. Bavin, Edith L. 2004. ‘Focusing on where,’ in S. Stromqvist & L. Verhoeven (eds.), Relating events in narrative: Typological and contextual perspectives: volume 2. Mahwah, NJ: Erlbaum, 17–​35. Bavin, Edith L. 2013. ‘The acquisition of ergativity in Warlpiri,’ in Edith L. Bavin & Sabine Stoll (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 107–​131. Bavin, Edith L. & Sabine Stoll (eds.). 2013. The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins. Beames, John. 1872–​1879. A comparative grammar of the modern Aryan languages of India. Delhi: Munshiram Manoharlal. [repub. 1966] Beavers, John. 2011. ‘On affectedness.’ Natural Language and Linguistic Theory 29: 335–​370. Beckman, Jill. 1998. ‘Positional faithfulness.’ PhD dissertation, University of Massachusetts, Amherst. Beckner, Clay, Richard Blythe, Joan L. Bybee, Morten H. Christiansen, William Croft, Nick C. Ellis, John H. Holland, Jinyun Ke, Diane Larsen-​Freeman, & Tom Schoenemann. 2009. ‘Language is a complex adaptive system: Position paper.’ Language Learning 59 (suppl. 1): 1–​26. Béjar, Susana & Milan Rezac. 2003. ‘Person licensing and the derivation of PCC effects.’ In Romance Linguistics: Theory and acquisition, eds. A. T. Pérez-Leroux and Y. Roberge. Amsterdam: John Benjamins, 49–62. Bell, Jeanie. 2003. ‘A sketch grammar of the Badjala language of Gari (Fraser Island).’ MA thesis, University of Melbourne. Belletti, Adriana & Luigi Rizzi. 1988. ‘Psych-​verbs and theta-​theory.’ Natural Language & Linguistic Theory 6: 291–​352. Bender, Marvin Lionel. 1996. The Nilo-​Saharan Languages:  A  comparative essay. Munich: Lincom Europa. Bentivoglio, Paola. 1994. ‘Spanish preferred argument structure across time and space.’ Revista de Documentaçao de Estudos em Lingüística Teórica e Aplicada (DELTA) 10: 227–​293. Bentivoglio, Paola. 1998. ‘The late acquisition of preferred argument structure in Venezuelan spoken Spanish,’ in Bernard Caron (ed.), Proceedings of the Xvith international congress of linguists. Amsterdam: Elsevier. Bentzen, Kristine. 2006. ‘Order and structure in embedded clauses in Northern Norwegian.’ PhD dissertation, CASTL, University of Tromsø.

 

References   1149 Benua, Laura. 1995. ‘Yup’ik antipassive.’ Chicago Linguistic Society. Papers from the Regional Meetings 31: 28–​44. Benveniste, Émile. 1952. ‘La construction passive du parfait transitif.’ Paris: C. Klincksieck. Benveniste, Émile. 1952/​1971. ‘The passive construction of the transitive perfect,’ in Émile Benveniste, Problems in general linguistics. Coral Gables: University of Miami Press, 153–​161. Benz, Anton & Jason Mattausch (eds.). 2011. Bidirectional optimality theory. Amsterdam & Philadelphia, PA: Benjamins. Berge, Anna. 2011. Topic and discourse structure in West Greenlandic agreement constructions. Lincoln, NE: University of Nebraska Press. Berger, Hermann. 1974. Das Yasin-​Burushaski (Werchikwar). Grammatik, Texte, Wörterbuch. Wiesbaden: Otto Harrassowitz. Berinstein, Ava. 1985. Evidence for multiattachment in K’ekchi Mayan. New York: Garland. Berman, Ruth & Dan I. Slobin. 1994. Relating events in narrative, volume 1. Hillsdale, NJ: Erlbaum. Berro, Ane. 2010. ‘Unergative predicates in Basque varieties: Consequences for ergative case assignment,’ in P. Salaburu, P. Goenaga, & I. Sarasola (eds.), DEA research work. Leioa: University of the Basque Country. Berro, Ane. 2012. ‘Three levels of root insertion in Basque intransitive verbs,’ in E. Carrilho & B. Fernández (eds.), Journal of Portuguese Linguistics (special issue) 11(1): 7–​22. Berro, Ane. 2015. ‘Breaking verbs: From event structure to syntactic categories in Basque.’ PhD dissertation, UPV-​EHU and Université de Bordeaux-​Montaigne. Berro, Ane. 2016. ‘On the relation between ergativity, stativity and the verbal configuration of Basque,’ in B. Fernández & J. Ortiz de Urbina (eds.), Microparameters in the grammar of Basque. Amsterdam & Philadelphia, PA: Benjamins, 39–66. Beyssade, Claire & Jean-​Marie Marandin. 2006. ‘The speech act assignment problem revisited: Disentangling speaker’s commitment from speaker’s call on addressee.’ Selected Papers of CSSP, 37–​68. Bhaskararao, Peri & K.V. Subbārāo. (eds.). 2004. Non-​nominative subjects, vols. 1–​ 2. Amsterdam & Philadelphia, PA: Benjamins. Bhat, Darbhe Narayana Shankara. 1991. Grammatical relations: The evidence against their necessity and universality. London & New York: Routledge. Bhat, Darbhe Narayana Shankara. & M.S. Ningomba. 1997. Manipuri grammar (Lincom Studies in Asian Linguistics, 4). Munich: Lincom Europa. Bhatia, Tej. 1993. Punjabi: A cognitive–​descriptive grammar. London: Routledge. Bhatt, Rajesh. 2003. ‘Causativisation:  Topics in the syntax of the modern Indo-​Aryan languages.’ Unpublished handout. Bhatt, Rajesh. 2005. ‘Long distance agreement in Hindi–​Urdu.’ Natural Language and Linguistic Theory 23: 757–​807. Bhatt, Rajesh & Elena Anagnostopoulou. 1996. ‘Object shift and specificity: Evidence from ko-​ phrases in Hindi,’ in Lise M. Dobrin, Kora Singer, & Lisa McNair (eds.), Papers from the main session of Chicago Linguistic Society. Chicago, IL: Chicago Linguistics Society, 32(1): 11–​22. Bianchi, Valentina. 2000. ‘The raising analysis of relative clauses: A reply to Borsley.’ Linguistic Inquiry 31: 123–​140 Bickel, Balthasar. 2000. ‘On the syntax of agreement in Tibeto-​Burman.’ Studies in Language 24(3): 583–​610. Bickel, Balthasar. 2003. ‘Belhare,’ in Graham Thurgood & Randy J. LaPolla (eds.), The Sino-​ Tibetan languages. New York: Routledge, 546–​570.

 

1150   References Bickel, Balthasar. 2004. ‘The syntax of experiencers in the Himalayas,’ in Peri Bhaskararao & K.V. Subbārāo (eds.), Non-​nominative subjects. Amsterdam & Philadelphia, PA: Benjamins, 77–​111. Bickel, Balthasar. 2008. ‘A refined sampling procedure for genealogical control.’ Sprachtypologie und Universalienforschung 61: 221–​233. Bickel, Balthasar. 2011. ‘Grammatical Relations Typology,’ in J.J. Song (ed.), The Oxford handbook of linguistic typology. Oxford: Oxford University Press, 399–​444. Bickel, Balthasar & Johanna Nichols. 2001. ‘Syntactic ergativity in light verb complements.’ Proceedings of the Berkeley Linguistics Society 27: 39–​52. Bickel, Balthasar & Johanna Nichols. 2009. ‘Case marking and alignment,’ in Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford:  Oxford University Press, 27–​43. Bickel, Balthasar & Alena Witzlack-​Makarevich. 2008. ‘Referential scales and case alignment: Reviewing the typological evidence.’ Scales 86: 1–​37. Bickel, Balthasar, Alena Witzlack-​ Makarevich, & Taras Zakharko. 2015. ‘Typological evidence against universal effects of referential scales on case alignment,’ in Ina Bornkessel-​Schlesewsky, Andrej Malchukov, & Marc Richards (eds.), Scales and hierarchies: A cross-​disciplinary perspective. Berlin: de Gruyter, 7–​45. Bielmeier, Roland, Felix Haller, Katrin Hassler, Brigitte Huber, & Marianne Volkart. 2008. ‘A short guide to the comparative dictionary of Tibetan dialects (CTDT).’ MS, University of Bern. Bierkandt, Lennart. 2006. ‘Kasusmorphologie des Diyari. Ein Ansatz im Rahmen der Distribuierten Morphologie,’ in G. Müller & J. Trommer (eds.), Subanalysis of argument encoding in distributed morphology, vol. 84 of Linguistische Arbeitsberichte. Universität Leipzig, 43–​62. Bierwisch, Manfred. 1967. ‘Syntactic features in morphology: General problems of so-​called pronominal inflection in German,’ To honor Roman Jakobson. The Hague & Paris: Mouton, 239–​270. Binder, Jeffrey R., Julie A. Frost, Thomas A. Hammeke, Robert W. Cox, Stephen M. Rao, & Thomas Prieto. 1997. ‘Human brain language areas identified by functional magnetic resonance imaging.’ Journal of Neuroscience: The Official Journal of the Society for Neuroscience 17(1): 353–​362. Birner, Betty J. & Gregory Ward. 1996. ‘A crosslinguistic study of postposing in discourse.’ Language and Speech 39: 111–​140. Birner, Betty J. & George Ward. 1998. Information status and noncanonical word order in English. Amsterdam & Philadelphia, PA: Benjamins. Bittner, Dagmar, Wolfgang Dressler, & Marianne Kilani-​Schoch (eds.). 2003. Development of verb inflection in first language acquisition:  A  cross-​ linguistic perspective. Berlin & New York: Mouton de Gruyter. Bittner, Maria. 1987. ‘On the semantics of the Greenlandic antipassive and related constructions.’ International Journal of American Linguistics 53: 194–​231. Bittner, Maria. 1994. Case, scope, and binding. Dordrecht: Kluwer. Bittner, Maria & Ken Hale. 1996a. ‘The structural determination of case and agreement.’ Linguistic Inquiry 27: 1–​68. Bittner, Maria & Kenneth Hale. 1996b. ‘Ergativity: Toward a theory of a heterogeneous class.’ Linguistic Inquiry 27: 531–​604. Blake, Barry J. 1976. ‘On ergativity and the notion of subject: Some Australian cases.’ Lingua 39: 281–​300.

 

References   1151 Blake, Barry J. 1977. Case marking in Australian languages. Canberra: Australian Institute of Aboriginal Studies. Blake, Barry J. 1979a. A Kalkatungu grammar (Pacific Linguistics Series B-​ 57). Canberra: Australian National University. Blake, Barry J. 1979b. ‘Degrees of ergativity in Australia,’ in Franz Plank (ed.), Ergativity: towards a theory of grammatical relations. London; New York: Academic Press, 291–​305. Blake, Barry J. 1979c. ‘Pitta-​Pitta,’ in R.M.W. Dixon & Barry J. Blake (eds.), Handbook of Australian languages. Volume 1. Canberra: Australian National University Press, 183–​242. Blake, Barry J. 1983. ‘Structure and word order in Kalkatungu: The anatomy of a flat language.’ Australian Journal of Linguistics 3: 143–​175. Blake, Barry J. 1987a. ‘The grammatical development of Australian languages.’ Lingua 71: 179–​210. Blake, Barry J. 1987b. Australian Aboriginal grammar. London: Croom Helm. Blake, Barry J. 1988. ‘Redefining Pama–​Nyungan: Towards the prehistory of Australian languages.’ Aboriginal Linguistics 1: 1–​90. Blake, Barry J. 1990a. ‘Languages of the Queensland/​Northern territory border: Updating the classification,’ in Peter Austin (ed.), Language and history: Essays in honour of Luise A. Hercus (Pacific Linguistics C-​116). Canberra: Australian National University, 49–​66. Blake, Barry J. 1990b. Relational grammar. London: Routledge. Blake, Barry J. 1994. Case. Cambridge: Cambridge University Press. Blake, Barry J. 2001. Case (2nd edn.). Cambridge: Cambridge University Press. Blevins, James P. 2005. ‘Thematic inversion in Georgian.’ MS, University of Cambridge. Blight, Ralph. 2004. ‘Head movement, passive, and antipassive in English.’ PhD dissertation, University of Texas at Austin. Bloch, Jules. 1906. ‘La phrase nominale en sanskrit.’ Mémoires de la Société Linguistique de Paris 14: 27–​96. Blom, Elma & Frank Wijnen. 2006. ‘Development need not be embarrassing: The demise of the root infinitive and related changes in Dutch child language.’ MS, University of Amsterdam/​ Utrecht University. Blom, Elma. 2007. ‘Modality, infinitives, and finite bare verbs in Dutch and English child language.’ Language Acquisition 14(1): 75–​113. Bloomfield, Leonard. 1917. Tagalog texts. Urbana, IL: University of Illinois Press. Blust, Robert. 1990. ‘Summary report: Linguistic change and reconstruction methodology in the Austronesian language family,’ in Philip Baldi (ed.), Linguistic change and reconstruction methodology. Berlin: Walter de Gruyter, 133–​154. Blust, Robert. 2002. ‘Notes on the history of “focus,” in Austronesian languages,’ in Fay Wouk & Malcolm Ross (eds.), The history and typology of Western Austronesian voice systems. Canberra: Pacific Linguistics, 63–78. Blust, Robert. 2013. The Austronesian languages (rev. edn.) Asia-​Pacific Linguistics. Canberra: Australia. Blust, Robert & Stephen Trussel. 2010. Austronesian comparative dictionary (web edn.). http://​ www.trussel2.com/​acd/​introduction.htm Blutner, Reinhard. 2000. ‘Some aspects of optimality in natural language interpretation.’ Journal of Semantics 17: 189–​216. Boas, Franz. 1911. ‘Tsimshian.’ Handbook of American Indian languages, vol. 40, pt. 1, 287–​422. Bobaljik, Jonathan David. 1993a. ‘On ergativity and ergative unergatives,’ in Collin Phillips (ed.), MIT working papers in linguistics 19:  Papers on case and agreement II. Cambridge, MA: MIT Working Papers in Linguistics. 45–​88.

 

1152   References Bobaljik, Jonathan David. 1993b. ‘Nominally absolutive is not absolutely nominative,’ Proceedings of the Eleventh West Coast Conference on Formal Linguistics. Stanford, CA: CSLI. Bobaljik, Jonathan David. 1995. ‘Morphosyntax:  The syntax of verbal inflection.’ PhD dissertation, MIT. Bobaljik, Jonathan David. 2007. ‘The limits of deponency:  A  Chuktoko-​centric perspective,’ in Matthew Baerman et  al. (eds.), Deponency and morpho-​syntactic mismatches. Oxford: Oxford University Press, 175–​201. Bobaljik, Jonathan David. 2008. ‘Where’s Phi? Agreement as a post-​syntactic operation,’ in Daniel Harbour, David Adger, & Susana Béjar (eds.), Phi-​theory: Phi features across interfaces and modules. Oxford: Oxford University Press, 295–​328. Bobaljik, Jonathan David & Phil Branigan. 2006. ‘Eccentric agreement and multiple case-​ checking,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity: Emerging issues. Springer: Dordrecht, 47–​77. Bobaljik, Jonathan David & Susi Wurmbrand. 2008. ‘Case in GB/​minimalism,’ in A. Malchukov & A. Spencer (eds.), The Oxford handbook of case. Oxford University Press, 44–​58. Boeder, Winfried. 1979. ‘Ergative syntax in language change: The South Caucasian languages,’ in Frans Plank (ed.), Ergativity: Towards a theory of grammatical relations. London: Academic Press, 435–​480. Boeder, Winfried. 2002. ‘Syntax and morphology of polysynthesis in the Georgian verb,’ in Nicholas Evans & Hans-​Jürgen Sasse (eds.), Problems of polysynthesis. Berlin:  Akademie Verlag, 87–​111. Boeder, Winfried (2004). ‘The South Caucasian languages.’ Lingua 115: 5–​89. Bogomolova, Natalja K. 2012. ‘Ličnoe soglasovanie v tabasaranskom jazyke. Konceptualizator i ego adresata v structure situacii.’ Voprosy Jazykoznanija 4: 101–​124. Bohnacker, Ute & Somayeh Mohammadi. 2012. ‘Acquiring Persian object marking: Balochi learners of L2 Persian.’ Orientalia Suecana 61: 59–​89. Bohnemeyer, Jürgen. 2015. ‘A practical epistemology for semantic elicitation in the field and elsewhere,’ in Ryan Bochnak & Lisa Matthewson (eds.), Methodologies in semantic fieldwork. Oxford: Oxford University Press, 13–​46. Bok-​Bennema, Reineke. 1991. Case and agreement in Inuit. Berlin & New York: Foris. Bolt, Janet E., William G. Hoddinott, & Frances M. Kofod. 1971. An elementary grammar of the Ngaliwuru language of the Northern Territory. Canberra: Australian Institute of Aboriginal Studies. Bond, Oliver, Kristine A. Hildebrandt, & Dubi Nanda Dhakal. 2013a. ‘Optional case marking:  What can be expressed by its absence?’ Presentation, 10th Biennial Association for Linguistic Typology Meeting, August, 15–​18, MPI EVA Leipzig, Germany. Bond, Oliver, Kristine A. Hildebrandt, & Dubi Nanda Dhakal. 2013b. ‘Probabilistic case in the languages of Manang.’ Paper presented at the Cambridge Group for Endangered Languages and Cultures, Cambridge, UK, December, 4. Bonet, Eulalia. 1991. ‘Morphology after syntax: Pronominal clitics in Romance.’ PhD dissertation, MIT. Booth, James R., Lydia Wood, Dong Lu, James C. Houk, & Tali Bitan. 2007. ‘The role of the basal ganglia and cerebellum in language processing.’ Brain Research 1133: 136–​144. Bopp, Franz. 1847. Die Kaukasischen Glieder des indoeuropäischen Sprachstamms. Berlin: Dümmler. Borer, Hagit. 1994. ‘The projections of arguments,’ in Elena Benedicto & Jeff Runner (eds.), Functional projections. Amherst, MA: GLSA, University of Massachusetts, 19–47.

 

References   1153 Borer, Hagit. 2005a. In name only. Oxford: Oxford University Press. Borer, Hagit. 2005b. The normal course of events. Oxford: Oxford University Press. Borer, Hagit. 2013. Taking form. Oxford: Oxford University Press. Borik, Olga. 2002. ‘Aspect and reference time.’ PhD dissertation, University of Utrecht. Bornkessel-​Schlesewsky, Ina & Angela D. Friederici. 2007. ‘Neuroimaging studies of sentence and discourse comprehension,’ in M. Gaskell (ed.), The Oxford handbook of psycholinguistics. Oxford: Oxford University Press, 407–​424. Bornkessel-​Schlesewsky, Ina & Matthias Schlesewsky. 2009. Processing syntax and morphology: A neurocognitive perspective. Oxford: Oxford University Press. Bornkessel-​Schlesewsky, Ina & Matthias Schlesewsky. 2015. ‘Scales in real-​time language comprehension: A review,’ in I. Bornkessel-​Schlesewsky, A. L. Malchukov, & M. Richards (eds.), Scales and hierarchies: A cross-​disciplinary perspective. Berlin: Mouton de Gruyter, 321–352. Bornkessel-​ Schlesewsky, Ina, Kamal Kumar Choudhary, Alena Witzlack-​ Makarevich, & Balthasar Bickel. 2008. ‘Bridging the gap between processing preferences and typological distributions: Initial evidence from the online comprehension of control constructions in Hindi.’ Scales [= Ling. Arbeits Berichte 86], 397–​436. Borsley, Robert & Jaklin Kornfilt. 2000. ‘Mixed extended projection,’ in R. Borsley (ed.), The nature and function of syntactic categories. London: Academic Press, 133–​166. Bošković, Zeljkǒ. 2005. ‘On the locality of left branch extraction and the structure of NP.’ Studia Linguistica 59: 1–​45. Bossong, Georg. 1985. Empirische Universalienforschung. Differentielle Objektmarkierung in der neuiranischen Sprachen. Tübingen: Narr Bossong, Georg. 1991. ‘Differential object marking in Romance and beyond,’ in D. Kibbee & D. Wanne (eds.), New analyses in Romance linguistics. Amsterdam & Philadelphia, PA: Benjamins, 143–​170. Bottari, Piero. 1992. ‘Romance passive nominals.’ Geneva Generative Papers 0(0): 66–​80. Bowden, John. 2001. Taba: A description of a South Halmahera language. Canberra: Pacific Linguistics. Bowerman, Melissa & Penelope Brown (eds.). 2008. Crosslinguistic perspectives on argument structure: Implications for learnability. Mahwah, NJ: Erlbaum. Bowern, Claire & Quentin Atkinson. 2012. ‘Computational phylogenetics and the internal structure of Pama–​Nyungan.’ Language 88: 817–​845. Bowern, Claire & Harold Koch (eds.). 2004. Australian languages: Classification and the comparative method (Current Issues in Linguistic Theory,  249). Amsterdam & Philadelphia, PA: Benjamins. Bowers, John. 2013. Arguments as relations. Cambridge, MA: MIT Press. Breen, J. Gavan. 1976. ‘Wangkumara,’ in Robert M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 336–​339. Bresnan, Joan, Shipra Dingare, & Christopher D. Manning. 2001. Soft constraints mirror hard constraints: Voice and person in English and Lummi. Proceedings of the LFG01 Conference. Stanford, CA: CSLI, 13–​32. Bresnan, Joan & Lioba Moshi. 1990. ‘Object asymmetries in comparative Bantu syntax.’ Linguistic Inquiry 21: 147–​85. Bricker, Victoria R. 1981. ‘The source of the ergative split in Yucatec Maya.’ Journal of Mayan Linguistics 2(2): 83–​127. Brito, Célia Maria Coélho. 1996. A transitividade verbal na língua portuguesa: Uma investigaçao de base funcionalista. São Paulo: Araraquara.

 

1154   References Brito, Célia Maria Coélho. 1998. ‘A transitividade verbal na língua portuguesa em diferentes tipos de gênero de texto.’ Veredas: Revista de Estudos Lingüísticos da Universidade Federal de Juiz de Fora 3(27–​35). Broadwell, Aaron. 2006. A Choctaw reference grammar. Lincoln, NE: University of Nebraska Press. Bromley, H. Myron. 1981. A grammar of Lower Grand Valley Dani. Pacific Linguistics, Series C-​ 63. Canberra: Research School of Pacific and Asian Studies, Australian National University. Brosset, Marie-Félicité. 1834. L’art libéral; ou, grammaire géorgienne. Paris: Roissy. Brouwer, Harm, Hartmut Fitz, & John Hoeks. 2012. ‘Getting real about semantic illusions:  Rethinking the functional role of the P600 in language comprehension.’ Brain Research 1446: 127–​143. Brown, Dunstan & Andrew Hippisley. 2012. Network morphology: A defaults-​based theory of word structure. Cambridge: Cambridge University Press. Brown, Jason. 2010. Gitksan phonotactics. Munich: Lincom. Brown, Penelope. 1998. ‘Early Tzeltal verbs: Argument structure and argument representation.’ Proceedings of the Annual Child Language Research Forum 29: 129–​140. Brown, Penelope. 2008. ‘Verb specificity and argument realization in Tzeltal child language,’ in Melissa Bowerman & Penelope Brown (eds.), Crosslinguistic perspectives on argument structure: Implications for learnability. Mahwah, NJ: Erlbaum, 167–​189. Brown, Penelope, Barbara Pfeiler, Lourdes de León, & Clifton Pye. 2013. ‘The acquisition of agreement in four Mayan languages,’ in Edith L. Bavin & Sabine Stoll (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 271–​305. Brown, Roger. 1973. A first language: The early stages. Cambridge, MA: Harvard University Press. Bruening, Benjamin. 2007. ‘On diagnostics of structural case and the nature of ergative case.’ MS, University of Delaware. Bruening, Benjamin. 2013. ‘By-​phrases in passives and nominals.’ Syntax 16: 1–​41. Bubenik, Vit. 1989. ‘On the origins and elimination of ergativity in Indo-​Aryan languages.’ Canadian Journal of Linguistics 34(4): 377–​398. Bubenik, Vit. 1996. The structure and development of middle Indo-​Aryan dialects. Delhi: Motilal Banarsidass. Bubenik, Vit. 1998. Historical syntax of late middle Indo-​Aryan (Apabhramsa). Amsterdam & Philadelphia, PA: Benjamins. Burling, Robbins. 2003. ‘The Tibeto-​Burman languages of northeastern India,’ in Graham Thurgood & Randy J. LaPolla (eds.), The Sino-​Tibetan languages. New York: Routledge, 167–​192. Burzio, Luigi. 1986. Italian syntax: A government–​binding approach. Dordrecht & Lancaster: Reidel. Buth, Randall. 1981. ‘Ergative word order:  Luwo is OVS.’ Occasional Papers in the Study of Sudanese Languages 1: 74–​90. Butt, Miriam. 1993a. ‘Hindi–​Urdu infinitives as NPs.’ South Asian Language Review 3(1): 51–​72. Butt, Miriam. 1993b. ‘Object specificity and agreement in Hindi/​Urdu,’ in K. Beals et al. (eds.), Papers from the main session of Chicago Linguistic Society (CLS) 29. Chicago, IL: Chicago Linguistic Society, 89–​103. Butt, Miriam. 1995. The structure of complex predicates in Urdu. Stanford, CA: CSLI. Butt, Miriam. 1997. ‘Complex predicates in Urdu,’ in A. Alsina, J. Bresnan, & P. Sells (eds.), Complex predicates. Stanford, CA: CSLI, 107–​149. Butt, Miriam. 2001. ‘A reexamination of the accusative to ergative shift in Indo-​Aryan,’ in Miriam Butt & Tracy Holloway King (eds.), Time over matter: Diachronic perspectives on morphosyntax. Stanford, CA: CSLI, 105–​141.

 

References   1155 Butt, Miriam. 2006a. ‘The dative–​ergative connection,’ in O. Bonami & P. Cabredo Hofherr (eds.), Empirical Issues in Syntax and Semantics 6: 69–​92. Butt, Miriam. 2006b. Theories of case. Cambridge: Cambridge University Press. Butt, Miriam. 2008. ‘From spatial expression to core case marker:  Ergative and dative/​ accusative.’ Handout for paper presented at Linguistics, RSPAS, Australian National University. Butt, Miriam & Tafseer Ahmed. 2011. ‘The redevelopment of Indo-​Aryan case systems from a lexical semantic perspective.’ Morphology 21(3): 545–​572. Butt, Miriam, Tafseer Ahmed, & Tikaram Poudel. 2008. ‘Development of case in South Asian languages.’ Paper presented at DGfS Workshop Sprachwandelvergleich, Bamberg. Butt, Miriam & Tracy Holloway King. 1991. ‘Semantic case in Urdu,’ in L. Dobrin, L. Nichols, & R.M. Rodriguez (eds.), Papers from the 27th regional meeting of the Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 31–​45. Butt, Miriam & Tracy Holloway King. 2003. ‘Case systems: Beyond structural distinctions,’ in Ellen Brandner & Heike Zinmeister (eds.), New perspectives on case theory. Stanford, CA: CSLI, 53–​87. Butt, Miriam & Tracy Holloway King. 2004. ‘The status of case,’ in Veneeta Dayal & Anoop Mahajan (eds.), Clause structure in South Asian languages. Dordrecht: Kluwer, 153–198. Butt, Miriam & Tracy Holloway King. 2007. ‘Urdu in a parallel grammar development environment.’ Language Resources and Evaluation 41: 191–​207. Special Issue on Asian Language Processing: State of the Art Resources and Processing. Butt, Miriam & Aditi Lahiri. 2013. ‘Diachronic pertinacity of light verbs.’ Lingua 135: 7–​29. Butt, Miriam & Tikaram Poudel. 2007. ‘Distribution of the ergative in Nepali.’ Talk held at the University of Leipzig. http://​ling.uni-​konstanz.de/​pages/​home/​tafseer/​leipzig07-​hnd.pdf Butt, Miriam & Gillian Ramchand. 2005. ‘Complex aspectual structure in Hindi/​Urdu,’ in N. Ertischik-​Shir & T. Rapoport (eds.), The syntax of aspect. Oxford:  Oxford University Press, 117–​153. Butt, Miriam & Jafar Rizvi. 2010. ‘Tense and aspect in Urdu,’ in Patricia Cabredo-​Hofherr & Brenda Laca (eds.), Layers of aspect. Stanford, CA: CSLI, 43–​66. Bybee, Joan. 1985. Morphology: A study of the relation between meaning and form. Amsterdam & Philadelphia, PA: Benjamins. Bybee, Joan. 2000. ‘Lexicalization of sound change and alternating environments,’ in M.D. Broe & Janet B. Pierrehumbert (eds.), Papers in laboratory phonology V: Acquisition and the lexicon. Cambridge: Cambridge University Press, 250–​268. Bybee, Joan. 2007. Frequency of use and the organization of language. Oxford: Oxford University Press. Bybee, Joan. 2010. Language usage and cognition. Cambridge: Cambridge University Press. Bybee, Joan, Revere Perkins, & William Pagliuca. 1994. The evolution of grammar: Tense, aspect and modality in the languages of the world. Chicago, IL: Chicago University Press. Bynon, Theodora. 1979. ‘The ergative construction in Kurdish.’ Bulletin of the School of Oriental and African Studies 42: 211–​224. Bynon, Theodora. 1980. ‘From passive to active in Kurdish via the ergative construction,’ in Elisabeth Traugott, Rebecca Labrum, & Susan Shepherd (eds.), Papers from the 4th International Conference on Historical Linguistics. Amsterdam & Philadelphia, PA: Benjamins, 151–​163. Bynon, Theodora. 2005. ‘Evidential, raised possessor, and the historical source of the ergative construction in Indo-​Iranian.’ Transactions of the Philological Society 103(1): 1–​72.

 

1156   References Çağlayan, Handan. 2014. Same home, different languages. Intergenerational language shift: Tendencies, limitations, opportunities. The case of Diyarbakır. Diyarbakır: DISA Publications. Caha, Pavel. 2009. ‘The nanosyntax of case.’ PhD dissertation, CASTL, University of Tromsø. Calabrese, Andrea. 2008. ‘On absolute and contextual syncretism: Remarks on the structure of case paradigms and how to derive them,’ in Asaf Bachrach & Andrew Nevins (eds.), Inflectional identity. Oxford: Oxford University Press, 156–​202. Campana, Mark. 1992. ‘A movement theory of ergativity.’ PhD dissertation, McGill University. Campbell, Lyle. 2012. ‘Classification of the indigenous languages of South America,’ in Verónica Grondona & Lyle Campbell (eds.), The indigenous languages of South America: A comprehensive guide. Berlin: de Gruyter, 59–​166 Capell, Arthur. 1962. Some linguistic types in Australia (Oceania linguistic monographs, 7). Sydney: University of Sydney. Capell, Arthur. 1964. ‘Verbal systems in Philippine languages.’ Philippine Journal of Science 93: 231–​249. Caponigro, Ivano & Maria Polinsky. 2011. ‘Relative embeddings: A Circassian puzzle for the syntax–​semantics interface.’ Natural Language and Linguistic Theory 29: 71–​122. Cardinaletti, Anna & Ur Shlonsky. 2004. ‘Clitic positions and restructuring in Italian.’ Linguistic Inquiry 35(4): 519–​557. Cardona, George. 1970. ‘The Indo-​Iranian construction Mana (Mama) Kriam.’ Language 46: 1–​12. Carreiras, Manuel, Jon Andoni Duñabeitia, Marta Vergara, Irene de la Cruz-​Pavía, & Itziar Laka. 2010. ‘Subject relative clauses are not universally easier to process:  Evidence from Basque.’ Cognition 115: 79–​92. Carrier, Julien. 2014. ‘Applicative in Inuktitut and transitivity.’ PhD dissertation, University of Toronto. Carstens, Vicki. 2005. ‘Agree and EPP in Bantu.’ Natural Language and Linguistic Theory 23: 219–​279. Cartier, Alice. 1979. ‘De-​ voiced transitive verb sentences in formal Indonesian,’ Ergativity: Towards a theory of grammatical relations. New York: Academic Press, 161–​183. Catford, John C. 1974. ‘Ergativity in Caucasian languages.’ MS, University of Michigan. Cena, Resty. 1979. ‘Tagalog counterexamples to the accessibility hierarchy.’ Studies in Philippine Linguistics 31: 119–​124. Cerin, Mark. 1994. ‘The pronominal system of Yaralde.’ BA (hons.) thesis, University of Melbourne. Chadwick, Neil. 1976. ‘Ergative, Locative, and Instrumental Case Inflections: The Western Barkly languages,’ in R.M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 390–​396. Chadwick, Neil. 1984. The relationship of Jingulu and Jaminjungan. Batchelor, NT: School of Australian Linguistics. Chadwick, Neil. 1997. ‘The Barkly and Jaminjung languages: a non-contiguous genetic grouping’, in T. Darrell T. and M. Walsh (eds.), Boundary rider: Studies in the lexicology and comparative linguistics of Australian languages. Canberra: Pacific Linguistics, 95–106. Chafe, Wallace. 1976. ‘Givenness, contrastiveness, definiteness, subjects, topics, and point of view,’ in Charles N. Li (ed.), Subject and topic. New York: Academic Press, 25–​56. Chafe, Wallace. 1980. The pear stories: Cognitive, cultural, and linguistic aspects of narrative production. Norwood, NJ: Ablex. Chang, Henry Yung-​li. 2010. ‘On the syntax of Formosan adverbial verb constructions,’ in Raphael Mercado, Eric Potsdam, & Lisa Travis (eds.), Austronesian and theoretical linguistics. Amsterdam & Philadelphia, PA: Benjamins, 183–​211.

 

References   1157 Chatterji, Suniti Kumar. 1926. The origin and development of the Bengali language, vol. 2. Calcutta: Mehra, Rupa. [1975 edn.] Chelliah, Shobhana. 1997. A grammar of Meithei. Berlin: Mouton de Gruyter. Chelliah, Shobhana. 2009. ‘Semantic role to new information in Meithei,’ in Jóhanna Barðdal & Shobhana Chelliah (eds.), The role of semantic, pragmatic, and discourse factors in the development of case. Amsterdam & Philadelphia, PA: Benjamins, 377–​400. Chelliah, Shobhana. 2016. ‘ “Responsive methodology”:  Perspectives on data gathering and language documentation for South Asia.’ Journal of South Asian Languages and Linguistics, vol. 3, 175–198. Chelliah, Shobhana & Gwendolyn Hyslop. 2011. ‘Introduction to special issue on optional case marking in Tibeto-​Burman.’ Linguistics of the Tibeto-​Burman Area 34(2): 1–​7. Chelliah, Shobhana & Gwendolyn Hyslop (eds.). 2011/​2012. Linguistics of the Tibeto-​Burman Area 34(2) & 35(1) (special issue on optional case marking, pts. I & II). Chen, Cheng-​Fu. 2008. ‘Aspect and tense in Rukai: Interpretation and interaction.’ PhD dissertation, University of Texas, Austin. Chen, Tingchun. 2010. ‘Restructuring in Mayrinax Atayal.’ BA (hons.) thesis, McGill University. Chen, Tingchun. 2014. ‘Restructuring in Atayal.’ MS, MIT. Chen, Lang, Hua Shu, Youyi Liu, Jingjing Zhao, & Ping Li. 2007. ‘ERP signatures of subject–​ verb agreement in L2 learning.’ Bilingualism: Language and Cognition 10(2): 161. Chierchia, Gennaro. 2004. ‘A semantics for unaccusatives and its syntactic consequences.’ In A. Alexiadou, E. Anagnostopoulou, & M. Everaert (eds.), The unaccusativity puzzle: Explorations of the syntax-lexicon interface. Oxford University Press, 22–59. Chikobava, Arnold. 1936. Č’anuris gramat’ik’uli analizi t’ekst’ebiturt [‘A grammatical analysis of Laz, with texts’]. Tbilisi: Mecniereba. Chikobava, Arnold. 1948. Ergat’iuli k’onst’rukciis p’roblema iberiul-​k’avk’asiur enebši, I [‘The problem of the ergative construction in the Ibero-​Caucasian languages, I’]. Tbilisi: Mecniereba. Chikobava, Arnold. 1965. Iberiul-​k’avk’asiur enata šesc’avlis ist’oria [‘The history of the study of the Ibero-​Caucasian languages’]. Tbilisi: Ganatleba. Chirkova, Katia. 2012. ‘The Qiangic subgroup from an areal perspective: A case study of languages of Muli.’ Language and Linguistics 13(1): 133–​170. Chomsky, Noam. 1970. ‘Remarks on nominalization,’ in R. Jacobs & P. Rosenbaum (eds.), Readings in English transformational grammar. Waltham, MA: Ginn, 184–​221. Chomsky, Noam. 1980. ‘On binding.’ Linguistic Inquiry 11: 1–​46. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht, Foris. Chomsky, Noam. 1993. ‘A Minimalist Program for linguistic theory,’ in Ken Hale & Samuel J. Keyser (eds.), The view from Building 20: Essays in honor of Sylvain Bromberger. Cambridge, MA: MIT Press, 1–​52. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. ‘Minimalist inquiries:  The framework,’ in Roger Martin, David Michaels, & Juan Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, MA: MIT Press, 89–​155. Chomsky, Noam. 2001. ‘Derivation by phase,’ in Michael Kenstowicz (ed.), Ken Hale: A life in language. Cambridge, MA: MIT Press, 1–​52. Chomsky, Noam. 2005. ‘Three factors in language design.’ Linguistic Inquiry 36(1): 1–​22. Chomsky, Noam. 2008. ‘On phases,’ in Robert Freidin, Carlos P. Otero, & Maria Luisa Zubizarreta (eds.), Foundational issues in linguistic theory: Essays in honor of J.-​R. Vergnaud. Cambridge, MA: MIT Press, 133–​166.

 

1158   References Chomsky, Noam. 2013. ‘Problems of projection.’ Lingua 130: 33–​49. Choudhary, Kamal Kumar, Matthias Schlesewsky, Dietmar Roehm, & Ina Bornkessel-​ Schlesewsky. 2009. ‘The N400 as a correlate of interpretively relevant linguistic rules: Evidence from Hindi.’ Neuropsychologia 47(13), 3012–​3022. Christianson, Kiel & Fernanda Ferreira. 2005. ‘Conceptual accessibility and sentence production in a free word order language (Odawa).’ Cognition 28: 105–​135. Chubinov, David Jesseevich. 1855. Kratkaja gruzinskaja grammatika. St. Petersburg: Imperatorskaja Akademia Nauk. Chung, Sandra. 1976. ‘On the subject of two passives in Indonesian,’ in Charles Li (ed.), Subject and topic. New York: Academic Press, 57–98. Chung, Sandra. 1978. Case marking and grammatical relations in Polynesian. Austin: University of Texas Press. Chung, Sandra. 1982. ‘Unbounded Dependencies in Chamorro Grammar.’ Linguistic Inquiry 13(1): 39–​77. Chung, Sandra. 1994. ‘Wh-​agreement and “referentiality” in Chamorro.’ Linguistic Inquiry 25: 1–​44. Chung, Sandra. 1998. The design of agreement:  Evidence from Chamorro. Chicago, IL, & London: University of Chicago Press. Chung, Sandra. 2008. ‘Indonesian clause structure from an Austronesian perspective.’ Lingua 118: 1554–​1582. Churchward, Clerk Maxwell. 1940. Rotuman grammar and dictionary. Sydney: Australasian Medical Publishing. Churchward, Clerk Maxwell. 1953. Tongan grammar. Oxford: Oxford University Press. Cinque, Guglielmo. 1980. ‘On extraction from NP in Italian.’ Journal of Italian Linguistics 5: 47–​99. Cinque, Guglielmo. 1999. Adverbs and functional heads:  A  cross-​ linguistic perspective. New York: Oxford University Press. Clahsen, Harald, Fraibet Aveledo, & Iggy Roca. 2002. ‘The development of regular and irregular verb inflection in Spanish child language.’ Journal of Child Language 29: 591–​622. Clahsen, Harald. 1999. ‘Lexical entries and rules of language:  A  multidisciplinary study of German inflection.’ Behavioral and Brain Sciences 22: 991–​1013. Clancy, Patricia M. 1993. ‘Preferred argument structure in Korean acquisition,’ in Eve V. Clark (ed.), The proceedings of the twenty-​fifth annual Child Language Research Forum. Stanford, CA: CSLI, 307–​314. Clancy, Patricia M. 1995. ‘Subject and object in Korean acquisition: Surface expression and casemarking,’ in Susumo Kuno, Ik-​Hwan Lee, John Whitman, Joan Maling, Young-​Se Kang, & Young-​joo Kim (eds.), Harvard studies in Korean linguistics. Cambridge, MA: Harvard University Press, 3–​17. Clancy, Patricia M. 1996. ‘Referential strategies and the co-​construction of argument structure in Korean acquisition,’ in Barbara A. Fox (ed.), Studies in anaphora. Amsterdam & Philadelphia, PA: Benjamins, 33–​68. Clancy, Patricia M. 1997. ‘Discourse motivations of referential choice in Korean acquisition,’ in Ho-​min Sohn & John Haig (eds.), Japanese/​Korean linguistics. Stanford, CA: CSLI, 639–​659. Clancy, Patricia M. 2003. ‘The lexicon in interaction: Developmental origins of preferred argument structure,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure:  Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 81–​108.

 

References   1159 Clark, Herbert & Susan Brennan. 1991. ‘Grounding in communication.’ Perspectives on socially shared cognition 13: 127–​149. Clark, Ross. 1973. ‘Transitivity and case in Eastern Oceanic languages.’ Oceanic Linguistics 12: 559–​605. Clark, Ross. 1976. Aspects of Proto-​Polynesian syntax. Auckland:  Linguistic Society of New Zealand. Clarke, Sarah. 2009. ‘The manifestation of viewpoint aspect in Inuktitut.’ Studia Linguistica 63: 292–​322. Clemens, Lauren Eby. 2014. ‘Prosodic noun incorporation and verb-​initial syntax.’ PhD dissertation, Harvard University. Clemens, Lauren Eby, Coon, Jessica, Pedro Mateo Pedro, Adam Milton Morgan, Maria Polinsky, Gabrielle Tandet, & Matthew Wagers. 2015. ‘Ergativity and the complexity of extraction: A view from Mayan.’ Natural Language and Linguistic Theory 33, 2: 417–467. Cleverly, John Robert. 1968. A preliminary study of the phonology and grammar of Djamindjung. Canberra: Australian Institute of Aboriginal Studies. Clynes, Adrian. 1995. ‘Topics in the phonology and morphosyntax of Balinese—​based on the dialect of Singaraja, North Bali.’ PhD dissertation, Australian National University. Cocchi, Gloria. 1999. ‘Nominative and ergative languages: Towards a unified account of case checking.’ Syntaxis 2: 103–​129. Coghill, Eleanor. 2016. The Rise and Fall of Ergativity in Aramaic: Cycles of Alignment Change. Oxford Studies in Diachronic and Historical Linguistics 21. Oxford: Oxford University Press. Colarusso, John (1992). A grammar of the Kabardian language. Calgary:  University of Calgary Press. Cole, Peter & Gabriella Hermon. 1994. ‘Is there LF wh-​ movement?’ Linguistic Inquiry 2: 219–​262. Cole, Peter & Gabriella Hermon. 1998. ‘The typology of wh movement: Wh questions in Malay.’ Syntax 1: 221–​258. Cole, Peter & Gabriella Hermon. 2005. ‘Subject and non-​subject relativization in Indonesian.’ Journal of East Asian Linguistics 14: 59–​88. Cole, Peter & Gabriella Hermon. 2008. ‘VP Raising in a VOS language.’ Syntax 11(2): 144–​197. Cole, Peter, Gabriella Hermon, & Yanti. 2008. ‘Voice in Malay/​Indonesian.’ Lingua 118: 1500–​1553. Collins, Chris. 2002. ‘The syntax of nonfinite complementation:  An economy approach.’ Lingua 112: 127–​133. Collins, Chris. 2003. ‘The internal structure of vP in Ju’hoansi and Hoan.’ Studia Linguistica 57: 1–​25. Collins, Chris. 2005. ‘A smuggling approach to the passive in English.’ Syntax 8(2): 81–​120. Compton, Richard. 2012. ‘The syntax and semantics of modification in Inuktitut: Adjectives and adverbs in a Polysynthetic language.’ PhD dissertation, University of Toronto. Compton, Richard. 2014. ‘An argument for genuine object agreement in Inuit.’ A talk presented at the Annual meeting of the Canadian Linguistics Association, Brock University, St. Catharines, May 24–​26. Compton, Richard & Christine Pittman. 2010. ‘Word-​formation by phase in Inuit.’ Lingua 120: 2167–​2192. Comrie, Bernard 1975. ‘Antiergative,’ in R.E. Grossman, L.J. San, & T.J. Vance (eds.), Papers from the 11th regional meeting of the Chicago Linguistic Society, 112–​121. Comrie, Bernard. 1978. ‘Ergativity,’ in Winfred P. Lehmann (ed.), Syntactic typology: Studies in the phenomenology of language. Austin: University of Texas Press, 329–​394.

 

1160   References Comrie, Bernard. 1979. ‘Degrees of ergativity: Some Chukchee evidence,’ in Frans Plank (ed.), Ergativity: Towards a theory of grammatical relations. London: Academic Press, 219–​240. Comrie, Bernard. 1980. ‘Inverse verb forms in Siberia: Evidence from Chukchee, Koryak, and Kamchadal.’ Folia Linguistica Historica 1: 61–​74. Comrie, Bernard. 1981a. ‘Ergativity and grammatical relations in Kalaw Lagaw Ya (Saibai Dialect).’ Australian Journal of Linguistics 1: 1–​42. Comrie, Bernard. 1981b. Language universals and linguistic typology: Syntax and morphology. Chicago, IL: University of Chicago Press. Comrie, Bernard. 1984. ‘Reflections on verb agreement in Hindi and related languages.’ Linguistics 22: 857–​864. Comrie, Bernard. 1985. ‘Reflections on subject and object control.’ Journal of Semantics 4: 47–​65. Comrie, Bernard. 1988a. ‘Passive and voice,’ in M. Shibatani (ed.), Passive and voice (Typological studies in language), vol. 16. Amsterdam & Philadelphia, PA: Benjamins, 9–​23. Comrie, Bernard. 1988b. ‘Topics, grammaticalized topics, and subjects,’ Proceedings of the fourteenth annual meeting:  General session and parassession on grammaticalization. Berkeley, CA: Berkeley Linguistics Society, 265–​279. Comrie, Bernard. 1989. Language universals and linguistic typology. Chicago, IL: University of Chicago Press. Comrie, Bernard. 2000. ‘Valency-​changing derivations in Tsez,’ in R.M.W. Dixon & Alexandra Aikhenvald (eds.), Changing valency: Case studies in transitivity. Cambridge: Cambridge University Press, 30–​83. Comrie, Bernard. 2004. ‘Oblique–​case subjects in Tsez,’ in Peri Bhaskararao & Kareumuri Venkata Subbārāo. (eds.), Non-​nominative subjects, vol. 1. Amsterdam & Philadelphia, PA: Benjamins, 113–​128. Comrie, Bernard. 2005. ‘Alignment of case marking of full noun phrases,’ in Martin Haspelmath, Matthew Dryer, David Gil, & Bernard Comrie (eds.), The world atlas of language structures. Oxford: Oxford University Press, 398–​403. Comrie, Bernard. 2013a. ‘Alignment of case marking of full noun phrases,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. Comrie, Bernard. 2013b. ‘Alignment of case marking of pronouns,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology. Comrie, Bernard. 2013c. ‘Ergativity: Some recurrent themes,’ in E. Bavin & S. Stolle (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 15–​33 Comrie, Bernard & Helma van den Berg. 2006. ‘Experiencer constructions in Daghestanian languages,’ in Ina Bornkessel, Matthias Schlesewsky, Bernard Comrie, & Angela D. Friederici (eds.), Semantic role universals and argument linking: Theoretical, typological and psycholinguistic perspectives. Berlin: de Gruyter, 127–​154. Comrie, Bernard & Edward L. Keenan. 1979. ‘Noun phrase accessibility revisited.’ Language 649–​664. Comrie, Bernard & Zaira Khalilova. 2013. ‘The antipassive alternation in Bezhta.’ Talk presented at workshop Valency classes in the world’s languages, MPI EVA Leipzig, August 14–​15. Comrie, Bernard, Diana Forker, & Zaira Khalilova. 2011. ‘Alignment typology, reflexives, and reciprocals in Tsezic languages.’ Proceedings of the 37th Annual Meeting of the Berkeley Linguistics Society 37: 32–​51.

 

References   1161 Comrie, Bernard, Diana Forker, & Zaira Khalilova. 2012. ‘Adverbial clauses in the Tsezic languages,’ in Holger Diessel & Volker Gast (eds.), Clause linkage in cross-​linguistic perspective. Berlin: de Gruyter, 157–​190 Comrie, Bernard, Diana Forker, & Zaira Khalilova. in press. ‘Affective constructions in Tsezic languages,’ in Jóhanna Barðdal, Stephen Mark Carey, Thórhallur Eythórsson, & Na’ama Pat-​ El (eds.), Non-​canonically case-​marked subjects within and across languages and language families: The Reykjavík–​Eyjafjallajökull papers. Amsterdam & Philadelphia, PA: Benjamins. Comrie, Bernard, Madzhid Khalilov, & Zaira Khalilova. 2015. ‘Valency and valency classes in Bezhta,’ in Bernard Comrie & Andrej Malchukov (eds.), Valency classes. Berlin: de Gruyter Mouton, 541–570. Condoravdi, Cleo & Ashwini Deo. 2014. ‘Aspect shifts in Indo-​Aryan and trajectories of semantic change,’ in Chiara Gianollo, Agnes Jäger, & Doris Penka (eds.), Language change at the syntax–​semantics interface. Berlin: Mouton de Gruyter, 261–​292. Cook, Claire. 2008. ‘The syntax and semantics of clause-​ typing in Plains Cree.’ PhD dissertation, UBC. Cook, Kenneth W. 1991. ‘The search for subject in Samoan,’ in Robert Blust (ed.), Current in Pacific linguistics: Papers on Austronesian languages and ethnolinguistics in honor of George W. Grace. Canberra: Pacific Linguistics, 77–​98. Cook, Conor & Alana Johns. 2009. ‘Determining the semantics of Inuktitut postbases,’ in M.-​A. Mahieu & N. Tersis (eds.), Variations on polysynthesis. Amsterdam & Philadelphia, PA: Benjamins, pp. 149–​170. Coon, Jessica. 2010a. ‘Rethinking split-​ergativity in Chol.’ International Journal of American Linguistics 76(2): 207–​53. Coon, Jessica. 2010b. ‘Complementation in Chol (Mayan): A theory of split ergativity.’ PhD dissertation, MIT. Coon, Jessica. 2012. ‘Split ergativity and transitivity in Chol.’ Lingua 122: 241–​256. Coon, Jessica. 2013a. Aspects of split ergativity. Oxford: Oxford University Press. Coon, Jessica. 2013b. ‘TAM split ergativity’ (pts. 1–​2). Language and Linguistics Compass 7: 171–​200. Coon, Jessica. 2014. ‘Predication, tenselessness, and what it takes to be a verb,’ in Hsin-​Lun Huang, Ethan Poole, & Amanda Rysling (eds.), NELS 43: Proceedings of the 43rd annual meeting of the North East Linguistics Society 77–​90. Coon, Jessica & Maayan Adar. 2013. ‘Ergativity,’ in Mark Aronoff (ed.), Oxford bibliographies in linguistics. New York: Oxford University Press. Coon, Jessica & Andrés Pablo Salanova. 2009. ‘Nominalization and predicate fronting: Two sources of ergativity,’ in Laurel MacKensie (ed.), Proceedings of the 32nd annual Penn Linguistics Colloquium. Philadelphia, PA:  University of Pennsylvania Working Papers in Linguistics, 45–​54. Coon, Jessica, Pedro Mateo Pedro, & Omer Preminger. 2014. ‘The role of case in A-​bar extraction asymmetries: Evidence from Mayan.’ Linguistic Variation 14: 179–​242. Coon, Jessica & Omer Preminger. 2012. ‘Taking ergativity out of split ergativity:  A  unified account of aspect and person splits.’ MS, McGill University, Syracuse University. Cooreman, Anne. 1988. ‘The antipassive in Chamorro: Variations on the theme of transitivity,’ in Masayoshi Shibatani (ed.), Passive and voice. Amsterdam & Philadelphia, PA: Benjamins, 561–​593. Cooreman, Ann. 1994. ‘A functional typology of antipassives,’ in B. Fox & P.J. Hopper (eds.), Voice: Form and function. Amsterdam & Philadelphia, PA: Benjamins, 49–​88.

 

1162   References Cooreman, Ann, Barbara A. Fox, & Talmy Givó́n. 1984. ‘The discourse definition of ergativity.’ Studies in Language 8: 1–​34. Corbett, Greville. 2005. ‘The canonical approach in typology,’ in Z. Frajzyngier, A. Hodges, & D. Rood (eds.), Linguistic diversity and language theories. Amsterdam & Philadelphia, PA: Benjamins, 25–​49. Corbett, Greville B. 2006. Agreement. Cambridge: Cambridge University Press. Corbett, Greville & Sebastian Fedden. 2014. ‘Multiple categorization: The case of Mian.’ MS, Surrey Morphology Group. Corston, Simon H. 1996. Ergativity in Roviana, Solomon Islands. Canberra: Pacific Linguistics. Corston-​Oliver, Simon H. 2003. ‘Core arguments and the inversion of the nominal hierarchy,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 273–​300. Coulson, Seana, Jonathan W. King, & Marta Kutas. 1998. ‘Expect the unexpected: Event-​related brain response to morphosyntactic violations.’ Language and Cognitive Processes 13(1): 21–​58. Coupe, Alexander R. 2008. A grammar of Mongsen Ao. Berlin: Mouton de Gruyter. Coupe, Alexander R. 2011a. ‘On core case marking patterns in two Tibeto-​Burman languages of Nagaland.’ Linguistics in the Tibeto-​Burman Area 34(2): 21–​47. Coupe, Alexander R. 2011b. ‘Pragmatic foundations of transitivity in Ao.’ Studies in Language 35(3): 492–​522. Cowper, Elizabeth. 1987. ‘What is a Subject? Non-​nominative Subjects in Icelandic,’ in J. Blevins & J. Carter (eds.), Proceedings of NELS 18. Amherst, MA: GLSA, 94–​108. Coyos, Jean-​Baptiste. 1999. Le parler basque souletin des Arbailles. Une approche de l’ergativité. Paris: L’Harmattan. Craig, Colette. 1977. A grammar of Jacaltec. Austin: University of Texas Press. Craig, Colette. 1979. ‘The antipassive and Jacaltec,’ in Laura Martin (ed.), Papers in Mayan linguistics. Colombia, MO: Lucas, 139–​165. Creider, Chet. 1978. The syntax of relative clauses in Inuktitut. Inuit Studies 2, 95–​110. Creissels, Denis. 2000. ‘Typology,’ in: Heine and Nurse (eds.), African languages: An introduction. Cambridge: Cambridge University Press, 231–​258. Creissels, Denis. 2008. ‘Person variations in Akhvakh verb morphology:  Functional motivation and origin of an uncommon pattern.’ Sprachtypologie und Universalienforschung 61: 309–​325. Creissels, Denis. 2010. ‘Specialized converbs and adverbial subordination in Axaxdərə Akhvakh,’ in Isabelle Bril (ed.), Clause linking and clause hierarchy: Syntax and pragmatics. Amsterdam & Philadelphia, PA: Benjamins, 105–​142. Creissels, Denis. 2014. ‘P-​lability and radical P-​alignment.’ Linguistics 52: 911–​944. Crisp, Simon. 1983. ‘Subject marking in some languages of Daghestan.’ Linguistics 16: 203–​216. Cristofaro, Sonia. 2014. ‘Competing motivation models and diachrony:  What evidence for what motivation?’ in Brian MacWhinney, Andrej L. Malchukov, & Edith Moravcsik (eds.), Competing motivations in grammar and usage. Oxford: Oxford University Press. Croft, William. 1990. Typology and universals. Cambridge: Cambridge University Press. Croft, William. 1998. ‘The structure of events and the structure of language,’ in Michael Tomasello (ed.), The new psychology of language: Cognitive and functional approaches to language structure, vol. 1. Mahwah, NJ & London: Erlbaum. Croft, William. 2001. Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press.

 

References   1163 Cumming, Susanna & Fay Wouk. 1987. ‘Is there “discourse ergativity,” in Austronesian languages?’ Lingua 71: 271–​296. Cummings, Thomas & Thomas G. Bailey. 1912. Panjabi Manual and grammar: A guide to the colloquial Panjabi of the Northern Panjab. Calcutta: Baptist Mission Press. Custis, Tonya. 2004. ‘Word order variation in Tongan: A syntactic analysis.’ PhD dissertation, University of Minnesota. Dahl, Östen. 2000. ‘Egophoricity in discourse and syntax.’ Functions of Language 7(1): 33–​77. Dahlstrom, Amy. 1983. ‘Agent–​patient languages and split case marking systems,’ in Amy Dahlstrom et al. (eds.), Proceeding of the ninth annual meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, 37–​46. D’Alessandro, Roberta & Ian Roberts. 2010. ‘Past participle agreement in Abruzzese:  Split auxiliary selection and the null-​subject parameter.’ Natural Language & Linguistic Theory 28: 41–​72. Dalrymple, Mary & Irina Nikolaeva. 2011. Objects and information structure. Cambridge: Cambridge University Press. Daniel, Michael. 2013. ‘Issues in Khinalugh syntax:  Building on corpus evidence,’ Working papers humanities, Higher School of Economics, Moscow. Daniel, Michael & Dmitry Ganenkov. 2009. ‘Case marking in Daghestanian: Limits of elaboration,’ in Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 68–​685. Daniel, Michael & Yuri Lander. 2010. ‘A girl of word, meat of a ram, and a life of longing: On peculiar cases of relativization in East Caucasian languages.’ Paper presented at Syntax of the World’s Languages IV, Lyon. Daniel, Michael & Yury Lander. 2011. ‘The Caucasian languages,’ in Bernd Kortmann & Johan van der Auwera (eds.), The languages and linguistics of Europe:  A  comprehensive guide. Berlin: de Gruyter, 125–​158. Daniel, Michael, Timur Maisak, & Solmaz Merdanova. 2012. ‘Causatives in Agul,’ in Pirkko Suihkonen, Bernard Comrie, & Valery Solovyev (eds.), Argument structure and grammatical relations: A crosslinguistic typology. Amsterdam & Philadelphia, PA: Benjamins, 55–​113. Davidiak, Elena & John Grinstead. 2004. ‘Root non-​finite forms in child Spanish.’ Poster presented at Generative Approaches to Language Acquisition. Davies, William D. 1984. ‘Antipassive: Choctaw evidence for a universal characterization,’ in David Perlmutter & Carol Rosen (eds.), Studies in relational grammar, vol. 2. Chicago, IL: University of Chicago Press, 331–​376. Davies, William D. 1986. Choctaw verb agreement and universal grammar. Dordrecht: Reidel. Davies, William D. & Stanley Dubinsky. 2003. ‘On extraction from NPs.’ Natural Language and Linguistic Theory 21: 1–​37. Davis, Henry & Jason Brown. 2011. ‘On A’-​dependencies in Gitksan,’ in John Lyon & Joel Dunham (eds.), Proceedings of the 46th international conference on Salish and neighbouring languages. Vancouver: UBCWPL, 43–​80. Davis, Henry & Clarissa Forbes. 2015. ‘Connect four! The morphosyntax of argument marking in Tsimshianic,’ in Zoe Lam Natalie Weber, Erin Guntly, & Sihwei Chen (eds.), Proceedings of the 50th international conference on Salish and neighbouring languages, vol. 40. Vancouver: University of British Columbia Working Papers in Linguistics, 154–​185. Davis, Henry & Hamida Demirdache. 2000. ‘On lexical verb meanings: evidence from Salish,’ in Carol Tenny & James Pustejovsky (eds.), Events as grammatical objects: The converging perspectives of lexical semantics and syntax, 100 (CSLI Lecture Notes). Stanford, CA: CSLI, 97–​142.

 

1164   References Davison, Alice. 1985. ‘Case and control in Hindi–​Urdu.’ Studies in the Linguistic Sciences 15(2): 9–​23. Davison, Alice. 1988. ‘Constituent structure and the realization of agreement features,’ in Lynn Macleod, Gary Larson, & Diane Brentari (eds.), Papers from the 24th regional meeting of the Chicagon Linguistics Society. Chicago, IL: Chicago Linguistic Society, 41–​53. Davison, Alice. 1999. ‘Ergativity,’ in Edith Moravcsik & Michael Noonan (eds.), Proceedings of the symposium on formalism and functionalism. Amsterdam & Philadelphia, PA: Benjamins, 177–​210. Davison, Alice. 2004a. ‘Non-​nominative subjects in Hindi–​Urdu:  VP structure and case parameters,’ in Peri Bhaskararao & K.V. Subbārāo (eds.), Non-​nominative subjects, vol. 1. Amsterdam & Philadelphia, PA: Benjamins, 71–​98. Davison, Alice. 2004b. ‘Structural case, lexical case and the verbal projection,’ in Veneeta Dayal & Anoop Mahajan (eds.), Clause structure in South Asian languages. Dordrecht:  Kluwer, 199–​225. Davison, Alice. 2010. ‘Dependent structural case and the role of functional projections,’ in R.M.Q. Rapaksha et al. (eds.), Dakkhina: Essays in honor of James Gair. Colombo: Godage, 183–​206 Dayal, Veneeta. 2003. ‘Bare nominals:  Non-​specific and contrastive readings under scrambling,’ in Simin Karimi (ed.), Word order and scrambling. Oxford: Blackwell, 67–​90. Dayal, Veneeta. 2011. ‘Hindi Pseudo-​incorporation.’ Natural Language and Linguistic Theory 29: 123–​167. Dayley, Jon P. 1981. ‘Voice and ergativity in Mayan languages.’ Journal of Mayan Linguistics 2: 3–​82. Dayley, Jon P. 1985. Tz’utujil grammar. Berkeley, CA: University of California Press. De Guzman, Videa P. 1976. ‘Syntactic derivation of Tagalog verbs.’ PhD dissertation, University of Hawai‘i. De Guzman, Videa P. 1988. ‘Ergative analysis for Philippine languages: An analysis,’ in Richard McGinn (ed.), Studies in Austronesian linguistics. Athens, Ohio: Ohio University Center for International Studies, 323–​345. de Hoop, Helen. 1996. Case configuration and noun phrase interpretation. New York: Garland. de Hoop, Helen. 1999. ‘Optimal case assignment.’ Linguistics in the Netherlands 16(1): 97–​109. de Hoop, Helen. 2009. ‘Case in optimality theory,’ The Oxford handbook of case. Oxford: Oxford University Press, 88–​101. de Hoop, Helen & Peter de Swart (eds.). 2008. Differential subject marking. Dordrecht: Springer. de Hoop, Helen & Peter de Swart. 2009. ‘Cross-​linguistic variation in differential subject marking,’ in H. de Hoop & P. de Swart (eds.), Differential subject marking. Dordrecht: Springer, 1–​16. de Hoop, Helen & Monique Lamers. 2006. ‘Incremental distinguishability of subject and object,’ in L. Kulikov, A. Malchukov, & P. de Swart (eds.), Case, valency, and transitivity. Amsterdam & Philadelphia, PA: Benjamins, 269–​287. de Hoop, Helen & Andrej Malchukov. 2007. ‘On fluid differential case marking: A bidirectional OT approach.’ Lingua 117: 1636–​1656. de Hoop, Helen & Andrej Malchukov. 2008. ‘Case-​marking strategies.’ Linguistic Inquiry 39: 565–​587. de Hoop, Helen & Bhuvana Narasimhan. 2005. ‘Differential case-​marking in Hindi,’ in M. Amberber & H. de Hoop (eds.), Competition and variation in natural languages: The case for case. Oxford: Elsevier, 321–​345.

 

References   1165 de León, Lourdes. 1999. ‘Verbs in Tzotzil (Mayan) early syntactic development.’ International Journal of Bilingualism 3(2–​3): 219–​239. de Oliveira, Christiane Cunha. 2005. ‘The language of the Apinajé people of Central Brazil.’ PhD dissertation, University of Oregon. de Rijk, Rudolf. 1981. ‘Euskal morfologiaren zenbait gorabehera,’ Euskal Linguistika eta Literatura: Bide Berriak. Bilbao: Universidad de Deusto/​Deustuko Unibertsitatea, 83–​101. de Rijk, Rudolf. 2007. Standard Basque: A progressive grammar. Cambridge, MA: MIT Press. de Smet, Hendrick. 2015. ‘How gradual change progresses.’ Paper presented at GLIMS [Ghent Research Team on Linguistic Meaning and Structure] Lecture Series. University of Ghent. de Swart, Peter. 2007. ‘Cross-​linguistic variation in object marking.’ PhD thesis, Radboud University. Deal, Amy Rose. 2010a. ‘Ergative case and the transitive subject:  A  view from Nez Perce.’ Natural Language & Linguistic Theory 28: 73–​120. Deal, Amy Rose 2010b. ‘Topics in the Nez Perce verb.’ PhD dissertation, University of Massachusetts, Amherst. Deal, Amy Rose. 2013. ‘Possessor raising.’ Linguistic Inquiry 44: 391–​432. Deal, Amy Rose. 2014. ‘Person-​based split ergativity in Nez Perce is syntactic.’ MS, University of California, Santa Cruz. Deal, Amy Rose. 2015. ‘Ergativity,’ in Tibor Kiss & Artemis Alexiadou (eds.), Syntax—​Theory and analysis: An international handbook, vol. 1. Berlin: de Gruyter Mouton, 654–​707. Deal, Amy Rose. 2016. ‘Person-based split ergativity in Nez Perce is syntactic,’ Journal of Linguistics 52:3. Déchaine, Rose-​Marie. 2007. ‘The evidential base.’ Paper presented at NELS 38. Déchaine, Rose-​Marie, Claire Cook, Jeff Muehlbauer, & Ryan Waldie. 2014. ‘(De-​)constructing evidentiality.’ MS, UBC. Déchaine, Rose-​Marie, & Martina Wiltschko. 2002. ‘Decomposing pronouns.’ Linguistic Inquiry 33: 409–​422. Dediu, Dan, Michael Cysouw, Stephen C. Levinson, Andrea Baronchelli, Morten H. Christiansen, William Croft, Nicholas Evans, Simon Garrod, Russell D. Gray, Anne Kandler, & Elena Lieven. 2013. ‘Cultural evolution of language,’ in Peter J. Richerson & Morten H. Christiansen (eds.), Cultural evolution:  Society, technology, language, and religion. Cambridge, MA: MIT Press. Deeters, Gerhard. 1930. Das kharthwelische Verbum: vergleichende Darstellung des Verbalbaus der südkaukasischen Sprachen. Leipzig: Markert und Petters DeLancey, Scott. 1981. ‘An interpretation of split ergativity and related patterns’. Language 57: 626–​657. DeLancey, Scott. 1984a. ‘Transitivity and ergative case in Lhasa Tibetan.’ Proceedings of the tenth meeting of the Berkley Linguistics Society. Berkley: Berkley Linguistics Society, 131–​140. DeLancey, Scott. 1984b. ‘Etymological notes on Tibeto-​Burman case particles.’ Linguistics of the Tibeto-​Burman Area 8(1): 59–​77. DeLancey, Scott. 1990. ‘Ergativity and the cognitive model of event structure in Lhasa Tibetan.’ Cognitive Linguistics 1(3): 289–​321. DeLancey, Scott. 1991. ‘Event construal and case role assignment.’ Berkeley Linguistic Society 17: 338–​353. DeLancey, Scott. 2006. ‘The blue bird of ergativity.’ Unpublished MS. DeLancey, Scott. 2011. ‘ “Optional” “ergativity” in Tibeto-​Burman Languages.’ Linguistics of the Tibeto-​Burman Area 34(2): 9–​20.

 

1166   References Delforooz, Behrooz Barjasteh. 2010. Discourse features in Balochi of Sistan (oral narratives). Uppsala: Studia Iranica Upsaliensa. Demiral, Seyfettin B., Matthias Schlesewsky, & Ina Bornkessel-​Schlesewsky. 2008. ‘On the universality of language comprehension strategies:  Evidence from Turkish.’ Cognition 106(1): 484–​500. Demirdache, Hamida & Miryam Uribe-​Etxebarria. 2000. ‘The primitives of temporal relations,’ in Roger Martin, David Michaels, & Juan Uriagereka (eds.), Step by step: Essays on minimalist syntax in honor of Howard Lasnik. Cambridge, MA: MIT Press, 157–​186. den Besten, Hans. 1981. Some remarks on the ergative hypothesis [mimeographed]. University of Amsterdam. Dench, Alan. 1982. ‘The development of an accusative case marking pattern in the Ngayarda languages of Western Australia.’ Australian Journal of Linguistics 2(1): 43–​59. Dench, Alan. 1991. ‘Panyjima,’ in R.M.W. Dixon & Barry J. Blake (eds.), Handbook of Australian languages 4. Oxford: Oxford University Press, 125–​243. Dench, Alan. 2006. ‘Case marking strategies in subordinate clauses in Pilbara languages—​ Some diachronic speculations.’ Australian Journal of Linguistics 26: 81–​105. Dench, Alan & Nicholas Evans. 1988. ‘Multiple case-​ marking in Australian languages.’ Australian Journal of Linguistics 8: 1–​47. Denwood, Philip. 1999. Tibetan. Amsterdam & Philadelphia, PA: Benjamins. Deo, Ashwini. 2012. ‘The imperfective–​perfective contrast in Middle Indic.’ Journal of South Asian Linguistics 5: 3–​33 Deo, Ashwini & Sharma, Devyani. 2006. ‘Typological variation in the ergative morphology of Indo-​Aryan languages.’ Linguistic Typology 10(3): 369–​418. Depraetere, Ilse. 1995. ‘On the necessity of distinguishing between (un)boundedness and (a)telicity.’ Linguistics and Philosophy 18: 1–​19. Devi, Jayantimala. 1986. ‘Ergativity:  A  historical analysis in Assamese.’ PhD dissertation, University of Delhi. DeWolf, Charles M. 1988. ‘Voice in Austronesian languages of the Philippine type: Passive, ergative, or neither?’ in Masayoshi Shibatani (ed.), Passive and voice (vol. 16 of Typological Studies in Language). Amsterdam & Philadelphia, PA: Benjamins, 143–​193. Díaz, Begoña, Núria Sebastián-​Gallés, Kepa Erdocia, Jutta L. Mueller, & Itziar Laka. 2011. ‘On the cross-​linguistic validity of electrophysiological correlates of morphosyntactic processing: A study of case and agreement violation in Basque.’ Journal of Neurolinguistics 24(3): 357–​373. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Diesing, Molly. 1996. ‘Semantic variables and object shift,’ in Höskuldur Thráinsson, Samuel D. Epstein, & Steve Peter (eds.), Studies in comparative Germanic syntax, vol. 2. Dordrecht: Kluwer, 66–​84. Diesing, Molly. 1997. ‘Yiddish VP order and typology of object movement in Germanic.’ Natural Language & Linguistic Theory 15: 369–​427. Dik, Simon C. 1997. The theory of functional grammar. Part 1: The structure of the clause. Berlin & New York: Mouton de Gruyter. Dillon, Brian, Andrew Nevins, Alison C. Austin, & Colin Phillips. 2012. ‘Syntactic and semantic predictors of tense in Hindi: An ERP investigation.’ Language and Cognitive Processes 27(3): 313–​344. Dimmendaal, Gerrit J. 2009. ‘Tima,’ in Gerrit J. Dimmendaal (ed.), Coding participant marking:  Construction types in twelve African languages. Amsterdam & Philadelphia, PA: Benjamins, 331–​353.

 

References   1167 Dimmendaal, Gerrit J. 2010a. ‘Differential object marking in Nilo-​Saharan.’ Journal of African Languages and Linguistics 31: 13–​46. Dimmendaal, Gerrit J. 2010b. ‘On the origin of ergativity in Tima,’ in Frank Floricic (ed.), Essais de typologie et de linguistique générale, Mélanges offerts à Denis Creissels. Paris: Presses Universitaires de l’École Normale Superieure, 233–​239. Dimmendaal, Gerrit J. 2014. ‘Marked nominative systems in eastern Sudanic and their historical origin.’ Afrikanistik Online 2014. Dirr, Adolf. 1912. ‘Rutulskij jazyk. Sbornik materialov dlja opisanija plemen Kavkaza,’ Tbilisi, 42(3): 1–​204. Dirr, Adolf. 1928. Einführung in das Studium der kaukasischen Sprachen. Leipzig: Verlag der Asia Major. Dixon, Robert M.W. 1972. The Dyirbal language of north Queensland (Cambridge Studies in Linguistics 9). Cambridge: Cambridge University Press. Dixon, Robert M.W. (ed.) 1976. Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies. Dixon, Robert M.W. 1977. A grammar of Yidiny. Cambridge: Cambridge University Press. Dixon, Robert M.W. 1979. ‘Ergativity.’ Language 55: 59–​138. Dixon, Robert M.W. 1980. The languages of Australia (Cambridge Language Surveys). Cambridge: Cambridge University Press. Dixon, Robert M.W. (ed.), 1987. Studies in ergativity. Amsterdam: North-​Holland. Dixon, Robert M.W. 1994. Ergativity. Cambridge: Cambridge University Press. Dixon, Robert M.W. 1997. The rise and fall of language. Cambridge: Cambridge University Press. Dixon, Robert M.W. 2002. Australian languages. Cambridge: Cambridge University Press. Dixon, Robert M.W. 2010. A grammar of Yidiɲ. (Digital re-edition of Dixon (1977). Cambridge: Cambridge University Press. Dobrushina, Nina. 2001. ‘Formy imperativnoj serii,’ in Aleksandr E. Kibrik (ed.), 319–​332. Dočekal, Mojmír & Dalina Kallulli. 2012. ‘More on the semantics of clitic doubling: Principal filters, minimal witnesses, and other bits of truth,’ in Christopher Piñón (ed.), Empirical issues in syntax and semantics 9. CSSP CNRS, 113–​128. Doliana, Aaron. 2013. ‘The super-​strong person–​case constraint:  Scarcity of resources by scale-​driven impoverishment,’ Rule interaction in grammar, vol. 90 of Linguistische Arbeits Berichte. Leipzig: Institut für Linguistik, Universität Leipzig, 177–​202. Donaldson, Tamsin. 1980. Ngiyambaa:  The language of the Wangaaybuwan. Cambridge: Cambridge University Press. Donohue, Cathryn. 1999. Optimizing fore case and word order. Stanford, CA: Stanford University. Donohue, Cathryn & Mark Donohue. 2010. ‘The case of possessors and “subjects,” ’ in Raphael Mercado, Eric Potsdam, & Lisa Travis (eds.), Austronesian and theoretical linguistics. Amsterdam & Philadelphia, PA: Benjamins, 103–​116. Donohue, Mark. 1996. ‘Bajau: A symmetrical Austronesian language.’ Language 72: 782–​793. Donohue, Mark. 1999. A grammar of Tukang Besi. Berlin: Mouton de Gruyter. Donohue, Mark. 2008. ‘Semantic alignment:  What’s what and what’s not,’ in Mark Donohue & Søren Wichmann (eds.), The typology of semantic alignment. Oxford: Oxford University Press. Donohue, Mark & Søren Wichmann (eds.). 2008. Typology of languages with semantic alignment. Oxford: Oxford University Press. Dorais, Louis-Jacques. 1988. Tukilik:  An Inuktitut grammar for all. Quebec:  Association Inuksiutiit Katimajiit.

 

1168   References Dorais, Louis-Jacques. 2003. Inuit Uqausiqatigiit—​Inuit languages and dialects (2nd edn.). Iqaluit, Nunavut: Nunavut: Nunavut Arctic College Dorleijn, Margreet. 1996. The decay of ergativity in Kurdish: Language internal or contact induced? Tilburg: Tilburg University Press. Doron, Edit & Geoffrey Khan. 2012. ‘The typology of morphological ergativity in Neo-​ Aramaic.’ Lingua 122: 225–​240 Dourado, Luciana. 2001. ‘Aspectos morfossintáticos da língua Panará (Jê).’ PhD dissertation, Universidade Estadual de Campinas, Campinas. Dowty, David. 1991. ‘Thematic proto-​roles and argument selection.’ Language 67: 547–​619. Dowty, David. 1979. Word meaning and Montague grammar: The semantics of verbsand times in generative semantics and in Montague’s PTQ. Dordrecht: Reidel. Dressler, Wolfgang. 1968. Studien zur verbalen Pluralität. Sitzungsberichte, Österreichische Akademie der Wissenschaften (Phil.-​hist. Klasse) Band 259, Abh 1. Dressler, Wolfgang. 2005. ‘Morphological typology and first language acquisition:  Some mutual challenges,’ in Geert Booij, Emiliano Guevara, Angela Ralli, Salvatore Sgroi, & Sergio Scalise (eds.), Morphology and linguistic typology. On-line proceedings of the fourth Mediterranean Morphology Meeting (MMM4). Dressler, Wolfgang & Annemarie Karpf. 1995. ‘The theoretical relevance of pre-​and protomorphology in language acquisition.’ Yearbook of Morphology 1994: 99–​122. Drossard, Werner. 1991. ‘Kasusmarkierung und die Zentralität von Partizipanten,’ in H. Seiler & W. Premper (eds.), Partizipation.  Das sprachliche Erfassen von Sachverhalten. Tübingen: Narr, 447–​480. Dryer, Matthew S. 1986. ‘Primary objects, secondary objects, and antidative.’ Language 62(4): 808–​845. Dryer, Matthew S. 1997. ‘On the six-​way word order typology.’ Studies in Language 21(1): 69–​103. Dryer, Matthew S. 2013a. ‘Determining dominant word order,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. Dryer, Matthew S. 2013b. ‘Order of subject and verb,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. Dryer, Matthew S. & Martin Haspelmath. (eds.) 2013. Wals online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. http://​wals.info Du Bois, John W. 1980. ‘The search for a cultural niche: Showing the Pear film in a Mayan community,’ in Wallace L. Chafe (ed.), The Pear stories: Cognitive, cultural, and linguistic aspects of narrative production. Norwood, NJ: Ablex, 1–​7. Du Bois, John W. 1981. ‘The Sacapultec language.’ Ph.D. dissertation, University of California, Berkeley. Du Bois, John W. 1985. ‘Competing motivations,’ in John Haiman (ed.), Iconicity in syntax. Amsterdam & Philadelphia, PA: Benjamins, 343–​365. Du Bois, John W. 1987a. ‘Absolutive zero:  Paradigm adaptivity in Sacapultec Maya.’ Lingua 71: 203–​222. Du Bois, John W. 1987b. ‘The discourse basis of ergativity.’ Language 63: 805–​855. Du Bois, John W. 2003a. ‘Argument structure: Grammar in use,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 11–​60. Du Bois, John. W. 2003b. ‘Discourse and grammar,’ in Michael Tomasello (ed.), The new psychology of language:  Cognitive and functional approaches to language structure, vol. 2. Mahwah, NJ, & London: Erlbaum, 47–​87.

 

References   1169 Du Bois, John W. 2006. ‘The Pear story in Sakapultek Maya: A case study of information flow and preferred argument structure,’ in Mercedes Sedano, Adriana Bolívar, & Martha Shiro (eds.), Haciendo Lingüística: Homenaje a Paola Bentivoglio. Caracas: Universidad Central de Venezuela, 189–​220. Du Bois, John W. 2014a. ‘Motivating competitions,’ in Brian MacWhinney, Andrej L. Malchukov, & Edith Moravcsik (eds.), Competing motivations in grammar and usage. Oxford: Oxford University Press, 263–​281. Du Bois, John W. 2014b. Representing discourse. Santa Barbara, CA: Linguistics Department, University of California, Santa Barbara. Du Bois, John W., Lorraine E. Kumpf, & William J. Ashby (eds.). 2003. Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins. Du Bois, John W. & Nicholas Lester [in progress]. ‘Cognitive containment and construction design: Grammaticizing the three-​place predicate.’ Du Bois, John W., Stephan Schuetze-​Coburn, Susanna Cumming, & Danae Paolino 1993. ‘Outline of discourse transcription,’ in Jane A. Edwards & Martin D. Lampert (eds.), Talking data: Transcription and coding in discourse research. Hillsdale, NJ: Erlbaum, 45–​89. Dukes, Michael. 1996. ‘On the nonexistence of anaphors and pronominals in Tongan.’ PhD dissertation, UCLA. Duñabeitia, Jon Andoni, Manuel Perea, & Manuel Carreiras. 2007. ‘Do transposed-​letter similarity effects occur at a morpheme level? Evidence for morpho-​orthographic decomposition.’ Cognition 105: 691–​703. Dunn, John A. 1979a. A reference grammar for the Coast Tsimshian language. Canadian Ethnology Service, Mercury Series Paper No. 55. Ottawa: National Museum of Man. Dunn, John A. 1979b. ‘Tsimshian connectives.’ International Journal of American Linguistics 45: 131–​140. Dunn, John A. 1990. ‘The government of oblique/​local grammatical relations in lower Tsimshian.’ Handout from a paper presented at Conference on American Indigenous Languages, New Orleans. Dunn, Michael John. 1999. ‘A grammar of Chukchi.’ PhD dissertation, Australian National University. Duranti, Alessandro. 2004. ‘Agency in language,’ in Alessandro Duranti (ed.), A companion to linguistic anthropology. Oxford: Blackwell, 451–​473. Duranti, Alessandro & Elinor Ochs. 1989. ‘Acquisition of genitive agents in Samoan.’ Papers & Reports on Child Language Development, Stanford University 28. Duranti, Alessandro & Elinor Ochs. 1990. ‘Genitive constructions and agency in Samoan discourse.’ Studies in Language 14(1): 1–​23. Durie, Mark. 1985. A grammar of Acehnese on the basis of a dialect of North Aceh. Dordrecht: Foris. Durie, Mark. 1988. ‘Preferred argument structure in an active language: Arguments against the category “intransitive subject.” ’ Lingua 74: 1–​25. Durie, Mark. 1994. ‘A case study of pragmatic linking.’ Text 14(4): 495–​529. Durie, Mark. 2003. ‘New light on information pressure: Information conduits, “escape valves,” and role alignment stretching,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 159–​196. Dutra, Rosalia. 1987. ‘The hybrid S category in Brazilian Portuguese: Some implications for word order.’ Studies in Language 11: 163–​180. Edmonson, Barbara W. 1988. ‘A descriptive grammar of Huasteco (Potosino dialect).’ PhD dissertation, Tulane University.

 

1170   References Ehrich, Veronika. 2002. ‘On the verbal nature of certain nominal entities,’ in Ingrid Kaufmann & Barbara Stiebels (eds.), More than words: A Festschrift for Dieter Wunderlich. Berlin: Akademie-​Verlag, 69–​89. Embick, David. 1998. ‘Voice systems and the syntax/​morphology interface,’ in Heidi Harley (ed.), MITWPL 32:  Papers from the UPenn/​MIT roundtable on argument structure and aspect. Cambridge, 41–​72. Embick, David. 2010. Localism and globalism in morphology and phonology. Cambridge MA: MIT Press. Embick, David & Rolf Noyer. 2001. ‘Movement operations after syntax.’ Linguistic Inquiry 32: 555–​595. Enç, Mürvet. 1991. ‘The semantics of specificity.’ Linguistic Inquiry 22(1): 1–​25. Endriss, Cornelia. 2009. Quantificational topics: A scopal treatment of exceptional wide scope phenomena a scopal treatment of exceptional wide scope phenomena. New York: Springer. England, Nora C. 1983. ‘Ergativity in Mamean (Mayan) languages.’ International Journal of American Linguistics 49: 1–​19. England, Nora C. 1991. ‘Changes in basic word order in Mayan languages.’ International Journal of American Linguistics 57(4): 446–​486. England, Nora C. & Laura Martin 2003. ‘Issues in the comparative argument structure analysis in Mayan narratives,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture. Amsterdam & Philadelphia, PA: Benjamins, 131–​157. Erckert, Roderich von. 1895. Die Sprachen des kaukasischen Stammes. Vienna: Niemeyer. Erdocia, Kepa, Itziar Laka, Anna Mestres-​ Missé, & Antoni Rodriguez-​ Fornells. 2009. ‘Syntactic complexity and ambiguity resolution in a free word order language: Behavioral and electrophysiological evidences from Basque.’ Brain and Language 109(1): 1–​17. Erdocia, Kepa, Itziar Laka, & Antoni Rodriguez-​Fornells. 2012. ‘Processing verb medial word orders in a verb final language,’ Case, word order and prominence. Dordrecht:  Springer, 217–​237. Erlewine, Michael Yoshitaka. 2016. ‘Anti-​locality and optimality in Kaqchikel agent focus.’ Natural Language & Linguistic Theory, 34(2): 429–479. Erlewine, Michael Yoshitaka. [to appear]. ‘Subject marking on non-​subjects in Squliq Atayal,’ Proceedings of the 20th meeting of the Austronesian Formal Linguistics Association (AFLA 20). Arlington, TX: University of Texas. Erlewine, Michael Yoshitaka, Theodore Levin, & Coppe van Urk. 2015. ‘What makes a voice system? On the relationship between voice marking and case,’ In Amber Camp, Yuko Otsuka, Claire Stabile and Nozomi Tanaka (eds.), AFLA 21: The Proceedings of the 21st Meeting of the Austronesian Formal Linguistics Association. Arlington, TX: University of Texas, 51–68. Estevam, Adriana Machado. 2011. ‘Morphosyntaxe du xavante. Langue Jê du Mato Grosso, Brésil.’ PhD dissertation, Université Paris 7. Estival, Dominique & John Myhill. 1988. ‘Formal and functional aspects of the development from passive to ergative systems,’ in Masayoshi Shibatani (ed.), Passive and voice. (Typological Studies in Language, 16.) Amsterdam & Philadelphia, PA: Benjamins, 441–​491. Etxepare, Ricardo. 2003. ‘Valency and argument structure in the Basque verb,’ in J.I. Hualde & J. Ortiz de Urbina (eds.), A grammar of Basque. Berlin: Mouton de Gruyter, 363–​425. Etxepare, Ricardo. 2013. ‘Basque spatial cases and the ergative–​absolutive syncretism,’ in R. Gómez, J. Gorrochategui, J.A. Lakarra, & C. Mounole (eds.), Proceedings of the 3rd Conference of the Luis Michelena Chair. Vitoria-​Gasteiz: UPV-​EHU.

 

References   1171 Etxepare, Ricardo. 2014. ‘Contact and change in a minimalist theory of variation,’ in C. Picallo (ed.), Linguistic variation in the minimalist framework. Oxford:  Oxford University Press, 108–​139. Etxepare, Ricardo & Bernard Oyharçabal. 2013. ‘Datives and adpositions in north-​eastern Basque,’ in B. Fernández & R. Etxepare (eds.), Variation in datives: A micro-​comparative perspective. Oxford: Oxford University Press. Etxepare, Ricardo & Miriam Uribe-​ Etxebarria. 2010. ‘Hitz hurrenkera eta birregituraketa euskaraz,’ in Ricardo Etxepare, Ricardo Gómez, & Joseba A. Lakarra (eds.), Beñat Oihartzabali gorazarre—​Festschrift for Bernard Oyharçabal. Anuario del Seminario de Filología Vasca Julio de Urquijo43(1–​2): 335–​356. Etxepare, Ricardo & Miriam Uribe-​Etxebarria. 2012. ‘Denominal necessity modals in Basque,’ in Urtzi Etxeberria, Ricardo Etxepare, & Myriam Uribe-​Etxebarria (eds.), DPs and nominalizations in Basque. Amsterdam & Philadelphia, PA: Benjamins, 283–​332. Evans, Nicholas. 1995. A grammar of Kayardild: With comparative notes on Tangkic. Berlin: Mouton. Evans, Nicholas (ed.). 2003. The non-​ Pama–​ Nyungan Languages of Northern Australia: Comparative studies of the continent’s most linguistically complex region (Pacific Linguistics 552/​Studies in Language Change). Canberra: Australian National University. Evans, Nicholas. 2007. ‘Insubordination and its uses,’ in Irina Nikolaeva (ed.), Finiteness. Oxford: Oxford University Press, 366–​431. Evans, Nicholas. 2013. ‘Language diversity as a resource for understanding cultural evolution,’ in Peter J. Richerson & Morten H. Christiansen (eds.), Cultural evolution: Society, technology, language, and religion. Cambridge, MA: MIT Press. Evans, Nicholas & Stephen C. Levinson. 2009. ‘The myth of language universals: Language diversity and its importance for cognitive science.’ Behavioral and Brain Sciences 32(5): 429–​ 448. [discussion: 448–​494] Everett, Caleb. 2009. ‘A reconsideration of the motivations for preferred argument structure.’ Studies in Language 33: 1–​24. Ezeizabarrena, Marijo. 1996. Adquisición de la morfología verbal en euskera y castellano por niños bilingües. Bilbao: Servicio Editorial de la Universidad del País Vasco. Ezeizabarrena, Maria José. 2012. ‘The (in)consistent ergative marking in early Basque: L1 vs. child L2.’ Lingua 127: 303–​317. Ezeizabarrena, Marijo & Pilar Larrañaga. 1996. ‘Ergativity in Basque: A problem for language acquisition?’ Linguistics 34: 955–​991. Faarlund, Jan Terje. 2012. A grammar of Chiapas Zoque. New York: Oxford University Press. Faarlund, Jan Terje. 2013. ‘Subject in Scandinavian,’ in Ilja A. Seržant & Leonid Kulikov (eds.), The diachronic typology of non-​prototypical subjects. Amsterdam & Philadelphia, PA: Benjamins, 187–​202. Falk, Yehuda N. 2006. Subjects and universal grammar: An explanatory theory. Cambridge & New York: Cambridge University Press Faller, Martina T. 2002. ‘Semantics and pragmatics of evidentials in Cuzco Quechua’. PhD dissertation, Stanford University. Farrell, Patrick. 1992. ‘Semantic relations vs. abstract syntactic relations: Evidence from Halkomelem,’ Proceedings of the eighteenth annual meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society, 76–​87. Farrell, Tim. 1995. ‘Fading ergativity? A study of ergativity in Balochi,’ in David C. Bennett (ed.), Subject, voice and ergativity: Selected essays. London: School of Oriental and African Studies, 218–​243.

 

1172   References Farrell, Tim. 2003. ‘Linguistic influences on the Baloch spoken in Karachi,’ in Carina Jahani & Agnes Korn (eds.) 2003. The Baloch and their neighbours. Ethnic and linguistic contact in Balochistan in historical and modern times. Wiesbaden: Reichert, 169–​210. Fassberg, Steven E. 2014. ‘The origin of the periphrastic preterite kəm/​qam-​qāṭəlle in North-​ Eastern Neo-​Aramaic,’ in G. Khan & L. Napiorkowska (eds.), Neo-​Aramaic and its linguistic context. Piscataway, NJ: Gorgias Press, 172–​186. Fauconnier, Stefanie & Jean-​Christophe Verstraete. 2010. ‘Distinguishing animacy effects for agents: A case study of Australian languages.’ Australian Journal of Linguistics 30: 183–​207. Fauconnier, Stefanie & Jean-​Christophe Verstraete. 2014. ‘A and O as each other’s mirror image? Problems with markedness reversal.’ Linguistic Typology 18(1): 3–​49. Feldman, Harry. 1986. A grammar of Awtuw. Canberra: Australian National University. Fernández, Beatriz. 1997. Egiturazko kasuaren erkaketa euskaraz. Bilbao: UPV-​EHU. Fernández, Beatriz. 2013. ‘Some reflexions on Basque auxiliary selection.’ Wedisyn’s Fourth Workshop on Syntactic Variation, Santiago de Compostela, Galiza, April 18–​19. Fernández, Beatriz & Pablo Albizu. 2000. ‘Ergative displacement in Basque and the division of labor between morphology and syntax,’ in J. Boyle, J. Lee, & A. Okrent (eds.), Chicago Linguistic Society 36, volume 2: The panels. Chicago, IL: Chicago Linguistic Society. Fernández, Beatriz & Jon Ortiz de Urbina. 2010. Datiboa hiztegian. Bilbao: UPV-​EHU. Fernández, Beatriz & Milan Rezac. 2016. ‘Differential object marking in Basque varieties,’ in B. Fernández & J. Ortiz de Urbina (eds.), Microparameters in the grammar of Basque. Amsterdam & Philadelphia, PA: Benjamins, 93–138. Ferreira, Marília. 2003. ‘Estudo morfossintático da língua Parkatêjê.’ PhD dissertation, Universidade Estadual de Campinas. Fiebach, Christian, Matthias Schlesewsky, & Angela D. Friederici. 2001. ‘Syntactic working memory and the establishment of filler–​gap dependencies: Insights from ERPs and fMRI.’ Journal of Psycholinguistic Research 30(3): 321–​38. Fiebach, Christian, Matthias Schlesewsky, & Angela D. Friederici. 2002. ‘Separating syntactic memory costs and syntactic integration costs during parsing: The processing of German WH-​questions.’ Journal of Memory and Language 47(2), 250–​272. Filimonova, Elena. 2005. ‘The noun phrase hierarchy and relational marking: Problems and counterevidence.’ Linguistic Typology 9: 22–​113. Filip, Hana. 2008. ‘Events and maximalization,’ in Susan Rothstein (ed.), Theoretical and crosslinguistic approaches to the semantics of aspect. Amsterdam & Philadelphia, PA: Benjamins, 217–​256. Filip, Hana & Susan Rothstein. 2006. ‘Telicity as a semantic parameter,’ Formal Approaches to Slavic Linguistics #14: The Princeton Meeting 2005. Ann Arbor, MI: Michigan Slavic Publications, 139–156. Fillmore, Charles J. 1968. ‘The case for case,’ in Emmon Bach & Richard Harms (eds.), Universals in linguistic theory. New York: Holt, Rinehart & Winston, 1–​90. Fillmore, Charles J. 1977. ‘The case for case reopened,’ in Peter Cole (ed.), Syntax and semantics vol. 8: Grammatical relations. New York: Academic Press, 59–​81. Finer, Daniel. 1994. ‘On the nature of two A’-​positions in Selyarese,’ in Norbert Corver & Henk C. Van Riemsdijk (eds.), Studies on scrambling: Movement and non-​movement approaches to free word order phenomena (Studies in Generative Grammar, 41). Berlin:  Mouton de Gruyter, 153–​184. Fintel, Kai von. 2003. ‘Epistemic modals and conditionals revisited.’ Paper presented at the UMass Linguistics colloquium.

 

References   1173 Fintel, Kai von. 2004. ‘A minimal theory of adverbial quantification,’ in Barbara Partee & Hans Kamp (eds.), Context dependence in the analysis of linguistic meaning (Current Research in Semantics/Pragmatics Interface, vol. 11). Amsterdam: Elsevier, 137–175. Fleck, David W. 2006. ‘Antipassive in Matses.’ Studies in Language 30: 551–​573. Fodor, Janet Dean & Atsu Inoue. 1994. ‘The diagnosis and cure of garden paths.’ Journal of Psycholinguistic Research 23: 407–​434. Foley, William A. 1986. The Papuan languages of New Guinea (Cambridge Language Surveys). Cambridge: Cambridge University Press. Foley, William A. 1998. ‘Symmetrical voice systems and precategoriality in Philippine languages.’ Paper presented at the 3rd LFG conference, Brisbane. Foley, William A. 2008. ‘The place of Philippine languages in a typology of voice systems,’ in Peter K. Austin & Simon Musgrave (eds.), Voice and grammatical relation in Austronesian languages (Studies in Constraint Based Lexicalism). Stanford, CA: CSLI, 22–​44. Foley, William A. & Robert Van Valin. 1984. Functional syntax and universal grammar. Cambridge: Cambridge University Press. Folli, Raffaela & Heidi Harley. 2007. ‘Causation, obligation, and argument structure: On the nature of little v.’ Linguistic Inquiry 38(2): 197–​238. Folli, Raffaela & Heidi Harley. 2004. ‘Consuming results: flavors of little v,’ in P. Kempchinsky & R. Slabakova (eds.), Aspectual inquiries. Dordrecht: Kluwer, 1–​25. Folmer, Margaretha L. 1995. The Aramaic language in the Achaemenid period: A study in linguistic variation. Orientalia Lovaniensia analecta 68. Leuven: Peeters Forbes, Clarissa. 2013. ‘Coordination and number in the Gitksan nominal domain.’ MA thesis, University of Toronto. Forker, Diana. 2010. ‘The biabsolutive construction in Nakh–​Daghestanian.’ MS, Max Planck Institute for Evolutionary Anthropology. Forker, Diana. 2011. ‘Grammatical relations in Hinuq,’ in Vittorio Tomelleri, Manana Topadze, & Anna Lukianowicz (eds.), Languages and cultures in the Caucasus: Papers from the international conference ‘Current advances in Caucasian studies,’ Macerata, January 21–​23, 2010. Berlin: Otto Sagner, 553–​567. Forker, Diana. 2012. ‘The bi-​absolutive construction in Nakh–​Daghestanian.’ Folia Linguistica 46: 75–​108. Forker, Diana. 2013a. A grammar of Hinuq. Berlin: de Gruyter. Forker, Diana. 2013b. ‘Microtypology and the Tsezic languages: A case study of syntactic properties of converbal clauses.’ SKY Journal of Linguistics 26: 21–​40. Forker, Diana. 2014. ‘Are there subject anaphors?’ Linguistic Typology 18: 51–​81. Forker, Diana. 2016. ‘Complementizers in Hinuq,’ in Kasper Boye & Petar Kehayov (eds.), Semantic functions of complementizers in European languages. Berlin: De Gruyter, 745–792. Forker, Diana. [in press] ‘Complementizers in Hinuq,’ in Kasper Boye & Petar Kehayov (eds.), Semantic functions of complementizers in European languages. Berlin: de Gruyter. Fortescue, Michael. 1984. West Greenlandic (Croom Helm Descriptive Grammars). London: Croom Helm. Fortescue, Michael. 1995. The historical source and typological position of ergativity in Eskimo languages. Inuit Studies 19(2): 61–​75. Fortescue, Michael., Steven. Jacobson, & Lawrence. Kaplan. 2010. Comparative Eskimo dictionary: 25 with Aleut cognates (2nd edn.). Fairbanks: Alaska Native Language Center, University of Alaska Fairbanks. Fox, Charles E. 1970. Arosi dictionary. Canberra: Pacific Linguistics.

 

1174   References Fox, Danny. 1995. ‘Economy and scope.’ Natural Language Semantics 3: 283–​341. Fox, Danny. 2000. Economy and semantic interpretation. Cambridge, MA: MIT Press. Fox, Samuel Ethan. 2009. The Neo-​Aramaic dialect of Bohtan. Piscataway, NJ: Gorgias Press. Frampton, John & Sam Gutmann. 2000. ‘Agreement is feature sharing.’ MS, Northeastern University. Franchetto, Bruna. 2010. ‘The ergativity effect in Kuikuro (Southern Carib, Brazil),’ in Spike Gildea & Francesc Queixalós (eds.), Ergativity in Amazonia (Typological Studies in Language, 89). Amsterdam & Philadelphia, PA: Benjamins, 121–​158. Frank, Paul S. 1985. ‘A grammar of Ika.’ PhD dissertation, Indiana University, Bloomington. Fraser, Bruce. 1974. ‘An examination of the performative analysis.’ Papers in Linguistics 7: 140. Freeze, Ray. 1992. ‘Existentials and other locatives.’ Language 68(3): 553–​595. Frey, Werner. 2000. ‘Über die syntaktische Position der Satztopiks im Deutschen,’ in Kerstin Schwabe et al. (eds.), Issues in topics (ZAS papers in linguistics 20). Berlin: ZAS, 137–​172. Fried, Mirjam. 2008. ‘Constructions and constructs:  Mapping a shift between predication and attribution,’ in Alexander Bergs & Gabriele Diewald (eds.), Constructions and language change. Berlin: de Gruyter, 47–​80. Friederici, Angela D. 1995. ‘The time course of syntactic activation during language processing:  A  model based on neuropsychological and neurophysiological data.’ Brain and Language 50: 259–​281 Friederici, Angela D. 2011. ‘The brain basis of language processing: From structure to function.’ Physiological Reviews 91(4): 1357–​1392. Friederici, Angela D., Axel Mecklinger, Kevin M. Spencer, Karsten Steinhauer, & Emanuel Donchin. 2011. ‘Syntactic parsing preferences and their on-​ line revisions:  A  spatio-​ temporal analysis of event-​related brain potentials.’ Brain Research: Cognitive Brain Research 11(2): 305–​323. Friederici, Angela D., Martin Meyer, & D. Yves von Cramon. 2000. ‘Auditory language comprehension: An event-​related fMRI study on the processing of syntactic and lexical information.’ Brain and Language 74(2): 289–​300. Frisch, Stefan & Matthias Schlesewsky. 2001. ‘The N400 reflects problems of thematic hierachizing.’ Basic and Climitical Neurophysiology 12(15): 3391–​3394. Frisch, Stefan & Matthias Schlesewsky. 2005. ‘The resolution of case conflicts from a neurophysiological perspective.’ Brain Research: Cognitive Brain Research 25(2): 484–​498. Furman, Yulia & Sergey Loesev. 2014. ‘Studies in the Ṭuroyo verb,’ in G. Khan & L. Napiorkowska (eds.), Neo-​ Aramaic and its linguistic context. Piscataway, NJ:  Gorgias Press, 1–​28. Gaby, Alice. 2008a. ‘Pragmatically case-​marked:  Non-​syntactic functions of the Thaayorre ergative suffix,’ in Ilana Mushin & Brett Baker (eds.), Discourse and grammar in Australian languages. Amsterdam & Philadelphia, PA: Benjamins, 111–​134. Gaby, Alice. 2008b. ‘Rebuilding Australia’s linguistic profile: Recent developments in research on Australian aboriginal languages.’ Language and Linguistics Compass 2(2): 211–​233. Gaby, Alice. 2010. ‘From discourse to syntax and back: The lifecycle of Kuuk Thaayorre ergative morphology.’ Lingua 120(7): 1677–​1692. Gagliardi, Annie. 2012. ‘Input and intake in language acquisition.’ PhD dissertation, University of Maryland. Gagliardi, Annie, Michael Goncalves, Maria Polinsky, & Nina Radkevich. 2014. ‘The biabsolutive construction in Lak and Tsez.’ Lingua 150: 137–​170. Gair, James W. & Kashi Wali. 1988. ‘On distinguishing AGR from agr: Evidence from South Asia.’ Chicago Linguistics Society 24: 87–​104.

 

References   1175 Galloway, Brent Douglas. 1977. A grammar of Chilliwack Halkomelem, vol. 1. University of California. Ganenkov, Dmitry. 2006. ‘Experiencer coding in Nakh–​Daghestanian,’ in Leonid Kulikov, Andrej Malchukov, & Peter de Swart (eds.), Case, valence and transitivity. Amsterdam & Philadelphia, PA: Benjamins, 179–​202. Ganenkov, Dmitry, Timur Maisak, & Solmaz Merdanova. 2008. ‘Non-​canonical agent marking in Agul,’ in Helen de Hoop Helen & Peter de Swart (eds.), Differential subject marking. Dordrecht: Springer, 173–​198. Garrett, Andrew. 1990. ‘The origin of NP split ergativity.’ Language 66(2): 261–​296. Gärtner, Hans Martin, & Markus Steinbach. 2006. ‘A skeptical note on the syntax of speech acts and point of view,’ Form, structure, grammar. Berlin: Akademie-​Verlag, 213–​222. Gathercole, Virginia C. Mueller, Eugenia Sebastián, & Pilar Soto. 1999. ‘The early acquisition of Spanish verbal morphology:  Across-​the-​board or piecemeal knowledge.’ International Journal of Bilingualism 3: 133–​182. Gawne, Lauren. 2013. ‘Lamjung Yolmo copulas in use: Evidentiality, reported speech and questions.’ PhD dissertation, University of Melbourne. Gazdar, Gerald. 1979. Pragmatics:  Implicature, presupposition and logical form. New  York: Academic Press. Genetti, Carol. 1991. ‘From postposition to subordinator in Newari,’ in Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization (Routledge Language Family Series, 2). Amsterdam & Philadelphia, PA: Benjamins, 227–​255. Genetti, Carol. 2007. A grammar of Dolakha Newar (Mouton Grammar Library 40). Berlin: Mouton de Gruyter. Genetti, Carol. 2015. ‘The Tibeto-​Burman language of South Asia: The languages, their histories, and genetic relationships,’ in Elena Bashir, Hans Henrich Hock, & K.V. Subbārāo (eds.), The languages and linguistics of South Asia: A comprehensive guide. Berlin: Mouton de Gruyter. Genetti, Carol & Laura Crain. 2003. ‘Beyond preferred argument structure: Sentences, pronouns, and given referents in Nepali,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure:  Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 197–​223. George, Leland & Jaklin Kornfilt. 1981. ‘Finiteness and boundedness in Turkish,’ in Frank Heny (ed.), Binding and filtering. Cambridge, MA: MIT Press, 105–​129. Georgi, Doreen. 2014. ‘Opaque interactions of merge and agree.’ PhD thesis, Universität Leipzig. Gerdts, Donna B. 1988a. Object and absolutive in Halkomelem Salish. New York: Garland. Gerdts, Donna B. 1988b. ‘Antipassives and causatives in Ilokano:  Evidence for an ergative analysis,’ in Richard McGinn (ed.), Studies in Austronesian Linguistics. Athens, OH: Ohio University Press, 295–​321. Gerdts, Donna B. & Thomas E. Hukari. 2005. ‘Multiple antipassives in Halkomelem Salish.’ Proceedings of the Twenty-​sixth Annual Meeting of the Berkeley Linguistics Society.  Berkeley: Berkeley Linguistics Society, 51–​62. Gerdts, Donna B. & Thomas E. Hukari. 2006. ‘The Halkomelem middle: A complex network of constructions.’ Anthropological Linguistics 48: 44–​81. Gibson, Jeanne & Stanley Starosta. 1990. ‘Ergativity East and West,’ in Philip Baldi (ed.), Linguistic change and reconstruction methodology. Berlin: Mouton de Gruyter, 195–​210. Gil, David. 1993. ‘Syntactic categories in Tagalog,’ in Sudapom Luksaneeyanawin (ed.), Pan-​ Asiatic linguistics: Proceedings of the third international symposium on language and linguistics Bangkok January 8–​10, 1991. Bangkok: Chulalongkorn University Press, 1136–​1150.

 

1176   References Gil, David. 1995. ‘Parts of speech in Tagalog,’ in Mark Alves (ed.), Papers from the third annual meeting of the Southeast Asian Linguistics Society. Tempe: Arizona State University, 67–​90. Gil, David. 2000. ‘Syntactic categories, cross-​linguistic variation and universal grammar,’ in Petra M. Vogel & Bernard Comrie (eds.), Approaches to the typology of word classes (Empirical Approaches to Language Typology, vol. 23). Berlin: de Gruyter. Gildea, Spike. 1998. On reconstructing grammar. New York: Oxford University Press. Gildea, Spike. 2000. Reconstructing grammar: Comparative linguistics and grammaticalization. Amsterdam & Philadelphia, PA: Benjamins. Gildea, Spike. 2004. ‘Are there universal cognitive motivations for ergativity?’ in Francesc Queixalós (ed.), L’ergativité en Amazonie, v. 2. Brasília: CNRS, IRD and the Laboratório de Línguas Indígenas, UnB, 1–​37. Gildea, Spike. 2008. ‘Explaining similarities between main clauses and nominalized phrases.’ Amérindia 32: 57–​75. Gildea, Spike & Flávia de Castro Alves. 2010. ‘Nominative–​absolutive:  Counter-​universal split ergativity in Jê and Cariban,’ in Spike Gildea & Francesc Queixalós (eds.), Ergativity in Amazonia. Amsterdam & Philadelphia, PA: Benjamins, 159–​199. Gillon, Carrie. 1999. ‘When wh-​words move and why: A case study in Inuktitut’. MA dissertation, University of Toronto. Giorgi, Alessandra. 2010. About the speaker: Towards a syntax of indexicality (Oxford Studies in Theoretical Linguistics). Oxford: Oxford University Press. Giorgi, Alessandra & Fabio Pianesi. 2001. ‘Ways of terminating,’ in Carlo Cecchetto, Gennaro Chierchia, & Maria T. Guasti (eds.), Semantic interfaces: Reference, anaphora and aspect. Stanford, CA: CSLI. Giridhar, Puttushetra Puttuswamy. 1980. Angami grammar. Mysore: Central Institute of Indian Languages. Givón, Talmy. 1979. On understanding grammar. New York: Academic Press. Givón, Talmy. 1980. ‘The drift away from ergativity.’ Folia Linguistica Historica 1: 41–​60. Givón, Talmy (ed.). 1983a. Topic continuity in discourse: A quantitative cross language study. Amsterdam & Philadelphia, PA: Benjamins. Givón, Talmy. 1983b. ‘Topic continuity in spoken English:  An introduction,’ in T. Givón (ed.), Topic continuity in discourse:  A  quantitative cross-​language study. Amsterdam & Philadelphia, PA: Benjamins, 1–​42. Givón, Talmy. 1984. Syntax: A functional–​typological introduction, vol. 1. Amsterdam & Philadelphia, PA: Benjamins. Givón, Talmy. 2001a. Syntax: An Introduction, vol. 1. Amsterdam & Philadelphia, PA: Benjamins. Givón, Talmy. 2001b. Syntax: An Introduction, vol. 2. Amsterdam & Philadelphia, PA: Benjamins. Givón, Talmy. 2009. The genesis of syntactic complexity: Diachrony, ontogeny, neuro-​cognition, evolution. Amsterdam & Philadelphia, PA: Benjamins. Glass, Amee & Dorothy Hackett. 2003. Ngaanytjarra & Ngaatjatjarra to English dictionary. Alice Springs, NT: IAD Press. Glennon, John J. 2014. ‘Syntactic ergativity in Nehan.’ MA thesis, Dallas, Graduate Institute of Applied Linguistics. Goddard, Cliff. 1982. ‘Case systems and case marking in Australian languages: A new interpretation.’ Australian Journal of Linguistics 2: 167–​196. Goddard, Cliff. 1986. Yankunytjatjara grammar. Alice Springs:  Institute for Aboriginal Development.

 

References   1177 Goenaga, Patxi. 1985. ‘Nahi eta Behar,’ in Jose Luis Melena (ed.), Symbolae Ludovico Mitxelena septuagenario oblatae. Veleia Anex 1. Vitoria-​Gasteiz: Instituto de Ciencias de la Antigüedad/​ Aitinate-​zientzien Instituta, Universidad del País Vasco/​Euskal Herriko Unibertsitatea UPV/​EHU, 943–​953. Goenaga, Patxi. 2006. ‘Behar-​en lekua hiztegian eta gramatikan,’ in Beatriz Fernández & Itziar Laka (eds.), Andolin gogoan: Essays in honour of Professor Eguzkitza. Bilbao: University of the Basque Country, 397–​416. Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago, IL: University of Chicago Press. Goldberg, Adele E. 2004. ‘Pragmatics and argument structure,’ in Laurence R. Horn & Gregory Ward (eds.), The handbook of pragmatics. Oxford: Blackwell, 427–​441. Goldenberg, Gideon. 1992. ‘Aramaic Perfects.’ Israel Oriental Studies 12: 113–​133. Goldin-​Meadow, Susan. 2003. ‘Thought before language:  Do we think ergative?’ in Dedre Gentner & Susan Goldin-​Meadow (eds.), Language in Mind: Advances in the study of language and thought. Cambridge, MA: MIT Press, 493–​522. Goldin-​Meadow, Susan, Elif Yalabik, & Lisa Gershkoff-​Stowe. 2000. ‘The resilience of ergative structure in language created by children and by adults,’ in S.C. Howell, S.A. Fish, & T. Keith-​Lucas (eds.), Proceedings of the 24th annual Boston University conference on language development, vol. 1. Somerville, MA: Cascadilla Press, 343–​353. Gómez, Ricardo & Koldo Sainz. 1995. ‘On the origin of the finite forms of the Basque verb,’ in J.I. Hualde, J. Lakarra & R.L. Trask (eds.), Towards a history of the Basque language. Amsterdam & Philadelphia, PA: Benjamins, 235–​273. Gonda, Jan. 1951. Remarks on the Sanskrit passive. Leiden: Brill. Goodall, Grant. 1993. ‘On case and the passive morpheme.’ Natural Language and Linguistic Theory 11: 31–​44. Goodall, Grant. 1999. ‘Passives and arbitrary plural subjects in Spanish,’ in J.-​Marc Authier, Barbara Bullock, & Lisa Reed (eds.), Formal perspectives on Romance linguistics:  Papers from the 28th linguistic symposium on Romance languages. Amsterdam & Philadelphia, PA: Benjamins, 135–​150. Gordon, Peter C. & Randall Hendrick. 2005. ‘Relativization, ergativity, and corpus frequency.’ Linguistic Inquiry 36(3), 456–​463. Granites, Robin Japanangka, Ken Hale, & David Odling-​Smee. 1976. Survey of Warlpiri syntax and morphology [typed version of Odling-​Smee MS ‘Survey of Warlpiri grammar’]. Cambridge MA: MIT, 65 pp. ts. Green, Ian & Rachel Nordlinger. 2004. ‘Revisiting Proto-​Mirndi,’ in Claire Bowern & Harold Koch (eds.), Australian languages: Classification and the comparative method. Amsterdam & Philadelphia, PA: Benjamins, 291–​311. Greenberg, Joseph H. 1963. The languages of Africa. The Hague: Mouton de Gruyter. Greenberg, Joseph H. 1978/​1990. ‘How does a language acquire gender markers?’ in Keith Denning & Suzanne Kemmer (eds.), On language: Selected writings of Joseph H. Greenberg. Stanford, CA: Stanford University Press, 241–​270. Gregersen, Edgar A. 1977. Language in Africa:  An introductory survey. New  York, Paris, & London: Gordon & Breach. Grewendorf, Günther. 1972. ‘Sprache ohne Kontext. Zur Kritik der performativen Analyse,’ in Dieter Wunderlich (ed.), Linguistische Pragmatik. Frankfurt/​M.: Athenäum, 144–​182 Griffin, Zenzi M. & Kathryn Bock. 2000. ‘What the eyes say about speaking.’ Psychological Science 11(4): 274–​279.

 

1178   References Grimm, Scott. 2011. ‘Semantics of dase,’ Morphology 21: 515–​544. Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press. Grimshaw, Jane & Armin Mester. 1985. ‘Complex verb formation in Eskimo.’ Natural Language and Linguistic Theory 3: 1–​19. Grinevald Craig, Colette. 1979. ‘The antipassive and Jacaltec,’ Papers in Mayan linguistics, vol. 7. Lucas Brothers, 139–​164. Grinevald, Colette & Marc Peake. 2012. ‘Ergativity and voice in Mayan languages: A functional–​ typological approach,’ in G. Authier & K. Haude (eds.), Ergativity, valency and voice. Berlin & Boston: Mouton de Gruyter, 15–​49 Grinstead, John. 2000. ‘Case, inflection and subject licensing in child Catalan and Spanish,’ Journal of Child Language 27: 119–​55. Grinstead, John. 2004. ‘Subjects and interface delay in child Spanish and Catalan,’ Language 80: 40–​72. Grohmann, Kleanthes. 2003. Prolific domains: On the anti-​locality of movement dependencies. Amsterdam & Philadelphia, PA: Benjamins. Grune, Dick. 1998. ‘Burushaski: An extraordinary language in the Karakoram mountains.’ MS, Vrije University. Gulya, János. 1966. Eastern Ostyak chrestomathy. Bloomington, IN: University of Indiana Press. Guasti, Maria Teresa. 1993. Causative and perception verbs:  A  comparative study. Turin: Rosenberg & Sellier. Guilfoyle, Eithne, Henrietta Hung, & Lisa Travis. 1992. ‘Spec of IP and Spec of VP: Two subjects in Austronesian languages.’ Natural Language and Linguistic Theory 10: 375–​414. Guillaume, Antoine. 2008. A grammar of Cavineña. Berlin: Mouton de Gruyter. Guillaume, Antoine. 2014. ‘The decay of ergativity in Tacana (Tacanan family, Amazonian Bolivia): A preliminary descriptive and comparative study.’ Paper presented at Syntax of the World’s Languages VI, Pavia, 8–​10, September. Guirardello-​Damian, Raquel. 2003. ‘Classes verbais e mudanças de valência em Trumai,’ in Spike Gildea & Francesc Queixalós (eds.), Proceedings of the workshop Ergatividade Na Amazônia. Paris:  Centre d’études des langues indigènes d’Amèrique (CNRS, IRD) and Laboratório de Línguas Indígenas, 195–​214. Guiradello-​Damian, Raquel. 2010. ‘Ergativity in Trumai,’ in Spike Gildea & Francesc Queixalós (eds.), Ergativity in Amazonia. Amsterdam & Philadelphia, PA: Benjamins, 203–​234. Gutiérrez-​Mangado, Maria Juncal. 2011. ‘Children’s comprehension of relative clauses in an ergative language: The case of Basque.’ Language Acquisition 18(3): 176–​201. Gutman, Ariel. 2009. ‘Reexamination of the bare preterite base in the Jewish Neo-​Aramaic dialect of Zakho.’ Aramaic Studies 6(1): 59–​84 Gzella, Holger. 2004. Tempus, Aspekt Und Modalität Im Reichsaramäischen. Veröffentlichungen der Orientalischen Kommission 48. Wiesbaden: Harrassowitz Hacquard, Valentine. 2006. ‘Aspects of modality.’ PhD dissertation, MIT. Haddican, Bill. 2005. ‘Two kinds of restructuring infinitives in Basque,’ in John Alderete, Chung-​ hye Han, & Alexei Kochetov (eds.), Proceedings of WCCFL 24. Somerville, MA: Cascadilla Press, 182–​190. Haegeman, Liliane. 1998. ‘Root infinitives, clitics and truncated structures’, in H. Clahsen (ed.), Generative perspectives on language acquisition. Amsterdam: John Benjamins, 271–307. Haegeman, Liliane & Virginia Hill. 2013. ‘The syntactization of discourse,’ in Raffaella Folli, Christina Sevdali, & Robert Truswell (eds.), Syntax and its limits. Oxford: Oxford University Press, 370–​390.

 

References   1179 Haegeman, Liliane & Terje Lohndal. 2010. ‘Negative concord and (multiple) agree:  A  case study of west Flemish.’ Linguistic Inquiry 41: 181–​211. Hagoort, Peter, Colin M. Brown, & Jolanda Groothusen. 1993. ‘The syntactic positive shift (SPS) as an ERP measure of syntactic processing.’ Language and Cognitive Processes 8: 439–​483. Hagoort, Peter, Colin M. Brown, & Lee E. Osterhout. 1999. ‘The neurocognition of syntactic processing,’ in C. Brown & P. Hagoort (eds.), The neurocognition of language. Oxford: Oxford University Press, 273–​316. Hagoort, Peter, Lea Hald, Marcel Bastiaansen, & Karl Magnus Petersson. 2004. ‘Integration of word meaning and world knowledge in language comprehension.’ Science 304(5669): 438–​441. Haig, Geoffrey. 1998. ‘On the interaction of morphological and syntactic ergativity: Lessons from Kurdish.’ Lingua 105: 149–​173. Haig, Geoffrey. 2004. Alignment in Kurdish:  A  diachronic perspective. Habilitationsschrift. Kiel: Christian-​Albrechts-​Universität zu Kiel. Haig, Geoffrey. 2006. ‘Turkish influence on Kurmanji: Evidence from the Tunceli dialect,’ in Lars Johanson & Christiane Bulut (eds.), Turkic–​Iranian contact areas: Historical and linguistic aspects. Wiesbaden: Harrassowitz, 283–​299. Haig, Geoffrey. 2008. Alignment change in Iranian languages. Berlin: Mouton de Gruyter. Haig, Geoffrey. 2010. ‘Alignment,’ in Silvia Luraghi & Vit Bubenik (eds.), Continuum companion to historical linguistics. New York: Continuum, 250–​268. Haig, Geoffrey & Ergin Öpengin. 2014. ‘Kurdish: A critical research overview.’ Kurdish Studies 2(2): 99–​122 Haig, Geoffrey & Ergin Öpengin. [forthcoming]. ‘Kurdish in Turkey: An overview of grammar, dialectal variation and status,’ in C. Bulut (ed.), Minority languages in Turkey. Wiesbaden: Harrassowitz. Haig, Geoffrey & Stefan Schnell. 2016. ‘The discourse basis of ergativity revisited.’ Language 92(3): 591–618. Hajičová, Eva, Barbara H. Partee, & Petr Sgall. 1998. Topic–​focus articulation, tripartite structures, and semantic content. Dordrecht, Boston, & London: Kluwer. Hale, Ken. 1959–​1960 and 1966–​1967. ‘Warlpiri language fieldnotes.’ MS deposited in Australian Institute of Aboriginal and Torres Strait Islander Studies (AIATSIS), Canberra ACT. Hale, Ken. 1968. ‘Review of Hohepa 1967.’ Journal of the Polynesian Society 77: 83–​99. Hale, Ken. 1970. ‘The passive and ergative in language change:  The Australian case,’ in S.A. Wurm & D.C. Laycock (eds.), Pacific linguistics studies in honour of Arthur Capell. Canberra: Australian National University, 757–7​81. Hale, Ken. 1972. ‘A new perspective on American Indian linguistics,’ in A. Ortiz (ed.), New perspectives on the Pueblos. Albuquerque: University of New Mexico Press, 87–​103. Hale, Ken. 1973a. ‘Person marking in Walbiri,’ in Stephen R. Anderson & Paul Kiparsky (eds.), A Festschrift for Morris Halle. New York: Holt, Rinehart, & Winston, 308–​344. Hale, Ken. 1973b. ‘Deep-​surface canonical disparities in relation to analysis and change: An Australian example,’ in A. Sebeok (ed.), Current trends in linguistics. The Hague: Mouton de Gruyter, 401–​458. Hale, Ken. 1976a. ‘The adjoined relative clause in Australia,’ in R.M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 78–​105. Hale, Ken. 1976b. ‘On ergative and locative suffixal alternations in Australian languages,’ in R.M.W. Dixon (ed.), 414–​417. Hale, Ken. 1981a. On the position of Walbiri in a typology of the base. Bloomington, IN: Indiana University Linguistics Club.

 

1180   References Hale, Ken. 1981b. ‘Preliminary remarks on the grammar of part-​whole relations in Warlpiri,’ in Jim Hollyman & Andrew Pawley (eds.), Studies in Pacific languages & cultures in honour of Bruce Biggs. Auckland: Linguistic Society of New Zealand, 333–​344. Hale, Ken. 1982. ‘Some essential features of Warlpiri verbal clauses,’ in Stephen Swartz (ed.), Papers in Warlpiri grammar: In memory of Lothar Jagst. Papers of SIL-​AAB, Series A, vol. 6. Darwin: Summer Institute of Linguistics, 217–​315. Hale, Ken. 1983. ‘Warlpiri and the grammar of nonconfigurational languages.’ Natural Language and Linguistic Theory 1: 1–​47. Hale, Ken. 2002. ‘Eccentric agreement,’ in Beatriz Fernández & Pablo Albizu (eds.), Kasu eta komunztaduraren gainean. Vitoria-​Gasteiz: Euskal Herriko Unibetsitatea, 15–​48. Hale, Ken & Samuel Jay Keyser. 1993. ‘On argument structure and the lexical expression of syntactic relations,’ in Ken Hale & Samuel Jay Keyser (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger. Cambridge, MA: MIT Press, 53–​110. Hale, Ken, Mary Laughren, & Jane Simpson 1995. ‘Warlpiri,’ in J. Jacobs, A. von Stechow, W. Sternefeld, & T. Venneman (eds.), Syntax: An international handbook of contemporary research, 2. Berlin: de Gruyter, 1430–​1451. Halle, Morris. 1997. ‘Distributed morphology:  Impoverishment and fission.’ MIT Working Papers in Linguistics 30: 425–​449. Halle, Morris & Ken Hale. 1997. ‘Chukchi transitive and antipassive constructions.’ MS, MIT. Halle, Morris & Alec Marantz. 1993. ‘Distributed morphology and the pieces of inflection,’ in Ken Hale & Samuel Jay Keyser (eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger. Cambridge, MA: MIT Press, 111–​176. Hallman, Peter. 2008. ‘Definiteness in Inuktitut.’ MS, University of Vienna. Halpert, Claire. 2012. ‘Argument licensing and agreement in Zulu.’ PhD dissertation, MIT. Hamann, Cornelia. 2002. From syntax to discourse: Pronominal clitics, null subjects and infinitives in child language. Dordrecht: Kluwer. Hamann, Cornelia, Luigi Rizzi, & Ulrich H. Frauenfelder. 1998. ‘On the acquisition of subject and object clitics in French,’ in H. Clahsen (ed.), Generative perspectives on language acquisition. Amsterdam & Philadelphia, PA: Benjamins, 309–​334. Han, Chung-​Hye. 2000. The structure and interpretation of imperatives: Mood and force in universal grammar. New York: Garland & Routledge. Handschuh, Corinna. 2014. A typology of marked-​S languages. Studies in Diversity Linguistics, 1. New York: Language Science Press. Harley, Heidi. 2002. ‘Possession and the double object construction.’ Linguistic Variation Yearbook 2: 29–​68. Harley, Heidi. 2013. ‘External arguments and the mirror principle: On the distinctness of voice and v.’ Lingua 125: 34–57. Harley, Heidi & Rolf Noyer. 1999. ‘State-​ of-​ the-​ article:  Distributed morphology.’ Glot International 4(4): 3–​9. Harley, Heidi & Elizabeth Ritter. 2002. Person and number in pronouns: A feature–​geometric analysis. Language 78: 482–​526. Harper, Kenn. 1974. Some aspects of the grammar of Eskimo dialects of Cumberland Peninsula and North Baffin Island. Ottawa, Ontario: National Museum of Canada. Harris, Alice C. 1981. Georgian syntax: A study in relational grammar. Cambridge: Cambridge University Press. Harris, Alice C. 1982. ‘Georgian and the unaccusative hypothesis.’ Language 58(2): 290–​306. Harris, Alice C. 1985. Diachronic syntax: The Kartvelian case. Orlando: Academic Press.

 

References   1181 Harris, Alice C. 1988. ‘art’ik’lisagan nac’armoebi brunvebi kartvelur enebši da enobrivi universialebi’ [‘Case desinences derived from articles in the Kartvelian languages and linguistic universals’], in E. Khintibidze (ed.), Proceedings of the 1st international symposium in Kartvelian studies. Tbilisi: TSUG, 64–​70. Harris, Alice C. 1990. ‘Georgian: A language with active case marking.’ Lingua 80: 347–​365. Harris, Alice C. 1994. ‘Ergative-​to-​accusative shift in agreement:  Tabasaran,’ in Howard I. Aaronson (ed.), Linguistic studies in the non–​Slavic languages in the Commonwealth of Independent states and the Baltic republics (NSL7). Chicago, IL: Chicago Linguistic Society, 113–​131. Harris, Alice C. 2002. Endoclitics and the origins of Udi morphosyntax. Oxford:  Oxford University Press. Harris, Alice C. 2006. ‘Active/​inactive marking,’ in Keith Brown (ed.), Encyclopedia of language and linguistics, vol. 1. Oxford: Elsevier, 40–​44. Harris, Alice C. 2008. ‘Reconstruction in syntax:  Reconstruction of patterns,’ in Gisella Ferraresi & Maria Goldbach (eds.), Principles of syntactic reconstruction. Amsterdam: Benjamins, 73–​96. Harris, Alice C. & Lyle Campbell. 1995. Historical syntax in cross-​ linguistic perspective. Cambridge: Cambridge University Press. Harris, John. 1986. Northern Territory pidgins and the origin of Kriol. Canberra:  Pacific Linguistics. Harris, Tony & Ken Wexler. 1996. ‘The optional infinitive stage in child grammars,’ in Harald Clahsen (ed.), Generative perspectives on language acquisition. Philadelphia, PA: Benjamins, 1–​42. Hartmann, Katharina & Malte Zimmermann. 2007. ‘In place—​out of place? Focus in Hausa,’ in Kerstin Schwabe & Susanne Winkler (eds.), On information structure, meaning and form. Amsterdam & Philadelphia, PA: Benjamins, 365–​403. Harves, Stephanie. 2008. ‘Intensional transitives and silent HAVE:  Distinguishing between Want and Need,’ in Natasha Abner & Jason Bishop (eds.), Proceedings of the 27th West Coast conference on formal linguistics, Somerville, MA: Cascadilla Proceedings Project, 211–​219. Harves, Stephanie & Richard S. Kayne. 2012. ‘Having “need” and needing “have.” ’ Linguistic Inquiry 43(1): 120–​132 Harvey, Mark. 2008. Proto Mirndi:  A  discontinuous language family in northern Australia. Canberra: Pacific Linguistics. Haspelmath, Martin. 1991. ‘On the question of deep ergativity: The evidence from Lezgian.’ Papiere zur Linguistik 44/​45: 5–​27. Haspelmath, Martin. 1993. A grammar of Lezgian. Berlin: Mouton de Gruyter. Haspelmath, Martin. 1996. ‘Complement clauses,’ in Aleksandr E. Kibrik (ed.), Godoberi. Munich: Lincom, 175–​197. Haspelmath, Martin. 1999. ‘Long distance agreement in Godoberi (Daghestanian) complement clauses.’ Folia Linguistica 33: 131–​151. Haspelmath, Martin. 2001. ‘Non-​canonical marking of core arguments in European languages,’ in A. Aikhenvald et al. (eds.), Non-​canonical marking of subjects and objects. Amsterdam & Philadelphia, PA: Benjamins, 53–​85. Haspelmath, Martin. 2005. ‘Universals of differential case marking.’ Handout, accessed at http://​email.eva.mpg.de/​~haspelmt/​2.DiffCaseMarking.pdf Haspelmath, Martin. 2006. ‘Review of preferred argument structure: Grammar as architecture for function.’ Language 83: 908–​912.

 

1182   References Haspelmath, Martin. 2013. ‘Argument indexing: A conceptual framework for the syntactic status of bound person forms,’ in Dik Bakker & Martin Haspelmath (eds.), Languages across boundaries: Studies in memory of Anna Siewierska. Berlin: Mouton. Hasselbach, Rebecca. 2013. Case in Semitic: Roles, relations, and reconstruction (Oxford studies in diachronic and historical linguistics, 3). Oxford: Oxford University Press Haupt, Friederike S., Matthias Schlesewsky, Dietmar Roehm, Angela D. Friederici, & Ina Bornkessel-​Schlesewsky. 2008. ‘The status of subject–​object reanalyses in the language comprehension architecture.’ Journal of Memory and Language 59(1): 54–​96. Haviland, John. 1979. ‘Guugu Yimidhirr,’ in R.M.W. Dixon & Barry J. Blake (eds.), Handbook of Australian languages 1. Amsterdam & Philadelphia, PA: Benjamins, 26–​180. Haviland, John. 1981. Sk’op sotz’leb. El tzotzil de San Lorenzo Zinacantán. Mexico, DF: UNAM. Hawkins, John A. 1994. A performance theory of order and constituency. Cambridge: Cambridge University Press. Hawkins, John A. 2004. Efficiency and complexity in grammars. Oxford: Oxford University Press. Hawkins, John A. 2014. Cross-​linguistic variation and efficiency. Oxford:  Oxford University Press. Hayashi, Midori. 2011. ‘The structure of multiple tenses in Inuktitut.’ PhD thesis, University of Toronto. Hayashi, Midori & Bettina Spreng. 2005. ‘Is Inuktitut tenseless?’ Proceedings of the 2005 annual conference of the Canadian Linguistic Association. Heath, Jeffrey. 1976. ‘Antipassivization: A functional typology.’ Proceedings of the second annual meeting of the Berkeley Linguistic Society, 202–​211. Heath, Jeffrey. 1980. ‘Whither ergativity? A review article.’ Linguistics 18, 877–​910. Heim, Johannes. 2015. ‘Expertise and common ground’. MS, UBC. Heim, Johannes, Hermann Keupdjio, Zoe Wai Man Lam, Adriana Osa-​Gómez, & Martina Wiltschko. 2014. ‘How to do things with particles.’ Proceedings of the CLA. Heine, Bernd. 2003. ‘Grammaticalization,’ in Brian D. Joseph & Richard D. Janda (eds.), The handbook of historical linguistics. Oxford: Blackwell, 575–​601. Heine, Bernd & Derek Nurse. 2000. African languages: An introduction. Cambridge: Cambridge University Press. Heine, Bernd & Heiko Narrog. 2011. ‘Introduction,’ in Heiko Narrog & Bernd Heine (eds.), The Oxford handbook of grammaticalization. Oxford: Oxford University Press, 1–​16. Heine, Bernd & Tania Kuteva. 2002. World lexicon of grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer. 1991. Grammaticalization: A conceptual framework. Chicago, IL: University of Chicago Press. Heine, Bernd. 2009. ‘Grammaticalizaton of cases’, in Andrej L. Malchukov and Andrew Spencer (eds.), The Oxford handbook of case. (Oxford Handbooks in Linguistics.) Oxford: Oxford University Press, 458–469. Helasvuo, Marja-​Liisa. 2003. ‘Argument splits in Finnish grammar and discourse,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure:  Grammar as architecture for function. Amsterdam & Philadelphia, PA:  Benjamins, 247–​272. Heine, Bernd. 2009. ‘Grammaticalizaton of cases,’ in Andrej L. Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 458–​469. Helmbrecht, Johannes. 1996. ‘The syntax of personal agreement in East Caucasian languages.’ Sprachtypologie und Universalienforschung 49: 127–​148.

 

References   1183 Hemmauer, Roland & Michael Waltisberg. 2006. ‘Zur Relationalen Verhalten Der Verbalflexion Im Ṭuroyo.’ Folia Linguistica Historica 27(1–​2): 19–​59. Henderson, Brent. 2006. ‘The syntax and typology of Bantu relative clauses.’ Dissertation, University of Illinois. Henderson, Brent. 2011. ‘Agreement, locality, and OVS in Bantu.’ Lingua 121: 742–​753. Henderson, Robert. 2012. Morphological alternations at the intonational phrase edge: The case of K’ichee.’ Natural language & linguistic theory 30: 741–​787. Hendriks, Petra & Helen de Hoop. 2001. ‘Optimality theoretic semantics.’ Linguistics and Philosophy 24: 1–​32. Hercus, Luise. 1982. The Bagandji language. Canberra: Pacific Linguistics. Herring, Susan. 1989. ‘Verbless presentation and the discourse basis of ergativity,’ in Bradley Music, Randolph Graczyk, & Caroline Wiltshire (eds.), Papers from the 25th annual regional meeting of the Chicago Linguistic Society 25, Part  2:  Parasession on language in context. Chicago, IL: Chicago Linguistics Society, 123–​137. Hewitt, Brian George. 1995. ‘Georgian: Ergative, active, or what?’ in D.C. Bennett, T. Bynon, & B.G. Hewitt (eds.), Subject, voice and ergativity. London: SOAS, 202–​217. Hewson, John & Vit Bubenik. 2006. From case to adposition: The development of configurational syntax in Indo-​European Languages. Amsterdam & Philadelphia, PA: Benjamins. Hill, Virginia. 2007. ‘Vocatives and the pragmatics: Syntax interface.’ Lingua 117: 2077–​2105. Hill, Virginia. 2013. ‘Features and strategies: The internal syntax of vocative phrases,’ in Barbara Sonnenhauser & Patrizia Noel Aziz Hanna (eds.), Vocatives! Addressing between system and performance. Berlin: Mouton, 79–​102. Hill, Virginia & Melita Stavrou. 2013. Vocatives: How syntax meets with pragmatics. Leiden: Brill. Himmelmann, Nikolaus P. 1987. Morphosyntax und Morphologie. Die Ausrichtungsaffixe im Tagalog, vol. 8 of Studien zur theoretischen Linguistik. Munich: Fink. Himmelmann, Nikolaus P. 1991. ‘The Philippine challenge to universal grammar.’ Institute for Linguistics, University of Köln Working Papers 15. Himmelmann, Nikolaus P. 2004. ‘Lexicalization and grammaticization: Opposite or orthogonal?’ in Walter Bisang, Nikolaus P. Himmelmann, & Björn Wiemer (eds.), What makes grammaticalization? A look from its fringes and its components (Trends in Linguistics. Studies and Monographs 158.) Berlin & New York: Mouton de Gruyter, 21–​42. Himmelmann, Nikolaus P. 2005. ‘Typological characteristics,’ in Alexander Adelaar & Nikolaus P. Himmelmann (eds.), The Austronesian languages of Asia and Madagascar. London: Routledge. Himmelmann, Nikolaus P. 2008. ‘Lexical categories and voice in Tagalog,’ Voice and grammatical functions in Austronesian languages. Stanford, CA: CSLI, 247–​293. Hiraiwa, Ken. 2001. ‘Multiple agree and the defective intervention constraint in Japanese.’ MIT Working Papers in Linguistics 40(40): 67–​80. Hock, Hans Heinrich. 1986. ‘ “P-​Oriented” constructions in Sanskrit,’ in Bh. Krishnamurti (ed.), South Asian languages: Structure, convergence and diglossia. Delhi: Motilal Banarsidass. Hock, Hans Henrich. 1991. Principles of historical linguistics. Berlin: de Gruyter. Hoddinott, William G., & Frances M. Kofod. 1976a. ‘The bivalent suffix -​ku: Djamindjungan,’ in R.M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 437–441. Hoddinott, William G., & Frances M. Kofod. 1976b. ‘Ergative, locative and instrumental case inflections: Djamindjungan,’ in R.M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 397–401.

 

1184   References Hoddinott, William G., & Frances M. Kofod. 1976c. ‘Simple and compound verbs: Conjugation by auxiliaries in Australian verbal systems:  Djamindjungan,’ in R.M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 698–704. Hoekstra, Teun & Nina Hyams. 1998. ‘Aspects of root infinitives.’ Lingua 106: 81–​112. Hoff, Erika & Letitia Naigles. 2002. ‘How children use input to acquire a lexicon.’ Child Development 73: 418–​33. Hoffmann, Dorothea. 2011. ‘Descriptions of motion and travel in Jaminjung and Kriol.’ PhD, University of Manchester, Manchester. Hofling, Charles A. 2003. ‘Tracking the deer: Nominal reference, parallelism and preferred argument structure in Itzaj Maya narrative genres,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 385–​410. Hohepa, Patrick W. 1969. ‘The accusative-​to-​ergative drift in Polynesian languages.’ Journal of the Polynesian Society 78: 295–​329. Holisky, Dee Ann. 1981. Aspect and Georgian medial verbs. Delmar, NY: Caravan Press. Holisky, Dee Ann. 1984. ‘Anomalies in the use of the ergative case in Tsova-​Tush (Batsbi).’ Folia Slavica 7: 181–​194. Holisky, Dee Ann. 1987. ‘The case of the intransitive subject in Tsova-​Tush.’ Lingua 71: 103–​132. Holisky, Dee Ann & Rusudan Gagua. 1994. ‘Tsova–​Tush (Batsbi),’ in Rieks Smeets (ed.), The Indigenous languages of the Caucasus, vol. 4. Delmar, NY: Caravan Press, 148–​212. Holmberg, Anders. 1986. ‘Word order and syntactic features in the Scandinavian languages and English.’ PhD dissertation, University of Stockholm. Holmberg, Anders. 1999. ‘Remarks on Holmberg’s Generalization.’ Studia Linguistica 53: 1–​39. Holmberg, Anders & David Odden. 2004. ‘Ergativity and role-​ marking in Hawrami.’ Conference handout. Holmberg, Anders & Thorbjörg Hróarsdóttir. 2000. ‘Agreement and movement in Icelandic raising constructions.’ Lingua 113: 997–​1019. Holmer, Arthur. 1999. ‘An active analysis of Basque ergativity.’ Fontes Linguae Vasconum 31(81) (May–​Aug.): 189–​225. Hook, Peter Edwin. 1974. The compound verb in Hindi. Michigan: Center for South and Southeast Asian Studies, University of Michigan. Hook, Peter Edwin. 1991. ‘On Identifying the Conceptual Restructuring of Passive to Ergative in Indo-​Aryan,’ in Madhav M. Deshpande & Saroja Bhate (eds.), Pāninian studies: Professor S.D. Joshi felicitation volume. Michigan: Center for South and Southeast Asian Studies, University of Michigan, 177–​199. Hook, Peter Edwin, Omkar N. Koul, & Ashok K. Koul. 1987. ‘Differential S-​marking in Marathi, Hindi–​ Urdu and Kashmiri,’ Papers from the twenty-​ third regional meeting. Chicago, IL: Chicago Linguistic Society, 148–​165. Hopkins, Simon. 1989. ‘Neo-​Aramaic dialects and the formation of the preterite.’ Journal of Semitic Studies 37: 74–​90. Hopper, Paul J. 1983. ‘Ergative, passive, and active in Malay narrative,’ in Flora Klein-​Andreu (ed.), Discourse perspectives on syntax. New York: Academic Press, 67–​88. Hopper, Paul J. 1987. ‘Emergent grammar.’ Berkeley Linguistics Society 13: 139–​157. Hopper, Paul J. & Sandra A. Thompson 1980. ‘Transitivity in grammar and discourse.’ Language 56: 251–​299. Hopper, Paul J. & Elizabeth Traugott. 1993. Grammaticalization. Cambridge:  Cambridge University Press.

 

References   1185 Hopper, Paul J. & Elizabeth Traugott. 2003. Grammaticalization (2nd edn.). Cambridge: Cambridge University Press. Hornstein, Norbert. 1995. Logical form: From GB to minimalism. London: Blackwell Hornstein, Norbert. 1999. ‘Movement and control.’ Linguistic Inquiry 30(1): 69–​96. Horrocks, Geoffrey & Melita Stavrou. 1987. ‘Bounding theory and Greek syntax: Evidence for wh-​movement in NP.’ Journal of Linguistics 23: 79–​108. Hou, Liwen. 2013. ‘Agent focus in Chuj reflexive constructions.’ BA thesis, McGill University. Hualde, José Ignacio. 1991. Basque phonology. London: Routledge. Hualde, José Ignacio. 2003. ‘Case and number inflection in noun phrases,’ in J.I. Hualde & J. Ortiz de Urbina (eds.), A grammar of Basque. Berlin: Mouton de Gruyter, 171–​179. Hualde, José Ignacio & Jon Ortiz de Urbina (eds.). 2003. A grammar of Basque. Berlin: Mouton de Gruyter. Huang, C.-​T. James. 1982. ‘Logical relations in Chinese and the theory of grammar.’ PhD dissertation, MIT. Huang, C.-​T. James. 1984. ‘On the distribution and reference of empty pronouns.’ Linguistic Inquiry 15: 531–​574. Huang, Lillian Meijin. 1994. ‘Ergativity in Atayal.’ Oceanic Linguistics 33: 129–​143. Huang, Lillian Meijin. 2000. Atayal reference grammar, vol. 1 of Austronesian Languages of Taiwan. Taipei: Yuanliu. Huang, Lillian Meijin, Marie M. Yeh, Elizabeth Zeitoun, Anna H. Chang, & Joy J. Wu. 1998. ‘A typological overview of nominal case marking systems of some Formosan languages,’ in Selected papers from the second international symposium on languages in Taiwan, 21–​48. Taipei: Crane. Huarte, Malder. 2007. ‘The acquisition of Basque ergative case: An experimental study.’ ASJU 41–​2: 131–​144. Hunkin, Galumalemana Afeleti L. 2009. Gagana Sāmoa:  A  Samoan language coursebook. Honolulu: University of Hawai‘i Press. Hunt, Katherine. 1993. ‘Clause structure, agreement and case in Gitksan.’ PhD dissertation, University of British Columbia. Hunt, Ruskin H. & Kathleen M. Tomas. 2008. ‘Magnetic resonance imaging methods in developmental science: A primer.’ Development and Psychopathology 20: 1029–​1051. Hutt, Michael & Subedi, Abhi. 1999. Nepali. London: Teach Yourself Books. Hyams, Nina. 1986. Language acquisition and the theory of parameters. Dordrecht: Reidel. Hyams, Nina. 1989. ‘The null subject parameter in language acquisition,’ in K. Safir & O. Jaeggli (eds.), The null subject parameter. Dordrecht: Reidel, 215–​238. Hyslop, Gwendolyn. 2010. ‘Kurtop case: The pragmatic ergative and beyond.’ Linguistics of the Tibeto-​Burman Area 33(1): 1–​40. Iatridou, Sabine, Elena Anagnastopoulou, & Roumyana Izvorski. 2001. ‘Observations about the form and meaning of the perfect,’ Ken Hale: A life in language. Cambridge, MA: MIT Press, 189–​238. Iemmolo, Giorgio. 2010. ‘Topicality and differential object marking: Evidence from Romance and beyond.’ Studies in Language 34: 239–​272. Iggesen, Oliver A. 2009. ‘Asymmetry in case marking: nominal vs. pronominal systems,’ in A. Malchukov & A. Spencer (eds.), The Oxford handbook of case. Oxford University Press, 246–​260. Iluridze, Ketevan. 2006. ‘saxelta bruneba XIX sauk’unis I naxevris kartuli enis gramat’ik’ebši’ [‘Noun declension in Georgian grammars from the first half of the 19th century’]. PhD thesis, Arn. Chikobava Institute of Linguistics, Tbilisi. Imanishi, Yusuke. 2014. ‘Default ergative.’ PhD dissertation, MIT.

 

1186   References Imedadze, Natela & Kevin Tuite. 1992. ‘The acquisition of Georgian,’ in Dan Slobin (ed.), The crosslinguistic study of language acquisition, Vol. III. Hillsdale, NJ: Erlbaum, 39–​109. Indefrey, Peter. 2007. ‘Brain-​imaging studies of language production,’ in M. Gaskell (ed.), The Oxford handbook of psycholinguistics. Oxford: Oxford University Press, 547–​564. Ionin, Tania. 2006. ‘This is definitely specific: specificity and definiteness in article systems.’ Natural Language Semantics 14: 175–​234. Ippolito, Michaela. 2000. ‘Remarks on the argument structure of causatives.’ MS, MIT. Jacobs, Joachim 2001. ‘The dimensions of topic-​comment.’ Linguistics 39: 641–​681. Jacques, Guillaume. 2010. ‘The inverse in Japhug Rgyalrong.’ Language and Linguistics 11(1): 127–​157. Jackendoff, Ray. 1991. ‘Parts and boundaries.’ Cognition 41: 9–​45. Jacobsen, Thomas. 2000. ‘Characteristics of processing morphological structural and inherent case in language comprehension.’ PhD dissertation, Max Planck Institute of Cognitive Neuroscience Leipzig. Jacobsen, William H. 1972. ‘Nominative–​ergative syncretism in Basque.’ Anuario del Seminario de Filología Vasca Julio de Urquijo 6: 67–​109. Jaeger, T. Florian & Neal Snider. 2008. ‘Implicit learning and syntactic persistence: Surprisal and cumulativity,’ Cognitive science conference proceedings. Washington, DC, 1061–​1066. Jäger, Gerhard. 2007. ‘Evolutionary game theory and typology:  A  case study.’ Language, 74–​109. Jahani, Carina. 2015. ‘Complex predicates and the issue of transitivity: The case of Southern Balochi.’ In Iván Szántó (ed.), From Aṣl to Zā’id: Essays in honour of Éva M. Jeremiás. Piliscsaba: The Avicenna Institute of Middle Eastern Studies, 79–105. Jahani, Carina & Agnes Korn. 2009. ‘Balochi,’ in Gernot Windfuhr (ed.), The Iranian languages. London: Routledge, 634–​692. Jahani, Carina & Agnes Korn (eds.), 2003. The Baloch and their neighbours: Ethnic and linguistic contact in Balochistan in historical and modern times. Wiesbaden: Reichert, 75–​111. Jakobson, Roman. 1936. ‘Beitrag zur allgemeinen Kasuslehre: Gesamtbedeutungen der russischen Kasus,’ Travaux du cercle linguistique de Prague VI, 240–​299. [repr. in E. Hamp et al. (eds.), 1966. Readings in linguistics II, 51–​89] Jamison, Stephanie. 1976. ‘Functional ambiguity and syntactic change: The Sanskrit accusative,’ in Papers from the parasession on diachronic syntax, 12th regional meeting of the Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 126–​135. Jamison, Stephanie. 1979a. ‘The case of the agent in Indo-​European.’ Die Sprache 25: 129–​143. Jamison, Stephanie. 1979b. ‘Remarks on the expression of agency with the passive in Vedic and Indo-​European.’ Zeitschrift für vergleichende Sprachforschung 93: 196–​219. Jamison, Stephanie. 2000. ‘Lurching towards ergativity: Expressions of agency in the Niya documents.’ Bulletin of the School of Oriental and African Studies 63: 64–​80. Jamison, Stephanie & Michael Witzel. 2002. ‘Vedic Hinduism.’ http://​www.people.fas.harvard. edu/​∼witzel/​vedica.pdf Janashvili, M. G. 1906. Kartuli gramat’ik’a. T’pilisi: Elektrombech’davi Shroma. Jastrow, Otto. 1985. Laut-​Und Formenlehre Des Neuaramäischen Dialekts von Mīdin Im Ṭūr ʿAbdīn, vol. 3. Wiesbaden: Harrassowitz Jastrow, Otto. 1988. Der Neuaramäische Dialekt Von Hertevin (Provinz Siirt). Semitica viva Bd. 3. Wiesbaden: O. Harrassowitz Jastrow, Otto. 1994. Der Neuaramäische Dialekt Von Mlaḥsô. Semitica viva 14. Wiesbaden: Harrassowitz

 

References   1187 Jegerski, Jill. 2014. ‘Self-​paced reading,’ in J. Jegerski & B. Van Patten (eds.), Research methods in second language psycholinguistics. New York: Routledge, 20–​49. Jelinek, Eloise. 1984. ‘Empty categories, case, and configurationality.’ Natural Language and Linguistic Theory 2: 39–​76. Jelinek, Eloise. 1993. ‘Ergativity and argument type,’ in J.D. Bobaljik & C. Phillips (eds.), Papers on case and agreement 1, MIT Working Papers in Linguistics 18: 15–​42. Jenny, Mathias, & San San Hnin Tun. 2013. ‘Differential subject marking without ergativity: The case of colloquial Burmese.’ Studies in Language 4: 693–​735. Jensen, John T. & Alana Johns. 1989. ‘The morphosyntax of Eskimo causatives,’ in D.B. Gerdts & K. Michelson (eds.), Theoretical perspectives on Native American languages. Albany, NY: State University of New York Press, 209–​229. Job, Michael. 1985. ‘Ergativity in Lezgian,’ in Frans Plank (ed.), Relational typology. Berlin: Mouton, 159–​173. Johanson, Lars. 2009. ‘Case and contact linguistics,’ in Andrej L. Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 494–​501. Johns, Alana. 1987. ‘Transitivity and Grammatical Relations in Inuktitut.’ PhD thesis, University of Ottawa. Johns, Alana. 1992. ‘Deriving ergativity.’ Linguistic Inquiry 23: 57–​88. Johns, Alana. 1996. ‘The occasional absence of anaphoric agreement in Labrador Inuttut,’ in J. Black & V. Motapanyane (eds.), Micro-​parametric syntax and dialect variation, current issues in linguistic theory 139. Amsterdam & Philadelphia, PA: Benjamins, 121–​143. Johns, Alana. 2000. Ergativity: A perspective on recent work. The first GLOT international state-​ of-​the-​art book. Berlin: Mouton de Gruyter, 47–​73. Johns, Alana. 2001a. ‘Ergative to accusative:  Comparing evidence from Inuktitut.’ In J.T. Faarlund (ed.), Grammatical relations in change. Amsterdam & Philadelphia, PA: Benjamins, 205–​222. Johns, Alana. 2001b. ‘An inclination towards accusative.’ Linguistica Atlantica 23: 127–​144. Johns, Alana. 2006. ‘Ergativity and change in Inuktitut,’ in A. Johns, D. Massam, & J. Ndayiragije (eds.), Ergativity: Emerging issues, studies in natural language & linguistic theory. Dordrecht: Springer, 293–​311. Johns, Alana. 2007a. ‘Ergativity in Inuktitut: Varying views.’ Presented at the MIT Ergativity Seminar, MIT. Johns, Alana. 2007b. ‘Restricting noun incorporation: Root movement.’ Natural Language and Linguistic Theory 25(3): 535–​576. Johns, Alana. 2013. ‘Ergativity lives: Eastern Canadian Inuktitut and clitic doubling.’ Paper presented at the Canadian Linguistic Association, Victoria, BC. Johns, Alana. [in press]. ‘Anaphoric arguments in Unangax and Eastern Canadian Inuktitut,’ Studies in Inuit linguistics: In honor of Michael Fortescue. Fairbanks: Alaska Native Language Center, University of Alaska, 91–103. Johns, Alana, Diane Massam, & Juvenal Ndayiragije. (eds.). 2006. Ergativity: Emerging issues. Dordrecht & Berlin: Springer. Johnson, Kyle. 1991. ‘Object positions.’ Natural Language & Linguistic Theory 9: 577–​636. Johnson, Rodney Charles. 1992. ‘The Limits of grammar:  Syntax and lexicon in spoken Burmese.’ PhD, thesis, University of Michigan. Joswig, Andreas. 2011. ‘A brief grammar of the Majang language.’ Unpublished MS. Jügel, Thomas. 2015. Die Entwicklung der Ergativkonstruktion im Alt- und Mitteliranischen − eine Korpusbasierte Untersuchung zu Kasus, Kongruenz und Satzbau. Wiesbaden: Harrassowitz.

 

1188   References Jukes, Anthony. 2013. ‘Voice, valence and focus in Makassarese.’ Proceedings of the workshop on Indonesian-​type voice systems, 54, NUSA. TUFS, 101–​121. Just, Marcel A. & Patricia A. Carpenter. 1980. ‘A theory of reading: From eye fixations to comprehension.’ Psychological Review 87: 329–​354. Kaan, Edith. 2007. ‘Event-​related potentials and language processing:  A  brief overview.’ Language and Linguistics Compass 1(6): 571–​591. Kaan, Edith & Tamara Y. Swaab. 2002. ‘The brain circuitry of syntactic comprehension.’ Trends in Cognitive Sciences. Kachru, Yamuna. 1978. ‘On ergativity in selected South Asian languages.’ Studies in the Linguistic Sciences 8(1): 111–​127. Kachru, Yamuna. 1981. ‘Transitivity and volitionality in Hindi.’ Studies in the Linguistic Sciences 11: 181–​193. Kachru, Yamuna. 1987. ‘Ergativity, subjecthood and topicality and Hindi–​ Urdu.’ Lingua 71: 223–​238. Kachru, Yamuna & Rajeshwari Pandharipande. 1978. ‘Ergativity in selected South Asian languages.’ Studies in the Linguistic Sciences 8: 111–​127. Kaldani, Maksime. 1978. ‘Aorist’is c’armoeba svanurši’ [‘Formation of the aorist in Svan’]. Ibero-​ k’avkasiuri enatmecniereba 20: 150–​161. Kalin, Laura. 2014. ‘Aspect and argument licensing in Neo-​Aramaic.’ PhD dissertation, UCLA. Kalin, Laura & Coppe van Urk. 2015. ‘Aspect splits without ergativity: Agreement asymmetries in Neo-​Aramaic.’ Natural Language & Linguistic Theory 33: 659–702. Kalmár, Ivan. 1979. Case and context in Inuktitut (Eskimo). Ottawa, Ontario: National Museums of Canada. Karimi, Simin, Vida Samiian, & Donald Stilo (eds.). 2008. Aspects of Iranian linguistics. Newcastle: Cambridge Scholars Publishing. Karimi, Yadgar. 2010. ‘Unaccusative transitives and the Person-Case Constraint effects in Kurdish.’ Lingua 120(3): 693–716. K’art’ozia, Guram, Rusudan Gersamia, Maia Lomia, & Taia Cxadaia. 2010. Megrulis lingvist’uri analizi. Tbilisi: Meridiani. Kaufman, Daniel. 2005. ‘Aspects of pragmatic focus in Tagalog,’ in I. Wayan Arka & Malcolm Ross (eds.), The many faces of Austronesian voice systems: Some new empirical studies (Pacific Linguistics, 57). Canberra: Research School of Pacific and Asian Studies, ANU, 175–​196. Kaufman, Daniel. 2008. ‘South Sulawesi pronominal clitics:  Form, function and position.’ Studies in Philippine Languages and Cultures 10-​ICAL Pronoun Papers: 13–​65. Kaufman, Daniel. 2009a. ‘Austronesian nominalism and its consequences:  A  Tagalog case study.’ Theoretical Linguistics 35: 1–​49. Kaufman, Daniel. 2009b. ‘Austronesian typology and the nominalist hypothesis,’ in Alexander Adelaar & Andrew Pawley (eds.), Austronesian historical linguistics and culture history: A festschrift for Robert Blust (Pacific Linguistics). Canberra: ANU Press. Kaufman, Daniel. 2009c. ‘On the Proto-​Austronesian morphemes *pa-​, * and *.’ Paper presented at ICAL XI, Aussois. Kaufman, Daniel. 2011. ‘Exclamatives and temporal nominalizations in Austronesian,’ in Foong Ha Yap & Janick Wrona (eds.), Nominalization in Asian languages: Diachronic and typological perspectives (Typological Studies in Language). Amsterdam & Philadelphia, PA: Benjamins. Kaufman, Daniel. 2012. ‘Pan *ka-​and predicate classes.’ Paper presented at AFLA 19, Academica Sinica, Taiwan. Kaufman, Daniel. [forthcoming]. ‘Response to commentators.’ Theoretical Linguistics.

 

References   1189 Kaufman, Terrence. 1976a. ‘Archaeological and linguistic correlations in Mayaland and associated areas of Meso-​America.’ World Archaeology 8: 101–​118. Kaufman, Terrence. 1976b. Proyecto de Alfabetos y Ortografías para Escribir las Lenguas Mayances. Guatemala: Ministerio de Educación. Kaufman, Terrence. 1990a. ‘Algunos rasgos estructurales de los idiomas Mayances,’ in N.C. England & S.R. Elliott (eds.), Lecturas sobre la lingüística Maya. Antigua, Guatemala: CIRMA, 59–​114. Kaufman, Terrence. 1990b. ‘Language history in South America: What we know and how to know more,’ in Doris Payne (ed.), Amazonian linguistics: Studies in lowland South American languages. Austin: University of Texas Press, 13–​73. Kaufman, Terrence & William M. Norman. 1984. ‘An outline of Proto-​Cholan phonology, morphology and vocabulary,’ in John. S. Justeson & Lyle R. Campbell (eds.), Phoneticism in Mayan hieroglyphic writing. Albany, NY: State University of New York, 77–​166. Kayne, Richard. 1975. French syntax. Cambridge, MA: MIT Press. Kayne, Richard. 1984. Connectedness and binary branching. Dordrecht: Foris. Kayne, Richard. 1993. ‘Toward a modular theory of auxiliary selection.’ Studia linguistica 47(1): 3–​31. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Kayne, Richard. 2000a, ‘Past participle agreement in French and Italian,’ in R.S. Kayne (ed.), Parameters and universals. New York: Oxford University Press, 10–​24. Kayne, Richard. 2000b. ‘A note on prepositions, complementizers and word order universals,’ in R.S. Kayne, Parameters and universals. New York: Oxford University Press, 314–​326. Kayne, Richard. 2010. ‘Why are there no directionality parameters?’ in M. Byram Washburn, K. McKinney-​Bock, E. Varis, A. Sawyer & B. Tomaszewicz (eds.), Proceedings of the 28th west coast conference on formal linguistics. Somerville, MA:  Cascadilla Proceedings Project, 1–​23. Kazenin, Konstantin. 1998. ‘On patient demotion in Lak,’ in Leonid Kulikov & Heinz Vater (eds.), Typology of verbal categories. Tübingen: Niemeyer, 95–​115. Kazenin, Konstantin. 2001a. ‘Deepričastnye konstrukcii,’ in Aleksandr E. Kibrik (ed.), 550–​594. Kazenin, Konstantin. 2001b. ‘The passive voice,’ in Martin Haspelmath et al. (eds.), Language typology and language universals: An international handbook. Berlin: de Gruyter, 899–​916. Kazenin, Konstantin I. 2001c. ‘Umen’šenie perexodnosti:  binominativnaja konstrukcija,’ in Aleksandr E. Kibrik (ed.), Bagvalinskij jazyk. Moscow: Nasledie, 394–​399 Kazenin, Konstantin. 2002. ‘Focus in Daghestanian and word order typology.’ Linguistic Typology 6: 289–​316. Kazenin, Konstantin. 2013. Sintaksis sovremennogo lakskogo jazyka. Makhachkala: Aleph. Kazenin, Konstantin & Jakov G. Testelec. 1999. ‘Porjadok slov i struktura sostavljajushchix,’ in Aleksandr E. Kibrik (ed.), Elementy caxurskogo jazyka v tipologishseskom osvjeshchenii. Moscow: Nasledie. Kazenin, Konstantin & Yakov Testelec. 2004. ‘Where coordination meets subordination:  Converb constructions in Tsakhur (Daghestanian),’ in Martin Haspelmath (ed.), Coordinating constructions. Amsterdam & Philadelphia, PA: Benjamins, 227–​239. Keen, Sandra. 1983. ‘Yukulta,’ in Robert M.W. Dixon & Barry J. Blake (eds.), Handbook of Australian languages 3. Canberra: Australian National University Press, 190–​304. Keenan, Edward L. 1972. ‘Relative clause formation in Malagasy,’ Paul Peranteau, Judith Levi, & Gloria Phares (eds.), The Chicago which hunt. Chicago: Chicago Linguistics Association, University of Chicago, IL, 169–​190.

 

1190   References Keenan, Edward L. 1976. ‘Towards a universal definition of subject,’ in C. Li (ed.), Subject and topic. New York: Academic Press, 305–​333. Keenan, Edward. 2000. ‘Morphology is structure:  A  Malagasy test case,’ in Ileana Paul, Vivianne Phillips, & Lisa Travis (eds.), Formal issues in Austronesian linguistics. Dordrecht: Kluwer, 27–​48. Keenan, Edward L. & Bernard Comrie. 1977. ‘Noun phrase accessibility and universal grammar.’ Linguistic Inquiry, 63–​99. Keine, Stefan. 2010. Case and agreement from fringe to core: A minimalist approach, vol. 536. Berlin: Walter de Gruyter. Keine, Stefan. 2012. ‘On the role of movement in Hindi/​Urdu long-​distance agreement,’ in S.K.S. Sloggett (ed.), Proceedings of NELS 42. Amherst, MA: GLSA. Keine, Stefan & Gereon Müller. 2011. ‘Non-​zero/​non-​zero alternations in differential object marking,’ in S. Lima, K. Mullin, & B. Smith (eds.), Proceedings of the 39th meeting of the NELS. Amherst, MA: GLSA, 441–​454. Keine, Stefan & Gereon Müller. 2015. ‘Differential argument encoding by impoverishment,’ in Ina Bornkessel-​Schlesewsky, Andrej Malchukov, & Marc Richards (eds.), Scales and hierarchies: A cross-​disciplinary perspective. Berlin: de Gruyter, 275–​297. Keller, Rudi. 1994. On language change: The invisible hand in language. London: Routledge. Kellogg, Samuel Henry 1893. Grammar of the Hindi language. Delhi: Munshiram Manoharlal. [2nd edn., 1990] Kennedy, Rod. 1984. ‘Semantic roles: The language speaker’s categories (in Kala Lagaw Ya),’ in Kathleen Glasgow et  al. (eds.), Papers in Australian linguistics, 16. Canberra:  Pacific Linguistics, 153–​170. Kent, Roland. 1953. Old Persian. New Haven, CT: American Oriental Society. [2nd rev. edn.] Khalilova, Zaira. 2009. A grammar of Khwarshi. Utrecht: LOT. Khan, Geoffrey. 2004. ‘Aramaic and the impact of languages in contact with it through the ages,’ in P. Bádenas de la Peña, S.T. Tovar, E.R. Luján, & M.A. Gallego (eds.), Lenguas En Contacto:  El Testimonio Escrito (Manuales y anejos de ‘Emerita’ 46). Madrid:  Consejo Superior de Investigaciones Científicas, 87–​108. Khan, Geoffrey. 2007a. ‘Ergativity in the north eastern Neo-​Aramaic dialects,’ in E. Cohen & Bar (eds.), Studies in Semitic and general linguistics in honor of Gideon Goldenberg. Münster: Ugarit-​Verlag, 147–​157. Khan, Geoffrey. 2007b. ‘North Eastern Neo-​Aramaic,’ in Y. Matras & J. Sakel (eds.), Grammatical borrowing in cross-​ linguistic perspective. Berlin:  Mouton de Gruyter, 197–​214 Khan, Geoffrey. 2008. The Neo-​Aramaic dialect of Barwar. Leiden: Brill Khan, Geoffrey. 2009. The Jewish Neo-​Aramaic dialect of Sanandaj. Piscataway, NJ: Gorgias Press. Khan, Geoffrey. [to appear]. ‘Jewish Neo-​Aramaic,’ in S. Benor and B. Hary (eds.), Handbook of Jewish languages. Khan, Geoffrey. [forthcoming]. Linguistic studies in North-Eastern Neo-Aramaic. Khan, Tafseer Ahmed. 2009. ‘Spatial expressions and case in South Asian languages.’ PhD dissertation, University of Konstanz. Khorounjaia, Ekaterina & Liliana Tolchinsky. 2004. ‘Discursive constraints on the lexical realization of arguments,’ in Ruth A. Berman (ed.), Language development across childhood and adolescence. Amsterdam & Philadelphia, PA: Benjamins, 83–​109. Kibrik, Aleksandr E. 1979. ‘Canonical ergativity and Daghestan languages,’ in Frans Plank (ed.), Ergativity. London: Academic Press, 61–​77.

 

References   1191 Kibrik, Aleksandr E. 1985. ‘Toward a typology of ergativity,’ in Johanna Nichols & A.C. Woodbury (eds.), Grammar inside and outside the clause: Some approaches to theory from the field. Cambridge: Cambridge University Press, 268–​323. Kibrik, Aleksandr E. 1994. ‘Khinalugh,’ in Rieks Smeets (ed.), The indigenous languages of the Caucasus, volume 4. Delmar: Caravan, 367–​406. Kibrik, Aleksandr E. 1997. ‘Beyond subject and object: Towards a comprehensive relational typology.’ Linguistic Typology 1: 279–​346. Kibrik, Aleksandr E. (ed.). 1999a. Èlementy caxurskogo jazyka v tipologičeskom osveščenii. Moscow: Nasledie. Kibrik, Aleksandr E. 1999b. ‘Soglasovanie,’ in Aleksandr E. Kibrik (ed.), Elementy caxurskogo jazyka v tipologičeskom osveščenii. Moscow: Nasledie, 354–​376. Kibrik, Aleksandr E. (ed.), 2001a. Bagvalinskij jazyk. Moscow: Nasledie. Kibrik, Aleksandr E. (ed.). 2001a. Bagvalinskij jazyk. Moscow: Nasledie. Kibrik, Aleksandr E. 2001b. ‘Subject–​oriented vs. subjectless languages,’ in Martin Haspelmath (ed.), Language typology and language universals:  An international handbook. Berlin:  de Gruyter, 1413–​1423. Kibrik, Aleksandr E. 2003. Konstanty i peremennye jazyka. St. Petersburg: Aletheia. Kibrik, Aleksandr E. 2007. ‘Principy i strategii klauzal’nogo sočinenija v dagestanskix jazykax.’ Voprosy Jazykoznanija 3: 78–​120. Kibrik, Andrej A. 1996. ‘Transitivity in Godoberi lexicon and grammar,’ in Aleksandr E. Kibrik (ed.), Godoberi. LINCOM Europa, 107–146. Kibrik, Andrej. 2011. Reference in discourse. Oxford: Oxford University Press. Kikusawa, Ritsuko. 2002. Proto central Pacific ergativity: Its reconstruction and development in the Fijian, Rotuman and Polynesian languages. Canberra: Pacific Linguistics. Kikusawa, Ritsuko. 2003a. ‘The development of some Indonesian pronominal systems,’ in Barry J. Blake & Burridge Kate (with Jo Taylor) (eds.), Historical linguistics 2001: Selected papers from the 15th international conference on historical linguistics, Melbourne, 13–​17 August 2001. Amsterdam & Philadelphia, PA: Benjamins, 237–​269. Kikusawa, Ritsuko. 2003b. ‘A new view of the Proto-​Oceanic pronominal system.’ Oceanic Linguistics 42(1): 161–​186. Kikusawa, Ritsuko. 2008. ‘Transitivity in the analysis of Northern Betsimisaraka Malagasy.’ Studies in Philippine Languages and Cultures 18: 44–​64. Kikusawa, Ritsuko. 2012. ‘Verb first or noun first? Putting “Nominalization” into a historical context.’ Paper presented at Linguistic Typology Research Group meeting, Institute of Linguistics, Academia Sinica, Taiwan (December 17). Kikusawa, Ritsuko. 2015. ‘The Austronesian language family,’ in Claire Bowern & Bethwyn Evans (eds.), The Routledge handbook of historical linguistics. London:  Routledge, 657–​674. Kikusawa, Ritsuko. [to appear, a]. ‘A diachronic typology of applicative verbs in western Austronesian languages: Toward a comparison and reconstruction (synopsis),’ in Leonid Kulikov & Seppo Kittila (eds.), Diachronic typology of voice and valency-​changing categories. Kikusawa, Ritsuko. [to appear, b]. ‘Typological generalisations and diachronic analyses: Actancy systems in Austronesian languages.’ Historical Linguistics in Japan 4. Kikusawa, Ritsuko. [to appear, c]. ‘Conducting syntactic reconstruction of languages with no written records,’ in Eugenio Luján, Jóhanna Barðdal, & Spike Guildea (eds.), Syntactic reconstruction: Applying the comparative method. Amsterdam: Brill. Kilham, Chirstine A. 1977. Thematic organisation of Wik-Munkan discourse. Canberra: Pacific Linguistics.

 

1192   References Kilian-​Hatz, Christa. (2008). A grammar of Modern Khwe (Central Khoisan). Cologne: Rüdiger Köppe. Kilian-​Hatz, Christa. (2013). ‘Syntax:  Kxoe subgroup:  Khwe’, in Rainer Vossen (ed.), The Khoesan languages. London & New York: Routledge, 356–​378. Killian, Don. 2015. ‘Complex verbal predicates in Uduk,’ in Angelika Mietzner & Anne Storch (eds.), Nilo-Saharan: Models and Descriptions. Cologne: Rüdiger Köppe, 177–197. Kim, Karl H.S., Norman R. Relkin, Kyoung-​Min Lee, & Joy Hirsch. 1997. ‘Distinct cortical areas associated with native and second languages.’ Nature 388: 171–​174. Kimball, George D. 1991. Koasati Grammar. Lincoln, NE: University of Nebraska Press. King, Jonathan W. & Marta Kutas. 1995. ‘Who did what and when? Using word-​and clause-​ level ERPs to monitor working memory usage in reading.’ Journal of Cognitive Neuroscience 7(3): 376–​95. Kinyalolo, Kasangati. 1991. ‘Syntactic dependencies and the SPEC–​head agreement hypothesis in KiLega.’ PhD disseration, UCLA. Kiparsky, Paul. 1973. ‘ “Elsewhere” in phonology,’ in Stephen Anderson & Paul Kiparsky (eds.), A Festschrift for Morris Halle. New York: Holt, Rinehart, & Winston, 93–​106. Kiparsky, Paul. 1982a. ‘Word formation and the lexicon,’ in F. Ingeman (ed.), Proceedings of the 1982 Mid-​America linguistic conference. Kiparsky, Paul. 1982b. Explanation in phonology. Dordrecht: Foris. Kiparsky, Paul. 1998a. ‘Partitive case and aspect,’ in Miriam Butt & William Geuder (eds.), Projecting from the lexicon. Stanford, CA: CSLI. Kiparsky, Paul. 1998b. ‘Aspect and event structure in Vedic.’ Yearbook of South Asian Languages and Linguistics 1: 29–​62. Kiparsky, Paul. 1999. ‘Analogy and OT: morphological change as emergence of the unmarked.’ MS, Stanford University. Kiparsky, Paul. 2001. ‘Structural case in Finnish.’ Lingua 111: 315–​376. Kiss, Katalin É. 1995. Discourse configurational languages. Oxford: Oxford University Press. Kittilä, Seppo. 2002. Transitivity: Towards a comprehensive typology (University of Turku publications in general linguistics, 5). Turku: Åbo Akademis Tryckeri. Kittilä, Seppo. 2005. ‘A Typology of involuntary agent Constructions.’ Word 56(3): 381–​419. Kiyosawa, Kaoru, & Donna Gerdts. 2010. Salish applicatives (Brill’s Studies in the Indigenous Languages of the Americas, vol. 1). Leiden: Brill. K’iziria, Ant’on. 1974. ‘Kvemdebare–​šemasmenlis urtiertoba kartuli enis dasavluri dialekt’ebis mixedvit’ [‘The subject–​predicate relation in the western Georgian dialects’]. Kartvelur enata st’rukt’uris sak’itxebi 4: 75–​91. K’iziria, Ant’on. 1982. Mart’ivi c’inadadebis šedgeniloba kartvelur enebši [‘The structure of the simple sentence in the Kartvelian languages’]. Tbilisi: Mecniereba. Klaiman, Miriam H. 1978. ‘Arguments against a passive origin of the IA ergative,’ Chicago Linguistic Society: Papers from the 14th regional meeting, 204–​216. Klaiman, M.H. 1980. ‘Bengali dative subjects.’ Lingua 51: 275–​295 Klaiman, M.H. 1987. ‘Mechanisms of ergativity in South Asia.’ Lingua 61–​102. Klein, Wolfgang. 1994. Time in language. New York: Routledge. Klimov, Georgiĭ Andreevich. 1960. ‘Kratkaja xarakteristiska sklonenija v zanskom jazyke,’ in V.M. Zhirmunskij (ed.), Voprosy grammatiki. Moscow:  Izadatel’stvo Akademija Nauk SSSR, 71–​80. Klimov, Georgiĭ Andreevich. 1962. Sklonenie v kartvel’skix jazykax v sravitel’no-​istoričeskom aspekte. Moscow: Nauka. Klimov, Georgiĭ Andreevich. 1964. Etimologičeskij slovar’ kartvel’skix jazykov. Moscow: Nauka.

 

References   1193 Klimov, Georgiĭ Andreevich. 1976. ‘Anomalii ergativnosti v lazskom (čanskom) jazike.’ Philo­ logia Orientalis 4: 150–​159. Klimov, Georgiĭ Andreevich. 1977. Tipologija jazykov aktivnogo stroja. Moscow: Nauka. Klimov, Georgiĭ Andreevich. 1979. ‘O donominativnom komponente kartvel’skoj jazykovoj struktury,’ in Sh. Dzidziguri (ed.), Saenatmecniero k’rebuli [‘Studies in linguistics’]. Tbilisi: Mecniereba, 130–​145. Klokeid, Terry J. 1976a. ‘Lardil,’ in R.M.W. Dixon (ed.), 550–​584. Klokeid, Terry J. 1976b. ‘Topics in Lardil grammar.’ PhD dissertation, MIT. Klokeid, Terry J. 1978. ‘Nominal inflection in Pama-Nyungan: A case study in relational grammar,’ in Werner Abraham (ed.), Valence, semantic case and grammatical relations. Amsterdam: Benjamins, 577–615. Kluender, Robert. 1998. ‘On the distinction between strong and weak islands: A processing perspective,’ in P.W. Culicover & L. McNally (eds.), The limits of syntax. New York: Academic Press, 241–​280. Koenig, Jean-​Pierre & Nuttanart Muansuwan. 2001. ‘How to end without ever finishing.’ Journal of Semantics 17: 147–​184. Kohn, Susan E. & Ana Cragnolino. 2003. ‘The role of preferred argument structure for understanding aphasic sentence planning,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure:  Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 339–​351. Kohnen, Bernardo. 1933. Shilluk grammar. Verona: Nigrizia School Press. Kolliakou, Dimitra. 2003. Nominal constructions in Modern Greek: Implications for the architecture of grammar. Stanford: CSLI Publications. König, Christa. 2006. ‘Marked nominative in Africa.’ Studies in Language 30(4): 655–​732. König, Christa. 2008a. ‘The marked-​nominative languages of eastern Africa,’ in Bernd Heine & Derek Nurse (eds.), A linguistic geography of Africa (Cambridge Approaches to Language Contact). Cambridge: Cambridge University Press, 251–​271. König, Christa. 2008b. Case in Africa. Oxford: Oxford University Press. König, Christa. 2009. ‘Marked nominatives,’ in Andrej L. Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 535–​548. König, Christa. 2011. ‘The grammaticalization of adpositions and case marking,’ in Heiko Narrog & Bernd Heine (eds.), The Oxford handbook of grammaticalization. Oxford: Oxford University Press, 511–​521. König, Christa. 2012. ‘On the rise of ergative structures in Africa,’ in Osamu Hieda (ed.), Studies in Nilotic linguistics, 5. Tokyo: Research Institute for Languages and Cultures of Asia and Africa, 27–​45. Konnerth, Linda. 2014. ‘Additive focus and additional functions in Karbi (Tibeto-​Burman) =tā’,’ in Kayla Carpenter, Oana David, Florian Lionnet, Christine Sheil, Tammy Start, & Vivian Wauters (eds.), Proceedings of the 38th annual meetings of the Berkeley Linguistics Society, 206–​222. Accessed from http://​escholarship.org/​uc/​item/​08v8j4qm Koopman, Hilda & Dominique Sportiche. 1991. ‘On the position of subjects,’ Lingua 85: 211–​258. Korn, Agnes. 2008. ‘Marking of arguments in Balochi ergative and mixed constructions,’ in Karimi et al. (eds.), Aspects of Iranian linguistics. Newcastle: Cambridge Scholars Publishing, 249–​276. Korn, Agnes. 2009. ‘The ergative system of Balochi from a typological perspective.’ Iranian Journal of Applied Language Studies 1(1): 43–​79. Kornfilt, Jaklin. 1984. ‘Case marking, agreement, and empty categories in Turkish.’ PhD dissertation, Harvard.

 

1194   References Kornfilt, Jaklin. 2003. ‘Subject case in Turkish nominalized clauses,’ in Uwe Junghanns & Luka Szucsich (eds.), Syntactic structures and morphological information. Berlin & New York: Mouton de Gruyter, 129–​215 Kornfilt, Jaklin. 2006. ‘Agreement: The (unique and local) syntactic and morphological licenser of subject case,’ in Joao Costa & Maria Cristina Figueiredo Silva (eds.), Studies on agreement. Amsterdam & Philadelphia, PA: Benjamins, 140–​171. Koptjevskaja-​Tamm, Maria. 1993. Nominalizations. London: Routledge. Koul, Omkar N. & Kashi Wali. 2006. Modern Kashmiri grammar. Springfield, VA: Dunwoody Press. Kozinsky, Isaac, Vladmir Nedjalkov, & Maria Polinskaja. 1988. ‘Antipassive in Chukchee,’ in Masayoshi Shibatani (ed.), Passive and voice. Amsterdam & Philadelphia, PA: Benjamins, 651–​706. Kramer, Ruth. 2014a. ‘Clitic doubling or object agreement: The view from Amharic.’ Natural Language and Linguistic Theory 32: 593–​634. Kramer, Ruth. 2014b. ‘Gender in Amharic: A morpho-​syntactic approach to natural and grammatical gender.’ Language Sciences 43: 102–​115. Kratzer, Angelika. 1995. ‘Stage-level and individual-level predicates,’ in Gregory N. Carlson & Francis Jeffry Pelletier (eds.), The generic book. Chicago, IL: University of Chicago Press, 125–​175. Kratzer, Angelika. 1996. ‘Severing the external argument from its verb,’ in J. Rooryck & L. Zaring (eds.), Phrase structure and the lexicon. Dordrecht, Kluwer, 109–​137. Krause, Cornelia. 2001. ‘On reduced relatives with genitive subjects.’ PhD dissertation, MIT. Krifka, Manfred. 1998. ‘The origins of telicity,’ in Susan Rothstein (ed.), Events and grammar. Dordrecht: Kluwer, 197–​236. Krifka, Manfred. 2006. ‘Basic notions of information structure,’ in C. Féry, G. Fanselow, & M. Krifka (eds.), The notions of information structure. Working papers of the SFB 632, Interdisciplinary studies on information structure (ISIS) 6. Potsdam:  Universitätsverlag Potsdam, 13–​56. Krifka, Manfred. 2013. ‘Response particles as propositional anaphors.’ Proceedings of SALT 23: 1–​18. Kroeger, Paul. 1993. Phrase structure and grammatical relations in Tagalog. Stanford University, CA: Center for the Study of Language and Information. Kruspe, Nicole. 1999. ‘A grammar of Semelai.’ PhD, University of Melbourne. Kučerová, Ivona. 2007. ‘The syntax of givenness.’ PhD dissertation, MIT. Kučerová, Ivona & Ad Neeleman. 2012. ‘Introduction,’ Contrasts and positions in information structure. Cambridge: Cambridge University Press, 1–​23. Kulikov, Leonid. 2001. ‘The Vedic -​ya-​presents.’ PhD dissertation, University of Leiden. Kulikov, Leonid. 2006. ‘Case systems in a diachronic perspective:  A  typological sketch,’ in Leonid Kulikov, Andrej Malchukov, & Peter de Swart (eds.), Case, valency and transitivity. Proceedings of the workshop held at University of Nijmegen, June 17–​19 2003 (Studies in Language Companion Series). Amsterdam & Philadelphia, PA: Benjamins. Kulikov, Leonid. 2009. ‘Evolution of case systems,’ in Andrej L. Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 439–​457. Kumagai, Yoshiharu. 2001. ‘Preferred argument structure and discourse ergativity in English.’ Mulberry:  Bulletin of the Department of English, Faculty of Letters, Aichi Prefectural University 50: 77–​90. Kumagai, Yoshiharu. 2006. ‘Information management in intransitive subjects: Some implications for the preferred argument structure theory.’ Journal of Pragmatics 38(5): 670–​694.

 

References   1195 Kumakhov, Mukhadin & Karina Vamling. 2009. Circassian clause structure. Malmö: Malmö University Department of International Migration and Ethnic Relations. Kumakhov, Mukhadin. 1971. Slovoizmenenie Adygskix Jazykov [‘Inflectional system of the Adygh languages’]. Moscow: Nauka. Kumakhov, Mukhadin. 1989. Sravnitel’no-​istoričeskaja Grammatika Adygsikix (čerkesskix) Jazykov [‘A comparative historical grammar of the Adygh (Circassian) languages’]. Moscow: Nauka. Kuperberg, Gina R. 2007. ‘Neural mechanisms of language comprehension: Challenges to syntax.’ Brain research 1146: 23–​49. Kuperberg, Gina R., Phillip J. Holcomb, Tatiana Sitnikova, Douglas Greve, Anders M. Dale, & David Caplan. 2003. ‘Distinct patterns of neural modulation during the processing of conceptual and syntactic anomalies.’ Journal of Cognitive Neuroscience 15(2): 272–​293. Kurebito, Tokusu. 2012. ‘An outline of valency-​reducing operations in Chukchi,’ in Wataru Nakamura & Ritsuko Kikusawa (eds.), Objectivization and subjectivization: A typology of voice systems. Suita, Osaka, Japan: National Museum of Ethnology, 177–​189. Kuroda, Shige-​Yuki. 1972. The categorical and the thetic judgement: Evidence from Japanese syntax. Foundations of Language 9: 153–​185. Kuryłowicz, Jerzy. 1975. ‘The evolution of grammatical categories.’ Esquisses linguistiques 2: 38–​54. Kutas, Marta & Kara Federmeier. 2000. ‘Electrophysiology reveals semantic memory use in language comprehension.’ Trends in Cognitive Sciences 4(12): 463–​470. Kutas, Marta & Kara Federmeier. 2007. ‘Event-​related brain potential (ERP) studies of sentence processing,’ in M. G. Gaskell (ed.), The Oxford handbook of psycholinguistics. Oxford: Oxford University Press, 385–​406. Kutas, Marta & Steven A. Hillyard. 1980. ‘Reading senseless sentences brain potentials reflect semantic incongruity.’ Science 207(4427): 203–​205. Kutas, Marta, Cyma K. Van Petten, & Robert Kluender. 2006. ‘Psycholinguistics electrified II 1994–​2005,’ in M.J. Traxler & M.A. Gernsbacher (eds.), Handbook of psycholinguistics. Amsterdam: Elsevier, 659–​724. Kutscher, Silvia. 2008. ‘Argument realization of psych-​verbs in an active language: The case of Laz (Ardeşen variety),’ in Syntax of the World’s Languages III. Berlin: Freie Universität. Kutscher, Silvia. 2009. Kausalität und Argumentrealisierung. Zur Konstruktionsvarianz bei Psychverben am Beispiel europäischer Sprachen. Tübingen: Niemeyer. Kwon, Nayoung, Peter C. Gordon, Yoonhyung Lee, Robert Kluender, & Maria Polinsky. 2010. ‘Cognitive and linguistic factors affecting subject/​object asymmetry: An eye-​tracking studs of prenominal relative clauses in Korean.’ Language 86: 546–​582. Kwon, Nayoung, Robert Kluender, & Maria Polinsky. 2006. ‘Subject preference in Korean,’ in D. Baumer & M. Scanlon (eds.), Proceedings of the 25th west coast conference on formal linguistics. Somerville, MA: Cascadilla Press, 1–​14. Kwon, Nayoung, Robert Kluender, Marta Kutas, & Maria Polinsky. 2013. ‘Subject/​object processing asymmetries in Korean relative clauses:  Evidence from ERP data.’ Language 89(3): 537–​585. Lacroix, René. 2009. ‘Description du dialecte laze d’Arhavi. Grammaire et textes.’ PhD dissertation, Université Lumière Lyon II. Ladusaw, William A. 2000. ‘Thetic and categorical, stage and individual, weak and strong,’ in Laurence R. Horn & Yasuhiko Kato (eds.), Negation and polarity: Syntactic and semantic perspectives. New York: Oxford University Press, 232–​242. Lafitte, Pierra. 1978 [1962]. Grammaire basque (Navarro–​Labourdin littéraire). Donostia: Elkar.

 

1196   References Lahne, Antje. 2008. ‘Excluding SVO in ergative languages.’ Linguistische Arbeitsberichte. Leipzig: University of Leipzig, 65–​80. Laka, Itziar. 1988. ‘Configurational heads in inflectional morphology:  The structure of the inflected forms in Basque.’ Anuario del Seminario de Filología Vasca Julio de Urquijo 22(2): 343–​365. Laka, Itziar. 1993a. ‘The structure of inflection:  A  case study in X Syntax,’ in J.I. Hualde & J. Ortiz de Urbina (eds.), Generative studies in Basque linguistics. Amsterdam & Philadelphia, PA: Benjamins, 21–​70. Laka, Itziar. 1993b. ‘Unergatives that Assign Ergative, Unaccusatives that Assign Accusative,’ in Jonathan Bobaljik & Colin Philipps (eds.), Papers on case and agreement I: Working Papers in Linguistics, 18. Cambridge, MA: MIT, 149–​172. Laka, Itziar. 1996. A brief grammar of Euskara, the Basque language: Open-access grammar. Vitoria-​ Gasteiz: Euskal Herriko Unibertsitatea (University of the Basque Country). http://​www.ei.ehu. es/​p289-​content/​eu/​contenidos/​informacion/​grammar_​euskara/​en_​doc/​index.html Laka, Itziar. 2000. ‘Thetablind case: Burzio’s generalization and its image in the mirror,’ in E. Reuland (ed.), Arguments and case:  Explaining Burzio’s generalization. Amsterdam & Philadelphia, PA: Benjamins, 103–​129. Laka, Itziar. 2006a. ‘Deriving Split-​ergativity in the progressive:  The case of Basque,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity:  Emerging issues. Dordrecht: Springer, 173–​195. Laka, Itziar. 2006b. ‘On the nature of case in Basque: Structural or inherent?’ in Hans Broekhuis, Norbert Corver, Jan Koster, Riny Huybregts, & Ursula Kleinhenz (eds.), Organizing grammar:  Linguistic studies in honor of Henk van Riemsdijk. Berlin & New  York:  Mouton de Gruyter, 374–​382. Laka, Itziar. 2012. ‘Merging from the temporal input: On subject–​object asymmetries and an ergative language,’ in R. Bervick & M. Piatelli-​Palmarini (eds.), Rich grammars from poor inputs. Oxford: Oxford University Press, 1–​21 Laka, Itziar & Lore Erriondo-​Korostola. 2001. ‘Aphasia manifestations in Basque.’ Journal of Neurolinguisitics 14(2–​4) (April–​Oct.): 133–​157. Laka, Itziar & Beatriz Fernández (eds.). 2012. ‘Accounting for ergativity.’ Lingua 122(3) (special issue). Laka, Itziar & Kepa Erdocia. 2012. ‘Linearization preferences given free word order: Subject preferences given ergativity: A look at Basque,’ in E. Torrego (ed.), Of grammar, words and verses. Amsterdam & Philadelphia, PA: Benjamins, 115–​142. Lakarra, Joseba Andoni. 2005. ‘Prolegómenos a la reconstrucción de segundo grado y al análisis del cambio tipológico en (proto)vasco.’ Paleohispánica 5: 407–​470. Lam, Zoe Wai Man. 2014. ‘A complex ForceP for speaker-​and addressee-​oriented discourse particles in Cantonese.’ Studies in Chinese Linguistics 35: 61–​80. Lambrecht, Knud. 1994. Information structure and sentence form: Topic, focus and mental representations of discourse referents. Cambridge: Cambridge University Press. Lambrecht, Knud. 2001. ‘Dislocations,’ in M. Haspelmath, E. König, W. Oesterreicher, & W. Raible (eds.), Language typology and language universals: An international handbook, vol. 2. Berlin: Walter de Gruyter, 1050–​1078. Landau, Idan. 2009. The locative syntax of experiencers. Cambridge, MA: MIT Press. Landau, Idan. 2010. ‘The explicit syntax of implicit arguments.’ Linguistic Inquiry 41: 357–​388. Lane, Nick. 2015. Life ascending: The ten great inventions of evolution. New York: Norton. LaPolla, Randy J. 1992. ‘ “Anti-​ergative” marking in Tibeto-​Burman.’ Linguistics of the Tibeto-​ Burman Area 15(1): 1–​9.

 

References   1197 LaPolla, Randy J. 1995a. ‘ “Ergative” marking in Tibeto-​Burman,’ in Y. Nishi, James Matisoff, & Yasuhiko Nagano (eds.), New horizons in Tibeto-​Burman morphosyntax (Senri Ethnological Studies 41). Osaka: National Museum of Ethnology, 189–​228. LaPolla, Randy J. 1995b. ‘On the Utility of the Concepts of Markedness and Prototypes in Understanding the Development of Morphological Systems.’ Bulletin of the Institute of History and Philology 66(4): 1149–​1185. LaPolla, Randy J. 2003a. A grammar of Qiang with annotated texts and glossary. Berlin: Mouton de Gruyter. LaPolla, Randy J. 2003b. ‘Dulong,’ in Graham Thurgood & Randy J. LaPolla (eds.), The Sino-​ Tibetan languages. New York: Routledge, 674–​682. LaPolla, Randy J. 2004. ‘On Nominal relational morphology in Tibeto-​Burman,’ in Ying-​jin Lin, Fang-​min Hsu, Chun-​chih Lee, Jackson T.-​S. Sun, Hsiu-​fang Yang, & Dah-​an Ho (eds.), Studies on Sino-​Tibetan languages: Papers in honor of professor Hwang-​cherng Gong on his seventieth birthday. Taipei: Institute of Linguistics, Academia Sinica, 43–​74. LaPolla, Randy J. and Chenglong Huang. 2008. A grammar of Qiang: With annotated texts and glossary. Berlin: Mouton de Gruyter. Larrañaga, Pilar. 1994. ‘La evolución del caso en euskera y castellano,’ in J. Meisel (ed.), La adquisición del vasco y del castellano en niños bilingües. Frankfurt: Vervuert, 113–​150. Larsen, Thomas W. 1988. ‘Manifestations of ergativity in Quiché grammar.’ PhD dissertation, Berkeley. Larsen, Thomas W. & William M. Norman. 1979. ‘Correlates of ergativity in Mayan grammar,’ in Frans Plank (ed.), Ergativity: Towards a theory of grammatical relations. New York: Academic Press, 347–​370. Larson, Richard K. 1982. ‘Restrictive modification: Relative clauses and adverbs.’ PhD dissertation, University of Wisconsin-Madison. Lau, Ellen F., Colin Phillips, & David Poeppel. 2008. ‘A cortical network for semantics: (de) constructing the N400.’ Nature Reviews NeuroScience 9(12): 920–​933. Laughren, Mary. 1988. ‘Towards a lexical representation of Warlpiri verbs,’ in Wendy Wilkins (ed.), Thematic relations (Syntax and Semantics 21). New York: Academic Press, 215–​242. Laughren, Mary. 1989. ‘The configurationality parameter and Warlpiri,’ in László Marácz & Pieter Muysken (eds.), Configurationality:  The typology of asymmetries (Studies in Generative Grammar 34). Dordrecht: Foris, 319–​353. Laughren, Mary. 1992. ‘Secondary predication as a diagnostic of underlying structure in Pama–​ Nyungan languages,’ in Iggy Roca (ed.), Thematic structure: Its role in grammar (Linguistic models series). Berlin & New York: Foris/​Walter de Gruyter, 199–​246. Laughren, Mary. 2001. ‘What Warlpiri “avoidance” registers do with grammar,’ in Jane Simpson, David Nash, Mary Laughren, Peter Austin, & Barry Alpher (eds.), Forty years on: Ken Hale and Australian languages (Pacific Linguistics 512). Canberra: Australian National University, 199–​225. Laughren, Mary. 2002. ‘Syntactic constraints in a “free word order” language,’ in Mengistu Amberber & Peter Collins (eds.), Language universals and variation. Westport, CT: Praeger, 83–​130. Laughren, Mary & Warlpiri Lexicography group. 2007. Warlpiri–​English Encyclopaedic dictionary. Electronic files, University of Queensland, deposited at, AIATSIS. Laughren, Mary. 2010. ‘Warlpiri verbs of change and causation: The thematic core,’ in Mengistu Amberber, Brett Baker and Mark Harvey (eds.), Complex predicates: Cross-linguistic perspectives on event structure. Cambridge: Cambridge University Press, 167–236. Law, Danny. 2009. ‘Pronominal borrowing among the Maya.’ Diachronica 26(2): 214–​252. Law, Danny. 2014. Language contact, inherited similarity and social difference. Amsterdam & Philadelphia, PA: Benjamins.

 

1198   References Law, Danny, John S. Robertson, & Stephen Houston. 2006. ‘Split Ergativity in the history of the Cholan branch of the Mayan language family.’ International Journal of American Linguistics 72(4): 415–​450. Lazard, Gilbert. 1994. L’actance. Paris: Presses Universitaires de France. Lazard, Gilbert. 1998. Actancy. Berlin: Mouton de Gruyter. Lecarme, Jacqueline. 1996. ‘Tense in the Nominal System: The Somali D,’ in Jacqueline Lecarme, Jean Lowenstamm, & Ur Shlonsky (eds.), Studies in Afroasiatic syntax. Amsterdam: Holland Academic Graphics, 159–​178. Lee, Hanjung. 2003. ‘Parallel optimization in case systems,’ in M. Butt & T.H. King (eds.), Nominals: Inside and out. Stanford, CA: CSLI, 15–​59. Lees, Robert. 1960. The grammar of English nominalizations. The Hague: Mouton de Gruyter. Legate, Julie Anne. 2001. ‘The configurational structure of a nonconfigurational language.’ Linguistic Variation Yearbook 1: 61–​104. Legate, Julie Anne. 2002. ‘Warlpiri: Theoretical implications.’ PhD dissertation. Legate, Julie Anne. 2003. ‘The configurational structure of a nonconfigurational language.’ Linguistic Variation Yearbook, 61–​104. Legate, Julie Anne. 2006. ‘Split absolutive,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity: Emerging issues, studies in natural language & linguistic theory. Dordrecht: Springer, 143–​171. Legate, Julie Anne. 2008. ‘Morphological and abstract case.’ Linguistic Inquiry 39: 55–​102. Legate, Julie Anne. 2012a. ‘Types of ergativity.’ Lingua 122(3): 181–​191. Legate, Julie Anne. 2012b. ‘Subjects in Acehnese and the nature of the passive.’ Language 88: 495–​525. Legate, Julie Anne. 2014a. ‘Split ergativity based on nominal type.’ Lingua 148: 183–​212. Legate, Julie Anne. 2014b. Voice and v: Lessons from Acehnese. Cambridge, MA: MIT Press. Legendre, Géraldine, William Raymond, & Paul Smolensky. 1993. An optimality–​theoretic typology of case and grammatical voice systems. Proceedings of the 19th meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society. Legendre, Géraldine, Paul Smolensky, & Colin Wilson. 1998. ‘When is less more? Faithfulness and minimal links in wh-​chains,’ in P. Barbosa, D. Fox, P. Hagstrom, M. McGinnis, & D. Pesetsky (eds.), Is the best good enough? Cambridge, MA: MIT Press & MITWPL, 249–​289. Lehmann, Christian. 1982. ‘Thoughts on grammaticalization: A  programmatic sketch.’ 1. Arbeiten des Kölner Universalien-​Projekts 48. Lehmann, Christian. 1993. ‘The genesis of auxiliaries in Yukatek Maya.’ Proceedings of the International Congress of Linguists 15(2): 313–​316. Lehmann, Christian. 2002. Thoughts on grammaticalization (2nd edn.). Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt, Nr. 9.  Erfurt:  Seminar für Sprachwissenschaft der Universität. Levin, Beth. 1983. ‘On the Nature of Ergativity.’ PhD dissertation, MIT. Levin, Beth. 1993. English verb classes and alternations: A preliminary investigation. Chicago, IL, & London: University of Chicago Press. Levin, Beth & Malka Rappaport-​Hovav. 1995. Unaccusativity: At the syntax–​lexical semantics interface. Cambridge, MA: MIT Press. Levin, Juliette & Diane Massam. 1985. ‘Surface ergativity: Case-​/t​ heta relations re-​examined,’ in Stephen Berman, Jae-​Woon Choe, & Joyce McDonough (eds.), Proceedings of NELS 15. GLSA, University of Massachusetts, Amherst, 286–​301. Levin, Theodore. 2015. ‘Licensing without case.’ PhD dissertation, MIT.

 

References   1199 Levin, Theodore & Omer Preminger. 2015. ‘Case in Sakha: Are two modalities really necessary?’ Natural Language and Linguistic Theory 33: 231–​250. Lewis, Richard L. 1998. ‘Leaping off the garden path: Reanalysis and limited repair parsing,’ in J. Fodor & F. Ferreira (eds.), Reanalysis in sentence processing. New York: Kluwer, 247–​268. Li, Chao. 2007. ‘Split ergativity and split intransitivity in Nepali.’ Lingua 117: 1462–​1482. Liao, Hsiu-​Chuan. 2004. ‘Transitivity and ergativity in Formosan and Philippine languages.’ PhD dissertation, University of Hawai‘i, Mānoa. Lichtenberk, František. 1996. ‘Patterns of Anaphora in To’aba’ita Discourse,’ in Barbara A. Fox (ed.), Studies in anaphora. Amsterdam & Philadelphia, PA: Benjamins, 379–​411. Lidz, Liberty. 2011. ‘Agentive marking in Yongning Na (Mosou).’ Linguistics of the Tibeto-​ Burman Area 34(2): 49–​72. Lightfoot, David. 1999. The development of language: Acquisition, change and evolution. Malden MA: Blackwell. Lindner, Thomas. 2014. ‘ “Ergative historiography” revisited.’ Historiographia Linguistica 41(1): 188–​192. Liu, Adlay Kun-​long. 2004. ‘On relativization in Squliq Atayal.’ MA thesis, National Tsing Hua University. Livingston, Ellen. 1989. ‘Conjoined arguments in Nisgha,’ Proceedings of the 24th international conference on Salish and Neighbouring languages. Ljutikova, Ekaterina A. 1997, ‘Refleksivy i èmfaza.’ Voprosy jazykoznanija 49–​74. Ljutikova, Ekaterina A. 1999. ‘Reflexives and emphasis in Tsaxur,’ in Zygmunt Frajzyngier & Traci S. Curl (eds.), Reflexives:  Form and function. Amsterdam & Philadelphia, PA: Benjamins, 227–​255. Ljutikova, Ekaterina A. 2001. ‘Uveličenie perexodnosti: Kauzativizacija,’ in Aleksandr E. Kibrik (ed.), Bagvalinskij jazyk. Moscow: Nasledie, 384–​394. Ljutikova, Ekaterina A. & Anastasija Bonč-​Osmolovskaja. 1999. ‘Aktantnye predloženija,’ in Aleksandr E. Kibrik (ed.), Elementy caxurskogo jazyka v tipologičeskom osveščenii. Moscow: Nasledie, 481–​536. Longenbaugh, Nicholas & Maria Polinsky 2016. ‘Subject/​object symmetry in Niuean: A theoretical and experimental investigation,’ in H. Hsieh (ed.), Proceedings of the 22nd meeting of the Austronesian Formal Linguistics Association. Sydney: Austronesian Formal Linguistics Association, 98–120. López Ixcoy, Candelaria Dominga. 1997. Gramática K’ichee’. Guatemala City: Cholsamaj. Lorimer, David L.R. 1935. The Burushaski language:  Introduction and grammar, vol. 1. Cambridge, MA: Harvard University Press. Loriot, James, Erwin Lauriault, & Dwight Day. 1993. Diccionario Shipibo–​Castellano. Lima, Peru: Instituto Linguistico de Verano. Lowe, Ronald. 1985. Kangiryuarmiut uquauhingita ilihautdjutikhangit: Basic Kangiryuarmiut Eskimo grammar. Inuvik, NWT: Committee for Original Peoples Entitlement. Luján, Eugenio, Jóhanna Barðdal, & Spike Guildea. [to appear]. Syntactic reconstruction: Applying the comparative method. Leiden: Brill. Lynch, John, Malcolm Ross, & Terry Crowley. 2002. ‘Proto Oceanic,’ in John Lynch, Malcolm Ross, & Terry Crowley (eds.), The Oceanic languages. New York: Curzon, 54–​91. MacDonald, Lorna. 1990. A grammar of Tauya (Mouton Grammar Library, 6). Berlin: Mouton de Gruyter. Macdonald, Catherine. 2014. ‘Functional projections and non-​local relations in Tongan nominal phrases.’ PhD dissertation, University of Toronto.

 

1200   References MacKenzie, David. 1961. Kurdish dialect studies vol. I. London: Oxford University Press. MacKenzie, David. 1962. Kurdish dialect studies vol. II. London: Oxford University Press. Maclachlan, Anna & Masanori Nakamura. 1997. ‘Case checking and specificity in Tagalog.’ Linguistic Review 14: 307–​333. MacLean, Edna Ahgeak. 2014. Iñupiatun Uqaluit Taniktun Sivuni–​Iñupiaq to English dictionary. Fairbanks: Alaska Native Language Center, University of Alaska. MacWhinney, Brian, Andrej Malchukov, & Edith Moravcsik (eds.). 2014. Competing motivations in grammar and cognition. Oxford: Oxford University Press. McConvell, Patrick. 1976. ‘Nominal hierarchies in Yukulta,’ in R.M.W. Dixon (ed.), 191–​200. McConvell, Patrick. 1980. ‘Hierarchical variation in pronominal clitic attachment in the Eastern Ngumbin languages,’ in Bruce Rigsby & Peter Sutton (eds.), Papers in Australian linguistics 13: 31–​117. Canberra: Pacific Linguistics. McConvell, Patrick. 1981. ‘How Lardil became accusative.’ Lingua 55: 141–​179. McConvell, Patrick & Mary Laughren. 2004. ‘The Ngumpin–​Yapa subgroup,’ in Claire Bowern & Harold Koch (eds.), Australian languages: Classification and the comparative method. Amsterdam & Philadelphia, PA: Benjamins, 151–​177. McConvell, Patrick & Felicity Meakins. 2005. ‘Gurindji Kriol: A mixed language emerges from code-​switching.’ Australian Journal of Linguistics 25(1): 9–​30. McFadden, Thomas. 2004. ‘The position of morphological case in the derivation: A study on the syntax–​morphology interface.’ PhD thesis, University of Pennsylvania. McFarland, Curtis D. 1974. ‘The dialects of the bikol area.’ PhD dissertation, Yale University. McFarland, Curtis D. 1978. ‘Definite objects and subject selection in Philippine languages.’ Studies in Philippine Linguistics 2: 139–​182. McGinn, Richard. 1988. ‘Government and case in Tagalog,’ Studies in Austronesian linguistics. Ohio: Ohio University Press. McGinnis, Martha. 1997. ‘Case and locality in L-​syntax:  Evidence from Georgian,’ in Heidi Harley (ed.), Papers from the UPenn/​MIT roundtable on argument structure and aspect. Cambridge, MA: MIT Working Papers in Linguistics 32: 139–​158. McGinnis, Martha. 2005. ‘On markedness asymmetries in person and number.’ Language 81: 699–​7 18. McGregor, William B. 1990. A functional grammar of Gooniyandi. Amsterdam & Philadelphia, PA: Benjamins. McGregor, William B. 1992. ‘The semantics of ergative marking in Gooniyandi.’ Linguistics 30: 275–​318. McGregor, William B. 1994. ‘The grammar of reported speech and thought in Gooniyandi.’ Australian Journal of Linguistics 14: 63–​92. McGregor, William B. 1997. Semiotic grammar. Oxford: Clarendon Press. McGregor, William B. 1998. ‘  “Optional” ergative marking in Gooniyandi revisited: Implications to the theory of marking.’ Leuvense Bijdragen 87: 491–​534. McGregor, William B. 1999a. ‘External possession constructions in Nyulnyulan languages,’ in Doris L. Payne & Immanuel Barshi (eds.), External possession (Typological Studies in Language, 39.) Amsterdam & Philadelphia, PA: Benjamins, 429–​448. McGregor, William B. 1999b. ‘The medio-​active construction in Nyulnyulan languages.’ Studies in Language 23(3): 531–​567. McGregor, William B. 2002a. ‘Ergative and accusative patterning in Warrwa,’ in Kristin Davidse & Béatrice Lamiroy (eds.), The nominative & accusative and their counterparts (Case and Grammatical Relations across Languages, 4). Amsterdam & Philadelphia, PA:  Benjamins, 285–​317.

 

References   1201 McGregor, William B. 2002b. ‘Structural changes in language obsolescence:  A  Kimberley (Australia) perspective.’ SKY Journal of Linguistics 15: 145–​185. McGregor, William B. 2002c. Verb classification in Australian languages. Berlin: Mouton de Gruyter. McGregor, William B. 2005. ‘Quantifying depictive secondary predicates in Australian languages,’ in Nikolaus P. Himmelmann & Eva Schultze-​Berndt (eds.), Secondary predication and adverbial modification: The typology of depictives. Oxford: Oxford University Press, 173–​200. McGregor, William B. 2006. ‘Focal and optional ergative marking in Warrwa (Kimberley, Western Australia).’ Lingua 116(4): 393–​423. McGregor, William B. 2007. ‘Ergative marking of intransitive subjects in Warrwa.’ Australian Journal of Linguistics 27: 201–​229. McGregor, William B. 2008. ‘Indexicals as sources of case markers in Australian languages,’ in Folke Josephson & Ingmar Söhrman (eds.), Interdependence of diachronic and synchronic analyses (Studies in Language Companion Series, 103). Amsterdam & Philadelphia, PA: Benjamins, 299–​321. McGregor, William B. 2009. ‘Typology of ergativity.’ Language and Linguistics Compass 3(1): 480–​508. McGregor, William B. 2010. ‘Optional ergative case marking systems in a typological–​semiotic perspective.’ Lingua 120: 1610–​1636. McGregor, William B. 2012. ‘Review of Spike Gildea & Francesc Queixalós eds., Ergativity in Amazonia.’ Journal of Linguistics 48(1): 238–​243. McGregor, William B. 2014. ‘Connate roles in Nyulnyul:  Non-​nuclear grammatical relations within the core,’ in Nicole Delbecque, Karen Lahousse, & Willy Van Langendonck (eds.), Nuclear and non-​nuclear cases (Case and Grammatical Relations across Language Boundaries, 6). Amsterdam & Philadelphia, PA: Benjamins, 67–​93. McGregor, William B. [forthcoming]. ‘Emergence of optional accusative case marking in Khoe languages’, in Ilja A. Seržant, Alena Witzlack-​Makarevich, and Kelsey Mann (eds.), The diachronic typology of differential argument marking. Berlin: Language Sciences Press. McGregor, William B. & Jean-​Christophe Verstraete. 2010. ‘Optional ergative marking and its implications for linguistic theory.’ Lingua 120(7): 1607–​1609. Maggio, Francisco-Maria. 1643. Syntagmata Linguarum Orientalium quæ in Georgiæ regionibus audiuntur. Rome: Typographia Sacræ Congregationis de Propaganda Fide. Magomedova, Patimat Turalovna & Rasidat Saxrudinovna Xalidova. 2001. Karatinsko–​russkij slovar. Makhachkala: Institut JaLI DNC RAN. Mahajan, Anoop. 1987. ‘Agreement and agreement phrases.’ Unpublished paper, MIT. Mahajan, Anoop. 1989. ‘Agreement and agreement phrases,’ in Itziar Laka & Anoop Mahajan (eds.), MIT working papers in linguistics 10:  Functional heads and clause structure. Cambridge, MA: MIT, 217–​252. Mahajan, Anoop. 1990. ‘The A/​A’ distinction and movement theory.’ PhD dissertation, MIT. Mahajan, Anoop. 1994. ‘The ergativity parameter: have–​be alternation, word order and split ergativity,’ in M. Gonzalez (ed.), Proceedings of NELS 24, GLSA. Amherst: University of Massachusetts, 317–​331. Mahajan, Anoop. 1997. ‘Universal grammar and the typology of ergative languages,’ in Artemis Alexiadou & Tracy Alan Hall (eds.), Studies on universal grammar and typological variation. Amsterdam & Philadelphia, PA: Benjamins, 35–​57. Mahajan, Anoop. 2000. ‘Oblique subjects and Burzio’s generalization,’ in Eric Reuland (ed.), Argument and case: Explaining Burzio’s generalization. Amsterdam & Philadelphia, PA: Benjamins, 79–​102.

 

1202   References Mahajan, Anoop. 2004. ‘On the origin of non-​nominative subjects,’ in Peri Bhaskararao & K.V. Subbārāo (eds.), Non-​nominative subjects. Amsterdam & Philadelphia, PA: Benjamins, 283–​300. Mahajan, Anoop. 2012. ‘Ergatives, antipassives and the overt light v in Hindi.’ Lingua 122: 204–​214. Mahalingappa, Laura J. 2013. ‘The acquisition of split-​ergative case-​marking in Kurmanji Kurdish,’ in Edith L. Bavin & S. Stolle (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 239–​269. Mahand, Mohammad Rasekh & Zaniar Naghshbandi. 2014. The effects of preferred argument structure and other discourse motivations on determination of split alignment in Hawrami. Mahmoudveysi, Parvin, Denise Bailey, Ludwig Paul, & Geoffrey Haig. 2012. The Gorani language of Gawraju (Gawrajuyi), a village of West Iran: Texts, grammar and lexicon. Wiesbaden: Reichert. Malchukov, Andrej L. 2005. ‘Case pattern splits, verb types and construction competition,’ in Mengistu Amberber & Helen de Hoop (eds.), Competition and variation in natural languages: The case for case (Perspectives on Cognitive Science). Amsterdam: Elsevier, 73–​118. Malchukov, Andrej. 2006. ‘Transitivity parameters and transitivity alternations: Constraining co-​variation,’ in Leonid Kulikov, Andrej Malchukov, & Peter de Swart (eds.), Studies on case, valency and transitivity. Amsterdam & Philadelphia, PA: Benjamins, 329–​359. Malchukov, Andrej. 2008a. ‘Animacy and asymmetries in differential case marking.’ Lingua 118: 203–​221. Malchukov, Andrej. 2008b. ‘Split intransitives, experiencer objects, and “transimpersonal” constructions: (Re-​)establishing the connection,’ in Donohue & Wichmann (eds.), 76–​100. Malchukov, Andrej. 2010. ‘Analyzing semantic maps:  A  multifactorial approach.’ Linguistic Discovery 8: 176–​198 Malchukov, Andrej. 2014. ‘Resolving alignment conflicts: A competing motivations approach,’ in Brian MacWhinney, Andrej Malchukov, & Edith Moravcsik (eds.), Competing motivations in grammar and cognition. Oxford: Oxford University Press, 17–​42. Malchukov, Andrej. 2015. ‘Towards a typology of split ergativity: A TAM hierarchy for alignment splits,’ in Ina Bornkessel-​Schlesewsky, Andrej Malchukov, & Marc Richards (eds.), Scales and hierarchies: A cross-​disciplinary perspective. Berlin: de Gruyter, 275–​297. Malchukov, Andrej & Helen de Hoop. 2011. ‘Tense, aspect and mood-​based differential object marking.’ Lingua 121: 35–​47. Malchukov, Andrej & Peter de Swart. 2009. ‘Differential case marking and actancy variation,’ in A. Malchukov & A. Spencer (eds.), The Oxford handbook of case. Oxford University Press, 339–​356. Malchukov, Andrej & Heiko Narrog. 2009. ‘Case polysemy,’ in Andrej L. Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 518–​534. Malchukov, Andrej & Akio Ogawa. 2011. ‘Towards a typology of impersonal constructions: A semantic map approach,’ in A. Malchukov & A. Siewierska (eds.), Impersonal constructions: A cross-​linguistic perspective. Amsterdam & Philadelphia, PA: Benjamins, 19–​56. Mallinson, Graham & Barry Blake. 1981. Language typology: Cross-​linguistic studies in syntax. Amsterdam & New York: North-​Holland. Manga, Louise Schieberl. 1996. ‘An explanation of ergative versus accusative languages: An examination of Inuktitut.’ PhD dissertation, University of Ottawa. Manning, Christopher D. 1994. ‘Ergativity: Argument structure and grammatical relations.’ PhD dissertation, Stanford University. Manning, Christopher D. 1996. Ergativity:  Argument structure and grammatical relations. Stanford, CA: Center for the Study of Language and Information.

 

References   1203 Manning, Christopher D. 2006. ‘Ergativity,’ in K. Brown (ed.), Encyclopedia of language and linguistics, vol. 4. Oxford: Elsevier, 210–​217. Manning, Christopher D. & Ivan A. Sag. 1999. ‘Dissociations between argument structure and grammatical relations,’ in G. Webelhuth, J.-​P. Koenig, & A. Kathol (eds.), Lexical and constructional aspects of linguistic explanation. Stanford, CA: CSLI, 236–​258. Manolessou, Io. 2000. ‘Greek noun phrase structure: A study in syntactic evolution.’ PhD dissertation, University of Cambridge. Manson. Ken. 2010. ‘A grammar of Kayan, a Tibeto-​Burman language.’ PhD dissertation, La Trobe University, Melbourne. Manterola, Julen. 2015. ‘Euskararen Morfologia Historikorako:  Artikuluak eta erakusleak.’ PhD dissertation, UPV-​EHU. Manzini, Rita & Leonardo Savoia. 2005. I dialetti italiani e romanci. Morfosintassi generativa, 3 vols. Alessandria: Edizioni dell’Orso. Manzini, Rita & Kenneth Wexler. 1987. ‘Parameters, binding theory, and learnability.’ Linguistic Inquiry 18: 413–​444. Marantz, Alec. 1984. On the nature of grammatical relations. Cambridge, MA: MIT Press. Marantz, Alec. 1991. ‘Case and licensing,’ in Germán F. Westphal, Benjamin Ao, & Hee-​Rahk Chae (eds.), Proceedings of the eighth eastern states conference on linguistics, 234–​253. Marinis, Theodore. 2003, ‘Psycholinguistic techniques in second language acquisition research.’ Second Language Research 19(2): 144–​161. Markman, Vita G. & Pavel Grashchenkov. 2012. ‘On the adpositional nature of ergative subjects.’ Lingua 122: 257–​266. Marr, Nicholas Yakovlevich. 1910. Grammatika čanskago (lazskago) jazyka. Materialy po jafetičeskomy jazykoznaniju II. St. Petersburg:  Tipografija Imperatorskoj Akademija Nauk. Marr, Nicholas Yakovlevich. 1925. Grammatika drevneliteraturnogo gruzinskogo jazyka. Materialy po jafetičeskomy jazykoznaniju XII. St. Petersburg:  Tipografija Akademija Nauk. Marr, Nicholas Yakovlevich & Maurice Brière. 1931. La langue géorgienne. Paris: Firmin-​Didot Martin, Laura. 2003. ‘Narrator virtuosity and the strategic exploitation of preferred argument structure:  Repetition and constructed speech in Mochó Narrative,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA: Benjamins, 411–​435. Martinet, André. 1979. ‘Shunting on to ergative or accusative,’ in F. Plank (ed.), Ergativity: Towards a theory of grammatical relations. London & New York: Academic Press, 39–​43. Martin-​ Loeches, Manuel, Francisco Muñoz, Pilar Casado, Angela Melcón, & Carlos Fernández-​Frias. 2005. ‘Are the anterior negativities to grammatical violations indexing working memory?’ Psychophysiology 42(5): 508–​519. Martínez Álvarez, Pedro Rosendo. 2012. ‘Las manifestaciones sintácticas, semánticas y discursivas de la agentividad en el tsotsil de Huixtán, Chiapas.’ MA thesis, CIESAS, Mexico, DF. Martínez Cruz, Victoriano. 2007. ‘Los adjetivos y conceptos de propiedad en Ch’ol.’ MA thesis, CIESAS, Mexico, DF. Mart’irosovi, Aram. 1964. nacvalsaxeli kartvelur enebši [‘The pronoun in the Kartvelian languages’]. Tbilisi: Mecniereba Mascaró, Juan. 1976. ‘Catalan phonology and the phonological cycle.’ PhD dissertation, MIT. Masica, Colin. 1976. Defining a linguistic area:  South Asia. Chicago, IL:  University of Chicago Press.

 

1204   References Masica, Colin. 1990. ‘Varied case marking in obligational constructions,’ in Manindra K. Verma & K.P. Mohanan (eds.), Experiencer subjects in South Asian Languages. Stanford, CA: CSLI, 335–​342. Masica, Colin. 1991. The Indo-​Arian languages. Cambridge: Cambridge University Press. Maslova, Elena & Giuliano Bernini. 2006. ‘Sentence topics in the languages of Europe and beyond,’ in G. Bernini & M.L. Schwarz (eds.), Pragmatic organization of discourse in the languages of Europe. Berlin: Mouton de Gruyter, 67–​120. Maslova, Elena & Tatiana Nikitina. 2007. ‘Stochastic universals and dynamics of cross-​ linguistic distributions: The case of alignment types.’ MS, Stanford University. Massam, Diane. 1985. ‘Case theory and the projection principle.’ PhD dissertation, MIT. Massam, Diane. 1998. ‘Instrumental Aki and the nature of Niuean transitivity.’ Oceanic Linguistics 37(1): 12–​28. Massam, Diane. 2000. ‘VSO and VOS: Aspects of Niuean word order,’ in Andrew Carnie & Eithne Guilfoyle (eds.), The Syntax of Verb Initial Languages. Oxford: Oxford University Press, 97–​116. Massam, Diane. 2001. ‘Pseudo noun incorporation in Niuean.’ Natural Language and Linguistic Theory 19: 153–​197. Massam, Diane. 2002. ‘Fully internal cases: surface ergativity can be profound,’ in A. Rakowski & N. Richards (eds.), Proceedings of AFLA 8, MITWPL. Cambridge, MA, 185–​196. Massam, Diane. 2003. ‘Questions and the left periphery in Niuean,’ Proceedings of the ninth annual meeting of the Austronesia Formal Linguistics Association. Ithaca, NY:  Cornell Working Papers in Linguistics, 94–​106. Massam, Diane. 2006. ‘Neither absolutive nor ergative is nominative or accusative,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity: Emerging issues, studies in natural language & linguistic theory. Dordrecht: Springer, 27–​46. Massam, Diane. 2009. ‘The structure of (un)ergatives.’ Proceedings of AFLA 16: 125–​135. Massam, Diane. 2013. ‘On definite and generic null arguments in Niuean.’ Paper presented 20th meeting of the Austronesian Formal Linguistics Association (AFLA 20), University of Texas: Arlington, TX. Masullo, Pascual José. 1992. ‘Antipassive constructions in Spanish,’ in Paul Hirschbuhler & Konrad Koerner (eds.), Romance languages and modern linguistic theory: Papers from the 20th linguistic symposium on Romance languages (LSRL XX). Ottawa, April 10–​14, 1990. Amsterdam & Philadelphia, PA: Benjamins, 175–​194. Mateo Pedro, Pedro. 2010. ‘Nominalization in Qanjobal (Maya),’ in Stephanie Lux & Pedro Mateo Pedro (eds.), Kansas Working Papers in Linguistics 31. University of Kansas: 46–​63. Mateo-​ Toledo, B’alam Eladio. 2008. ‘The family of complex predicates in Q’anjob’al (Maya): Their syntax and meaning.’ PhD dissertation, University of Texas Austin. Mateo-​Toledo, B’alam Eladio. 2003. ‘Ergatividad mixta en Q’anjobal (Maya):  un reanálisis,’ Proceedings of the conference of indigenous language of Latin America 1. Matsumoto, Kazuko. 1997. ‘NPs in Japanese conversation.’ Pragmatics 7(2): 163–​181. Matsumoto, Kazuko. 2000. ‘Intonation units, clauses and preferred argument structure in conversational Japanese.’ Language Sciences 22: 63–​86. Matthews, Peter H. 2007. The concise Oxford dictionary of linguistics (2nd edn.). Oxford: Oxford University Press. Matushansky, Ora. 2006. ‘Head movement in linguistic theory.’ Linguistic Inquiry 37: 69–​109. Matzke, Mike, Heinke Mai, Wido Nager, Jascha Rüsseler, & Thomas Münte. 2002. ‘The costs of freedom: An ERP—​study of non-​canonical sentences.’ Clinical Neurophysiology: Official Journal of the International Federation of Clinical Neurophysiology 113(6): 844–​52. Mayr, Ernst. 2001. What evolution is. New York: Basic Books.

 

References   1205 Mazaudon, Martine. 2003. ‘Tamang,’ in Graham Thurgood & Randy LaPolla (eds.), The Sino-​ Tibetan languages. London: Routledge, 289–​313. Meakins, Felicity. 2009. ‘The case of the shifty ergative marker: A pragmatic shift in the ergative marker of one Australian mixed language,’ in J. Barðdal & S.L. Chelliah (eds.), The Role of semantic, pragmatic, and discourse factors in the development of case. Amsterdam & Philadelphia, PA: Benjamins, 59–​92. Meakins, Felicity. 2011. Case-​marking in contact: The development and function of case morphology in Gurindji Kriol (Creole Language Library, 39). Amsterdam & Philadelphia, PA: Benjamins. Meakins, Felicity. 2014a. ‘Language contact varieties,’ in Harold Koch & Rachel Nordlinger (eds.), The languages and linguistics of Australia: A comprehensive guide. Berlin & Boston, MA: Mouton, 365–​416. Meakins, Felicity. 2014b. ‘Nominals as adjuncts or arguments:  Further evidence from language mixing,’ in Rob Pensalfini, Myfany Turpin, & Diana Guillemin (eds.), Language description informed by theory (Studies in Language Companion Series 147). Amsterdam & Philadelphia, PA: Benjamins, 283–​315. Meakins, Felicity. 2015. ‘From absolutely optional to only nominally ergative: The life cycle of the Gurindji ergative suffix,’ in Francesco Gardani, Peter Arkadiev, & Nino Amiridze (eds.), Borrowed morphology. Berlin: Mouton de Gruyter, 189–​218. Meakins, Felicity & Carmel O’Shannessy. 2010. ‘Ordering arguments about: Word order and discourse motivations in the development and use of the ergative marker in two Australian mixed languages.’ Lingua 120: 1693–​1713. Medová, Lucie. 2010. ‘Reflexive clitics in the Slavic and Romance languages: A comparative view from an antipassive perspective.’ PhD dissertation, Princeton University. Megerdoomian, Karine. 2000. ‘Aspect and partitive objects in Finnish,’ Proceedings of the west coast conference on formal linguistics 19: 316–​328. Mejias-​Bikandi, Errapel 1999. ‘Unaccusative and antipassive constructions in Basque and Spanish,’ in Jon Franco, Alazne Landa, & Juan Martin (eds.), Grammatical analyses in Basque and Romance linguistics:  Papers in honor of Mario Saltarelli. Amsterdam & Philadelphia, PA: Benjamins, 165–​177. Mel’čuk, Igor A. 1979. ‘The predication construction in Dyirbal,’ in I.A. Mel’čuk (ed.), Studies in dependency syntax. Ann Arbor, MI: Karoma, 23–​90. Mel’čuk, Igor A. 1988. ‘Is there an ergative construction in Lezgian?’ in I. Mel’čuk (ed.), Dependency syntax. Albany, NY: State University of New York Press, 207–​249. Meltzer, Jed A. & Allen R. Braun. 2013. ‘P600-like positivity and left anterior negativity responses are elicited by semantic reversibility in nonanomalous sentences.’ Journal of Neurolinguistics 26(1): 129–​148. Mendívil-​Giró, José Luis. (2012) ‘Ergativity as transitive unaccusativity,’ Folia Linguistica 46(1): 1–​40. Mengozzi, A lessandro. 2005. ‘Neo-​Aramaic and the so-​called “decay of ergativity” in Kurdish,’ in P. Fronzaroli & P. Marrassini (eds.), Proceedings of the 10th meeting of Hamito-​Semitic (Afroasiatic) linguistics. Florence: Department of Linguistics, 239–​256. Mercado, Raphael, Eric Potsdam, & Lisa deMena Travis. (eds.). 2010. Austronesian and theoretical linguistics. Amsterdam & Philadelphia, PA: Benjamins. Merchant, Jason. 2004. ‘Fragments and ellipsis.’ Linguistics and Philosophy 27: 661–​738. Merchant, Jason. 2009. ‘Polyvalent case, geometric hierarchies, and split ergativity,’ in J. Bunting et al. (eds.), Proceedings of the 42nd annual meeting of the Chicago Linguistic Society (CLS 42), vol. 2. Chicago, IL: Chicago Linguistics Society, 57–​76.

 

1206   References Merchant, Jason. 2011. ‘Aleut case matters,’ in Tista Bagchi Etsuyo Yuasa & Katharine P. Beals (eds.), Pragmatics and autolexical grammar: In honor of Jerry Sadock, vol. 176. Amsterdam & Philadelphia, PA: Benjamins, 382–​411. Merlan, Francesca. 1985. ‘Split intransitivity:  Functional oppositions in intransitive inflections,’ in Johanna Nichols & Anthony C. Woodbury (eds.), Grammar inside and outside the clause:  Some approaches to theory from the field. Cambridge:  Cambridge University Press, 324–​362. Merlan, Francesca. 1994. A grammar of Wardaman: A language of the Northern Territory of Australia. Berlin & New York: Mouton de Gruyter. Meyer, Martin, Angela D. Friederici, & D. Yves von Cramon. 2000. ‘Neurocognition of auditory sentence comprehension: Event related fMRI reveals sensitivity to syntactic violations and task demands.’ Cognitive Brain Research 9(1): 19–​33. Michelena, Luis. 1987. Palabras y textos. Vitoria-​Gasteiz: UPV-​EHU. Michelena, Luis. 1990. Diccionario general vasco–​Orotariko Euskal Hiztegia, vol. 4. Bilbao: Desclée de Brouwer. Miller, Cynthia L. & Leoma G. Gilley. 2001. ‘Evidence for ergativity in Shilluk.’ Journal of African Languages and Linguistics 22(1): 33–​68. Milner, George. 1973. ‘It is aspect (not voice), which is marked in Samoan,’ Oceanic Linguistics 12:  621–​ 639. [Papers of First International Conference on Comparative Austronesian Linguistics] Milner, George. 1976. Samoan dictionary (Samoan–​English and English–​Samoan). Oxford: Oxford University Press. Milner, Jean-​Claude. 1982. Ordres et raisons de langue. Paris: Éditions du Seuil. Mitchell, Don C. 1994. ‘Sentence parsing,’ in M.A. Gernsbacher (ed.), Handbook of psycholinguistics. New York: Academic Press, 375–​409. Mithun, Marianne. 1984. ‘The evolution of noun incorporation.’ Language 60: 847–​893. Mithun, Marianne. 1991a. ‘The role of motivation in the emergence of grammatical categories: The grammaticalization of subjects,’ in Elizabeth C. Traugott & Bernd Heine (eds.), Approaches to grammaticalization, vol. 2. Amsterdam & Philadelphia, PA:  Benjamins, 159–​184. Mithun, Marianne. 1991b. ‘Active/​ agentive case marking and its motivations.’ Language 67(3): 510–​546. Mithun, Marianne. 1999. The languages of Native North America. Cambridge:  Cambridge University Press. Mithun, Marianne. 2000. ‘Valency-​changing derivations in central Alaskan Yup’ik,’ in R.M.W. Dixon & Alexandra Aikhenvald (eds.), Changing valency:  Case studies in transitivity. Cambridge: Cambridge University Press, 84–​114. Mithun, Marianne. 2003. ‘Functional perspectives on syntactic change,’ in Richard Janda & Brian Joseph (eds.), Handbook of Historical Linguistics. Oxford: Blackwell’s, 552–572. Mithun, Marianne. 2006. ‘Voice without subjects, objects, or obliques: Manipulating argument structure in agent/​patient systems (Mohawk),’ in T. Tsunoda & T. Kageyama (eds.), Voice and grammatical relations. Amsterdam & Philadelphia, PA: Benjamins, 195–​216. Mithun, Marianne. 2008. ‘The emergence of agentive systems in core argument marking,’ in Donohue & Wichmann (eds.), 297–​333. Mithun, Marianne & Wallace Chafe. 1999. ‘What are S, A, and O?’ Studies in Language 23(3): 569–​596. Mittwoch, Anita. 1977. ‘How to refer to one’s own words: Speech–​act modifying adverbials and the performative analysis.’ Journal of Linguistics 13: 177–​189.

 

References   1207 Miyagawa, Shigeru. 2001. ‘The EPP, scrambling, and Wh-​in-​situ,’ in Michael Kenstowicz (ed.), Ken Hale: A life in language. Cambridge, MA: MIT Press, 293–​338. Miyagawa, Shigeru. 2005. ‘On the EPP,’ in Martha McGinnis & Norvin Richards (eds.), Perspectives on phase, MIT working papers in linguistics 49. Cambridge, MA: MIT Working Papers in Linguistics, 201–​235. Miyagawa, Shigeru. 2012. ‘Agreements that occur mainly in the main clause,’ in Aelbrecht Loebeke, Liliane Haegeman, & R. Dye (eds.), Main clause phenomena: New horizons. Amsterdam: John Benjamins, 79–112. Miyamoto, Edson T. 2008. ‘Processing sentences in Japanese,’ in S. Miyagawa & M. Saito (eds.), The Oxford handbook of Japanese linguistics. Oxford: Oxford University Press, 217–​249. Mohanan, Tara. 1990. ‘Arguments in Hindi.’ PhD dissertation, Stanford University. Mohanan, Tara. 1994a. Argument structure in Hindi. Stanford, CA: CSLI. Mohanan, Tara. 1994b. ‘Case OCP:  A  constraint on word order in Hindi.’ Theoretical Perspectives on Word Order in South Asian Languages 185: 216. Mohanan, Tara. 1995. ‘Wordhood and lexicality:  noun Incorporation in Hindi.’ Natural Language and Linguistic Theory 13: 75–​134. Molinaro, Nicola, Horacio A. Barber, & Manuel Carreiras. 2011. ‘Grammatical agreement processing in reading: ERP findings and future directions.’ Cortex: A Journal Devoted to the Study of the Nervous System and Behavior 47(8): 908–​30. Molinaro, Nicola, Horacio A. Barber, Sendy Caffarra, & Manuel Carreiras. 2014. ‘In the left anterior negativity (LAN):  The case of morphosyntactic agreement.’ Cortex:  A  Journal Devoted to the Study of the Nervous System and Behavior 4–​7. Molnár, Valéria. 1993. Zur Pragmatik und Grammatik des TOPIK-​Begriffes [‘On pragmatics and grammar of the term topic’]. In Wortstellung und informationsstruktur [‘Word order and information structure’], ed. Marga Reis, volume 306. Tübingen: Linguistische Arbeiten, 155–​202. Molochieva Zarina & Alena Witzlack-​Makarevich. 2008. ‘Grammatical relations in Chechen.’ Talk presented at the University of Leipzig. Mondloch, James. 1981. ‘Voice in Quiche-​Maya.’ PhD dissertation, SUNY at Albany. Montaut, Annie. 2003. ‘Oblique main arguments in Hindi/​ Urdu as localizing predications: Questioning the category of subject,’ in Peri Bhaskarao & K.V. Subbārāo (eds.), Non-​ nominative subjects. Amsterdam & Philadelphia, PA: Benjamins, 33–​56. Montaut, Annie. 2006. ‘The evolution of the tense–​aspect system in Hindi/​Urdu: The status of the ergative alignment,’ in Miriam Butt & Tracy Holloway King (eds.), Proceedings of the LFG06 conference. Stanford, CA: CSLI, 365–​385. Montaut, Annie. 2009. ‘Ergative and pre-​ergative patterns in Indo-​Aryan as predications of localization: A diachronic view of past and future systems,’ in A.R. Faithi (ed.), Language vitality in South Asia. Aligarh: Aligarh Muslim University, 295–​325. Mora-​Marín, David F. 2004. ‘The preferred argument structure of classic lowland Mayan texts,’ in Søren Wichmann (ed.), The linguistics of Maya writing. Salt Lake City: University of Utah Press, 339–​364. Moravcsik, Edith. 1978a. ‘On the case marking of objects,’ in J. Greenberg (ed.), Universals of human language, vol. 4. Stanford, CA: Stanford University Press, 249–​289. Moravcsik, Edith. 1978b. ‘On the distribution of ergative and nominative patterns.’ Lingua 45: 233–​279. Morey, Stephen. 2012. ‘The Singpho agentive–​functions and meanings.’ Linguistics of the Tibeto-​Burman Area 35(1): 1–​14. Morgenroth, Lisa & Martin Salzmann. 2013. ‘Reanalyse syntaktischer Ergativität.’ MS, Universität Leipzig.

 

1208   References Morphy, Frances. 1983. ‘Djapu, a Yolngu dialect,’ in R.M. Dixon & B.J. Blake (eds.), Handbook of Australian languages, vol. 3. Amsterdam & Philadelphia, PA: Benjamins, 1–​189. Mosel, Ulrike. 1987. ‘Subject in Samoan,’ in Donald C. Laycock & Werner Winter (eds.), A world of language: Papers presented to professor S.A. Wurm on his 65th birthday. Canberra: Pacific Linguistics, 455–​479. Mosel, Ulrike & Even Hovdhaugen. 1992. Samoan reference grammar. Oslo:  Scandinavian University Press. Mounole, Céline. 2012. ‘The evolution of transitive verbs in Basque and the emergence of dative-​marked patients,’ in G. Authier & K. Haude (eds.), Ergativity, valency and voice. Berlin: Mouton de Gruyter, 355–​380. Moyse-​Faurie, Claire. 1983. Le drehu, langue de Lifou (Iles Loyauté). Paris: SELAF. Mueller, Jutta L., Anja Hahne, Yugo Fuji, & Angela D. Frederici. 2005. ‘Native and nonnative speakers’ processing of a miniature version of Japanese as revealed by ERPs.’ Journal of Cognitive Neuroscience 17(8): 1229–​1244. Mueller, Jutta L., Masako Hirotani, & Angela D. Friederici. 2007. ‘ERP evidence for different strategies in the processing of case markers in native speakers and non-​native learners.’ BMC Neuroscience 8(1): 18. Mulder, Jean G. 1994. Ergativity in Coast Tsimshian (Sm’algyax), vol. 124. Berkeley, CA: University of California Press. Müller, Friedrich Max. 1887. The science of thought. New York: Longmans. Müller, Gereon. 2009. ‘Ergativity, accusativity, and the order of Merge and Agree,’ in K.K. Grohmann (ed.), Explorations of phase theory: Features and arguments. Berlin: Mouton de Gruyter, 269–​308. Müller, Gereon. 2011. ‘Regeln oder Konstruktionen? Von verblosen Direktiven zur sequentiellen Nominalreduplikation,’ in Stefan Engelberg et al. (eds.), Sprachliches Wissen zwischen Lexikon und Grammatik. Berlin: de Gruyter, 211–​249. Müller, Gereon & Daniela Thomas. 2014. ‘Three-​way systems do not exist.’ MS, University of Leipzig. Munarriz, Amaia, Maria-​ José Ezeizabarrena, & M. Juncal Gutiérrez-​ Mangado. 2014. ‘Differential and selective morpho-​syntactic impairment in Spanish–​Basque bilingual aphasia.’ Bilingualism: Language and Cognition 1–​24. Munshi, Sadaf. 2006. ‘Jammu and Kashmir Burushaski language, language contact, and change.’ PhD dissertation, University of Texas Austin. Munshi, Sadaf. [n.d.]. ‘A grammatical sketch of Hunza Burushaski.’ MS, University of North Texas. Münte, Thomas F., Hans-​Jochen Heinze, & George R. Mangun. 1993. ‘Dissociation of brain activity related to syntactic and semantic aspects of language.’ Journal of Cognitive Neuroscience 5: 335–​344. Murasugi, Kumiko. 1991. ‘The role of transitivity in ergative and accusative languages:  The cases of Inuktitut and Japanese.’ Paper presented at Association of Canadian Universities for Northern Studies, Ottawa. Murasugi, Kumiko. 1992. ‘Crossing and nested paths: NP movement in accusative and ergative languages.’ PhD dissertation, MIT. Murasugi, Kumiko. 2014. ‘A hierarchical view of the ergative and antipassive in Inuktitut.’ A talk presented at Annual meeting of the Canadian Linguistics Association, May 24–​26, Brock University. Musgrave, Simon. 2001a. ‘Non-​subject arguments in Indonesian.’ PhD dissertation, University of Melbourne. Mushin, Ilana & Jane Simpson. 2008. ‘Free to bound to free? Interactions between Pragmatics and syntax in the development of Australian pronominal systems.’ Language 84: 566–​596.

 

References   1209 Mutzafi, Hezy. 2004. The Jewish Neo-​ Aramaic dialect of Koy Sanjaq [Iraqi Kurdistan]. Wiesbaden: Harrassowitz Verlag. Mutzafi, Hezy. 2014. ‘Jewish Neo-​Aramaic,’ Encyclopedia of the Jews of the Islamic world (online edn.). Leiden: Brill. Mylne, Tom. 2000. ‘Argument structure and the status of the complement.’ PhD dissertation, University of Queensland. Næss, Åshild. 2004. ‘What markedness marks: The markedness problem with direct objects.’ Lingua 114(9–​10): 1186–​1212. Næss, Åshild. 2007. Prototypical transitivity. Amsterdam & Philadelphia, PA: Benjamins. Næss, Åshild. 2015. ‘The Äiwoo verb phrase:  Syntactic ergativity without pivots.’ Journal of Linguistics 51(1): 75–​76. Nagano, Yasuhiko. 2003. ‘Tamang,’ in Graham Thurgood & Randy LaPolla (eds.), The Sino-​ Tibetan languages. London: Routledge, 467–​489. Narasimhan, Bhuvana. 2005. ‘Splitting the notion of “agent”:  Case-​marking in early child Hindi.’ Journal of Child Language 32: 787–​803. Narasimhan, Bhuvana. 2013. ‘Ergative case-​marking in Hindi child-​caregiver speech,’ in Edith L. Bavin & Sabine Stolle (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 209–​237. Narasimhan, Bhuvana, Nancy Budwig, & Lalita Murty. 2005. ‘Argument realization in Hindi caregiver–​child discourse.’ Journal of Pragmatics 37(4): 461–​495. Narita, Hiroki. 2011. ‘Phasing in full interpretation.’ PhD dissertation, Harvard University. Narrog, Heiko. 2014. ‘The grammaticalization chain of case functions –​extension and reanalysis of case marking vs. universals of grammaticalization,’ in Silvia Luraghi & Heiko Narrog (eds.), Perspectives on semantic roles (Typological Studies in Language). Amsterdam & Philadelphia, PA: Benjamins, 71–​99. Nash, David. 1982. ‘Warlpiri preverbs and verb roots,’ in S. Swartz (ed.), Papers in Warlpiri grammar: In memory of Lothar Jagst [Work Papers of SIL–​AAB: Series A, 6]. Darwin: SIL, 165–​216. Nash, David. 1986. Topics in Warlpiri grammar. [Outstanding dissertations in linguistics]. New York: Garland. [Published version of 1980 PhD dissertation, MIT.] Nash, David. 1996. ‘Pronominal clitic variation in the Yapa languages: Some historical speculations,’ in William McGregor (ed.), Studies in Kimberley languages in honour of Howard Coate. Munich: Lincom, 117–​138. Nash, Léa. 1995. ‘Portée argumentale et marquage casuel dans les langues SOV et dans les langues ergatives: l’example du géorgien.’ PhD dissertation, Université de Paris VIII. Nash, Léa. 1996. ‘The internal ergative subject hypothesis,’ in Kiyomi Kusumoto (ed.), Proceedings of the North East Linguistic Society 26. Amherst: GLSA, 195–​210. Nash, Léa. 1997. ‘La partition personnelle dans les langues ergatives,’ in Anne Zribi-​Hertz (ed.), Les pronoms. Morphologie, syntaxe et typologie. Saint-​Denis: Presses Universitaires de Vincennes. Nash, Léa. 2016. ‘Aktionsart of unergative verbs in Georgian.’ MS, Paris 8 & CNRS. Nash-​Haran, Léa. 1992. ‘La catégorie AGR et l’accord en géorgien.’ Recherches Linguistiques de Vincennes 21: 65–​79. Naylor, Paz. 1980. ‘Linking, relation-​marking, and Tagalog syntax,’ in Paz B. Naylor (ed.), Austronesian studies:  Papers from the 2nd eastern conference on Austronesian Languages. Ann Arbor, MI: University of Michigan, 33–​49. Nebieridze, Givi. 1987. ‘Arsebobs tu ara ergat’iuli da dat’iuri k’onst’rukciebi kartulši? (Do ergative and dative constructions exist in Georgian?)’ Macne 3: 177–​191.

 

1210   References Nedjalkov, Vladimir P. 1976. ‘Diathesen und Satzstruktur im Tschuktschischen.’ Studia Grammatica 13: 181–​213. Nedjalkov, Vladimir P. 1979. ‘Degrees of ergativity in Chukchee,’ in Frans Plank (ed.), Ergativity:  Towards a theory of grammatical relations. New  York:  Academic Press, 241–​262. Nedjalkov, Vladimir P. 1980. ‘Čukotskij antipassiv i vtoričnaja tranzitivacija.’ Proceedings of the XIV Pacific scientific congress. Committee L., vol. 2. Moscow: USSR Academy of Sciences, 266–​268. Neeleman, Ad. & Kriszta Szendroi. 2007. ‘Radical pro drop and the morphology of pronouns.’ Linguistic Inquiry 38(4): 671–​7 14. Neville, Helen, Janet L. Nicol, Andrew Barss, Kenneth I. Forster, & Merrill F. Garrett. 1991. ‘Syntactically based sentence processing classes: evidence from event-​related brain potentials.’ Journal of Cognitive Neuroscience 3(2): 151–​65. Nevins, Andrew. 2007. ‘The representation of third person and its consequences for person–​case effects.’ Natural Language & Linguistic Theory 25: 273–​313. Nevins, Andrew. 2011. ‘Multiple agree with clitics: Person complementarity vs. omnivorous number.’ Natural Language & Linguistic Theory 29: 939–​971. Nevins, Andrew, Brian Dillon, Shishuka Malhotra, & Colin Phillips. 2007. ‘The role of feature-​ number and feature-​type in processing Hindi verb agreement violations.’ Brain Research 1164: 81–​94. Newmeyer, Fritz. 1986. Linguistic theory in America. New York: Academic Press. Nichols, Johanna. 1980. ‘Control and ergativity in Chechen.’ Chicago Linguistic Society 16: 259–​268. Nichols, Johanna. 1982. ‘Ingush transitivization and detransitivization.’ Proceedings of the eighth annual meeting of the Berkeley Linguistics Society: 445–​462. Nichols, Johanna. 1986. ‘Head-​ marking and Dependent-​ marking Grammar.’ Language 62: 56–​119. Nichols, Johanna. 1992. Linguistic diversity in space and time. Chicago, IL:  University of Chicago Press. Nichols, Johanna. 1993. ‘Ergativity and linguistic geography.’ Australian Journal of Linguistics 13: 39–​89. Nichols, Johanna. 1997. ‘Nikolaev and Starostin’s North Caucasian etymological dictionary and the methodology of long-​range comparison: An assessment.’ Paper presented at 10th Biennial Non-​Slavic Languages Conference, Chicago, May 8–​10. Nichols, Johanna. 2003. ‘The Nakh–​ Daghestanian consonant correspondences,’ in Dee Ann Holisky & Kevin Tuite (eds.), Current trends in Caucasian, East European, and inner Asian linguistics:  Papers in honor of Howard I.  Aronson. Amsterdam & Philadelphia, PA: Benjamins, 207–​251. Nichols, Johanna. 2008. ‘Case in Ingush syntax,’ in Greville G. Corbett & Michael P. Noonan (eds.), Case and grammatical relations: Studies in honor of Bernard Comrie. Amsterdam & Philadelphia, PA: Benjamins, 57–​74. Nichols, Johanna. 2011. Ingush grammar. Berkeley, CA: University of California Press. Nichols, Johanna & Balthasar Bickel. 2013a. ‘Locus of marking in the clause,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology. Nichols, Johanna & Balthasar Bickel. 2013b. ‘Locus of marking: Whole-​language typology,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig: Max Planck Institute for Evolutionary Anthropology.

 

References   1211 Nieuwland, Mante S., Andrea E. Martin, & Manuel Carreiras. 2012. ‘Brain regions that process case: Evidence from Basque.’ Human Brain Mapping 33(11): 2509–​20. Nikolaeva, Irina. 1999. ‘Object agreement, grammatical relations, and information structure.’ Studies in Language 23(2): 331–​376. Nikolaeva, Irina. 2001. ‘Secondary topic as a relation in information structure.’ Linguistics 39 (1): 1–​49. Nilsen, Øystein. 1997. ‘Adverbs and A-​shift.’ Working Papers in Scandinavian Syntax 59: 1–​31. Nöldeke, Theodor. 2001. Compendious Syriac grammar. Winona Lake, IN: Eisenbrauns Noonan, Michael. 2003. ‘Nar-​Phu,’ in Graham Thurgood & Randy LaPolla (eds.), The Sino-​ Tibetan languages. London: Routledge, 336–​352. Nordlinger, Rachel. 1998a. A grammar of Wambaya, Northern Territory (Australia). Canberra: Pacific Linguistics. Nordlinger, Rachel. 1998b. Constructive case: Evidence from Australian languages. Stanford, CA: CSLI. Nordlinger, Rachel. 2000. ‘Australian case systems: Towards a constructive solution,’ in Miriam Butt & Tracy Halloway King (eds.), Argument realization. Stanford, CA: CSLI, 41–​7 1. Nordlinger, Rachel. 2006. ‘Spearing the emu drinking: Subordination and the adjoined relative clause in Wambaya.’ Australian Journal of Linguistics 26(1): 5–​29. Norman, William M. & Lyle R. Campbell. 1978. ‘Toward a Proto-​Mayan syntax: A comparative perspective on grammar,’ in Nora C. England (ed.), Papers in Mayan linguistics. Columbia, MO: University of Missouri, 136–​156. Ochs, Elinor, 1982. ‘Ergativity and word order in Samoan child language.’ Language 58: 646–671. Ochs, Elinor. 1988. Culture and language development:  Language acquisition and language socialization in a Samoan village (Studies in the Social and Cultural Foundations of Language 6). Cambridge: Cambridge University Press. Oda, Kenji. 2002. ‘Wh-​questions in V-​initial languages.’ MA thesis, University of Toronto. O’Connor, Mary C. 1987. ‘Topics in northern Pomo grammar.’ PhD dissertation, University of California, Berkeley. O’Grady, Geoffrey N. 1979. ‘Preliminaries to a proto nuclear Pama–​Nyungan stem list,’ in S.A. Wurm (ed.), Australian linguistic studies (Pacific Linguistics C-​54). Canberra:  Australian National University, 107–​139. O’Grady, Geoffrey N., Charles F. Voegelin, & Florence M. Voegelin. 1966. Languages of the world: Indo-​Pacific fascicle, 6. Anthropological Linguistics 8: 1–​199. O’Shannessy, Carmel. 2005. ‘Light Warlpiri: A new language.’ Australian Journal of Linguistics 25: 31–​57. O’Shannessy, Carmel. 2006. ‘Language contact and children’s bilingual language acquisition:  Learning a mixed language and Warlpiri in northern Australia.’ PhD dissertation, University of Sydney. O’Shannessy, Carmel. 2013. ‘The role of multiple sources in the formation of an innovative auxiliary category in Light Warlpiri, a new Australian mixed language.’ Language 89: 328–​353. Ochs, Elinor. 1982. ‘Ergativity and word order in Samoan child language.’ Language 58: 646–​671. Odria, Ane. 2014. ‘Differential object marking and the nature of dative case in Basque varieties.’ Linguistic Variation 14–​2: 288–​316. Oehrle, Richard. 1976. ‘The grammatical status of the English dative alternation.’ PhD thesis, MIT. Ojima, Shiro, Hiroki Nakata, & Ryusuke Kakigi. 2005. ‘An ERP study of second language learning after childhood: Effects of proficiency.’ Journal of Cognitive Neuroscience 17(8): 1212–​1228. Okell, John. 1969. A reference grammar of colloquial Burmese. London:  Oxford University Press.

 

1212   References Öpengin, Ergin. 2012. ‘Adpositions and argument indexing in the Mukri dialect of Central Kurdish: Focus on ditransitive constructions.’ Orientalia Suecana 61: 187–​198. Öpengin, Ergin. 2013. ‘Clitic/​affix interactions: A corpus-​based study of person marking in the Mukri variety of central Kurdish.’ PhD thesis, Paris III/​Bamberg. Öpengin, Ergin & Geoffrey Haig. 2014. ‘Regional variation in Kurmanji: A preliminary classification of dialects.’ Kurdish Studies 2(2): 143–​176 Ordóñez, Francisco. 1995. ‘The antipassive in Jacaltec:  A  last resort strategy.’ CUNYForum: Papers in Linguistics 19: 95–​108. Ortiz de Urbina, Jon. 1989. Parameters in the grammar of Basque. Providence: Foris. Osterhout, Lee & Phillip J. Holcomb. 1992. ‘Event-​related brain potentials elicited by syntactic anomaly.’ Journal of Memory and Language 31(6): 785–​806. Otsuka, Yuko. 2000. ‘Ergativity in Tongan.’ PhD dissertation, University of Oxford. Otsuka, Yuko. 2005a. ‘Two derivations of VSO: A comparative study of Niuean and Tongan,’ in Andrew Carnie, Heidi Harley, & Sheila A. Dooley (eds.), Verb first: On the syntax of verb-​ initial languages. Amsterdam & Philadelphia, PA: Benjamins, 281–​302. Otsuka, Yuko. 2005b. ‘Scrambling and information focus: VSO–​VOS alternation in Tongan,’ in Joachim Sabel & Mamoru Saito (eds.), The free word order phenomenon: Its syntactic sources and diversity. Berlin: Mouton de Gruyter, 243–​279. Otsuka, Yuko. 2006. ‘Syntactic ergativity in Tongan:  Resumptive pronouns revisited,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity:  Emerging issues. Dordrecht: Springer, 79–​107. Otsuka, Yuko. 2010a. ‘NP ellipsis in Tongan: Is syntactic ergativity real?’ Natural Language and Linguistic Theory 28(2): 315–​342. Otsuka, Yuko. 2010b. ‘Genitive relative constructions and agent incorporation in Tongan,’ in Raphael Mercado, Eric Potsdam, & Lisa Travis (eds.), Austronesian and theoretical linguistics. Amsterdam & Philadelphia, PA: Benjamins, 117–​140. Otsuka, Yuko. 2011a. ‘Neither accusative nor ergative: An alternative analysis of case in eastern Polynesian,’ in Claire Moyse-​Faurie & Joachim Sabel (eds.), Topics in Oceanic morphosyntax. Berlin: Mouton de Gruyter, 289–​318. Otsuka, Yuko. 2011b. ‘PRO vs. null SE: Case, tense, and empty categories in Tongan.’ Syntax 14(3): 265–​296. Otsuka, Yuko. 2014. ‘Relative clauses in Polynesian and the role of D.’ Paper presented at 21st meeting of the Austronesian Formal Linguistics Association. May, Honolulu, University of Hawai‘i. Ouali, Hamid. 2006. ‘On C-​to-​T ɸ-​transfer: The nature of agreement and anti-​agreement in Berber,’ in Roberta D’Alessandro, Susann Fischer, & Gunnar Hrafn Hrafnbjargarson (eds.), Agreement restrictions. Berlin: Mouton de Gruyter, 159–​180. Oyharçabal, Bernard. 1992. ‘Inherent and structural case-​ marking:  Ergaccusativity in Basque,’ in Joseba A. Lakarra & Jon Ortiz de Urbina (eds.), Syntactic theory and Basque syntax. Supplements of the Anuario del Seminario de Filología Vasca Julio de Urquijo 27. Donostia: Gipuzkoako Foru Aldundia, 309–​342. Oyharçabal, Bernard. 1993. ‘Les premières analyses des particularités morphosyntaxiques du basque au 17ème siècle.’ ASJU 27(1): 265–​284. Oyharçabal, Bernard. 1994. ‘Contribution de la comparaison typologique à une analyse des rapports ergativité–​(in)transitivité en basque,’ in J.B. Orpustan (ed.), La langue basque parmi les autres. Baigorri: Izpegi, 115–​148. Ozerov, Pavel. 2014. ‘The System of information packaging in Colloquial Burmese.’ PhD dissertation, La Trobe University, Melbourne.

 

References   1213 Ozerov, Pavel. 2015. ‘Information structure without topic and focus: Differential object marking in Burmese.’ Studies in Language 39(2): 386–​423. Öztürk, Balkız & Markus A. Pöchtrager. 2011. Pazar Laz. Munich: Lincom Europa. Page, Scott E. 2010. Diversity and complexity. Princeton, NJ: Princeton University Press. Palancar, Enrique L. 2002. The origin of agent markers [Studia Typologica 5]. Berlin: Akademie Verlag. Palmer, Frank R. 1994. Grammatical roles and relations. Cambridge:  Cambridge University Press. Pandharipande, Rajeshwari. 1981. ‘Transitivity in Hindi.’ Studies in the Linguistic Sciences 11: 161–​179. Pandharipande, Rajeshwari & Yamuna, Kachru. 1977. ‘Relational grammar, ergativity and Hindi–​Urdu.’ Lingua 41: 217–​238. Pascual, Adán. 2007. ‘Transitividad y dependencia sintáctica y discursiva en q’anjob’al.’ MA thesis, CIESAS, Mexico, DF. Paslawska, Alla & Arnim von Stechow. 2003. ‘Perfect readings in Russian,’ in Artemis Alexiadou (ed.), Perfect explorations. Berlin: Mouton de Gruyter. Paul, Daniel. 2011. ‘A comparative dialectal description of Iranian Taleshi.’ PhD thesis, University of Manchester. Paul, Ileana. 1996. ‘The Malagasy genitive,’ in Matthew Pearson & Ileana Paul (eds.), The structure of Malagasy (UCLA Occasional Papers in Linguistics, 20). Los Angeles: UCLA. Paul, Ileana. 2001. ‘Concealed pseudo-​clefts.’ Lingua 111: 707–​727. Paul, Ileana & Lisa Travis. 2006. ‘Ergativity in Austronesian Languages,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity. Dordrecht: Springer, 315–​335. Paul, Waltraud. 2014. ‘Why particles are not particular: Sentence-​final particles in Chinese as heads of a split CP.’ Studio Linguistica 68: 77–​115 Pawley, Andrew. 2001. ‘Proto Polynesian *CIA,’ in Joel Bradshaw & Kenneth L. Rehg (eds.), Issues in Austronesian Morphology. Canberra: Pacific Linguistics, 193–​216. Payne, Doris L. 1987. ‘Information structuring in Papago narrative discourse.’ Language 63: 783–​804. Payne, John R. 1980. ‘The decay of ergativity in Pamir languages.’ Lingua 51: 147–​186. Payne, John R. 1995. ‘Inflecting postpositions in Indic and Kashmiri,’ in Frans Plank (ed.), Double case: Agreement by Suffixaufnahme. Oxford: Oxford University Press, 283–​298. Payne, Thomas E. 1982. ‘Role and reference related subject properties and ergativity in Yup’ik Eskimo and Tagalog.’ Studies in Language 6(1): 75–​106. Payne, Thomas E. 1993. The twins stories:  Participant coding in Yagua narrative. Berkeley, CA: University of California Press. Payne, Thomas E. & Thomas Laskowske. 1997. ‘Voice in Seko Padang,’ in Joan Bybee, John Haiman, & Sandra A. Thompson (eds.), Essays on language function and language type: Dedicated to T. Givón. Amsterdam & Philadelphia, PA: Benjamins, 423–​436. Pearson, Matthew. 2001. ‘The clause structure of Malagasy: A minimalist approach.’ PhD, dissertation, UCLA. Pearson, Matthew. 2005. ‘The Malagasy subject/​topic as an A′-​element.’ Natural Language and Linguistic Theory 23: 381–​457. Penke, Martina, Helga Weyerts, Matthias Gross, Elke Zander, Thomas F. Mü nte, & Harald Clahsen. 1997. ‘How the brain processes complex words: An event-​related potential study of German verb inflections.’ Cognitive Brain Research 6: 37–​52. Pensalfini, Rob. 1999. ‘The rise of case suffixes as discourse markers in Jingulu: A case study of innovation in an obsolescent language.’ Australian Journal of Linguistics 19(2): 225–​240.

 

1214   References Pensalfini, Rob. 2004. ‘Towards a typology of configurationality.’ Natural Language & Linguistic Theory 22: 359–​408. Perlmutter, David M. 1978. ‘Impersonal passives and the unaccusative hypothesis.’ Proceedings of the 4th Annual Meeting of the Berkeley Linguistics Society 157–​190. Perlmutter, David M. 1980. ‘Relational grammar,’ in Edith Moravcsik & Jessica R. Wirth (eds.), Syntax and semantics: Current approaches to syntax. New York: Academic Press, 195–​229. Perlmutter, David M. & Paul M. Postal. 1984. ‘The I-​advancement exclusiveness law,’ in D.M. Perlmutter & C. Rosen (eds.), Studies in relational grammar, 2. Chicago, IL: University of Chicago Press, 81–​125. Pesetsky, David. 1995. Zero syntax: Experiencers and cascades. Cambridge, MA: MIT Press. Pesetsky, David & Esther Torrego. 2001. ‘T-​to-​C:  Causes and Consequences,’ in Michael Kenstowicz (ed.), Ken Hale: A life in language. Cambridge, MA: MIT Press, 355–​426. Peterson, David A. 2002. ‘Ergativity strata in Kuki-​Chin.’ Unpublished MS, Max Planck Institute for Evolutionary Anthropology. Peterson, David A. 2011. ‘Core participant marking in Khumi.’ Linguistics of the Tibeto-​Burman Area 34(2): 73–​100. Peterson, John M. 1998. Grammatical relations in Pāli and the emergence of ergativity in Indo-​ Aryan. Munich: Lincom. Peterson, Tyler. 2004. ‘The reorganization of semantic roles in Tsimshian connectives,’ in J.C. Brown & T. Peterson (eds.), ICSNL XXXIX: Proceedings of the 39th international conference on Salish and neighbouring languages. Vancouver: UBCWPL, 323–​340. Peterson, Tyler. 2006a. ‘Issues of morphological ergativity in the Tsimshian languages: Determiners, agreement and the reconstruction of case,’ in A. Malchukov, L. Kulikov & P. de Swart (eds.), Case, valency and transitivity (Studies in Language Companion Series, 77). Amsterdam & Philadelphia, PA: Benjamins, 65–​90. Peterson, Tyler. 2006b. ‘Phi-​ determiners, agreement and case in Interior Tsimshianic.’ Qualifying paper, University of British Columbia. Pfeiler, Barbara. 2003. ‘Early acquisition of the verbal complex in Yukatek Maya,’ in D. Bittner, W.U. Dressler, & M. Kilani-​Schoch (eds.), Development of verb inflection in first language acquisition. Berlin: Mouton de Gruyter, 379–​99. Pfeiler, Barbara & Enrique Martín Briceño. 1997. ‘Early verb inflection in Yucatec Maya.’ Papers and Studies in Contrastive Linguistics 33: 117–​125. Phillips, Colin, Nina Kazanina, & Shani H. Abada. 2005. ‘ERP effects of the processing of syntactic long-​distance dependencies.’ Brain Research: Cognitive Brain Research 22(3): 407–​28. Phillips, Maxwell. 2013. ‘Ergative case attrition in central Indo-​Aryan: NP-​splits and the referential hierarchy.’ Studies in Language 196–​216. Picallo, Carme. 1991. ‘Nominals and nominalizations in Catalan.’ Probus 3: 279–​316. Pickering, Martin J. & Roger P. G. van Gompel. 2006. ‘Syntactic parsing,’ Handbook of psycholinguistics. Oxford: Elsevier, 455–​503. Pinker, Steven. 1984. Language learnability and language development. Cambridge, MA: Harvard University Press. Pinker, Steven. 1989. Learnability and cognition:  The acquisition of argument structure. Cambridge, MA: MIT Press. Pischel, Richard. 1900. Grammatik der Prākrit-​Sprachen. Delhi: Motilal Banarsidass. [trans. from German by Subhadra Jha, 1981]

 

References   1215 Plank, Frans. 1979. ‘Ergativity, syntactic typology and universal grammar: Some past and present viewpoints,’ in Frans Plank (ed.), Ergativity. Towards a theory of grammatical relations. London: Academic Press, 3–​36. Plank, Frans. 1985. ‘The extended accusative/​restricted nominative in perspective,’ in Frans Plank (ed.), Relational typology. Berlin: Mouton de Gruyter, 269–​311. Platzack, Christer. 2009. ‘Towards a Minimal Argument Structure,’ in Petra Bernardini, Verner Egerland, & Jonas Granfeldt (eds.), Mélanges plurilingues offerts à Suzanne Schlyter à l’occasion de son 75ème anniversaire. Lund, Sweden:  Språk- och litteraturcentrum, Lund University, 353–​371. Platzack, Christer. 2011. ‘Towards a minimal argument structure.’ MS. Polian, Gilles. 2013. Gramática del tseltal de Oxchuc. Mexico, DF: CIESAS. Polich, John. 2007. ‘Updating P300:  An integrative theory of P3a and P3b.’ Clinical Neurophysiology 118: 2128–​2148. Polinskaja, Maria & Vladimir Nedjalkov. 1987. ‘Contrasting the absolutive in Chukchee: Syntax, semantics, pragmatics.’ Lingua 71: 239–​270. Polinsky, Maria. 1994. ‘Relativization in Chukchee,’ in Howard Aronson (ed.), Linguistic studies in the non-​Slavic languages of the Commonwealth of Independent States and the Baltic republics. Chicago, IL: Chicago Linguistic Society, 241–​262. Polinsky, Maria. 2003. ‘Non-​canonical agreement is canonical.’ Transactions of the Philological Society 101: 279–​312. Polinsky, Maria. 2005. ‘Antipassive constructions,’ in Mathew Dryer & Martin Haspelmath (eds.), The world atlas of language structures. Oxford: Oxford University Press, 438–​441. Polinsky, Maria. 2007. ‘The relationship between coordination and subordination:  Clause chains in Tsez and beyond.’ Talk presented at Conference on Languages of the Caucasus, Leipzig, December. Polinsky, Maria. 2013. ‘Antipassive Constructions,’ in Mathew Dryer & Martin Haspelmath, (eds.), The world atlas of language structures online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. http://​wals.info/​chapter/​108 Polinsky, Maria. 2015. ‘Tsez syntax: A description’. Unpublished MS. Harvard University, Cambridge, MA. Polinsky, Maria. 2016. Deconstructing ergativity: two types of ergative languages and their features. New York, NY: Oxford University Press. Polinsky, Maria. [in press]. ‘Syntactic ergativity,’ The companion to syntax (2nd edn.). Polinsky, Maria & Eric Potsdam. 2001. ‘Long-​distance agreement and topic in Tsez.’ Natural Language and Linguistic Theory 19: 583–​646. Polinsky, Maria & Eric Potsdam. 2002. ‘Backward control.’ Linguistic Inquiry 33: 245–​282. Polinsky, Maria & Omer Preminger. 2014. ‘Case and grammatical relations,’ in Andrew Carnie, Daniel Siddiqi, & Yosuke Sato (eds.), The Routledge handbook of syntax. London: Routledge, 150–​166. Polinsky, Maria, Carlos Gomez Gallo, Peter Graff, & Ekaterina Kravtchenko. 2012. ‘Subject preference and ergativity.’ Lingua 122(3); 267–​277. Pollock, Jean-​Yves. 1989. ‘Verb movement, universal grammar and the structure of IP.’ Linguistic Inquiry 20: 365–​424. Popjes, Jack & Jo Popjes. 1986. ‘Canela-​Krahô,’ in Desmond Derbyshire & Geoffrey Pullum (eds.), Handbook of Amazonian languages, vol. 1. Berlin: Mouton de Gruyter, 128–​199. Porizka, Vincenc. 1967. ‘On the perfective verbal aspect in Hindi.’ Archiv Orientalni 35: 64–​88, 208–​231.

 

1216   References Porizka, Vincenc. 1968. ‘On the perfective verbal aspect in Hindi.’ Archiv Orientalni 36: 233–​251. Porizka, Vincenc. 1969. ‘On the perfective verbal aspect in Hindi.’ Archiv Orientalni 37: 19–​47, 345–​364. Portner, Paul. 2005. ‘The semantics of imperatives within a theory of clause types,’ in K. Watanabe & R.B. Young (eds.), Proceedings of SALT, 14. Ithaca, NY: CLC. Portner, Paul. 2006. ‘Comments on Faller’s Paper.’ Talk presented at Workshop on Philosophy and Linguistics, University of Michigan, Ann Arbor, November 3–​5. Postal, Paul. 1977. ‘Antipassive in French.’ Linguisticae Investigationes 1: 333–​374. Potsdam, Eric. 2006. ‘More concealed pseudoclefts in Malagasy and the clausal typing hypothesis.’ Lingua 116: 2154–​2182. Potsdam, Eric. 2009. ‘Austronesian verb-​initial languages and wh-​question strategies.’ Natural Language and Linguistic Theory 27: 737–​771. Potsdam, Eric & Maria Polinsky. 2011. ‘Questions and word order in Polynesian,’ in Claire Moyse-​Faurie & Joachim Sabel (eds.), Topics in oceanic morphosyntax. Berlin: Mouton de Gruyter, 83–​109. Pott, August Friedrich. 1873. ‘Unterschied eines transitiven und intransitiven Nominativs.’ Bieträge zur vergleichenden Sprachforschung 7: 71–​94. Pray, Bruce R. 1976. ‘From Passive to Ergative in Indo-​Aryan,’ in Manindra K. Verma (ed.), The notion of subject in Indo-​Aryan languages (South Asian Studies, Publication series 2). Madison, WI: University of Wisconsin, 195–​211. Preminger, Omer. 2009. ‘Breaking agreements: Distinguishing agreement and clitic doubling by their failures.’ Linguistic Inquiry 40: 619–​666. Preminger, Omer. 2011a. ‘Agreement as a fallible operation.’ PhD dissertation, MIT. Preminger, Omer. 2011b. ‘Asymmetries between person and number in syntax: A commentary on Baker’s SCOPA.’ Natural Language & Linguistic Theory 29(4): 917–​937 Preminger, Omer. 2012. ‘The absence of an implicit object in unergatives: New and old evidence from Basque.’ Lingua 122: 278–​88. Preminger, Omer. 2014. Agreement and its failures (Linguistic Inquiry Monograph 68). Cambridge, MA: MIT Press. Preminger, Omer & Jaklin Kornfilt. 2013. ‘Nominative as no case at all: An argument from raising-​ to-​ accusative in Sakha.’ Paper presented at 9th Workshop on Altaic Formal Linguistics, Ithaca, NY. Prince, Alan & Paul Smolensky. 2004. Optimality theory: Constraint interaction in generative grammar. Malden, MA: Blackwell. Prince, Ellen F. 1992. ‘The ZPG letter:  Subjects, definiteness, and information-​status,’ in William C. Mann & Sandra A. Thompson (eds.), Discourse description: Diverse linguistic analyses of a fund-​raising text. Amsterdam & Philadelphia, PA: Benjamins, 295–​326. Pucilowski, Anna. 2006. ‘Split-​ergativity in Māori.’ MA thesis, University of Canterbury. Pustejovsky, James. 1995. The generative lexicon. Cambridge, MA: MIT Press. Pye, Clifton. 1980. ‘Person marker acquisition in Quiché Mayan.’ Papers and Reports on Child Language Development, Stanford, 19: 53–​9. Pye, Clifton. 1983. ‘Mayan telegraphese: Intonational determinants of inflectional development in Quiché Mayan.’ Language 59: 583–​604. Pye, Clifton. 1990. ‘The acquisition of ergative languages.’ Linguistics 28: 1291–​1330 Pye, Clifton. 1992. ‘The acquisition of K’iche’ (Maya), in D.I. Slobin (ed.), The crosslinguistic study of language acquisition, vol. 3. Hillsdale, NJ: Erlbaum, 221–​308. Pye, Clifton & Barbara Pfeiler. 2014. ‘The comparative method of language acquisition research: A Mayan case study.’ Journal of Child Language 41: 382–​415.

 

References   1217 Pye, Clifton, Barbara Pfeiler, & Pedro Mateo Pedro. 2013. ‘The acquisition of extended ergativity in Mam, Q’anjob’al and Yucatec,’ in E.L. Bavin & S. Stolle (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 307–​335. Pye, Clifton, Barbara Pfeiler, Lourdes de León, Penelope Brown, & Pedro Mateo Pedro. 2007. ‘Roots or edges? Explaining variation in children’s early verb forms across five Mayan languages,’ in Barbara Pfeiler (ed.), Learning indigenous languages: Child language acquisition in Mesoamerica. Berlin & New York: Mouton de Gruyter, 15–​46. Pylkkänen, Liina. 1999. ‘Causation and external arguments,’ in Liina Pylkkänen, Angeliek van Hout, & Heidi Harley (eds.), Papers from the Penn/​MIT roundtable on the lexicon (MITWPL 35). Cambridge, MA. Pylkkänen, Liina. 2002. ‘Introducing arguments.’ PhD dissertation, MIT. . . . Pylkkänen, Liina. 2008. Introducing arguments. Cambridge, MA: MIT Press. Qìngxìa, Dài & Lon Diehl. 2003. ‘Jinghpo,’ in Graham Thurgood & Randy J. LaPolla (eds.), The Sino-​Tibetan languages. New York: Routledge, 401–​408. Queixalós, Francesc. 2007. ‘Relaciones gramaticales en katukina-​kanamari.’ Forma y Función 20: 95–​146. Queixalós, Francesc. 2008. ‘Incorporation nominale en sikuani et en katukina-​kanamari.’ Queixalós, F. (resp.) Relations grammaticales dans les langues d’Amazonie, Amerindia 31: 61–​86. Queixalós, Francesc. 2010. ‘Grammatical relations in Katukina-Kanamari,’ in Spike Gildea & Francesc Queixalós (eds.), Ergativity in Amazonia (Typological Studies in Language, 89). Amsterdam & Philadelphia, PA: Benjamins, 235–284. Queixalós, Francesc. 2012. ‘Nominalization in Sikuani,’ in A.C. Bruno, F. Pacheco, F. Queixalós, S. Telles, & L. Wetzels (resps.), La structure des langues amazoniennes II, Amerindia 35: 153–​186. Queixalós, Francesc. 2013. L’ergativité est-​elle un oiseau bleu? Munich: Lincom, LSLT 26. Queixalós, Francesc. 2014. ‘Incremental valence changes in an ergative vs. an accusative language: A comparison between Katukina-​Kanamari and Sikuani,’ in F. Queixalós, S. Telles, & A.C. Bruno (eds.), Valence increase in Amazonian languages. Bogota: Instituto Caro & Cuervo, 295–​309. Queixalós, Francesc. [forthcoming]. ‘Valence in Katukina-​Kanamari noun phrases and the nature of genitive classifiers,’ in F. Queixalós & D. Gomes (eds.), A estrutura do sintagma nominal nas línguas amazônicas. Brasilia: Pontes. Queixalós, Francesc & Spike Gildea. 2010. ‘Manifestations of ergativity in Amazonia,’ in Spike Gildea & Francesc Queixalós (eds.), Ergativity in Amazonia (Typological Studies in Language, 89). Amsterdam & Philadelphia, PA: Benjamins, 1–​25. Quesada, Juan Diego. 1999. ‘Ergativity in Chibchan.’ Sprachtypologie und Universalienforschung 52: 22–​51. Quick, Philip. 1997. ‘Active and inverse voice selection criteria in Pendau, a western Austronesian language,’ in Cecilia Odé & Wim Stokhof (eds.), Proceedings of the seventh international conference on Austronesian linguistics, Leiden 22–​27 August 1994. Amsterdam, Atlanta, GA: Rodopi, 461–​481. Quick, Philip. 2007. A grammar of the Pendau language. Canberra: Pacific Linguistics. Rackowski, Andrea. 2002. ‘The structure of Tagalog: Specificity, voice, and the distribution of arguments.’ PhD dissertation, MIT. Rackowski, Andrea & Norvin Richards. 2005. ‘Phase edge extraction: A Tagalog case study.’ Linguistic Inquiry 36: 565–​599. Rackowski, Andrea & Lisa Travis. 2000. ‘V-​initial languages: X or XP movement and adverbial placement,’ in Andrew Carnie & Eithne Guilfoyle (eds.), The syntax of verb initial languages. Oxford: Oxford University Press, 117–​142.

 

1218   References Radford, Andrew & Ingrid Ploennig-​Pacheco. 1995. ‘The morphosyntax of subjects and verbs in child Spanish: A case study.’ Essex Reports in Linguistics 5: 23–​67. Ramchand, Gillian Catriona. 1997. Aspect and predication. Oxford: Oxford University Press. Ramchand, Gillian Catriona. 2008. ‘Perfectivity as aspectual definiteness: Time and the event in Russian.’ Lingua 118: 1690–​1715. Ramchand, Gillian Catriona. 2011. Verb meaning and the lexicon:  A  first phase syntax. Cambridge: Cambridge University Press. Ramchand, Gillian Catriona. 2013. ‘The event domain.’ Talk given at little v Workshop, Leiden. Ramchand, Gillan Catriona & Peter Svenonius. 2014. ‘Deriving the functional hierarchy.’ Language sciences. Oxford: Oxford University Press, 152–​174. Ramirez, Henri. 2003. ‘Ergatividade em yanomami.’ Queixalós, F.  (resp.) Ergativity in Amazonia III. Paris: Centre d’études des langues indigènes d’Amérique (CNRS, IRD)/​ Laboratório de Línguas Indígenas (UnB). Randal, Scott. 2000. ‘Tennet’s ergative origins.’ Occasional papers in the study of Sudanese languages 8: 67–​80. Nairobi: Summer Institute of Linguistics. Rappaport Hovav, Malka, & Beth Levin. 2010. ‘Reflections on manner/​result complementarity,’ in Edith Doron, Malka Rappaport Hovav, & Ivy Sichel (eds.), Syntax, lexical semantics, and event structure. Oxford: Oxford University Press, 21–​38. Räsänen, Sanna Heini Maria, Ben Ambridge, & Julian M. Pine. 2014. ‘Infinitives or bare stems? Are English-​speaking children defaulting to the highest-​frequency form?’ Journal of Child Language 41: 756–​79. Rebuschi, Georges. 1984. Structure de l’enoncé en basque. Paris: SELAF. Rebuschi, Georges. 1999. ‘Le complexe verbal basque:  un regard universaliste.’ Lapurdum 4: 199–​222. Reh, Mechthild 1996. Anywa language:  Description and internal reconstructions (Nilo-​ Saharan: Linguistic Analyses and documentation, 11). Cologne: Rüdiger Köpke. Reid, Lawrence A. 1966. ‘An Ivatan syntax. Sydney: Oceanic Linguistics Special Publications, 1–​160. Reid, Lawrence A. 1978. ‘Problems in the Reconstruction of Proto-​Philippine Construction Markers,’ in S.A. Wurm & Lois Carrington (eds.), Second international conference on Austronesian linguistics: Proceedings, fascicle I —​Western Austronesian. Canberra: Pacific Linguistics. Reinhart, Tanya. 1981. ‘Pragmatics and linguistics: An analysis of sentence topics.’ Philosophica 27: 53–​94. Reinhart, Tanya. 2003. ‘The theta system—​An overview.’ Theoretical Linguistics 28(3): 229–​290. Reinhart, Tanya. 1995. ‘Focus—​The PF interface.’ MS, University of Utrecht. Reinhart, Tanya. 2006. Interface strategies:  Optimal and costly computations. Cambridge, MA: MIT Press. Reinhart, Tanya & Eric Reuland. 1993. ‘Reflexivity.’ Linguistic Inquiry 24: 657–​720. Reinöhl, Uta. 2015. Grammaticalization and configurationality: The emergence of postpositional phrases in Indo-​Aryan. Oxford: Oxford University Press. Reis Silva, M.A. 1996. ‘El aspecto en Mẽbengokre. Consideraciones sobre una construcción progresiva,’ Actas de las segundas Jornadas de Etnolingüística. Rosario, Argentina. Reis Silva, M.A. 2001. ‘Pronomes, ordem e ergatividade em Mẽbengokre.’ MA thesis, Universidade Estadual de Campinas. Reis Silva, Maria Amélia, & Andrés Pablo Salanova. 2000. ‘Verbo y ergatividad escindida en Mẽbengokre,’ in Hein van der Voort & Simon van de Kerke (eds.), Indigenous languages of lowland South America. Leiden:  Research School of Asian, African, and Amerindian Studies, Leiden University, 225–​242.

 

References   1219 Reuland, Eric (ed.). 2000. Arguments and case: Explaining Burzio’s generalization. Amsterdam & Philadelphia, PA: Benjamins. Rezac, Milan. 2003. ‘The fine structure of cyclic Agree.’ Syntax 6: 156–​182. Rezac, Milan. 2004. ‘The EPP in Breton: An unvalued categorial feature,’ in A. Triggers, H. Breitbarth, & H. van Riemdijk. Berlin: Mouton de Gruyter. Rezac, Milan. 2008. ‘The syntax of eccentric agreement: The person case constraint and absolutive displacement in Basque.’ Natural Language and Linguistic Theory 26: 61–​106. Rezac, Milan. 2011. Phi-​features and the modular architecture of language. Dordrecht: Springer. Rezac, Milan, Pablo Albizu, & Ricardo Etxepare. 2014. ‘The structural ergative of Basque and the theory of case.’ Natural Language and Linguistic Theory 32(4): 1273–​1330. Rhee, Jaemin & Ken Wexler. 1995. ‘Optional infinitives in Hebrew,’ in Carson Schütze, Jennifer Ganger, & Kevin Broihier (eds.), Papers on language processing and acquisition. MIT working papers in linguistics 26: 383–​402. Richards, Norvin. 1997. ‘What moves where when in which language?’ PhD dissertation, MIT. Richards, Norvin. 2000. ‘Another look at Tagalog subjects,’ in Ileana Paul, Vivianne Philipps, & Lisa Travis (eds.), Formal issues in Austronesian linguistics. Dordrecht: Kluwer, 105–​116. Richards, Norvin. 2009. ‘Nouns, verbs, and hidden structure in Tagalog.’ Theoretical Linguistics 35: 139–​152. Richards, Norvin. 2010. Uttering trees. Cambridge, MA: MIT Press. Richards, Norvin. 2013. ‘Tagalog anaphora,’ in Lisa Lai-​Shen Cheng & Norbert Corver (eds.), Diagnosing syntax (Oxford Studies in Theoretical Linguistics, vol. 46). Oxford:  Oxford University Press, 412–​433. Rigsby, Bruce. 1986. ‘Gitksan grammar.’ Unpublished MS, University of Queensland, Australia. Rigsby, Bruce & John Ingram. 1990. ‘Obstruent voicing and glottalic obstruents in Gitksan.’ International Journal of American Linguistics 56: 251–​263. Rigsby, Bruce & Noel Rude. 1996. ‘Sketch of Sahaptin, a Sahaptian language,’ in Ives Goddard (ed.), Handbook of North American Indians, vol. 17. Smithsonian Institution: Washington DC, 666–​692. Rizzi, Luigi. 1986. ‘Null objects in Italian and the theory of pro.’ Linguistic Inquiry 17: 501–​558. Rizzi, Luigi. 1990a. ‘On the anaphor–​agreement effect.’ Revista di Linguistica 2: 27–​42. Rizzi, Luigi. 1990b. Relativized minimality. Cambridge, MA: MIT Press. Rizzi, Luigi. 1994. ‘Early null subjects and root null subjects,’ in T. Hoekstra & B. Schwartz (eds.), Language acquisition studies in generative grammar. Amsterdam & Philadelphia, PA: Benjamins. Rizzi, Luigi. 1997. ‘The fine structure of the left periphery,’ in Liliane Haegeman (ed.), Elements of grammar: Handbook of generative syntax. Dordrecht: Kluwer, 281–​337. Rizzi, Luigi. 2004. ‘Locality and left periphery,’ in Adriana Belletti (ed.), Structures and beyond:  The cartography of syntactic structures. Oxford:  Oxford University Press, 223–​251. Roberts, Craige. 1998 [1996]. ‘Information structure in discourse: Towards an Integrated formal theory of pragmatics,’ in J.H. Yoon & A. Kathol (eds.), Papers in semantics: Ohio State University working papers in linguistics, vol. 49. Roberts, Ian G. 2007. Diachronic syntax. Oxford: Oxford University Press. Roberts, Ian G. 2012. ‘On the nature of syntactic parameters: A programme for research,’ in Charlotte Galves, Sonia Cyrino, Ruth Lopes, & Juanito Avelar (eds.), Parameter theory and linguistic change. Oxford: Oxford University Press, 319–​334 Robertson, John S. 1983. ‘From symbol to icon: The evolution of the pronominal system from common Mayan to modern Yucatecan.’ Language 59(3): 529–​540.

 

1220   References Robertson, John S. 1992. The history of tense/​aspect/​mood/​voice in the Mayan verbal complex. Austin: University of Texas Press. Roeper, Thomas. 1987. ‘Implicit arguments and the head–​complement relation.’ Linguistic Inquiry 18: 267–​310. Rooth, Mats. 1992. ‘A theory of focus interpretation.’ Natural Language Semantics 1: 75–​116. Rosen, Carol. 1984. ‘The interface between semantic roles and initial grammatical relations,’ in David Perlmutter & Carol Rosen (eds.), Studies in relational grammar 2. Chicago, IL: University of Chicago Press, 38–​77. Ross, John Robert. 1967. ‘Constraints and variables in syntax.’ PhD dissertation, MIT. Ross, John Robert. 1970. ‘On declarative sentences,’ in Roderick A. Jacobs & Peter S. Rosenbaum (eds.), Readings in English transformational grammar. Waltham, MA: Ginn, 222–​272. Ross, Malcolm. 2002. ‘The history and transitivity of western Austronesian voice and voice marking,’ in Fay Wouk & Malcolm Ross (eds.), The history and typology of western Austronesian voice systems. Canberra: Pacific Linguistics, Research School of Pacific Asian Studies, ANU, 17–​63. Ross, Malcolm. 2006. ‘Reconstructing the case-​marking and personal pronoun systems of Proto-​Austronesian,’ in Yongli Zhang (ed.), Streams converging into an ocean:  Festschrift in honor of Professor Paul Jen-​kuei Li on his 70th birthday (Language and Linguistics Monographs, vol. 5). Institute of Linguistics, Academica Sinica, 521–​563. Ross, Malcolm. 2009. ‘Proto Austronesian verbal morphology: A reappraisal,’ in K. Alexander Adelaar & Andrew Pauley (eds.), Austronesian historical linguistics and culture history: A festschrift for Robert Blust (Pacific Linguistics 601). Canberra: Research School of Pacific and Asian Studies, Australian National University, 295–​326. Ross, Malcolm & Stacy Fang-​ching Teng. 2005. ‘Formosan languages and linguistic typology.’ Language and Linguistics 6: 739–​781. Rude, Noel Emerson. 1985. ‘Studies in Nez Perce grammar and discourse.’ PhD thesis, University of Oregon. Rude, Noel. 1988. ‘Ergative, passive, and antipassive in Nez Perce,’ in M. Shibatani (ed.), Passive and voice. Amsterdam & Philadelphia, PA: Benjamins, 547–​560. Rude, Noel. 1991. ‘On the origin of the Nez Perce ergative NP suffix.’ International Journal of American Linguistics 57(1): 24–​50. Rude, Noel. 1997. ‘On the history of nominal case in Sahaptian.’ International Journal of American Linguistics 63(1): 113–​143. Rudenko, B. T. 1940. Grammatika gruzinskogo jazyka. Moscow: Nauka. Ruffolo, Roberta. 2004. ‘Topics in morpho-​syntax of Ibaloy, Northern Philippines.’ PhD thesis, Canberra, Australian National University. Rumsey, Alan. 2010. ‘ “Optional” ergativity and the framing of reported speech.’ Lingua 120: 1652–​1676. Rumsey, Alan, Lila San Roque, & Bambi B. Schieffelin 2013. ‘The acquisition of ergative marking in Kaluli, Ku Waru and Duna (Trans New Guinea),’ in Edith L. Bavin & Sabine Stoll (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA:  Benjamins, 133–​182. Sabar, Yona. 1984. Homilies in the Neo-​Aramaic of the Jews of Kurdistan for the biblical portions of Wayhi (Genesis), Beshallah and Yitro (Exodus). Jerusalem: Israel Academy of Sciences and Humanities. Sadock, Jerrold. 1969. ‘Hypersentences.’ Papers in Linguistics 1: 283–​370. Sadock, Jerrold M. (1980). ‘Noun incorporation in Greenlandic: A case of syntactic word formation.’ Language 56: 300–​319.

 

References   1221 Sadock, Jerrold M. (1999). ‘The nominalist theory of Eskimo: A case study in scientific self-​ deception.’ International Journal of American Linguistics 65(4): 383–​406. Sadock, Jerrold M. (2003). A grammar of Kalaallisut (West Greenlandic Inuttut). Languages of the World/​Materials. Munich: Lincom. Saddy, Doug. 1991. ‘Wh-​scope mechanisms in Bahasa Indonesia,’ in Lisa Lai-​Shen Cheng & Hamida Demirdache (eds.), More papers on wh-​movement. MIT Working Papers in Linguistics, 183–​218. Salanova, Andrés Pablo. 2007. ‘Nominalizations and aspect.’ PhD dissertation, MIT. Salanova, Andrés Pablo. 2008a. ‘Nominalizations and ergativity.’ Proceedings of the 13th workshop on the structure and constituency of the languages of the Americas (WSCLA). Salanova, Andrés Pablo. 2008b. ‘Uma análise unificada das construções ergativas do Mẽbengokre.’ Amérindia 32: 109–​134. Salanova, Andrés Pablo. 2011a. ‘A flexão de terceira pessoa nas línguas Jê.’ LIAMES 11: 75–​114. Salanova, Andrés Pablo. 2011b. ‘Relative clauses in Mẽbengokre,’ in Pieter Muysken, Rik van Gijn, & Katharina Haude (eds.), Subordination in South American languages. Amsterdam & Philadelphia, PA: Benjamins, 45–​78. Salas, Naymé. 2010. ‘Developmental and discursive underpinnings in the expression of Spanish NPs.’ University of Barcelona. Samvelian, Pollet. 2006. ‘What Sorani Kurdish absolute prepositions tell us about cliticization,’ in Frederic Hoyt, Nikki Seifert, Alexandra Teodorescu, & Jessica White (eds.), Texas Linguistics Society IX: The morphosyntaxe of understudied languages. Stanford, CA: CSLI Online Publications. Samvelian, Pollet. 2007. ‘A lexical account of Sorani Kurdish prepositions,’ in S. Muller (ed.), Proceedings of the 14th international conference on head-​driven phrase structure grammar. Stanford, CA: CSLI, 235–​249. Sandefur, John. 1984. ‘A language coming of age: Kriol of North Australia.’ MA dissertation, University of Western Australia, Perth. dos Santos, Ludoviko. 1997. ‘Desçar̃ao da morfossintaxe da língua Suyá/​Kĩsêdjê (Jê).’ PhD dissertation, Universidade Federal de Santa Catarina, Florianópolis. San Martin, Itziar & Juan Uriagereka. 2002. ‘Infinitival complementation in Basque,’ in X. Artiagoitia, P. Goenaga, & J.A. Lakarra (eds.), Erramu Boneta. Festschrift for Rudolf P. de Rijk. Bilbao: UPV-​EHU, 597–​610. Sands, Kristina. 1996. The ergative in Proto Australian. Munich: Lincom Europa. Santazilia, Ekaitz. 2013. ‘Noun morphology,’ in M. Martinez Areta (ed.), Basque and Proto-​ Basque:  Language-​ internal and typological approaches to linguistic reconstruction. New York: Peter Lang, 223–​281. Sasse, Hans-​Jürgen. 1996. ‘Theticity.’ Working Papers 27, Institute of Linguistics, University of Cologne, 1–​55. Sawada, Hideo. 1995. ‘On usages and functions of particles –​kou/​ka in colloquial Burmese.’ Studies in Language 37(4): 693–​735. Sawada, Hideo. 2012. ‘Optional marking of NPs with core case functions P and A in Lhaovo.’ Linguistics of the Tibeto-​Burman Area 35(1): 15–​34. Saxena, Anju. 1989. ‘Ergative or nominative–​accusative? The Tibeto-​Burman case.’ Proceedings of the fourth meeting of the Pacific linguistics conference. Oregon: University of Oregon, 420–​447. Say, Sergey. 2005. ‘Antipassive  –​sja verbs in Russian:  Between inflection and derivation,’ in Wolfgang U. Dressler et al. (eds.), Morphology and its demarcations: Selected papers from the 11th morphology meeting, Vienna, February 2004. Amsterdam & Philadelphia, PA: Benjamins, 253–​275.

 

1222   References Scancarelli, Janine. 1985. ‘Referential strategies in Chamorro narratives: Preferred clause structure and ergativity.’ Studies in Language 9(3): 335–​362. Schachter, Paul. 1976. ‘The subject in Philippine languages: Topic, actor, actor-​topic, or none of the above,’ in Charles Li (ed.), Subject and topic. New York: Academic Press, 491–​518. Schachter, Paul. 1996. ‘The subject in Tagalog: Still none of the above,’ UCLA occasional papers in linguistics, vol. 15. Los Angeles: Department of Linguistics, UCLA, 1–​61. Schachter, Paul & Fe T Otanes. 1982. Tagalog reference grammar. University of California Press. Schäfer, Florian. 2008. The syntax of (anti-​)causatives: External arguments in change-​of-​state contexts. Amsterdam & Philadelphia, PA: Benjamins. Schäfer, Florian. 2012a. ‘Local case, cyclic agree and the syntax of truly ergative verbs,’ in A. Alexiadou, T. Kiss, & G. Müller (eds.), Local modelling of non-​local dependencies in syntax. Tübingen: Niemeyer, 273–​304. Schäfer, Florian. 2012b. ‘Two types of external argument licensing: The case of causers.’ Studia Linguistica 66(2): 128–​180. Schieffelin, Bambi B. 1981. ‘A developmental study of the pragmatic appropriateness of word order and casemarking in Kaluli,’ in W. Deutsch (ed.), The child’s construction of language. London: Academic Press, 105–​120. Schieffelin, Bambi B. 1985. ‘The acquisition of Kaluli,’ in D.I. Slobin (ed.), The crosslinguistic study of language acquisition, vol. 1. Hillsdale, NJ: Erlbaum, 525–​593. Schikowski, Robert F., Netra P. Paudyal, & Balthasar Bickel. [July 2012 draft]. ‘Flexible valency in Chintang,’ accessed from http://​www.clrp.uzh.ch/​downloads/​schikowski-​chintang-​ valency.pdf Schmidt, Annette. 1985. Young people’s Dyirbal: An example of language death from Australia. Cambridge: Cambridge University Press. Schmidt, Bodil Kappel. 2003. ‘West Greenlandic antipassive.’ Nordlynd 31(2): 385–​399. Schneider, Lucien. 1972. Inuktituorutit: grammaire purement esquimaude. Québec: Ministère des Richesses Naturelles. Schuchardt, Hugo. 1895. ‘Über den passiven Charakter des Transitivs in den kaukasischen Sprachen.’ Sitzungsberichte der philosophisch-historischen Classe der Kaiserlichen Akademie der Wissenschaften. Wien 133: 1–91. Schuchardt, Hugo. 1896a. ‘Kharthwelische Sprachwissenschaft I.’ Wiener Zeitschrift für die Kunde des Morgenlandes 10: 115–​128 Schuchardt, Hugo. 1896b. ‘Kharthwelische Sprachwissenschaft II.’ Wiener Zeitschrift für die Kunde des Morgenlandes 10: 309–​322 Schuchardt, Hugo. 1896c. ‘Georgisches -​qe,’ Mélanges Charles de Harlez. Leiden: Brill, 278–​280. Schuchardt, Hugo. 1897. ‘Kharthwelische Sprachwissenschaft III.’ Wiener Zeitschrift für die Kunde des Morgenlandes 11, 167–​180. Schulman, Bari Wieselman, Carolyn Mylander, & Susan Goldin-​Meadow. 2001. ‘Ergative structure at sentence and discourse levels in a self-​generated communication system,’ in Anna H.-​J. Do, Laura Domínguez, & Aimee Johansen (eds.), Proceedings of the 25nd annual Boston University conference on language development, vol. 2. Somerville, MA: Cascadilla Press, 815–​824. Schultze-​Berndt, Eva. 2000. ‘Simple and complex verbs in Jaminjung: A study of event categorisation in an Australian language.’ PhD thesis, Catholic University of Nijmegen. Schultze-​ Berndt, Eva. 2001. ‘Ideophone-​ like characteristics of uninflected predicates in Jaminjung (Australia),’ in F.K.E. Voeltz & C. Kilian-​Hatz (eds.), Ideophones. Amsterdam & Philadelphia, PA: Benjamins, 355–​373.

 

References   1223 Schultze-​Berndt, Eva. 2006a. ‘Sketch of a Jaminjung grammar of space,’ in S.C. Levinson & D.P. Wilkins (eds.), Grammars of space:  Explorations in linguistic diversity. Cambridge: Cambridge University Press, 63–​114. Schultze-​Berndt, Eva. 2006b. ‘Towards a semanto-​pragmatic account of optional ergativity in Jaminjung/​Ngaliwurru.’ Paper presented at Second European Workshop on Australian Languages, Somlószöllös, Hungary, April 23–25. Schultze-​Berndt, Eva. 2007. ‘Making sense of complex verbs: On the semantics and argument structure of closed-​class verbs and coverbs in Jaminjung,’ in M. Bowerman & P. Brown (eds.), Cross-​linguistic perspectives on argument structure:  Implications for learnability. Hillsdale, NJ: Erlbaum, 69–​88. Schultze-​Berndt, Eva. 2010. ‘Grammatical properties and classification of three-​participant predicates in Jaminjung,’ in A. Malchukov, M. Haspelmath, & B. Comrie (eds.), Studies in ditransitive constructions. Berlin: Mouton de Gruyter, 510–​528. Schultze-​Berndt, Eva. 2012. Pluractional posing as progressive: A construction between lexical and grammatical aspect. Australian Journal of Linguistics 32(1): 7–​39. Schultze-​Berndt, Eva. 2015. ‘Complex verbs, simple alternations: Valency and verb classes in Jaminjung,’ in A. Malchukov & B. Comrie (eds.), Valency classes in the world’s languages, volume 2: Case studies from New Guinea, Australia, and the Americas, and theoretical outlook. Berlin: Mouton de Gruyter, Schultze-​Berndt, Eva, Felicity Meakins, & Denise Angelo. 2013. ‘Kriol,’ in S. Michaelis, P. Maurer, M. Huber, & M. Haspelmath (eds.), The atlas of pidgin and creole language structures (APiCS). Oxford: Oxford University Press. Schultze-​Berndt, Eva & Candide Simard. 2012. Constraints on noun phrase discontinuity in an Australian language:  The role of prosody and information structure. Linguistics 50(5): 1015–​1058. Schulze, Wolfgang. 1994. ‘The Udi language: A grammatic sketch’ in Rieks Smeets (ed.), The Indigenous Languages of the Caucasus 4: 447–​514. Schulze, Wolfgang. 1988. ‘Noun classification and ergative construction in East-​Caucasian languages,’ in Fridrik Thordarson (ed.), Proceedings of the third Caucasian colloquium, Oslo, July 1986. Oslo: Norwegian University Press, 251–​274.A north Caucasian etymological dictionary.’ Diachronica Schulze, Wolfgang. 2000. ‘Towards a typology of the accusative ergative continuum: The case of East Caucasian.’ General Linguistics 37: 71–​155. Schulze, Wolfgang. 2007. ‘Personalität in den ostkaukasischen Sprachen’. Munich Working Papers in Cognitive Typology 4, IATS University of Munich. Schulze, Wolfgang. 2011. ‘The Lak language.’ Nová fiilologiická revue 3: 11–​36. Schütze, Carson. 2001. ‘On the nature of default case.’ Syntax 4(3): 205–​238. Schwartz, Arthur. 1972. ‘The VP-​constituent of SVO-​languages.’ Syntax and Semantics 1: 213–​236. Schwarzschild, Roger. 1999. ‘GIVENness, AvoidF and other constraints on the placement of accent.’ Natural Language Semantics 7: 141–​177. Scott, Graham. 1978. The Fore language of Papua New Guinea. Canberra: Pacific Linguistics. Scott, Graham. 1986. ‘On ergativity in Fore and other Papuan language.’ Papers in New Guinea Linguistics 24: 167–​175. Searle, John R. 2013. ‘Theory of mind and Darwin’s legacy.’ Proceedings of the National Academy of Sciences 110: 10343–​10348. Seely, Jonathan. 1977. ‘An ergative historiography.’ Historiographia Linguistica 4: 191–​206. Seiler, Hansjakob. 1989. The dimension of participation. Función 7, Universidad de Guadalajara. [transl. of 1984 MS Die Dimension der Partizipation]

 

1224   References Seiler, Walter. 1985. Imonda, a Papuan Language. (Pacific Linguistics, Series B, 93.) Canberra: Australian National University. Seiler, Hansjakob. (1989) The dimension of participation. Función 7, Universidad de Guadalajara. [transl. of 1984 MS Die Dimension der Partizipation] Seiter, William. 1980. Studies in Niuean Syntax. New York: Garland. Serdobolskaya, Natalia. 2009. ‘Towards the typology of raising:  A  functional approach,’ in Alexander Arkhipov & Patience Epps (eds.), New challenges in typology. Berlin: de Gruyter, 245–​270. Sériot, Patrick (ed.). 2005. Un paradigme perdu:  La linguistique marriste [= Cahiers de l’Institut de Linguistique et des Sciences du Langage, 20]. Lausanne: Université de Lausanne, 263–​294 Seržant, Ilja. 2013. ‘Rise of canonical subjecthood,’ in Ilja A. Seržant & Leonid Kulikov (eds.), The diachronic typology of non-​ prototypical subjects. Amsterdam & Philadelphia, PA: Benjamins, 283–​310. Shackle, Christopher. 1976. The Siraiki language of central Pakistan:  A  reference grammar. Oxford: School of Oriental and African Studies. Shanidze, Ak’ak’i. 1953. kartuli gramat’ik’is sapudzvlebi, I: morpologia [‘The fundamentals of Georgian grammar I: morphology’]. Tbilisi: TSUG. Shanshovani, Zurab. 1737. mok’le grammat’ik’a kartuli enisa [‘Concise grammar of the Georgian language’] (repr. 1881). St. Petersburg: Imperatorskaja Akademija Nauk. Sharadzenidze, Tinatin. 1954. ‘saxelta mravlobiti ricxvis c’armoeba svanurši balszemouri k’ilos mixedvit’ [‘Formation of the plural number of the noun in the Upper Bal dialect of Svan’]. Ibero-​k’avkasiuri enatmecniereba 6: 189–​203. Sheehan, Michelle. 2014a. ‘Portuguese, Russian and the theory of control,’ in Hsin-​Lun, Huang, Ethan Poole, & Amanda Rysling (eds.), Proceedings of the 43rd annual meeting of the North East Linguistic Society (NELS 43). Amherst, MA: GLSA, 115–​26. Sheehan, Michelle. 2014b. ‘Towards a parameter hierarchy for alignment,’ in Robert E. Santana-​ LaBarge (ed.), Proceedings of the 31st West Coast conference on formal linguistics. Somerville, MA: Cascadilla Press, 399–​408. Sheehan, Michelle. [to appear]. ‘Case and ergativity,’ in Antonio Fábregas, Jaume Mateu, & Mike Putnam (eds.), Contemporary linguistic parameters. London: Bloomsbury. Sherkina-​Lieber, Marina. 2004. ‘Focus fronting in wh-​questions in Inuktitut.’ Toronto Working Papers in Linguistics 23(1): 119–​132. Shibasaki, Reijirou. 2006. ‘The evolution of preferred argument structure in English:  With focus on referential forms and information status.’ English Linguistics 23(1): 1–​26. Shibatani, Masayoshi. 1977. ‘Grammatical relations and surface cases.’ Language 53: 789–​807. Shibatani, Masayoshi. 1988. ‘Voice in Philippine languages,’ in Masayoshi Shibatani (ed.), Passive and voice. (Typological Studies in Language, vol. 16.) Amsterdam & Philadelphia, PA: Benjamins, 85–​142. Shibatani, Masayoshi. 1991. ‘Grammaticization of topic into subject,’ in Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization, vol. 2. Amsterdam & Philadelphia, PA: Benjamins, 93–​133. Shibatani, Masayoshi. 2008. ‘Relativization in Sasak and Sumbawa, Eastern Indonesia.’ Language and Linguistics 9: 865–​916. Shimamura, Koji. 2014a ‘Anticausativisation in Icelandic via –​st: Dative–​nominative alternation and reflexivization.’ MS, UConn.

 

References   1225 Shimamura, Koji. 2014b. ‘Syntactically ergative = morphologically accusative.’ Paper presented at Glow in Asia X, National Tsinghua University, Taiwan. Shiohara, Asako. 2013. ‘Voice system in Sumbawa,’ Proceedings of the workshop on Indonesian-​ type voice system. TUFS, 131–​142. Shirani, Mahmood. 1987. ‘urdu ki shakh haryani zaban mein talifat,’ in Mazhar Mahmood Shirani (ed.), maqaalaat-​e-​hafiz mahmood shirani [‘Papers of Hafiz Mahmood Shirani’]. Lahore: Majilis-​e-​Taraqqi-​e-​Adab. Shojai, Behrooz. 2005. ‘The Kurdish of Gilan.’ MA thesis, Uppsala University. Siegel, Laura. 2000. ‘Ergativity and semantic bootstrapping.’ Presented at the LSA annual meeting, Chicago, January. Siewierska, Anna. 2004. Person. Cambridge: Cambridge University Press. Siewierska, Anna. 2005 ‘Alignment of person agreement marking,’ in M. Haspelmath, M.S. Dryer, D. Gil, & B. Comrie (eds.), World atlas of language structure. Oxford: Oxford University Press, 406–​409. Siewierska, Anna. 2011. ‘Person marking,’ in J. J. Song (ed.), The Oxford handbook of linguistic typology. Oxford: Oxford University Press, 322–​345. Siewierska, Anna. 2013. ‘Alignment of verbal person marking,’ in Matthew S. Dryer & Martin Haspelmath (eds.), The world atlas of language structures online. Leipzig:  Max Planck Institute for Evolutionary Anthropology. Siewierska, Anna & Dik Bakker. 2009. ‘Case and alternative strategies: Word order and agreement marking,’ in Andrej Malchukov & Andrew Spencer (eds.), The Oxford handbook of case. Oxford: Oxford university Press, 290–​303. Sigurðsson, Halldór Ármann. 2002. ‘To be an oblique subject: Russian vs. Icelandic.’ Natural Language and Linguistic Theory 20: 691–​724. Silverstein, Michael. 1972. ‘Chinook Jargon: Language contact and the problem of multi-​level generative systems, I.’ Language 48: 378–​406. Silverstein, Michael. 1976. ‘Hierarchy of features and ergativity,’ in Robert M.W. Dixon (ed.), Grammatical categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 112–​171. Silverstein, Michael. 1981. ‘Case marking and the nature of language.’ Australian Journal of Linguistics 1: 227–​244. Silverstein, Michael. 1986. ‘Hierarchy of features and ergativity,’ in P. Muysken & H.C. van Riemsdijk (eds.), Features and projections. Dordrecht: Foris, 163–​232. Simard, Candide. 2010. ‘The prosodic contours of Jaminjung, a language of Northern Australia.’ PhD dissertation, University of Manchester, Manchester. Simard, Candide. 2013a. ‘Marking boundaries:  Intonation units and prosodic sentences.’ Paper presented at proceedings of conference Prosody–​ Discourse Interface, Leuven, September 11–​13. Simard, Candide. 2013b. ‘Prosody and function of “iconic lengthening” Jaminjung.’ SOAS working papers in linguistics, 16. Simard, Candide. 2014. ‘Another look at right detached NPs’. Paper presented at proceedings of conference Language Documentation and Linguistic Theory 4, SOAS, London. Simpson, Jane. 1988. ‘Case and complementiser suffixes in Warlpiri,’ in Peter Austin (ed.), Complex sentence constructions in Australian languages (Typological studies in language, 15). Amsterdam & Philadelphia, PA: Benjamins, 205–​218. Simpson, Jane. 1991. Warlpiri morpho-​syntax: A lexicalist approach. Dordrecht: Kluwer. [published version of 1983 PhD dissertation, MIT]

 

1226   References Simpson, Jane. 2005. ‘Depictives in English and Warlpiri,’ in Nikolaus P. Himmelmann & Eva Schultze-​Berndt (eds.), Secondary predication and adverbial modification. Oxford: Oxford University Press, 69–​106. Simpson, Jane. 2007. ‘Expressing pragmatic constraints on word order in Warlpiri,’ in Annie Zaenen, Jane Simpson, Tracy Holloway King, Jane Grimshaw, & Joan Maling (eds.), Architectures, rules, and preferences:  Variations on themes by Joan W.  Bresnan. Stanford, CA: CSLI, 403–​427. Simpson, Jane. 2009. ‘Pama–​Nyungan: Warlpiri,’ in Rochelle Lieber & Pavol Stekauer (eds.), The Oxford handbook of compounding. New York: Oxford University Press, 609–​622. Simpson, Jane. 2012. ‘Information structure, variation and the referential hierarchy,’ in F. Seifart, G. Haig, N.P. Himmelmann, D. Jung, A. Margetts, & P. Trilsbeek (eds.), Potentials of language documentation: Methods, analyses, and utilization. Language Documentation & Conservation Special Publication No. 3: 73–​82. Simpson, Jane. 2014 ‘Pama–​Nyungan,’ in Rochelle Lieber & Pavol Stekauer (eds.), The Oxford handbook of derivational morphology. Oxford: Oxford University Press, 651–​668. Simpson, Jane & Joan Bresnan. 1983. ‘Control and obviation in Warlpiri.’ Natural Language and Linguistic Theory 1: 49–​64. Simpson, Jane, Rob Amery, & Mary-​Anne Gale (eds.). 2008. ‘I could have saved you linguists a lot of time and trouble:  180  years of research and documentation of South Australia’s Indigenous languages, 1826–​2006,’ in William B. McGregor (ed.), Encountering Aboriginal languages:  Studies in the history of Australian linguistics. Canberra:  Pacific Linguistics, 85–​144. Singh, Jag Deva. 1970. A descriptive grammar of Bangru. Kurukshetra: Kurukshetra University Press. Singh, Mona. 1998. ‘On the Semantics of the Perfective Aspect.’ Natural Language Semantics 6: 171–​199. Sinnemäki, Kaius. 2014. ‘A typological perspective on differential object marking.’ Linguistics 52(2): 281–​313. Sirk, Ülo. 1996. The Buginese language. Moscow: Nauka. Skjærvø, Prods Oktor. (2009). ‘Old Iranian,’ in Gernot Windfuhr (ed.), The Iranian languages. London: Routledge, 43–​196. Skopeteas, Stavros, Gisbert Fanselow, & Rusudan Asatiani. 2012. ‘Case inversion in Georgian: Syntactic properties and sentence processing,’ in M. Lamers & H. de Swart (eds.), Case, Word Order and Prominence. Dordrecht: Springer, 145–​171 Skorik, Piotr Ja. 1948. Očerki po sintaksisu čukotskogo jazyka:  inkorporacija. Leningrad: Nauka. Skorik, Piotr Ja. 1961. Grammatika čukotskogo jazyka, čast’ I: Fonetika i morfologija imennyx častej reči. Leningrad: Izd. AN SSSR. Skorik, Piotr Ja. 1977. Grammatika čukotskogo jazyka, čast’ II: Glagol, narečie, služebnye slova. Leningrad: Izd. AN SSSR. Slobin, Dan. 1973. ‘Cognitive prerequisites for the development of grammar,’ in Charles Ferguson & Dan Slobin (eds.), Studies of child language development. New  York:  Holt, Rinehart & Winston, 241–​249. Slobin, Dan. (ed.) 1985. The crosslinguistic study of language acquisition, vols. 1–​2. Hillsdale, NJ: Erlbaum. Slobin, Dan. (ed.) 1992. The crosslinguistic study of language acquisition, vol. 3. Hillsdale, NJ: Erlbaum. Slobin, Dan. (ed.) 1997. The crosslinguistic study of language acquisition, vols. 4–​5. Mahwah, NJ: Erlbaum.

 

References   1227 Smeets, Rieks (ed.) 1994. The indigenous languages of the Caucasus, volume 4: The North East Caucasian languages (part 2). Delmar: Caravan. Smith, Carlota. 1991. The parameter of aspect (2nd edn.). Dordrecht. Kluwer. Smith, Lawrence R. 1982a. ‘Labrador Inuttut (Eskimo) and the theory of morphology.’ Studies in Language 6: 221–​244. Smith, Lawrence R. 1982b. ‘An analysis of affixal derivation and complementation in Labrador Inuttut.’ Linguistic Analysis 10(2): 161–​189. Smith-​Stark, Thomas. 1978. ‘The Mayan antipassive: Some facts and fictions,’ Nora C. England (ed.), Papers in Mayan linguistics. Columbia, MO: University of Missouri, 169–​87. Smolensky, Paul. 1995. ‘On the internal structure of con, the constraint component of UG.’ MS, Johns Hopkins University. Smolensky, Paul. 2006. ‘Harmonic Completeness, Local Constraint Conjunction, and Feature Domain Markedness,’ in P. Smolensky & G. Legendre (eds.), The harmonic mind, vol. 2. Cambridge, MA: MIT Press, 27–​160. Sneddon, James Neil. 1996. Indonesian: A comprehensive grammar. New York: Routledge. Soe, Myint. 1999. ‘A grammar of Burmese.’ PhD dissertation, University of Oregon. Soh, Hooi Ling. 1996. ‘Certain restrictions on A-​bar movement in Malay,’ in Matthew Pearson (ed.), Proceedings of the third and fourth meetings of Austronesian Formal Linguistics Association (UCLA Occasional Papers in Linguistics, 21). Los Angeles:  Department of Linguistics, UCLA, 295–​307. So-​Hartmann, Helga. 2009. A descriptive grammar of Daai Chin. STEDT monograph series. Berkeley, CA: University of California. Solnit, David. 1997. Eastern Kayah Li: Grammar, texts, glossary. University of Hawai‘i Press. Song, Jae Jung. 2001. Linguistic typology: Morphology and syntax. London: Longman, Pearson Education. Sorace, Antonella. 2000. ‘Gradients in Auxiliary Selection with Intransitive Verbs.’ Language 76(4): 850–​90. Speas, Margaret. 2004. ‘Evidentiality, logophoricity and the syntactic representation of pragmatic features.’ Lingua 114: 255–​276. Speas, Margaret & Carol Tenny. 2003. ‘Configurational properties of point of view roles,’ in Anna-​Maria Di Sciullo (ed.), Asymmetry in grammar. Amsterdam & Philadelphia, PA: Benjamins, 315–​343. Speijer, Jakob Samuel. 1886. Sanskrit syntax. Delhi: Motilal Banarsidas. [repr. 1973] Spencer, Andrew. 1991. Morphological theory: An introduction to word structure in generative grammar. Oxford: Blackwell. Spencer, Andrew. 1995. ‘Incorporation in Chukchi.’ Language 71(3): 439–​489. Spencer, Andrew & Ana R. Luís. 2012. Clitics:  An introduction. Cambridge:  Cambridge University Press. Spencer, Andrew. 2006. ‘Syntactic vs. morphological case:  implications for morphosyntax,’ in L. Kulikov, A. Malchukov, & P. de Swart (eds.), Case, valency and transitivity: A cross-​ linguistic perspective. Amsterdam & Philadelphia, PA: Benjamins. Spreng, Bettina. 2006. ‘Antipassive morphology and case assignment in Inuktitut,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity:  Emerging issues. Dordrecht: Springer, 247–​270. Spreng, Bettina. 2010. ‘On the conditions for antipassives.’ Language and Linguistics Compass 4(7): 556–​575, Spreng, Bettina. 2012. ‘Viewpoint aspect in Inuktitut: The syntax and semantics of antipassives.’ PhD dissertation, University of Toronto.

 

1228   References Starosta, Stanley. 1999. ‘Transitivity, ergativity, and the best analysis of Atayal case marking,’ Selected papers from the eighth international conference on Austronesian linguistics. Canberra: Research School of Pacific and Asian Studies, Australian National University. Starosta, Stanley, Andrew K. Pawley, & Lawrence A. Reid. 1982/​2009. ‘The evolution of focus in Austronesian,’ in Amran Halim, Lois Carrington, & S.A. Wurm (eds.), Papers from the third international conference on Austronesian linguistics. Vol. 2: Tracking the travellers (Pacific Linguistics C-​65). Canberra:  Research School of Pacific and Asian Studies, Australian National University, 145–​ 170. [repub. in Elizabeth Zeitoun (ed.), Formosan linguistics: Stanley Starosta’s contributions. Vol. 2: Publications on Formosan languages (Language and Linguistics Monograph Series C6-​65). Taipei: Institute of Linguistics, Academia Sinica, 297–​328, with an expanded version of the paper, with the same title, published for the first time in the same volume, 329–​481] Stebbins, Tonya. 2003. Fighting language endangerment:  Community directed research on Sm’algyax (Coast Tsimshian). Endangered Languages of the Pacific Rim Project. Steedman, Mark. 2001. The syntactic process. Cambridge, MA: MIT Press. Stepanov, Artur. 2004. ‘Ergativity, case, and the minimal link condition,’ in Artur Stepanov, Gisbert Fanselow, & Ralf Vogel (eds.), Minimality effects in syntax. Berlin:  Mouton de Gruyter, 367–​399. Stiebels, Barbara. 2002. Typologie des Argumentlinkings:  Ökonomie und Expressivität. Berlin: Akademie Verlag. Stiebels, Barbara. 2006. ‘Agent focus in Mayan languages.’ Natural Language and Linguistic Theory 24: 501–​570. Stilo, Donald 2008. ‘Two sets of mobile verbal person agreement markers in the Northern Talyshi language,’ in Karimi et al. (eds.), Aspects of Iranian linguistics. Newcastle: Cambridge Scholars Publishing, 363–​390. Stilo, Donald. 2009. ‘Case in Iranian: From reduction and loss to innovation and renewal,’ in A. Malchukov & A. Spencer (eds.), The Oxford handbook of case. Oxford: Oxford University Press, 700–​7 16. Stirtz, Timothy M. 2012 A grammar of Gaahmg: A Nilo-​Saharan language of Sudan. Utrecht: LOT. Stirtz, Timothy M. 2013 ‘Ergative, antipassive and other verb derivational morphemes in Gaahmg.’ [MS, 25 pp.] Stokes, Bronwyn & William B. McGregor. 2003. ‘Classification and subclassification of the Nyulnyulan languages,’ in Nicholas Evans (ed.), The non-​Pama–​Nyungan languages of northern Australia: Comparative studies of the continent’s most linguistically complex region (Studies in Language Change). Canberra: Pacific Linguistics, 29–​74. Stoll, Sabine & Balthasar Bickel. 2013. ‘The acquisition of ergative case in Chintang,’ in E.L. Bavin & S. Stoll (eds.), The acquisition of ergativity. Amsterdam & Philadelphia, PA: Benjamins, 183–​207. Stowell, Timothy. 1996. ‘The phrase structure of tense,’ in J. Rooryck & L. Zaring (eds.), Phrase Structure and the lexicon. Kluwer, Dordrecht. Stowell, Timothy Angus. 1981. ‘Origins of phrase structure.’ PhD dissertation, MIT. Stromme, Kari. 1991. Cerita Bahasa Mamuju. Ujung Pandang: UNHAS-​SIL. Stromme, Kari. 1994. ‘Person marking in the mamuju language.’ NUSA 91–​113. Stroński, Krzysztof. 2009. ‘Approaches to ergativity in Indo-​Aryan.’ Lingua Posnaniensis LI, 77–​118. Subbārāo, Kārumūri V. 1999. ‘Agreement in South Asian languages and Minimalist inquiries:  The framework,’ Proceedings of the international symposium on language contact,

 

References   1229 convergence and typology in South Asia. Tokyo:  Institute for the Study of Language and Cultures of Asia and Africa. Subbārāo, Kārumūri V. 2012. South Asian languages:  A  syntactic typology. Cambridge: Cambridge University Press. Sulkala, Helena & Merja Karjalainen. 1992. Finnish. London: Routledge. Sumbatova, Nina. 2011. ‘Person hierarchies and the problem of person marker origin in Dargwa: Facts and diachronic problems,’ in Gilles Authier & Timur Maisak (eds.), Tense, aspect, modality and finiteness in East Caucasian languages. Bochum: Brockmeyer, 313–​160. Sumbatova, Nina. 2013. ‘Tipologičeskoe i diaxroničeskoe issledovanie morfosintaksisa (na primere jazykov darginskoj gruppy).’ PhD dissertation, RGGU, Moscow. Sumbatova Nina & Rasul Mutalov. 2003. A grammar of Icari Dargwa. Munich: Lincom. Sun, Jackson. 2003. ‘Tani languages,’ in Graham Thurgood & Randy J. LaPolla (eds.), The Sino-​ Tibetan languages. New York: Routledge, 456–​466. Suñer, Margarita. 1998. ‘Resumptive restrictive relatives: A crosslinguistic perspective.’ Language 74: 335–​364. Surtani, Nitesh, Khushboo Jha, & Soma Paul. 2011. ‘Issues with the unergative/​unaccusative classification of the intransitive verbs,’ Proceedings of the international conference on Asian language processing, 55–​58. Surtani, Nitesh & Soma Paul. 2012. ‘Automatic tripartite classification of intransitive verbs,’ Proceedings of the 26th Pacific Asia conference on language, information and computation, 446–​455. Suter, Edgar. 2010. ‘The optional ergative in Kâte,’ in J. Bowden, N.P. Himmelmann, & M. Ross (eds.), A journey through Austronesian and Papuan linguistic and cultural space. Papers in honour of Andrew Pawley. Canberra: Pacific Linguistics, 423–​437. Sutherland-​Smith, Wendy. 1996. ‘Spoken narrative and preferred clause structure: Evidence from modern Hebrew discourse.’ Studies in Language 20(1): 163–​189. Suzuki, Hiroyuki. 2012. ‘Ergative marking in Nyagrong-​Minyag (Xinlong, Sichuan).’ Linguistics of the Tibeto-​Burman Area 35(1): 35–​48. Svenonius, Peter. 2004. ‘On the edge,’ in David Adger, Cecile de Cat, & George Tsoulas (eds.), Peripheries: Syntactic edges and their effects. Dordrecht: Kluwer, 259–​287. Svenonius, Peter. 2006. ‘The emergence of Axial Parts,’ in P. Svenonius (ed.), Trømso working papers on language and linguistics:  Nordlyd 33(1) (special issue on adpositions). Trømso: University of Trømso. Swartz, Stephen M. (ed.). 1982. Papers in Warlpiri grammar:  In memory of Lothar Jagst (Work Papers of SIL-​AAB, series A, vol. 6). Berrimah, NT, Australia: Summer Institute of Linguistics. Swartz, Stephen M. 1991. Constraints on zero anaphora and word order in Warlpiri narrative text (SIL-​AAIB Occasional Papers No. 1). Berrimah, NT: SIL-​AAIB. [published version of 1988 MA dissertation, Pacific College of Graduate Studies, Melbourne] Swartz, Stephen. 1996. ‘Warlpiri dictionary.’ [MS, 2nd edn] http://​ausil.org.au/​node/​3717 Szabolcsi, Anna. 1983. ‘The possessor that ran away from home.’ Linguistic Review 3. Tada, Hiroaki. 1993. ‘A/​A-​bar partition in derivation.’ PhD dissertation, MIT. Takeuchi, Tsuguhito & Yoshiharu Takahashi. 1995. ‘Split ergative patterns intransitive and intransitive sentences in Tibetan:  A  reconsideration,’ in Yoshio Nishi, James Matisoff, & Yasuhiko Nagano (eds.), New horizons in Tibeto-​Burman morphosyntax (Senri Ethnological Studies 41). Osaka: National Museum of Ethnology, 277–​288. Tanangkingsing, Michael & Shuanfan Huang. 2007. ‘Cebuano passives revisited.’ Oceanic Linguistics 46: 554–​584.

 

1230   References Tanner, Darren. 2014. ‘On the left anterior negativity (LAN) in electrophysiological studies of morphosyntactic agreement.’ Cortex 1–​7. Tanner, Darren & Janet G. Van Hell. 2014. ‘ERPs reveal individual differences in morphosyntactic processing.’ Neuropsychologia 56: 289–​301. Taplin, George. 1872. ‘Notes on a comparative table of Australian languages.’ Journal of the Anthropological Institute of Great Britain and Ireland 1: 84–​88. Taplin, George. 1880. Grammar of the Narrinyeri tribe of Australian Aborigines. Adelaide: E. Spiller, Government Printer. Taraldsen, Knut Tarald. 2000. ‘Derivations leading to VO-​order,’ in P. Svenonius (ed.), The derivation of VO and OV (Linguistik Aktuell/​Linguistics Today 31). Amsterdam & Philadelphia, PA: Benjamins, 97–​122. Taraldsen, Knut Tarald. 2010. ‘Case alignment and verb placement,’ in Jan-​Wouter Zwart & Mark de Vries (eds.), Structure preserved: Studies in syntax for Jan Koster (Linguistik Aktuell/​ Linguistics Today 164). Amsterdam & Philadelphia, PA: Benjamins, 341–​346. Tarpent, Marie-​Lucie. 1987. ‘A grammar of the Nisgha language.’ PhD dissertation, University of Victoria. Tarpent, Marie-​Lucie. 1997. ‘Tsimshianic and Penutian: Problems, methods, results, and implications.’ International Journal of American Linguistics 65–​112. Tatevosov, Sergey. 2008. ‘Subevental structure and non-​culmination.’ Empirical Issues in Syntax and Semantics 7: 393–​422. Tchekhoff, Claude. 1987. ‘ “Antipassif ”: Aspect imperfectif et autonomie du sujet.’ Bulletin de la Société de Linguistique de Paris 82: 43–​67. Teng, Stacy Fang-​ching. 2008. A reference grammar of Puyuma. Canberra: Pacific Linguistics. Teo, Amos. 2012. ‘Sumi agentive and topic markers: no and ye,’ Linguistics of the Tibeto-​Burman Area 35(1): 49–​74. Terrill, Angela. 1997. ‘The development of antipassive constructions in Australian languages.’ Australian Journal of Linguistics 17: 71–​88. Tessitori, Luigi. 1913. ‘On the origin of the dative and genitive postpositions in Gujarati and Marwari.’ Journal of the Royal Asiatic Society of Great Britain and Ireland 553–​567 Tessitori, Luigi. 1914. ‘Notes on the grammar of the Old Western Rajasthani with special reference to Apabhramça and to Gujarati and Marwari.’ Indian Antiquary 43: 181–​216, 225–​236. Testelec, Yakov & Peter Arkadiev. 2014. ‘Differential nominal marking:  The pervasive case alternation in Circassian,’ Syntax of the World’s Languages VI, Pavia. Thackston, Wheeler M. 2006. Kurmanji Kurdish reference grammar. http://​www.fas.harvard. edu/​iranian/​Kurmanji Thoma, Sonja. 2016. ‘Discourse particles and the syntax of discourse: Evidence from Miesbach Bavarian’. PhD dissertation, UBC. Thomas, Daniela. 2013. ‘Split ergativity in subordination,’ Rule Interaction in Grammar, vol. 90 of Linguistische Arbeits Berichte. Institut für Linguistik, Universität Leipzig, 1–​21. Thomas, Daniela. 2015. ‘Three-​way systems and harmonic grammar.’ MA thesis, Universität Leipzig. Thompson, Sandra A. & Anthony Mulac. 1991. ‘A quantitative perspective on the grammaticization of epistemic parentheticals in English,’ in Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization, Vol. 2:  focus on types of grammatical markers. Amsterdam & Philadelphia, PA: Benjamins, 313–​329. Thráinsson, Höskuldur. 2001. ‘Object shift and scrambling,’ in Mark Baltin et al. (eds.), A handbook of contemporary syntactic theory. New York: Blackwell.

 

References   1231 Thurgood, Graham. 1978. ‘Thematicization and aspects of the verbal morphology in Burmese:  The principles of organization.’ Proceedings of the 4th annual meeting of the Berkeley Linguistics Society, 254–​67. Thurgood, Graham & Randy LaPolla (eds.). 2003. The Sino-​Tibetan languages. London: Routledge. Tiffou, Etienne & Yves-​Charles Morin. 1982. ‘A note on split ergativity in Burushaski.’ Bulletin of the School of African and Oriental Studies 45: 88–​94. Tiffou, Etienne & Jurgen Pesot. 1989. Contes du Yasin: Introduction au bourouchaski du Yasin avec grammaire et dictionnaire analytique. Paris: Peteers/​SELAF. Tollan, Rebecca. 2013. ‘Case marking and sensitivity of little v: Evidence from dialectal variation in Basque (and beyond).’ Paper presented at the little v Workshop, Leiden, October. Tomasello, Michael. 2003. Constructing a language. Cambridge, MA: Harvard University Press. Tonhauser, Judith. 2003. ‘On the syntax and semantics of content questions in Yucatec Maya.’ Proceedings of the 6th workshop on American Indian languages (WAIL) (Santa Barbara Papers in Linguistics, vol. 14.), 106–​121. Tonoike, Shigeo. 2008. ‘DP movement analysis of relativization.’ MS, Honolulu, University of Hawai‘i. Topuria, Varlam. 1967. svanuri ena, 1: zmna [‘the Svan language, 1: verb’]. Tbilisi: Mecniereba. Tournadre, Nicolas. 1991. ‘The rhetorical use of the Tibetan ergative.’ Linguistics of the Tibeto-​ Burman Area 14(1): 93–​107. Tournadre, Nicolas. 1995. ‘Tibetan ergativity and the trajectory model,’ in Y. Nishi, J.A. Matisoff, & Y. Nagano (eds.), New horizons in Tibeto-​Burman morphosyntax. Osaka:  National Museum of Ethnology, 261–​275. Tournadre Nicolas. 2014. ‘The Classical Tibetan cases and their transcategoriality. From sacred grammar to modern linguistics.’ Himalayan Linguistics Journal 9(2): 87–​125. Trager, George L. 1946. ‘An outline of Taos grammar,’ in C. Osgood (ed.), Linguistic structures of Native America (Viking Fund Publications in Anthropology, vol. 6). New York: Viking Fund, 184–​221. Trask, Robert L. 1977. ‘Historical syntax and Basque verbal morphology:  Two hypotheses,’ in W.A. Douglass, R.W. Etulain, & W.H. Jacobsen (eds.), Anglo-​American contributions to Basque studies: Essays in honor of Jon Bilbao. Reno: Desert Research Institution Publications on the Social Sciences, 13: 203–​217. Trask, Robert L. 1979. ‘On the origins of ergativity,’ in Frans Plank (ed.), Ergativity: Towards a theory of grammatical relations. London: Academic Press, 385–​404. Trask, Robert L. 1997. The history of Basque. London: Routledge. Trask, Robert L. 1998. ‘The typological position of Basque: Then and now.’ Language Sciences 20(3): 313–​324. Traugott, Elizabeth Closs & Graeme G. Trousdale. 2014. Constructionalization and constructional change. Oxford: Oxford University Press. Travis, Lisa. 2000. ‘Event structure in syntax,’ in Carol Tenny & James Pustejovsky (eds.), Events as grammatical objects: The converging perspectives of lexical semantics and syntax. Stanford, CA: CSLI, 145–​185. Traxler, Matthew J. & Morton A. Gernsbacher (eds.). 2006. Handbook of psycholinguistics. Amsterdam: Elsevier. Trechsel, Frank R. 1982. ‘A categorial fragment of Quiche.’ PhD thesis, University of Texas at Austin. Trick, Douglas. 2006. ‘Ergative control of syntactic processes in Sama Southern.’ 10th International Conference on Austronesian Linguistics, 17–​20 January, Palawan.

 

1232   References Trommer, Jochen. 2006. ‘Person and number agreement in Dumi.’ Linguistics 44: 1011–​1057. Truckenbrodt, Hubert. 2012. ‘Semantics of intonation,’ in Claudia Maienborn, Klaus von Heusinger, & Paul Portner (eds.), Semantics: An international handbook of natural language meaning, vol. 3. Berlin: de Gruyter, 2039–​2969. Trueswell, John C. & Albert E. Kim. 1998. ‘How to prune a garden path by nipping it in the bud: Fast priming of verb argument structure.’ Journal of Memroy and Language 39(1): 102–​123. Trumpp, Ernest (1872). Grammar of the Sindhi language. Osnabrück: Biblio Verlag. [repr. 1970] Tryon, Darrell T. 1967. Dehu grammar (Pacific Linguistics, Series B, 7). Canberra: Australian National University. Tsagareli, Alexander. 1873. O grammatike literaturnago gruzinskago jazyka. St Peterburg: Akad. Nauk. Tschenkéli, Kita. 1958. Einführung in die georgische Sprache. Zürich: Amirani Verlag. Tsunoda, Tasaku. 1981a. The Djaru language of Kimberley, Western Australia. Canberra: Pacific Linguistics. Tsunoda, Tasaku. 1981b. ‘Split case-​marking patterns in verb-​types and tense/​aspect/​mood.’ Linguistics 19: 389–​438. Tsunoda, Tasaku. 1985. ‘Remarks on transitivity.’ Journal of Linguistics 21: 385–​396. Tsunoda, Tasaku. 1988. ‘Antipassives in Warrungu and other Australian languages,’ in Masayoshi Shibatani (ed.), Passive and voice. Amsterdam & Philadelphia, PA: Benjamins, 595–​650. Tucker, Archibald N. & Margaret A. Bryan. 1966. Linguistic analyses: The Non-​Bantu languages of North-​Eastern Africa. London, New York, & Cape Town: Oxford University Press and International for the African Institute. Tuite, Kevin. 1987. ‘Indirect transitives in Georgian.’ Proceedings of the Berkeley Linguistics Society 13: 296–​309 Tuite, Kevin. 1996. ‘Paradigm recruitment in Georgian,’ in H. Aronson (ed.), NSL 8: Linguistic studies in the non-​Slavic languages of the Commonwealth of Independent States and the Baltic republics. Chicago, IL: Chicago Linguistic Society, 375–​387. Tuite, Kevin. 1998. Kartvelian morphosyntax:  Number agreement and morphosyntactic orientation in the South Caucasian languages (Studies in Caucasian Linguistics, 12). Munich: Lincom Europa. Tuite, Kevin. 2002. ‘Deponent verbs in Georgian,’ in Wolfram Bublitz, Manfred von Roncador, & Heinz Vater (eds.), Philologie, Typologie und Sprachstruktur:  Festschrift für Winfried Boeder zum 65. Geburtstag. Frankfurt am Main: Peter Lang Verlag, 375–​389. Tuite, Kevin. 2004. ‘Early Georgian,’ in Roger D. Woodard (ed.), Cambridge encyclopedia of the world’s ancient languages. Cambridge: Cambridge University Press, 967–​987. Tuite, Kevin. 2008. ‘The rise and fall and revival of the Ibero-​ Caucasian hypothesis.’ Historiographia Linguistica 35(1): 23–​82. Tuite, Kevin. 2009. ‘Agentless transitive verbs in Georgian.’ Anthropological Linguistics 51(3–​4): 269–​295. Tuite, Kevin. 2011. ‘The reception of Marr and Marrism in the Soviet Georgian academy,’ in Florian Mühlfried & Sergey Sokolovsky (eds.), Exploring the edge of empire:  Soviet Era anthropology in the Caucasus and Central Asia. Halle: LIT Verlag, 197–​214 Tuite, Kevin J., Asif Agha, & Randolph Graczyk. 1985. ‘Agentivity, transitivity, and the question of active typology,’ in W.H. Eilfort, P.D. Kroeber, & K.L. Peterson (eds.), Papers from the parasession on causatives and agentivity at the 21st regional meeting of the Chicago Linguistic Society. Chicago, IL: Chicago Linguistic Society, 252–​270. Udaar, Usha, Gurmeet Kaur, & Pritha Chandra. 2014. ‘Revisiting morphological ergativity in Hindi–​Urdu.’ Acta Linguistica 8(1).

 

References   1233 Ueno, Mieko & Susan M. Garnsey. 2008. ‘An ERP study of the processing of subject and object relative clause in Japanese.’ Language and Cognitive Processes 23: 646–​688. Ueno, Mieko & Robert Kluender. 2009. ‘On the processing of Japanese wh-​questions: An ERP study.’ Brain Research 1290: 63–​90. Uhlenbeck, Christianus Cornelius. 1916. ‘Het passieve Karakter van het Verbum transitivum of van het Verbum actionis in Taalen van Noord-​Amerika.’ Verslagen en Mededeelingen der Koninklijke Akademie van Wetenschappen: Afd. Letterkunde 5(2): 187–​216. Ullman, Michael T. 2001. ‘A neurocognitive perspective on language: The declarative/​procedural model.’ Neuroscience 2(10): 717–​726. Ullman, Michael T. 2004. ‘Contributions of memory circuits to language: The declarative/​procedural model.’ Cognition 92(1–​2): 231–​270. Unseth, Peter. 1989a. ‘An initial comparison and reconstruction of case suffixes in Surmic languages.’ Journal of Ethiopian Studies 22: 97–​104. Unseth, Peter. 1989b. ‘Sketch of Majang syntax,’ in Marvin Lionel Bender (ed.), Topics in NiloSaharan Linguistics. (Nilo-Saharan: Linguistic Analyses and Documentation, vol. 3. Series editor Franz Rottland.) Hamburg: Helmut Buske Verlag, 97–​127. Ura, Hiroyuki. 2000. Checking theory and grammatical functions in universal grammar. New York: Oxford University Press. Ura, Hiroyuki. 2001. ‘Case,’ in Mark Baltin & Chris Collins (eds.), The handbook of contemporary syntactic theory. Oxford: Blackwell. Ura, Hiroyuki. 2006. ‘A parametric syntax of aspectually conditioned split-​ergativity,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity:  Emerging issues. Dordrecht: Springer, 111–​141. Urban, Greg. 1985. ‘Ergativity and accusativity in Shokleng (Gê).’ International Journal of American Linguistics 51: 164–​87. Uribe-​Etxebarria, Miriam. 1989. ‘Some notes on the Structure of IP in Basque.’ MS, UConn, Storrs. Uridia, O. 1960. ‘megrulis sint’aksuri taviseburebani kartultan mimartebit’ [‘Syntactic characteristics of Mingrelian in comparison to Georgian’], Tbilisis Saxelmc’ipo Universit’et’is Shromebi 93: 167–​178. Vajda, Edward. 2007. ‘Losing semantic alignment:  from Proto-​Yeniseic to Modern Ket,’ in Mark Donohoe & Sören Wichman (eds.), Semantic alignment. Oxford: Oxford University Press, 140–​161 Valenzuela, Pilar. 1997/​2003. ‘Basic verb types and argument structures in Shipibo-​Conibo.’ PhD dissertation, University of Oregon. Valenzuela, Pilar. 2003. ‘Transitivity in Shipibo-​Konibo grammar.’ PhD dissertation, University of Oregon. Valenzuela, Pilar. 2011. ‘Argument encoding and pragmatic marking of the transitive subject in Shiwilu (Kawapanan).’ International Journal of American Linguistics 77: 91–​120. Vamarasi, Marit. 2002. Rotuman. Languages of the World/Materials 415. Munchen: LINCOM Europa. van Breugel, Seino. 2014. A grammar of Atong. Leiden: Brill. van de Velde, Freek. 2014. ‘The discourse motivation for split-​ergative alignment in Dutch nominalisations (and elsewhere).’ Pragmatics 24(2). van de Visser, Mario. 2006. The marked status of ergativity. Utrecht: LOT Publications. van den Berg, Helma. 1995. A grammar of Hunzib. Munich: Lincom. van den Berg, Helma. 1999. ‘Gender and person agreement in Akusha Dargi.’ Folia Linguistica 33: 153–​168.

 

1234   References van den Berg, Helma. 2004. ‘Coordinating constructions in Daghestanian,’ in Martin Haspelmath (ed.), Coordinating constructions. Amsterdam & Philadelphia, PA: Benjamins, 197–​226. van den Berg, Helma. 2005. ‘The East Caucasian language family.’ Lingua 115: 147–​190. van Everbroeck, Ezra. 2003. ‘Language type frequency and learnability from a connectionist perspective.’ Linguistic Typology 7: 1–​50. van Urk, Coppe. 2015. ‘A uniform syntax for phrasal movement: A Dinka Bor case study.’ PhD dissertation, MIT. van Urk, Coppe & Norvin Richards. 2015. ‘Two components of long-​ distance extraction: Successive cyclicity in Dinka.’ Linguistic Inquiry 46: 113–​155. Van Valin, Jr., Robert D. 1985. ‘Case marking and the structure of the Lakhota clause,’ in J. Nichols & A. Woodbury (eds.), Grammar inside and outside the clause. Cambridge: Cambridge University Press, 363–​413. Van Valin, Jr., Robert D.  1990. ‘Semantic parameters of split intransitivity.’ Language 66(2): 221–​260. Van Valin, Jr., Robert D. 1992. ‘An overview of ergative phenomena and their implications for language acquisition,’ in D. Slobin (ed.), The cross-​linguistic study of language acquisition, vol. 3. Hillsdale, NJ: Erlbaum, 15–​38. Van Valin, Jr., Robert D. 2013. ‘Head-​marking languages and linguistic theory,’ in B. Bickel, L.A. Grenoble, D.A. Peterson, & A. Timberlake (eds.), Language typology and historical contingency: In honor of Johanna Nichols. Amsterdam & Philadelphia, PA: Benjamins, 91–​123. Van Valin, Jr., Robert D. & Randy Lapolla. 1997. Syntax:  Structure, meaning and function. Cambridge: Cambridge University Press. Van Valin, Robert D. & David Wilkins. 1996. ‘The case for “effector”: Case roles, agents, and agency revisited,’ in M. Shibatani & S.A. Thomas (eds.), Grammatical constructions: Their form and meaning. Oxford: Clarendon Press, 289–​322. Vapnarsky, Valentina, Cédric Becquey, & Aurore Monod Becquelin. 2012. ‘Ergativity and passive in three Mayan languages,’ in Gilles Authier & Katharina Haude (eds.), Ergativity, valency, and voice. Berlin: de Gruyter, 51–​110. Vázquez Álvarez, Juan Jesús. 2002. ‘Morfología del verbo de la lengua Ch’ol de Tila, Chiapas.’ MA thesis, Centro de Investigaciones y Estudios Superiores en Antropología Social, México, DF. Vázquez Álvarez, Juan Jesús & Roberto Zavala. 2013. ‘La estructura argumental preferida en chol, una lengua agentiva.’ In Memorias del VI Congreso de Idiomas Indígenas de Latinoamérica, October 24–26, Universidad de Texas en Austin. http://​www.ailla.utexas. org/​site/​events.html Velázquez-​Castillo, Maura. 2002. ‘Grammatical relations in active systems:  The case of Guaraní.’ Functions of Language 9(2): 133–​167. Vendler, Zeno. 1967. Linguistics in philosophy. Ithaca, NY: Cornell University Press. Verbeke, Saartje. 2013a. Alignment and ergativity in new Indo-​Aryan languages. Berlin: Mouton de Gruyter. Verbeke, Saartje. 2013b. ‘Differential subject marking in Nepali: The agent marker le in imperfective constructions.’ Linguistics 51(3): 585–​610. Vergnaud, Jean Roger. 2008. ‘Letter to Noam Chomsky and Howard Lasnik on “Filters and Control April 17, 1977,’ in Robert Freidin, Carlos Otero, & Maria Luisa Zubizarreta (eds.), Foundational issues in linguistic theory. Cambridge, MA: MIT Press, 3–​15. Verhaar, John W. M. 1988. ‘Syntactic ergativity in contemporary Indonesian,’ in Richard McGinn (ed.), Studies in Austronesian linguistics. Athens, OH: Ohio University Press, 347–​384.

 

References   1235 Verhoeven, Elisabeth. 2007. Experiential constructions in Yukatek Maya:  A  typologically based analysis of a functional domain in a Mayan language. Amsterdam & Philadelphia, PA: Benjamins (SLCS 87). Verma, Mahendra K. & K. P. Mohanan (eds.). 1990. Experiencer subjects in South Asian languages. Stanford, CA: CSLI. Verstraete, Jean-​Christophe. 2010. ‘Animacy and information structure in the system of ergative marking in Umpithamu.’ Lingua 120: 1637–​1651. Vikner, Sten. 1991. ‘Verb movement and the licensing of NP-​positions in the Germanic Languages.’ PhD dissertation, University of Geneva. Vikner, Sten. 1997. ‘The interpretation of object shift, optimality theory, and minimalism.’ Working Papers in Scandinavian Syntax 60: 1–​24. Vincent, Nigel. 2013. ‘Conative.’ Linguistic Typology 17: 269–​289. Vogt, Hans. 1947. ‘Le système des cas en géorgien ancien.’ Norsk tidsskrift for sprogvidenskap 14: 98–​140 Vogt, Hans. 1971. Grammaire de la langue géorgienne. Oslo: Universitetsforlaget. Vollman, Ralf. 2008. Descriptions of Tibetan ergativity:  A  historiographical account. Graz: Leykam Buchverlagsgesellschaft. Wagers, Matt, Manuel F. Borja & Sandra Chung. 2015. ‘The real-​time comprehension of wh-​ dependencies in a Wh-​Agreement language.’ Language 91: 88–​101. Waikhom, Pinky. 2014. ‘A descriptive grammar of Inpui.’ PhD thesis, Manipur University. Wali, Kashi & Omkar N. Koul. 1997. Kashmiri:  A  cognitive–​descriptive grammar. London: Routledge. Walker, Neil Alexander. 2013. ‘A grammar of Southern Pomo:  An indigenous language of California.’ PhD dissertation, University of California, Santa Barbara. Wallace, William. 1982. ‘The evolution of ergative syntax in Nepali.’ Studies in the Linguistic Sciences 12: 147–​209. Wallace, William. 1985. ‘Subjects and subjecthood in Nepali: An analysis of Nepali clause structure and its challenges to Relational Grammar and Government & Binding.’ PhD dissertation, University of Illinois, Urbana-​Champaign. Walter, Sven. 2009. ‘Epiphenomenalism,’ Encyclopedia of neuroscience. Berlin:  Springer, 1137–​1139. Wartenburger, Isabell, Hauke R. Heekeren, Jubin Abutalebi, Stefano F. Cappa, Arno Villringer, & Daniela Perani. 2003. ‘Early setting of grammatical processing in the bilingual brain.’ Neuron 37(1): 159–​70. Watkins, Laurel. 1984. A grammar of Kiowa. Lincoln, NE: University of Nebraska Press. Watters, David E. 2002. A grammar of Kham. Cambridge: Cambridge University Press. Weber, Elizabeth. 2003. ‘Nominal information flow in the talk of two boys with autism,’ in John W. Du Bois, Lorraine E. Kumpf, & William J. Ashby (eds.), Preferred argument structure: Grammar as architecture for function. Amsterdam & Philadelphia, PA:  Benjamins, 353–​383. Weber-​Fox, Christine M. & Helen J. Neville. 1996. ‘Maturational constraints on functional specializations for language processing ERP and behavioral evidence in bilingual speakers.’ Journal of Cognitive Neuroscience 8(3): 231–​256 Wechsler, Stephen & I Wayan Arka. 1998. ‘Syntactic ergativity in Balinese: An argument structure based theory.’ Natural Language & Linguistic Theory 5(1): 387–​442. Weckerly, Jill & Marta Kutas. 1999. ‘An electrophysiological analysis of animacy effects in the processing of object relative sentences.’ Psychophysicalogy 36(5): 559–​570.

 

1236   References Weissenborn, Jürgen. 1991. ‘Functional categories and verb movement:  The acquisition of German syntax reconsidered,’ in M. Rothweiler (ed.), Spracherwerb und Grammatik. Linguistische Untersuchungen zum Erwerb von Syntax und Morphologie. Linguistische Berichte, Sonderheft 3: 190–​224. Westermann, Diedrich. 1912. The Shilluk people:  Their language and folklore. Glückstadt: Augustin. Wexler, Ken. 1994. ‘Optional infinitives, head movement, and the economy of derivations,’ in D. Lightfoot & N. Hornstein (eds.), Verb movement. Cambridge: Cambridge University Press, 305–​350. Wexler, Ken. 1995. ‘Feature interpretability and optionality in early child grammar.’ Paper presented at Workshop on Optionality, University of Utrecht. Wexler, Ken. 1998. ‘Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive stage.’ Lingua 106: 23–​79. Wharram, Douglas. 2003. ‘On the interpretation of (un)certain indefinites in Inuktitut and related languages.’ PhD thesis, University of Connecticut. Wheatley, Julian. 1982. ‘Burmese: A grammatical sketch.’ PhD thesis, Berkeley, University of California. Whitman, John. 2000. ‘Relabelling,’ in Susan Pintzuk, George Tsoulas, & Anthony Warner (eds.), Diachronic syntax:  Models and mechanisms. Oxford:  Oxford University Press, 220–​238. Whitman, John & Yuko Yanagida. 2012. ‘The formal syntax of alignment change,’ in Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, & Juanito Avelar (eds.), Diachronic syntax:  Parameter theory and linguistic change. Oxford:  Oxford University Press, 177–​195. Widmer, Manuel. 2015. ‘A descriptive grammar of Bunan.’ Dissertation, University of Bern. Wierzbicka, Anna. 1981. ‘Case marking and human nature.’ Australian Journal of Linguistics 1: 43–​80. Wiesemann, Ursula. 1972. ‘Die phonologische und grammatische Struktur der Kaingang-​ Sprache.’ Janua Linguarum, Series Practica. The Hague: Mouton de Gruyter. Wiesemann, Ursula. 1986a. ‘Aspect and mood as a nine-​cell matrix,’ in Benjamin Elson (ed.), Language in global perspective. Dallas: SIL, 471–​506. Wiesemann, Ursula. 1986b. ‘The pronominal systems of some Jê and Macro-​Jê languages,’ in U. Wiesemann (ed.), Pronominal systems. Tübingen: Gunter Narr Verlag, 359–​380. Williams, Edwin. 1987. ‘The thematic-​structure of derived nominals.’ Proceedings of the Chicago Linguistic Society. Williams, Edwin. 2003. Representation theory. Cambridge, MA: MIT Press. Willis, Christina. 2011. ‘Optional case marking in Darma (Tibeto-​Burman).’ Linguistics of the Tibeto-​Burman Area 34(2): 101–​132. Willson, Stephen. 1996. ‘Verb agreement and case marking in Burushaski.’ Work Papers of the Summer Institute of Linguistics North Dakota 40: 1–​7 1. Wiltschko, Martina. 2003. ‘On the interpretability of tense on D and its consequences for case theory.’ Lingua 113: 659–​696. Wiltschko, Martina. 2006. ‘On ergativity in Halkomelem Salish,’ in Alana Johns, Diane Massam, & Juvenal Ndayiragije (eds.), Ergativity: Emerging issues. Dordrecht: Springer, 197–​227. Wiltschko, Martina. 2008. ‘Person–​hierarchy effects without a person hierarchy,’ in Roberta D’Alessandro, Susan Fischer, & Gunnar Hrafn Hrafnbjargarson (eds.), Agreement restrictions. Berlin: Mouton de Gruyter, 281–​314.

 

References   1237 Wiltschko, Martina. [in press]. ‘Response particles beyond answering,’ in Bailey, L. and M. Sheehan (eds.) Order and Structure in Syntax. Language Science Press Wiltschko, Martina & Johannes Heim. 2016. ‘The syntax of confirmationals. A neo-performative analysis’. In Gunther Kaltenböck, Evelien Keizer, & Arne Lohmann (eds.), Outside the clause: Form and function of extra-clausal constituent. London: John Benjamins, 303–340. Windfuhr, Gernot (ed.) 2009. The Iranian languages. London: Routledge. Windfuhr, Gernot. 2009. ‘Dialectology and topics,’ in Windfuhr (ed.), 5–​42. Witzel, Michael. 1999. ‘Substrate languages in Old Indo-​Aryan (Ṛgvedic, Middle and Late Vedic.’ Electronic Journal of Vedic Studies 5(1): 1–​67 Wolff, John U. 1973. ‘Verbal inflection in Proto-​Austronesian,’ in Andrew Gonzalez (ed.), Parangal kay Cecilio Lopez (Philippine Journal of Linguistics (special monograph) 4: 71–​91). Quezon City: Linguistic Society of the Philippines. Wolff, John U., Maria Theresa C. Centeno, & Der-​Hwa V. Rau. 1991. Pilipino through self-​ instruction. Ithaca, NY: Cornell University Press. Wood, Jim. 2014. ‘Reflexive -​st verbs in Icelandic.’ Natural Language and Linguistic Theory 32: 1387–​1425. Woodbury, Anthony C. 1981. ‘Study of the Chevak Dialect of Central Yup’ik Eskimo.’ PhD dissertation, University of California, Berkeley. Woodbury, Anthony & Jerrold Sadock. 1986. ‘Affixial verbs in syntax: A reply to Grimshaw and Mester.’ Natural Language and Linguistic Theory 4: 229–​244. Woolford, Ellen. 1997. ‘Four-​way case systems: Ergative, nominative, objective, and accusative.’ Natural Language and Linguistic Theory 15: 181–​227. Woolford, Ellen. 1999. ‘More on the anaphor agreement effect.’ Linguistic Inquiry 30: 257–​287. Woolford, Ellen. 2000. ‘Ergative agreement systems.’ University of Maryland Working Papers in Linguistics 10: 157–​191. Woolford, Ellen. 2001. ‘Case patterns,’ in Géraldine Legendre, Jane Grimshaw, & Sten Vikner (eds.), Optimality-​theoretic syntax. Cambridge, MA: MIT Press, 509–​545. Woolford, Ellen. 2006. ‘Lexical case, inherent case, and argument structure.’ Linguistic Inquiry 37: 111–​130. Woolford, Ellen. 2008. ‘Differential subject marking at argument structure, syntax, and PF,’ in H. de Hoop, & P. de Swart (eds.), Differential subject marking. Dordrecht: Springer, 17–​40. Woolford, Ellen. 2010. ‘Active–​stative agreement in Choctaw and Lakota.’ ReVEL 8: 6–​46. [Special issue, 4: Optimality Theoretic Syntax, ed. Gabriel de Ávila Othero & Sérgio de Moura Menuzzi] Woolford, Ellen. 2013. ‘Aspect splits and parasitic marking,’ in Hans Broekhuis & Ralf Vogel (eds.), Linguistic derivations and filtering:  Minimalism and optimality theory. London: Equinox, 166–​192. Woolford, Ellen. 2015. ‘Ergativity and transitivity.’ Linguistic Inquiry 46(3): 489–​531. Wouk, Fay & Malcolm Ross (eds.). 2002. The history and typology of Western Austronesian voice systems. Canberra: Pacific Linguistics, Research School of Pacific Asian Studies, ANU. Wunderlich, Dieter. 1997. ‘Cause and the structure of verbs.’ Linguistic Inquiry 27: 27–​68. Wunderlich, Dieter. 2006. ‘Towards a structural typology of verb classes,’ in D. Wunderlich (ed.), Advances in the theory of the lexicon. Berlin: Mouton de Gruyter, 57–​166. Wunderlich, Dieter & Renate Lakämper. 2001. ‘On the interaction of structural and semantic case.’ Lingua 111: 377–​418. Wurmbrand, Susi. 1999. ‘Modal verbs must be raising verbs,’ in Sonya Bird, Andrew Carnie, Jason D. Haugen, & Peter Norquest (eds.), Proceedings of the 18th West Coast conference on formal linguistics. Somerville, MA: Cascadilla Press, 599–​612.

 

1238   References Wurmbrand, Susi. 2003. Infinitives:  Restructuring and Clause Structure. Berlin: Mouton de Gruyter. Wurmbrand, Susi. 2015. ‘Complex predicate formation via voice incorporation.’ In Léa Nash & Pollet Samvelian (eds.), Approaches to Complex Predicates. Leiden: Brill, 248–290. Xundadze, Silov. 1901. Salit’erat’uro kartuli. Kutaisi: K’iladzisa da Xeladzis st’amba. Yamada, Hisanari. 2013. ‘Reciprocal constructions in Avar.’ Lingua 126: 150–​171. Yanagida, Yuko. 2006. ‘Word order and clause structure in early old Japanese.’ Journal of East Asian Linguistics 15: 37–​67. Yanagida, Yuko. 2012. ‘The syntactic reconstruction of alignment and word order: The case of old Japanese,’ in Ans van Kemenade & Nynke de Haas (eds.), Historical linguistics 2009. Amsterdam & Philadelphia, PA: Benjamins, 107–​128. Yanagida, Yuko & John Whitman. 2009. ‘Alignment and word order in old Japanese.’ Journal of East Asian Linguistics 18(2): 101–​144. Yanti. 2010. ‘A reference grammar of Jambi Malay.’ PhD dissertation, University of Delaware. Yip, Moira, Joan Maling, & Ray Jackendoff. 1987. ‘Case in tiers.’ Language 63: 217–​250. Yoshioka, Noboru. 2012. ‘A reference grammar of Eastern Burushask.’ PhD dissertation, Tokyo University of Foreign Studies. Yuan, Michelle. 2013. ‘A phasal account of ergativity in Inuktitut.’ MA thesis, University of Toronto. Zabala, Igone. 2003. ‘Nominal predication,’ in Jon Ortiz de Urbina & José Ignacio Hualde (eds.), A grammar of Basque. Berlin: de Gruyter, 426–​447. Zaenen, Annie & Joan Maling. 1984. ‘Unaccusative, passive, and quirky case,’ in Mark Cobler, Susannah MacKaye, & Michael T. Wescoat (eds.), Proceedings of the third West Coast conference on formal linguistics. Stanford, CA:  Stanford Linguistics Association, 317–​329. Zaenen, Annie, Joan Maling, & Höskuldur Thráinsson. 1985. ‘Case and grammatical functions: The Icelandic passive.’ Natural Language and Linguistic Theory 3: 441–​83. Zakharyin, Boris. 1979. ‘On the formation of ergativity in Indo-​Aryan and Daric.’ Osmania Papers in Linguistics 5: 50–​7 1. Zanuttini, Raffaela. 1997. Negation and clausal structure: A comparative study of Romance languages. Oxford: Oxford University Press. Zanuttini, Raffaella. 2008. ‘Encoding the addressee in the syntax:  Evidence from English imperative subjects.’ Natural Language & Linguistic Theory 185–​218. Zanuttini, Raffaella & Paul Portner. 2003. ‘Exclamative clauses: At the syntax–​semantics interface.’ Language 39–​81. Zavala, Roberto. 1992. El kanjobal de San Miguel Acatán. Mexico City: Universidad Nacional Autónoma de México. Zavala Maldonado, Roberto. 1994. ‘Inverse alignment in Huastec.’ Función 15/​16: 27–​81. Zawiszewski, Adam, Eva Gutiérrez, Beatriz Fernández, & Itziar Laka. 2011. ‘Language distance and non-​ native syntactic processing:  Evidence from event-​ related protentials.’ Bilingualism: Language and Cognition 14(3): 400–​411. Zeevat, Henk, & Gerhard Jäger. 2002. ‘A reinterpretation of syntactic alignment,’ in D. de Jongh, H. Zeevat, & M. Nilsenova (eds.). Proceedings of the third and fourth international Tbilisi symposium on language, logic and computation. Amsterdam:  University of Amsterdam. Zeisler, Bettina. 2004. Relative tense and aspectual values in Tibetan languages. Berlin: Mouton de Gruyter.

 

References   1239 Zeisler, Bettina. 2007. ‘Case patterns and pattern variation in Ladakhi:  A  field report,’ in Roland Bielmeier & Felix Haller (eds.), Linguistics of the Himalayas and beyond (Trends in Linguistics: Studies and Monographs, 196). Berlin: Mouton de Gruyter, 399–​425. Zeisler, Bettina. 2012. ‘Practical issues of pragmatic case marking variations in the Kenhat varieties of Ladakh.’ Linguistics of the Tibeto-​Burman Area 35(1): 75–​106. Zhghent’i, Sergi. 1936. guruli k’ilo. gamok’vleva, t’ekst’ebi, leksik’oni [‘Gurian dialect. Analysis, texts, lexicon’]. T’pilisi: Mecniereba. Ziv, Yael. 1994. ‘Left and right dislocations:  Discourse functions and anaphora.’ Journal of Pragmatics 22: 629–​645. Zorc, David Paul. 1977. The Bisayan dialects of the Philippines:  Subgrouping and reconstruction (Pacific linguistics: Series C, vol. 40). Canberra, Australia: Research School of Pacific Studies, Australian National University. Zorell, Franz. 1930. Grammatik zur altgeorgischen Bibelübersetzung. Rome:  Pontificium Institutum Biblicum. Zu, Vera. 2015. ‘Probing for conversation participants: The case of Jingpo,’ Proceedings of the 49th annual regional meeting of Chicago Linguistic Society. Zuazo, Koldo. 1998. ‘Euskalkiak, gaur.’ Fontes Linguae Vasconum 78: 191–​233. Zubizarreta, Maria Luisa. 1987. Levels of representation in the lexicon and in the syntax. Dordrecht: Foris. Zubizarreta, Maria Luisa. 1998. Prosody, focus, and word order. Cambridge, MA: MIT Press.

 

 

Author Index

Agha, A. 930 Ahmed, T.  548, 550–​1, 817 Aikhenvald, A.Y.  344 (fn14) Aissen, J.  255, 284 Albizu, P.  790–​1 Albright, A. 647 Aldai, G. 793 Aldridge, E.  160, 205, 386 (fn20), 593, 749, 1000–​1 Alexiadou, A. 62 Anagnostopoulou, E.  401, 412 Andersen, P.K.  549 Andersen, T.  452, 902–​3, 918 Anderson, N.  214–​15 Anderson, S.R.  503–​4 Andvik, E.  147–​9 Ariel, M. 50 Aronson, H.  1119, 1132 Arregi, C.  652–​4, 660 Ashby, W. 39 Baerman, M. 209 Baker, M.  176, 188, 235, 236 (fn11), 241, 319, 368–​9 Ball, D.  587, 993 (fn6, 7), 996 (fn8), 1001 (fn14), 1003 (fn16) Barker, C.  603 (fn14) Bashir, E. 819 Basilico, D.  320, 414 Bejar, S. 243 Bentivoglio, P. 39 Benveniste, E. 508 Berge, A.  407, 410 Beyssade, C. 430 Bhatt, R. 816 Bickel, B.  638, 641, 468 Bittner, M.  129, 221, 223, 335, 345, 749, 839–​40, 999 (fn11)

Blake, B. 918 Bobaljik, J.  88, 173, 196, 327, 730–​1 Bohnacker, U. 486 Borer, H.  362, 367 Bornkessel-​Schlesewsky, I.  711–​13 Bowden, J. 583 Branigan, P. 327 Brown, P. 638 Brown, R. 637 Bruening, B.  322, 367–​8 Bubenik, V. 551 Burzio, L. 62 Butt, M.  213–​14 Bybee, J. 647 Bynon, T.  451, 508, 532–​3, 535–​6, 549 Caha, P. 337 Campbell, L.  46–​8 Carreiras, M.  700, 718–​20 Chomsky, N.  173, 501, 767, 778 Choudhary, K.K.  698 Chung, S. 505 Clemens, L.E.  725–​6 Cocchi, G. 245 Compton, R.  402–​3 Comrie, B.  127, 255, 263, 634, 716 (fn5), 882, 946, 1047, 1053 Coon, J.  197, 206, 210–​13, 366, 673, 679, 749 Coupe, A.R.  449, 943 Creissels, D. 902 Crowley, T. 578 Custis, T.  990 (fn2) Dahlstrom, A.  206, 218 Dalrymple, M.  264 (fn4) Davis, H.  613, 1031, 1033 Davison, A.  211, 812, 816, 819 De Hoop, H.  943

 

1242   Author Index De Rijk, R.  166–​7 De Swart, P.  255 DeLancey, S.  207, 267–​8, 635, 942–​3 Deal, A.R.  140, 142 (fn11), 205. 223, 305 Déchaine, R.-​M.  1017 Demirdache, H. 613 Dench, A.  458–​9 Diaz, B. 699 Dillon, B.  702, 727 Dimmendaal, G.  451, 452 (fn4), 459–​60, 903, 910–​12 Dixon, R.M.W.  26, 127, 206, 213–​14, 218–​19, 228, 248, 860, 881, 884, 1007, 1117 Dočekal, M.  405, 412, 412 (fn23) Dorais, L.-​J.  840 (fn13) Dowty, D. 24 DuBois, J.W.  264, 737 Dukes, M.  993 (fn7) Dunn, J.  1033 Durie, M.  45–​6 Endriss, C.  410, 417 England, N. 40 Estival, D. 449 Etxepare, R.  162–​5, 167, 172 Evans, N. 25 Ezeizabarrena, M.-​J.  706 Ezeizabarrena, M. 654 Falk, Y.N.  1047 Faller, M.T.  435 Farrell, P. 317 Fernández, B. 791 Fillmore, C. 24 Foley, W.  451, 611 Folli, R. 61 Forbes, C.  1028, 1031, 1033 Fox, D.  408 (fn19) Frank, P.S.  263 Gagua, R. 856 Garrett, A.  450–​1, 511–​12 Genetti, C. 461 Gerdts, D.B.  318 Gibson, J. 584 Gildea, S.  51, 450 Gilley, L.G.  902–​5

Goddard, C.  209, 245 Goenaga, P.  166–​7 Goldberg, A.E.  49 Goldin-​Meadow, S.  37 Greenberg, J.H.  917 Grimshaw, J.  356 (fn1), 366 Grinstead, J. 668 Guillaume, A. 463 Gutiérrez-​Mangado, M.J.  706 Haddican, B. 166 Haig, G.  52–​3, 505, 508 Hale, K.  129, 221, 223, 335, 345, 423, 505, 749, 839–​40, 999 (fn11) Halle, M. 660 Hallman, P.  408 (fn18) Harley, H. 61 Harris, A.  46–​8, 132 (fn16), 1119 Hartmann, K. 756 Haspelmath, M.  48–​9, 71–​2, 946 Hayashi, M.  840, 847 Heine, B.  448, 461 Hewson, J. 551 Hnin Tun, S.S.  938 Hock, H.H.  536 Hoekstra, T.  668–​70 Hohepa, P.W.  505 Holisky, D.A.  136–​7, 856 Holmberg, A. 341 Hopper, P.J.  448, 884, 924–​5 Hovdhaugen, E. 578 Hukari, T.E.  318 Hutt, M. 826 Hyamas, N.  668–​70 Jahani, C.  484, 488 Jamison, S. 543 Jelinek, E. 195 Jenny, M. 938 Johns, A.  512–​13, 602, 604, 845–​7 Joswig, A. 915 Jügel, T.  470, 473 Kalin, L.  234 (fn9) Kallulli, D.  405, 412, 412 (fn23) Kaufman, D.  516, 569 Kayne, R.  341, 346

 

Author Index   1243 Keenan, E.L.  715 (fn5) Keine, S.  283–​4 Keyser, S.J.  423 Kibrik, A.E.  866, 868 Killian, D.  912–​13 Kimball, G.D.  275 King, J.  811, 817, 820 Kiparsky, P. 646 Klaiman, M.H.  505 Klimov, G.A.  1119 Kolliakou, D. 363 König, C. 452 Korn, A.  484–​5, 487 Kramer, R. 406 Kratzer, A.  176–​7 Kuteva, T. 448 LaPolla, R.J.  925, 945 Lahne, A. 349 Laka, I.  197, 212, 230, 790–​1 Lam, Z.W.M.  432 Larrañaga, P. 654 Larsen, T.  681, 737–​8 Laughren, M.  1096 Legate, J.A.  89, 107, 114–​15, 208–​9, 283, 304, 749, 964 Legendre, G.  273 (fn10) Lehmann, C. 673 Levin, B.  329–​30, 793, 795 Levinson, S.C.  25 Lightfoot, D. 473 Lorimer, D.  769, 776 Lowe, R. 842 Lynch, J. 578 MacDonald, L. 457 Macdonald, K.  991 (fn3) Mahajan, A.  341–​2, 509–​10, 820 Malchukov, A.  943, 1026 Mallinson, G. 918 Manning, C. 628 Marandine, J.-​M.  430 Marantz, A.  114, 130, 193, 348, 368–​9, 773 Martin, L. 40 Massam, D.  120 (fn7), 238, 372 McGinn, R.  386 (fn21) McGregor, W.  206, 264, 635

Meakins, F. 463 Michelena, L.  166–​7 Miller, C.  902–​5 Mohammadi, S. 486 Mohanan, T.  68, 91, 815, 819 Molnár, V. 409 Moravcsik, E, 226 Morin, Y.-​C.  776 Mosel, U.  578, 591 Mulder, J.G.  1020, 1022, 1025, 1031 Munarriz, A. 706 Munshi, S.  765, 772 Murasugi, K.  196, 411 Myhill, J. 449 Næss, A.  50, 268, 278 (fn9) Narasimhan, B. 635 Nash, L.  132, 250, 369 Nevins, A.  402, 652–​4, 660 Nichols, J.  329, 866 Nieuwland, M.S.  703 Nikolaeva, I.  264 (fn4) Norman, W.M.  737–​8 O’Connor, M.C.  129 O’Shannessy, C. 463 Ochs, E. 634 Odden, D. 341 Ordóñez, F.  749, 751 Ortiz de Urbina, J.  160, 166 Otsuka, Y.  576, 584–​5 Oyharçabal, B. 160 Ozerov, P.  935, 938 Palancar, E.  450–​1 Paul, D. 493 Pawley, A. 506 Pearson, M.  604 (fn17) Pensalfini, R. 462 Perlmutter, D. 422 Peterson, J.M.  536 Picallo, C. 356 Pinker, S. 633 Pittman, C. 402 Plank, F. 811 Polinsky, M.  73, 701 Postal, P. 329

 

1244   Author Index Poudel, T.  213–​14, 826 Pray, B. 504 Preminger, O.  124–​5 (fn9), 131, 197, 652 Pye, C. 641 Pylkkänen, L. 598 Queixalós, F. 450 Quick, P.  564, 572 Rackowski, A.  595, 598–​600 Ramchand, G.  198–​9 Randal, S.  915–​16 Rappaport-​Hovav, M.  795 Rezac, M.  124–​5, 243, 341, 791, 803–​5 Richards, N.  589 (fn2), 598–​600 Rigsby, B.  1017 Roberts, I. 70 Roeper, T. 367 Ross, J.R.  425–​7 Ross, M.  519, 578 Rude, N. 457 Ruffolo, R. 567 Rumsey, A. 634 Salanova, A.  210, 358, 366 San Roque, L.  634 Sands, K.  449 (fn1) Sasse, H.-​J.  933 Schachter, P.  386 (fn21) Schäfer, F. 62 Schieffelin, B.B.  634, 639–​40 Schmidt, A. 462 Schnell, S.  52–​3 Schuchardt, H.  1118 Schütze, C. 651 Seiler, W. 129 Shanidze, A.  1123–​24 Silverstein, M.  207–​9, 222, 255, 310, 511 Simpson, J. 962 Smith, C.  182–​3 Speas, M.  427–​8 Spreng, B.  233, 836 (fn6), 840, 844–​5, 847–​8

Starosta, S. 518 Stiebels, B.  756–​7 Stirtz, T.M.  913–​15 Stoll, S.  638, 641 Subedi, A. 826 Svenonius, P. 198 Tada, H. 750 Taplin, G. 453 Tarpent, M.-​L.  1027 Tenny, C.  427–​8 Thompson, S.A.  884, 924–​5 Tiffou, E. 776 Tonoike, S.  1001 (fn13) Trask, R.L.  449 Traugott, E. 448 Tsunoda, T.  213, 233–​4 Urban, G.  1073 Uribe-​Etxebarria, M.  162–​5, 167, 172 Van Everbroeck, E.  713–​14 Van Urk, C.  234 (fn9) Verbeke, S.  825, 827 Vinokurova, N.  176, 235, 236 (fn11), 241 Wade, M.  214–​15 Wexler, K.  668–​70 Wharram, D. 840 Whitman, J.  510, 514–​15 Wiesemann, U.  1973 Willson, S.  761, 765, 772, 776, 776 (fn17) Wiltschiko, M.  1017 Woolford, E.  124, 195, 245, 673, 679 Wurmbrand, S.  168, 173 Yanagida, Y.  510, 514–​15 Yuan, M. 404 Zeisler, B.  928, 930, 942 Zimmermann, M. 756

 

Language Index

Abruzzese 249 Adyghe 233, 325 Afroasiatic 901 Agul  274, 853, 867–​8 Akatek  744–​5 Akhvakh  854–​5, 857, 870 Albanian 405 Anatolian  450, 510–​1 Antekerrepenhe 282 Anywa  452, 903, 908–​10, 920 Apinayé 1070, 1074 Aramaic  873–​4 Aranda  208–​9 Ariellese 245 (fn17) Arosi 506 Australian languages  209, 245, 253, 268, 326, 462, 948–​52, 955, 1096 Austronesian  373–​9, 385, 516, 519, 559, 590, 624 languages  519, 526, 553–​7, 561, 610 (fn23) Proto–​  373 (fn3), 622 Avar  701–​2, 720–​724, 726, 857–​8 Anzuq 869 Standard 863 Avestan  Old  469–​70 Young 470, 473 Awtuw  255–​6 Bagvalal  857, 860–​2, 864, 870 Bahasa Indonesia  379, 387–​91 Bajau  620 (fn32), 627 Balinese  379, 387–​94, 573–​4 Balochi  246, 466–​7, 469, 484–​91, 498 Sarawani  489–​90 Sistan 490 Western Balochi  489 Bandjalang 208

Basque  64, 65–​7, 70, 124–​5, 130–​1, 160–​73, 175, 197–​8, 212, 229–​30, 242–​3, 341–​2, 372, 427, 451, 647–​664, 699–​701, 703–​7, 710, 717–​20, 782–​805, 1095 eastern dialect  792–​5, 797–​8 western and central dialect  792–​5, 797–​8 Bengali  542–​3, 549, 829–​30 Old 542 Bezhta  314, 856–​9, 862, 865, 869–​70 Budukh  856, 866–​7, 870 Burmese  924–​5, 937–​9 Burushaski  760–​80 Cantonese  429–​30 Cashinawa 248 Catalan 357 Ch’ol  672, 674–​8, 680–​7, 725, 737 (fn1), 742–​3 Chol  39 (fn7, 8), 45, 211–​13, 359, 594 (fn7), 725 Chamalal  861–​2, 870 Chamorro  74–​7, 80, 83–​4, 316–​7, 391 (fn28) Chechen  857–​8, 861, 867 Chintang  636–​8, 641, 924–​5, 941 Chirag Dargi  868 Choctaw  217, 219–​20, 328 Chukchi  118, 120, 122–​3, 173, 270, 313–​16, 319, 323–​4, 327–​8, 730, 771 Coast Tsimshian  777–​8; see also Tsimshianic languages Cuzco Quechua  435 Czech 416 Darma 636 Dhirari 208 Dinka  379–​85, 906 Diyari  313–​14 Djapu  293–​5, 304 Dolakha Newar  924–​5, 939–​41

 

1246   Language Index Drehu 128 Duna 634 Dutch 231 Dyirbal  239–​41, 243–​5, 247–​9, 258, 262, 298–​303, 335 (fn2), 378, 462, 510–​11, 950–​2, 960, 982, 1053 East Sudanic languages  918–​19 Eastern Ostyak  237 Eastern Sudanic  921 English  39, 61–​3, 99 (fn21), 198–​9, 234, 325, 329–​30, 334, 356–​7, 360–​2, 364–​5, 370, 410, 415–​16, 430, 512, 523, 598 (fn10), 607, 610 Eskimo-​Aleut language family  397 (fn1), 832–​3 Finnish 266, 503 Folopa  214–​16 Fore 258 Formosan languages  375, 383 (fn14), 518–​9 French  39, 334, 358, 509, 666, 668–​9, 678, 685 Gaahmg  913–​14, 921–​2 Georgian  128, 132, 175–​99, 250, 457, 710, 1114, 1117–​20, 1122, 1127–​36 Modern Standard  1133 Old  1115–​16, 1120, 1125–​6, 1132–​3, 1135–​6 German  124, 342, 440, 731 Germanic language  222, 257 (fn1) Gitksan  1007–​14, 1017, 1023, 1027–​8 Gooniyandi  263–​4, 461 Gorokan languages  451 Greek  355–​6, 360, 362–​5, 370 Gujarati  544–​5, 829 Gunwinjuwan 455 Gurindji Kriol  463, 644; see also Kriol Guugu-​Yimidhirr  951 Halkomelem  314–​15, 317–​18 Upriver  297–​300 Haryani  547–​8, 550, 829 Hawrami 341 Hindi  64, 67–​70, 86–​108, 128, 130, 150–​2, 175, 254–​5, 258–​9, 268, 341, 503–​4, 509, 541–​2, 545, 551, 635, 698–​700, 702, 706, 711–​12, 727–​9, 732 /​Urdu  462, 807, 809–​21; see also Urdu, /​Hindi

Hinuq  856, 863, 867, 870 Hittite 511 Hunzib  854, 856–​7, 868, 870 Ibaloy  556, 562–​77 Icari Dargwa  861 Icelandic  113, 124, 128, 130, 222, 257 (fn1), 349–​50 Ika 263 Imonda  129–​30 Indo-​Aryan  462, 465, 503–​4, 508, 530, 534–​6, 547, 551, 808, 828 Middle (MIA)  151, 536, 538–​40, 544, 549 New (NIA)  530–​1, 533–​6, 540, 544–​5, 547–​8, 550 Old (OIA)  151, 530–​1, 534–​7, 548–​9 Indonesian 556, 624 Ingush  857, 858, 866 Inuit  206 221–​2, 314, 320, 324–​5, 397–​403, 405, 407–​8, 411–​14, 529, 833, 842 Baker Lake  400 Kalaallisut  117–​18, 409; see also West Greenlandic Labrador  308–​9 Labrador Inuttitut  400–​1, 402 (fn12), 403, 406, 847 North Baffin  405, 834, 845 South Baffin  403, 833, 838 (fn11), 844, 847, 849 Inuktitut  512–​3, 602, 604, 608, 832–​50 Arctic Quebec  635, 639–​41 Kangiryuarmiut  843, 846 (fn17) Kivallirmiut 837, 849 Nunavik 836, 842 Iranian  451, 465–​6, 468–​9, 474 (fn6), 504, 507, 535–​6, 873–​4 Middle  471–​3, 497 Old  154, 469–​70, 472–​4 Italian 210, 358 Standard  339 (fn5), 340 (fn6) Jakaltec 324 Jaminjung  451, 453, 1089–​1100, 1102–​12 Japanese 124, 731 Modern 514 Old  514–​5 Jê languages  1065–​68, 1070–​2, 1077, 1086–​7

 

Language Index   1247 Jingulu  453, 462–​3, 950, 1102 Jur-​Luwo  908 K’iche’  673–​681, 683–​7, 745–​8, 753–​5 K’ichean 40 Kabardian 457 Besleney Kabardian  1136 Kaingang 1073 Kala Lagau Ya  461 Kala Lagaw  456 Kalkutungu  950, 960, 966–​7 Kaluli  636, 639–​40 Karata 852 Kartvelian  1114, 1117–​24, 1126, 1132 Proto-​Kartovelian  1135–​8 Kashmiri 256, 333 Katukina-​Kanamari (KatKan)  1036–​52, 1054–​64 Kham  246 (fn18), 289, 291–​2, 304 Kherwada Wagdi  547–​8, 550–​1 Khinalugh  869–​70 Khwarshi 866 Kija 451 Kĩseêdjé’ 1074 Koasati 275 Kriol  462–​3, 987, 1090 Kryz 865 Ku Waru  634 Kurdish  153, 466, 475, 484, 896, 898 (fn13) Central  481–​3, 498–​9 Kurmanji  153–​4, 644; see also Kurmanji Modern Northern  507–​8 Mukuri  482–​3 Norhern  475–​81, 498–​9 Kurmanji  475–​6, 478 Northern  479–​80 Southern 476 Lak  854–​7, 865, 871 Laz  132, 218, 220, 1114, 1116, 1119, 1122, 1124, 1130–​1, 1133, 1135–​6 Lezgian  70–​4, 267, 274, 853, 857, 860, 867, 870 Lhasa Tibetan  128, 930, 942–​3 Majang  915–​6 Malagasy  394–​5, 627 Betsimisaraka 554, 557

Malay 448 Mamuju  616–​23 Maori  312–​13, 505–​6, Māori  584–​5 Marathi  207–​8, 210, 544, 550 Matses 310 Mayan languages  39–​40, 42, 230 (fn6), 326, 359, 529, 635–​6, 638, 670, 673, 677–​9, 681, 725–​6, 737–​42, 746, 750, 754, 756 Mẽbengokre  358, 1067–​9, 1071–​2, 1074–​86 Meitei  925, 931–​6, 938, 942, 945 Mingrelian  460, 1114, 1122, 1124, 1131–​3, 1135 Mongsen Ao  216–​7, 449, 943, 945 Nakh-​Daghestanian  851–​4, 856–​64, 866–​7, 1095 Nam 635 Neo-​Aramaic  234 (fn9), 873–​6, 895, 897 North Eastern Neo-​Aramaic (NENA)  873–​876, 881–​4, 892–​5, 897 C. Barwar  881–​2, 885 C. Bohtan  884, 891 C. Hertevin  891, 895, C. Qaraqosh  891–​2 C. Urmi  882, 886, 892 J. Arbel  886 J. Sanandaj  876–​9, 887–​9 J. Urmi  883, 886, 890, 893 Nepali  213–​14, 824–​8 Newari  264, 460–​1 Nez Perce  139–​42, 206, 222–​3, 283–​4, 295–​6, 304–​6, 337, 457 Ngaliwurru  451, 1090–​1, 1112 Ngarrindjeri 453 Ngayarda languages  458 Nilo-​Saharan languages  901, 918 Nilotic languages  451, 459–​60, 906, 919 Nisga’a  1007, 1012–​4, 1015 (fn6), 1017; see also Tsimshianic languages, Interior Tsimshianic Niuean  89 (fn6), 120, 206, 222, 238, 359, 372, 723–​4, 771 Northern Lwoo languages  903, 906, 918, 920–​1 Nungali 448, 453 Nyulnyul 462 Nyulnyulan languages  453, 459 (fn8) Oceanic languages  506, 558, 577–​8, 587

 

1248   Language Index Paakantyi  454–​6 Paarruntyi  455–​6 Pama-​Nyungan  314, 458, 949 Proto-​Pama-​Nyungan  454, 458, 460 Panará  1066–​9, 1073–​4 Panyjima  458–​9, 984–​7 Papuan languages  451 Päri  452, 903, 906–​8, 920 Paumari 347 Pendau  562–​74 Persian  254, 256, 466, 474, 498, 535 Middle Persian  471–​2 Old Persian  48, 470–​3, 474 (fn4, 5) 505, 508, 535 Philippine languages  375, 518–​9, 572, 569 (fn9), 584, 591–​5, 603–​4 Philippine-​type languages  568, 589, 591, 593, 595, 604, 606–​7, 618 Pitta-​Pitta  460 Polynesian languages  316, 507, 583–​5 Proto-​Polynesian  457 Pomo  129–​30, 262 Popti’ 748, 771 Proto-​Malayo-​Polynesian  561, 618 (fn28), 622 (fn34) Punjabi  822–​4 Puyuma  518–​20, 526 (fn10) Q’anjob’al  308–​10, 725, 748, 750–​3 Q’eqchi’  741, 743–​5 Qiang 257, 1103 Rajasthani  550–​1 Rotuman  558–​9 Roviana 457 Rukai  518–​9, 523, 526–​8 Rudai 522 Tanan 526 Rutul 867 Sahaptin  223–​4, 458 Sakapultek  26–​7, 46 Sakha  235–​8, 241 Samoan  175, 211, 231–​2, 259, 262, 506–​7, 575, 578–​81, 583–​5, 634 Sanskrit  450, 504–​5, 532–​4, 536, 550 Classical 504

Epic  532–​3, 537–​8 Vedic  509, 537, 548–​9 Sanzhi Dargi  854–​5, 867 Scandinavian language  417 Mainland Scandinavian  334 Seediq 502 Selayarese  620, 624, 626–​7 Semelai  265, 777–​8 Semetic languages  873 Sgüüx̠  1008, 1028, 1030; see also Tsimshianic languages, Coast Tsimshianic Shilluk  451, 902–​5, 920 Shipibo  111, 115–​17, 119, 766 Shipibo-​Conibo  271 Sm’algya̠x  1008–​9, 1011, 1024, 1028–​31; see also Tsimshianic languages, Coast Tsimshianic Spanish  37, 39, 358, 657, 665–​6, 668, 678, 685, 704, 726 Squliq Atayal  374–​7, 385 St’át’imcets 613 Sumbawa Besar  644–​5 Surmic  451, 459–​60, 901, 921–​2 Svan  1116, 1119, 1121–​2, 1124–​6, 1129–​30, 1133, 1135 Taba  581–​3 Tabasaran  853–​6, 871 Tacana 463 Tagalog  80–​1, 385–​6, 516–​8, 520–​1, 525–​6, 568, 589 (fn2), 590–​600, 602–​15, 617 Taleshi  466, 469, 491, 495–​6, 498 Anbarane dialect (Anb)  491–​2, 494 Asâlem dialect (Asâ)  494–​5 Mâsal dialect (Mâs)  492–​3, 495 Tauya 457 Thai  426–​7 Tennet  459, 915–​16 Tibetan  268, 924, 928, 931, 942, 945 Tibeto-​Burman  635, 924–​7, 941, 944–​5 Tima  910–​2, 917–​18 Timbira  1072, 1074, 1076–​7 Tongan  312, 506, 575–​9, 581, 584–​5, 989–​1004 Trans New Guinea languages  634–​5 Trumai  74–​5, 78–​80, 84 Tsakhur  853–​4, 857, 867, 869–​70 Tseltal 741

 

Language Index   1249 Tsez  70–​4, 858, 860, 863, 867 Tshangla  146–​50 Tsimshianic languages  1007–​11, 1020, 1024, 1026–​7, 1030 Coast Tsimshianic (CT)  1008, 1011, 1018–​26; see also Sm’algya̠x; Sgüüxs Interior Tsimshianic (IT)  1010–​18; see also Gitksan; Nisga’a Tsotsil  737 (fn1), 740–​2, 745–​6, 748 Tsova-​Tush  129, 136–​9, 142 (fn12), 854, 856, 871 Tukang Besi  620 Ṭuroyo  873–​4, 893–​5, 898 Tz’utujil  741, 745, 753–​5

Warlpiri  142–​6, 173, 205, 220–​1, 272, 312, 463, 632–​3, 636, 641–​2, 730, 732, 772, 948, 950–​87 Light Warlpiri  463, 987 Warrungu 272 Warrwa 635, 1110 Wastek  671, 674–​81, 683–​8 West Greenlandic  83, 320, 324, 839–​40; see also Inuit Western Greenlandic  407, 417; see also Inuit Western Nilotic languages  902, 906 Wik-​Mungkan  456

Udi  853–​4, 856, 858 Uduk  910, 912–​3, 922 Ukranian  308–​9 Umpithamu 455 Urdu 549 /​Hindi  545, 551; see also Hindi, /​Urdu

Xavante  1067–​9, 1071–​5 Xerente 1070, 1074 Xokleng  1068–​9, 1071–​3, 1075

Vafsi 1026 Wangkumara  454, 456, 460 Wardaman 455

Yidiny 950 Yup’ik  119, 770–​1 Yukatec 635 Yukatek  671–​2, 674–​681, 683–​8 Yukulta  154–​7, 326, 950 Zoque 347

 

 

Subject Index

1st/​2nd person pronoun  192, 208, 240–​8, 882–​3, 941, 1103, 1105; see also participant A-​bar (Aʹ, Ā) extraction  74, 374–​6, 379, 715–​16, 1000; see also extraction movement  323, 328, 385, 517, 527, 528 (fn14), 529, 742, 748–​9, 1001; see also extraction A-​bar (Ā) position  394, 595, 604 (fn17) ablative  451, 1089, 1109–​10 ABS=DEF(ault) 107, 849 ABS=NOM 107 ABS-​OBL pattern  232–​4 absolutive  280, 285, 290, 293, 295, 299, 302, 650, 653, 811 -​absolutive verb (ABS-​ABS)  766, 768–​9 argument  89, 378, 623, 627, 653, 791, 849, 853, 866 position 29, 1060 case  79, 121, 173, 336, 348–​50, 409, 652–​3, 838–​9, 849, 867 high ABS  79–​81 marker  288, 679, 683–​4, 687, 1066 mixed ABS  79–​81 licensing 321 object  60, 143, 313, 320, 324, 398, 407, 409–​12, 701, 725 -​object gap  702, 716, 719, 721–​5; see also object, gap person marker  679, 681–​2, 684–​6 PRO 1053 subject 65, 144 -​subject gap  702, 716–​17, 721–​6; see also subject, gap abstract case  173 961, 983, 986 accusative  140–​1, 280, 285, 297–​8, 302, 421 (fn1), 501, 583–​5, 595, 597, 810, 812–​13 alignment  41, 270, 1076

case  lack of 503 unavailability of 90 discourse profile  42 versus ergative alignment, relative learnability of 713 licensing 321 accusativity, typological preference for  711 acquisition  74, 82, 637–​9, 657, 662 of absolutive person marking  679, 682 of ergative marking  634, 639 of ergative person marking  673 of person marking  665, 667–​70 active  alignment 132, 514 case 783 case  pattern  126–​7, 133 class verbs  1123, 1126–​7, 1129–​31 ergative  206, 219–​20 language  45, 127–​30, 206, 217–​20; see also split-​S voice  960, 969, 975, 984–​5 Activity Condition  527 actor  45, 452, 455–​6, 459–​62, 564, 583, 633, 1093 voice  374 (fn3), 564, 569, 595, 609, 621, 624; see also voice system adjacency  393–​4, 510 head-​head  392, 393 (fn33), 395 linear 393 adjunction  394–​5 adpositions  346–​7 advanced Tongue Root (ATR)  902 adverb insertion  675, 681 agent  40, 61–​4, 66, 69–​70, 114, 126, 141, 187, 355–​9, 450–​1, 455–​6, 459–​60, 534–​5, 547, 633, 944, 1089, 1095–​7, 1100–​12 forgrounding 316 marking, see agentive marking

 

1252   Subject Index Agent Focus (AF)  326, 738, 742–​9, 753–​6 agented passive  472, 913–​15 agentive marking  933–​4, 942–​5 agentivity  259, 268, 793, 816 reduced 812 Agree  297, 369–​70, 527–​8, 1000, 1004–​5 agreement  133, 173, 210–​11, 220, 297, 315, 423–​4, 426, 468–​9, 476, 482–​3, 488–90, 498, 540, 544, 740, 744, 746, 754, 762–​5, 769, 773–​5, 787–​91, 813–​16, 840–​4, 849 and case  marking, interaction between  807 morphology, dissociation of  172 double  398–​405, 401–​2, 412, 765, 769 inversion of 876 marker 482, 560 markers and pronominal clitics, distinction between 667 object  140–​2, 401, 762–​5, 775, 839–​40 patterns  337–​9 as post-​syntactic operation  327 subject  762–​4, 839–​40, 855 alignment  parameter 70 hierarchy  70, 74–​5, 81–​2 systems, competition between  713 allocutive 788 agreement 427, 789 allomorphy 283, 301 reduction of 462 animacy  245–​6, 254, 262, 543, 548, 817, 1061, 1105 hierarchy  255, 258, 486, 498; see also person, hierarchy; Silverstein hierarchy scale 285, 946 anticausative  61, 314, 536 anti-​locality  75 antipassive  123, 310–​11, 399–​401, 406, 417, 516, 592–​5, 617–​18, 620, 625, 640, 744, 749, 844–​8, 858–​9, 966, 1056, 1095 construction  233, 272–​3, 309, 316, 322, 377, 382 ~imperfective correlation  316 marker  314–​15 morphology  311, 382, 594 object  310, 321–​2, 324, 625 spurious  327–​8 super-​  619

anti-​topic  316, 407, 410 aorist  178–​84 series  1115–​17, 1119 A-​position  595, 604 (fn17) applicative  115–​20, 193–​4, 377, 382 (fn12), 388, 414–​15, 519, 594–​7, 619–​20, 1059, 1063 high  116, 596, 619 low  194, 414, 596, 598, 619 argument  encoding 279, 291 structure  26, 419, 421–​2, 436, 611, 859, 1093 ergativity 423, 428 aspect  128, 150, 182–​3, 416–​18 category 183 neutral  182–​3 /​aspectual split  210–​13, 228–​31, 234 aspectual properties of verb  847 aspectual splits, fixed directionality of 228, 231 AspP  361, 614, 778 associative clause  974–​977, 979–​80 atelic event  362 atelicity 316 auxiliary  212, 230, 341–​2, 380, 382, 509–​10, 648, 652–​3, 787, 797, 799–​801, 1044–​5 backgrounding 316, 942 backward control  867 bare nominal  799, 815 bare noun  757, 787, 796 bi-​absolutive (bi-​nominative) construction 313 (fn5) biabsolutive construction  857–​8 biclausal configuration  197–​9 biclausal construction  212, 450, 464 binary split  239, 243–​4, 246 binding  62, 747, 1000 non-​local  1004 bleeding 305 blood oxygen-​level dependent contrast (BOLD) 696 body part  461, 953, 958, 960, 972 bound pronoun  453–​4, 949–​52, 1003 (fn16) Burzio’s generalization  83, 276, 187 Call on Addressee (CoA)  430–​1 canonical ergativity, deviation from  478–​9

 

Subject Index   1253 case  absorption 318 agreement  analysis  595–​7, 599, 601 approach 383, 392, alternation  124–​5, 259, 384, 805, 836 (fn8) assignment  281, 303, 338, 366, 412, 415, 761, 765 -​competition  344–​6; see also dependent case competitor  132, 194, 322, 335, 345, 839; see also dependent case concord 190, 952 error 656, 660 feature  175, 501, 527, 797, 999,deletion  286, 291–​4 preservation 286, 291 valuation 1000; see also case, valuation; feature, valuation language 900, 919 neutralization  905, 916–​17 polysemy 550 realization hierarchy  368 relations  92, 101, 105 scale  286 290, 292–​7, 299 -​structure ergativity  423 valuation  59, 195; see also case, feature, valuation; feature, valuation case theory  160; see also Dependent Case Theory; Inherent Case Theory causation 795, 1058 causative  77–​8, 125–​6, 140, 611, 615, 836–​8, 858–​9, 1058, 1063 -​inchoative alternation  118 causativization 858 Causer  61–​4, 66, 69–​70, 73 child language  36–​7 class marker  790 clausal bifurcation  231 clausal conjunction  863, 866 cleft construction  452, 456, 460 clitic  43, 155–​6, 242, 401–​7, 471, 481–​2, 632, 653–​4, 667–​9, 675–​9, 681–​2, 685–​7, 741, 876, 993; see also pronominal clitic; enclitic pronoun climbing 172 doubling  401, 405–​6, 412, 666–​7, 762 (fn3) nominative  667, 676, 678–​9, 682, 685–​6 pronoun  574–​8, 581, 992–​93, 1003

co-​reference relashionship  966 cognate object  954 comment  409–​11, 414 communicative function  24 complement  clause  866–​70, 915–​16 presence of a  148 complementizer  961–​5, 967–​72, 974–​5, 980–​1, 983–​4 complex predicate  148, 186, 796, 800, 1093–​4 N-​V  815, 819 V-​V  818 compound verb  859, 1121 conative  955–​6, 959 alternation  234, 325, 329 /​locative-​type alternation  118, 122 concealed transitive  130–​2, 423 concord  302–​3, 305 Condition on Extraction Domain  518 conjoined construction  879 conjunction reduction  866, 871; see also coordination, reduction connective  1011, 1027–​31 constituent order  917 change 920 contrastive alignment  1007, 1023 control  163, 172, 198, 324, 367, 504, 532, 539, 633, 866–​70, 961–​6, 1047–​9, 1052, 1085, 1098–​100 clause 712 /​ownership  214–​15; see also volitionality stucture 711 converb  863–​6 conveyance voice  590 (fn3) coordination  305–​6, 996–​8, 1002–​3, 1048 clausal 505 reduction  995, 1005; see also conjunction reduction copula  214, 820, 885–​90, 991 construction  516–​19, 523 identification 602 sentence 170 coreference  864–​5, 867–​9, 995–​998, 1002–​4, 1042, 1048, 1085 case constraint on  998 counter-​bleeding  303 C-​T inheritance  527

 

1254   Subject Index cueing  715, 727–​8; see also semantic cue; syntactic cue agreement-​based  725–​6 case-​  716–​18, 721, 723–​4, 726 cyclic head movement  999 dative  124, 130, 151, 472–​4, 458–​9, 596, 650, 653, 655, 761, 767, 772–​6, 810, 813, 819–​20, 825, 939, 950, 968 agreement  596, 660, 662 /​ergative alternation  820, 827 person split  210 datives and ergatives, relationship between 829 default 646 agreement  403–​4, 711 case  349–​51, 651 form  646–​7 inflection 647 morphological 379, 732 defective intervention  350 definiteness  245–​6, 263–​5, 416, 486, 600–​1, 610 marker 452 scale  285, 290, 294, 946 /​specificity effect  609 demonstrative  304, 453–​5, 635, 785 dependent case  112, 124 176, 193, 195, 236, 238, 241, 249–​50, 344–​6, 368–​9, 731, 767–​8, 770, 775–​7, 820 approach 135, 424 Dependent Case Theory (DCT)  112, 115, 120, 760, 772 dependent marker  1010–​11, 1013, 1015, 1020, 1023 dependent-​marking  115, 127, 713–​14, 717, 726, 949 derived nominal  364–​5, 370–​1 derived subject  114–​15, 120–​1, 195, 341, 804 desiderative 156 determiner  360–​1, 371, 785–​6, 1008, 1017–​18, 1027–​32 detransitivization  640, 743–​4 Diff  260–​2, 265 differential agent marking  1089, 1100 Differential Argument Marking  235–​9, 244, 248 Generalization 246

Differential Case Marking (DCM)  269, 546, 548, 550–​1, 767, 808–​9, 812, 816–​17, 820, 826, 1111 asymmetric type  266–​9 ergative/​nominative  818, 826 fluid type  268–​9 split type  268–​9 symmetric type  266–​9, 274 differential marking  925, 929–​31, 940–​1 hierarchy 946; see also animacy, hierarchy; person hierarchy; Silverstein hierarchy Differential Object Marking (DOM)  92–​3, 102, 106–​7, 130, 235–​6, 245–​8, 250, 253–​6, 261–​2, 265–​6, 270, 277 (fn12), 326, 465, 486–​7, 490, 493–​5, 503, 544–​5, 548, 814 (fn4), 845, 878, 882 optimality-​theoretic account of  255 Differential Subject Marking (DSM)  235, 245–​50, 253, 257, 261–​5, 267, 270, 274, 326, 547, 549, 760, 925 direct case  783 directional element  457 disambiguation  938, 943–​4 discourse  function 28 participant 192; see also participant -​pragmatic factor  990 profile  26, 29, 36, 45–​6 prominence  315, 950, 1134 dislocation  537, 651, 1045 distinguishability  255–​6, 262, 269 Distributed Morphology  209, 285, 415, 646, 652, 660 ditransitive  242, 771, 836–​7 double agreement, see agreement double case violation  699 double oblique, see oblique downstep 904 DP  370–​1 case scale  286 dyadic sentence  557–​8, 569–​73 dyadic verb  766, 768, 771 Dynamic-​Stative  883–​4, 886, 895–​896, 898 early posterior negativity  728 economy  74, 1111–​12 constraint  255, 270–​1

 

Subject Index   1255 effectiveness 233, 1103 Effectivness Condition (EF-​CON)  233 effector  1095–​7, 1112 electroencephalogram (EEG)  694–​5 ellipsis  637–​41, 651 Elsewhere Principle  339 Elsewhere Condition  646, 660 enclitic pronoun  956 EPP  74–​5, 80, 82–​3, 510, 526, 600, 999–​1001 ergative-​absolutive alternation  802; see also exceptional case marking ergative-​absolutive syncretism  785–​7 ergative agreement  76–​7, 544–​6, 617–​19, 624–​6, 804–​5 pattern 842 production of 654 ergative analysis  377, 382–​4, 390, 589, 591–​2, 594, 599–​601, 622 ergative case  assignment  115, 123, 136, 138, 140–​4, 146, 150, 152, 154, 346, 771 distribution of  60, 64, 69–​70, 74 on internal arguments  115 licensing of 89 marker  44–​9, 451–​2, 456–​9, 461–​3, 640–​1, 657, 902, 910, 913, 920 marking  450, 452, 562, 592, 634–​6, 817, 916, 1094 Ergative Case Generalization (ECG)  114–​15, 117, 120–​1 ergative/​dative alternation, see dative, /​ergative alternation ergative-​dative verb (ERG-​DAT)  771, 775, 798 ergative discourse profile  25, 29, 32, 36–​40, 42–​4 ergative displacement (ED)  790–​1 ergative extraction  605, 622, 624, 742, 745–​51, 753–​6; see also extraction Ergative Extraction Constraint (EEC)  738, 742, 746–​9, 750, 753–​6 ergative hypothesis  376, 378–​9; see also ergative analysis ergative language  high-​  135, 153 low 136 ergative-​marked complement  960 ergative-​marked instrument  972–​3

ergative-​marked manner adjunct  972 ergative marking  discourse-​pragmaic function of  1102–​3 of instrument  957, 1096 loss of 543 production of 677 ergative morphology  631, 634–​8, 737–​8, 749 ergative person marker  673, 675–​9, 686 ergative-​subject gap  716, 719, 721–​2, 724; see also subject, gap ergativity  classic definition of  468–​9 discourse motivation for  42 morphological  160, 740, 758, 871, 879, 895, 1002 origin of 920 passive origins of  46 syntactic  74–​5, 206, 301, 333, 344, 377–​9, 387, 390–​1, 468, 633, 716–​17, 724, 849, 941, 998, 1005, 1098 event  176–​7, 185–​9, 192–​3, 196–​7, 199 -​denoting predicate  593, 603–​4, 616, 612, 623–​4 -​denoting root  611, 613–​14, 617 related potential (ERP)  693–​5, 698–​702, 704–​5, 707, 713, 720, 726–​8 evidential verb  683 exceptional case marking (ECM)  60, 113–​14, 125, 126 (fn11) exclamatives 440 experiencer  63, 66–​7, 69, 73, 76–​8, 118, 145, 148, 161, 164–​5, 276, 479, 768–​9, 940, 1082 extended ergative  881–​3, 885–​6, 890, 895–​8; see also active; marked nominative; Split-​S marking  681–​2 extended intransitive  557, 559, 567, 569 (fn9), 570–​1, 584, 852, 857, 861, 870 external argument  113–​14, 121, 144, 152, 162–​5, 177, 212–​13, 221, 223, 319, 321, 343–​5, 359, 364, 367–​7 1, 503, 510, 513, 595–​6, 802, 1044–​5 extraction  76, 381–​92, 394, 518, 523, 525, 605–​8, 621–​7, 738, 744, 756; see also A-​bar, extraction; ergative extraction marking  383–​4, 392 object  725–​6 restriction  378, 390, 525, 527, 601 (fn13)

 

1256   Subject Index extraction (cont.) restriction on  595, 607, 622 subject  725–​6 wh-​  388–​90 Ezafe (EZ) construction  476 faithfulness  756–​7, 1111 constraint  260, 286–​8 feature  deletion  286, 289, 291–​3 valuation 999, 1004 filler-​gap dependency  723 fine structure  43–​4 finite clause  952, 962–​3, 975, 977, 983–​4, 986 finiteness  96, 97 (fn17, 18), 160–​1, 173, 178, 190, 195, 748, 1077 fluid A marking, see optional ergative, marking fluid-​S  214 focal ergativity  264 focus  451, 456, 929, 931, 933, 935, 1106 argument  1109–​10 broad  933–​6, 938, 1106 clauses 905 contrastive  451, 930, 936 (FOC) feature  756–​7 narrow 1107 focused pronoun  640 foregrounding 311, 942 functional magnetic resonance imaging (fMRI) 696, 707 event-​related  703 future tense  776–​80 gender  marking  362–​3 /​number agreement  853–​5, 866 genitive  355–​7, 359, 370, 375, 381–​6, 451, 472–​4, 514–​16, 549, 560, 569, 573–​5, 577–​8, 601, 605–​8, 612, 636, 811, 914, 921 interrogative 606, 608 possessor 535 predicate 606 pronoun  48, 565, 588 gerund  360–​2 nominal  360–​2, 370 verbal  360–​2, 366, 370–​1

Given A Constraint  40, 49, 52 given information, see information, structure grammaticalization  275, 447, 667, 912, 920–​1, 1026 grammaticization 42 gross structure  43–​4 grounding  430–​2; see also speech act, structure structure  433–​6 harmonic alignment  284, 286, 290 HAVE-​BE alternation  509 head  -​marking  42, 127, 717, 726, 949 movement  318–​20, 394, 653, 783, 999 headedness 724 holder  63–​4, 69 72–​3, 800; see also experiencer; psych-​predicate homophony  718, 813, 848 imperative  183–​4, 434–​5, 859–​60, 982–​3 imperfective  103–​7, 180–​6, 189, 196–​8, 212–​13, 316, 874, 1095, 1105 implicit argument  194, 319, 367; see also reflexive, -​implicit argument marker implicit object hypothesis  796 impoverishment rule  285, 288, 292 inactive class verb  1123. 1127–​28 inalienable possession  461, 1066, 1103 incorporation  164, 342–​3, 509–​10, 751, 796–​7, 815–​16 noun (NI)  122, 164–​5, 313–​14, 319–​20, 833, 844 (fn16), 1059–​63 pseudo noun (PNI)  139, 312–​14, 393 (fn33) independent accusative phrase  880, 882–​3 Index  260–​2, 265 indexical  452–​6 indexing function  260, 262 indirect syntax  1126 individual-​level predicate  213–​14, 216 individual-​level predication  826 infinitive  820–​1, 964, 970–​1, 974–​7, 977–​9, 981–​3, 986 inflectional error  656, 658–​61 information  and discourse structure  925, 927, 946 given  28, 43, 455, 1136 new  28–​9, 32–​4, 42, 411, 455, 990, 1000–​1, 1102, 1134

 

Subject Index   1257 processing 26 packaging  409, 938–​9 structure  29, 42, 401, 411–​12, 756, 827, 925, 939, 941, 1089, 1092, 1106, 1108–​11, 1113 inherent case  59–​60, 113, 124, 195, 196 (fn12), 221, 341, 502–​3, 731, 798 approach 424 Inherent Case Theory (ICT)  111, 115, 123 inherent ergative hypothesis  159–​60 input frequency  684 instrument/​instrumental  62, 66, 69, 71, 78, 450–​1, 460, 532–​4, 537, 549, 636, 641, 810, 812, 912, 957–​8, 1096–​97 instrumental adjunct  512 instrumental agent  534 instrumental-​to-​ergative hypothesis  534 internal argument  114, 118–​19, 121–​2, 144, 503 interrogative 623 intransitive  954–​5 inverse  function 327 marking 326 voice  564, 572; see also voice system irrealis perfect  889 L2 acquisition  705, 707 labile verb  859 language  contact  461–​2 use  24, 46, 52, 54 learnability 714, 730 left anterior negativity (LAN)  695–​696, 698, 706, 720 lexeme-​based splits  233 lexical arguments  29, 31, 34 lexical effect  151 lexical selection  138–​9, 146 lexical semantic approach to case  817 lexical(ist) approaches to antipassive  317 lexically saturated implicit objects (IMPs)  319–​20 lexico-​semantic process  698 light verb  796, 818–​19 construction  152, 774 (fn15) linguistic realism  54 local conjunction  284, 286–​99 local person  208–​9; see also 1st/​2nd person pronoun; participant

locality restriction  525, 529 locative construction  231 locative prepositional phrase  28 long-​distance agreement (LDA)  711, 815–​16, 867–​7 long-​distance dependency  715–​16 macro-​comparison  561 Mahajan’s Generalization  332 marked intransitive  458 marked nominative  452, 459–​60, 900–​1, 903, 906–​7, 915, 919, 922; see also extended ergative /​ergative split  906 markedness  209, 255, 257, 259, 263, 266–​7, 473, 546–​7 maximized interpretation  417 medio-​active construction  460 (fn8) meteorological verb  593 micro-​comparison  561 minimal search  413 (fn25) modality 549, 1105 modifier  302–​5 mood 402 marker 843 Moravcsik’s genralization  227 morphological case  173 morphological complexity  658 morphological ergativity, see ergativity morphological markedness  877, 883, 895 morphological object, see object morphological subject see subject orientation, see split, ergative orientation morphologically ergative language  749 morphosyntcatic orientation  1132 morphosyntactic processing  699 Multiple Case Condition  370–​1 n  359, 368–​9, 528, 608, 613–​14 N400  695–​6, 698, 704, 706 narrative 638, 643 native vs. non-​native language processing  704–​5 N-​bonding  394–​5 Network Morphology Framework  647 neural network  713–​14 neutral alignment  1007

 

1258   Subject Index neutral aspect, see aspect neutral languages  337–​9 neutralization of case  901 new information, see information, structure new referent, introduction of  34, 42 Niya documents  543 nominal antipassive analysis  319–​20 nominal inflection  785 nominal layer  360 nominal possessor  636, 740; see also possessor nominal predication  520–​1 nominalization  190, 197–​8, 366–​7, 512–​13, 516, 518–​20, 528–​9, 601–3, 608–​10, 628, 850, 1052, 1057 clausal, reanalysis of  359 oriented 1049 nominalizer  366, 518, 521, 528 nominative  178, 280, 285, 297, 349–​50, 451, , 501–​2 557, 561, 566–​7, 570, 572–​5, 577–​8, 595, 811–​12, 815 -​accusative  231, 251, 270, 273, 366, 632, 879, 885, 900, 948, 950–​1, 1007 /​accusative alternation  814 agreement 596, 669 marker  667, 676, 678–​9, 682, 685 clitic, see clitic /​ergative alternation  812 interpretation 137 orientation 1133 person marking  669 pronominal clitic, see pronominal clitic non-​active voice  188 Non-​Canonical Subject construction  473 non-​control infinitival complement  169 non-​culmination of events  180–​1 non-​finite clause  961–​73, 976, 980, 983–​4, 986 non-​finite context  79, 651 non-​perfective aspects  231 non-​pivot subjects  392–​3 noun incorporation, see incorporation nP  359, 366, 369–​7 1 NP  hierarchy 511; see also Silverstein hierarchy; animacy, hierarchy split  207, 209, 510–​11; see also person, split null subject  960–​2, 966, 971 number agreement  191 703–​4

object 1053, 1064 agreement  140–​2, 242 demotion 234 double marking of  893 gap  702, 716, 719–​26 grammatical 420, 422 marking, clitic nature of  405 morphological 1117 shift  221–​3, 248, 250, 402, 417, 595, 599–​60, 610, 767; see also short object movement voice 377, 389 Obligatory Case Parameter  88 Obligatory Case Principle  196 oblique  79, 106, 138, 153, 221, 232–​4, 259, 311–​12, 315–​19, 325, 404, 408–​11, 463, 474–​81, 485, 493–​5, 499, 592, 596, 602, 744–​5, 755, 846, 856, 859 double  478–​9, 499, 1026 obviative  967–​9, 973–​4, 983–​4 omnivorous number  405 Optimality Theory (OT)  255, 258, 263, 284, 292, 808, 1102 architecture  756–​7 optional ergative  55, 644, 1089, 1094 marking  463, 925, 1100–​1 overgenralization 640 P300 713 P600  695–​6, 698–​701, 703–​4, 706–​7, 720, 728–​9 participant  243–​6, 249 nominalization 601 participle  470, 474, 504, 508–​9, 532–​6, 874, 884–​9 ta  473, 534–​7 passive  190, 198, 273, 329, 364–​5, 367–​9, 449–​50, 464, 503–​6, 532, 534, 619, 640, 902, 912, 941 analysis 1118 -​to-​ergative analysis  504–​5 -​to-​ergative hypothesis  503, 533–​5 -​to-​ergative reanalysis  531 participle  874–​6, 878, 895–​6 voice  975, 984–​5 passivization  366–​7 past copula  889, 898 past perfective  894 clause 1105

 

Subject Index   1259 past reference  942 Patient Proto-​Role  793 perception verb  63, 69, 72, 124, 131, 801–​3, 1082 perfective aspect  179, 181, 196, 537, 547, 941 perfectivizing preverb  181 performative hypothesis  425 neo-​  426 person  agreement  788, 853–​4, 871 exceptionality 753, 755 feature 138, 142 hierarchy  207, 256, 258, 855, 1021; see also animacy, hierarchy marking  665–​7 and number features, separation of  789 -​number hierarchy  753 restriction on 880 scale  285, 292, 296–​7, 299; see also animacy, hierarchy split  129, 207–​10, 226, 228, 235, 239, 511, 824; see also NP, split Person Case Constraint (PCC)  242–​3, 403–​4, 654 Person Licensing Condition (PLC)  243–​4 PF deletion  1002–​3, 1005 phase  122–​3, 176, 199, 289, 349, 369, 398, 402, 412–​14, 767, 772, 777–​80, 1004–​5 soft 768, 778 Phase Impenetrability Condition (PIC)  75, 349, 522, 778 phasehood  244–​6 phi-​ (ɸ-​)  agree  402–​6 feature  243, 401, 501–​2, 527–​9, 774, 756–​7, 791, 841–​3, 1004 picture-​matching experiment  721, 723 pivot  375, 375 (fn4) plural agreement  478, 488–​9 plural marker  786 plural morphology  361–​2 polar question  439 possessive  agreement  402 (fn12), 403–​4 construction  212, 508–​9, 535–​6 possessor  140, 194, 376–​9, 404, 451, 473, 508–​9, 516, 535, 603–​9, 613–​16, 636, 740–​1, 746, 768, 840, 896, 1066, 1080

agreement  834, 842–​3 extraction 605 (fn18) raising 140, 536 post-​syntactic operation  327, 726 post-​syntatic optimization  284 postposition  190, 344–​6, 772, 1037–​8, 1047, 1055 PP  102, 122–​3, 365, 367, 370, 772 pragmatic constraint  32 pragmatic factor  643 pragmatic functions  635 predicate nominal  522–​3 Preferred Argument Structure (PAS)  29, 31–​2, 35, 39, 41–​3, 45–​50, 737 (fn1) constraint  455–​6 preposition  187, (fn6), 341, 347, 448, 451, 509, 601 (fn13), 749, 896, 902, 920 (P-​) incorporation, see incorporation presentative 435 Primary Actant Immunity Principle (PAIP)  270–​3 privative clause  981–​3 PRO  79, 802, 961–​3, 966, 969–​70, 980, 982–​4, 1047, 1053; see also control pro  192, 194, 401 (fn9), 987, 1015 -​drop  141, 143–​4, 400, 661, 834, 838, 866, 1015 Probe-​Goal relation  173, 297, 791 processing  cost 712 load 717 progressive  197–​9, 212, 229–​30, 867, 1076, 1085–​5, 1095 prominence  265, 273, 925, 1102 scales  255, 284–​5; see also animacy, hierarchy; Silverstein hierarchy pronominal clitic  28, 220, 472, 481, 489, 495–​6, 498, 666, 669; see also clitic nominative 669 pronoun  403–​4, 568, 575–​81, 823–​4, 879, 991–​2, 1012, 1018, 1023–​5 prosodic detatchment  1109 proximate/​obviate distinction  223 pseudo-​cleft  198, 608, 994 pseudo-​noun incorporation, see incorporation pseudo-​progressive  1095, 1097 psych-​predicate  63–​4, 66, 72, 77, 799–​800 purposive clause  970–​4

 

1260   Subject Index quantity constraint  32–​4, 42, 49 raising  867, 1045, 1047, 1057 -​to-​ergative  160, 170–​1, 341 modal  168–​70 of object  749 of subject  414 verb  803–​4 realis mood  155–​7 reciprocal construction  862–​3, 956 reduced transitivity, see transitivity referential address  409–​13 referentiality 265, 271 reflexive  314, 319, 330, 746–​7, 751–​2, 756, 860–​2, 868–​70, 956, 1071, 1093 -​implicit argument marker (RIAM)  179 (fn2), 188–​9, 191 relative case  399–​400 relative clause  302, 304, 391, 516–​19, 522–​9, 623, 627, 850, 1001, 1041, 1080–​1 object  697, 700–​2, 706; see also object, gap prenominal  94–​108 subject  697, 700–​2, 706; see also subject, gap relativization  378, 391, 526–​7, 605 (fn18), 621, 623, 716 (fn5), 849, 994, 1000–​2, 1005 Relativized Minimality  338, 513 Relevance Principle  271–​2 response structure  433, 436–​41; see also speech act, structure restructuring  177, 186–​7 resultative participle  884–​5, 887, 895 resultative perfect  883–​4, 886, 896, 898 resultative semantics  531 resumptive pronoun  455, 567–​8, 878, 994, 1001 reversed case marking  861–​3 right-​lateralized anterior negativity (RAN)  703, 706, 728 role constraint  32, 34, 42; see also Preferred Argument Structure Romanian supine  366 (fn4), 370 rule-​based linguistic knowledge  698, 706 S/​A agreement  826 sampling 637 scale  -​based differential object marking  283 -​driven impoverishment  284

scope  324–​5, 840, 959 ambiguity 325 economy 408 (fn19) narrow  324, 408, 411 wide  324, 399, 406–​8, 410, 840 scrumbling 1000 secondary predicate  192, 461, 960 self-​paced reading  693–​4, 700–​1, 713, 718, 721 semantic bootstrapping  633–​4 semantic control  177, 186 semantic cue  703, 727; see also cueing semantic processing  703–​4 semantic role  260, 547, 854, 862, 1008, 1028, 1044–​5, 1057–​8, 1063 semantic split  793 semi-​transitive  954–​5 short object movement  236, 240–​1, 244, 248; see also object, shift Silverstein hierarchy  254, 880, 882–​3, 890–​1, 897 small clause  164–​5, 169, social context  215, 634, 639 soft constraint  32 spatial case  852, 857–​8, 860–​1, 867 specificity  235–​8, 245–​6, 599–​601; see also definiteness speech act  424, 428 modification 430 participant  457–​8, 1105 phrase 427 (SA) structure  419, 426, 428–​32, 436 spell out  120, 369, 767–​8, 1001 (fn15) split-​alignment language  711, 714, 727 split ergative  alignment  503, 507, 635 case pattern  205 language  90, 128, 205–​6, 219, 223, 760, 809, 823, 906, 910, 1032 orientation 1132 system  639, 1101–​02, 1105 split ergativity  160–​1, 176, 206, 217–​21, 223–​4, 226, 228–​9, 248, 250–​1, 253, 397, 463, 533, 547, 895, 941, 993, 1017, 1022 definition of  206, 218–​19 tense-​aspect governed  777 split intransitivity  275–​6

 

Subject Index   1261 split-​S  161, 206, 219, 340, 792, 875, 877, 887, 890, 895–​6, 898; see also active splits, universal directionality of  227, 251 spurious antipassive, see antipassive stage level  predicate  213–​14, 216 predication 826 status suffix  744 stem syncretism  786–​7 stranding  521–​3 Strict Cycle Condition  767–​8 structural case  60, 80–​3, 88, 93, 99–​102, 111–​13, 135, 160, 189, 196, 280–​2, 313, 322, 502, 512, 515, 620, 731–​2, 820, 839, 849, 999 subject 29, 41 control 711 gap  702, 716–​17, 719, 721–​6 grammatical  124, 413, 420–​2 morphological 1117 -​in-​situ generalization  370–​1 notion of  1053 -​object asymmetry  715 obligatorily controlled  983–​4 syntactic  309, 878, 1117, 1126, 1132–​3 thematic  114, 157, 421–​2 voice  377, 381, 383, 388–​9 Subject Processing Advantage (SPA)  715–​18, 721, 724–​6 subjecthood  473, 1052–​4, 1117 subjunctive  183–​4, 448, 833 Subset Principle  288, 660 subset problem  83 Superset Principle  337, 339 SVO order  332–​3 TP-​internal SVO order  334–​5 switch-​reference  840, 1099–​100 syncretism  209, 275, 314, 335, 337, 339–​40, 377, 529, 539–​40, 608, 785–​7; see also ergative-​absolutive syncretism syntactic accusativity  870–​1 syntactic cue  703, 727; see also cueing syntactic ergative case  340 syntactic ergativity, see ergativity syntactic and morphological case, distinction between 951 syntactic split  904

syntactically ergative language  76–​7, 80, 529, 717, 748–​9, 1047 T  80, 88–​9, 91–​4, 96, 100–​1, 103–​5, 117, 176, 195, 280–​2, 327, 366, 501–​2, 527, 763, 999, 1004–​5 TAM split  163, 210, 229 (fn3), 233, 299 telicity 416, 817 telic event  362 tense  128, 154, 160, 170–​1, 177–​9, 193, 197–​8, 210, 371, 497, 778–​80, 839–​40, 1105 /​aspect processing  702 ergative  178, 184–​190; see also aorist; perfective aspect invariance 402 nominative  178, 184–​5, 188–​9; see also imperfective thematic role  420–​1, 611, 633 thematic suffix (TS)  185–​7 theta-​ (θ-​)  position  138–​9, 144, 149, 157, 195 role  59–​61, 113–​15, 121–​3, 160, 195 Theta Theory  160 three-​way system  281–​4, 288, 293, 295, 297–​8, 306 three-​way systems, morphological reanalysis of 283 tone 901, 919 topic  405, 409–​14, 417–​18, 756, 931, 934–​5, 937–​8, 1107–​9 changing construction  301–​2 /​comment structure  414 construction 912 continuity  41–​2, 1099 contrastive  412, 931, 933–​5, 937 grammaticization of 41 movement 413 sentential  398, 407, 409–​14 topicalization  391, 452, 594, 621, 623, 626 topicalized constituent  567–​8 topicalized object  930 TotalErg hypothesis  159, 173 TP 366 transimpersonal construction  275–​6 transitive split  1049 transitivity  60, 121, 126, 138, 244 (fn15), 376–​8, 719, 726, 884, 924–​5, 939–​42 condition  123–​4, 345 parameter 271

 

1262   Subject Index transitivity (cont.) reduced  156–​7, 229, 595 requirement 322 scale  284–​6, 290, 293, 295, 297, 299 tripartite languages  336–​7, 340 tripartite system  65 (fn8), 572; see also three-​way system unaccusative  137, 365, 367, 422, 648, 764–​5, 769, 817, 836, 845, 875, 877, 881, 888 Unaccusative Hypothesis  817 undergoer  45, 452, 455–​6, 458–​9, 462, 563–​4, 583, 1093 unergative  137, 191, 648, 764–​5, 810, 816–​17, 826, 836, 845, 888 nominalization  356 (fn1), 372 predicate  191, 792, 796–​7 unexpressed object alternation  330 Unique Checking Constraint (UCC)  668–​9 universal processing strategies  700 v (little v)  59. 62, 65, 67–​8, 72, 74, 81, 88–​9, 93, 100, 105–​6, 113, 175, 189, 280–​2, 341, 502–​3, 608, 617, 621

valency  328, 593, 611, 616, 858–​9, 913 verb serialization  479, 940 (fn13) verbal agreement and case marking, relation between 654 vocabulary insertion  285, 287–​9, 652, 732 voice  alternation 273, 602 morphology  374–​8, 382, 386, 596, 601, 611 paradigm 590 system  373–​5 (fn4), 380, 391, 589; see also actor, voice; inverse, voice Voice  176–​7, 187–​9, 364–​5 VoiceP  359, 364–​5, 367–​70, 372, 614 volition  68–​9 volitionality  259, 268, 928, 943, 1095 vP  136, 176, 189, 341, 359–​60, 366, 369, 600 VP 369 Wackernagel clitic  574 Wackernagel position  471, 481 wh-​extraction, see extraction word order  42, 222, 315, 452, 554, 562, 566, 569, 573, 639–​40, 713–​14, 834–​5, 951 change  562, 569, 572

 

OXFORD HANDBOOKS IN LINGUISTICS THE OXFORD HANDBOOK OF APPLIED LINGUISTICS Second edition Edited by Robert B. Kaplan

THE OXFORD HANDBOOK OF ARABIC LINGUISTICS Edited by Jonathan Owens

THE OXFORD HANDBOOK OF CASE Edited by Andrej Malchukov and Andrew Spencer

THE OXFORD HANDBOOK OF COGNITIVE LINGUISTICS Edited by Dirk Geeraerts and Hubert Cuyckens

THE OXFORD HANDBOOK OF COMPARATIVE SYNTAX Edited by Gugliemo Cinque and Richard S. Kayne

THE OXFORD HANDBOOK OF COMPOSITIONALITY Edited by Markus Werning, Wolfram Hinzen, and Edouard Machery

THE OXFORD HANDBOOK OF COMPOUNDING Edited by Rochelle Lieber and Pavol Štekauer

THE OXFORD HANDBOOK OF COMPUTATIONAL LINGUISTICS Edited by Ruslan Mitkov

THE OXFORD HANDBOOK OF CONSTRUCTION GRAMMAR Edited by Thomas Hoffman and Graeme Trousdale

THE OXFORD HANDBOOK OF CORPUS PHONOLOGY Edited by Jacques Durand, Ulrike Gut, and Gjert Kristoffersen

THE OXFORD HANDBOOK OF DERIVATIONAL MORPHOLOGY Rochelle Lieber and Pavol Štekauer

THE OXFORD HANDBOOK OF DEVELOPMENTAL LINGUISTICS Edited by Jeffrey Lidz, William Snyder, and Joe Pater

THE OXFORD HANDBOOK OF ERGATIVITY Edited by Jessica Coon, Diane Massam, and Lisa deMena Travis

THE OXFORD HANDBOOK OF GRAMMATICALIZATION Edited by Heiko Narrog and Bernd Heine

THE OXFORD HANDBOOK OF HISTORICAL PHONOLOGY Edited by Patrick Honeybone and Joseph Salmons

THE OXFORD HANDBOOK OF THE HISTORY OF ENGLISH Edited by Terttu Nevalainen and Elizabeth Closs Traugott

THE OXFORD HANDBOOK OF THE HISTORY OF LINGUISTICS Edited by Keith Allan

THE OXFORD HANDBOOK OF INFLECTION Edited by Matthew Baerman

 

THE OXFORD HANDBOOK OF INFORMATION STRUCTURE Edited by Caroline Féry and Shinichiro Ishihara

THE OXFORD HANDBOOK OF JAPANESE LINGUISTICS Edited by Shigeru Miyagawa and Mamoru Saito

THE OXFORD HANDBOOK OF LABORATORY PHONOLOGY Edited by Abigail C. Cohn, Cécile Fougeron, and Marie Hoffman

THE OXFORD HANDBOOK OF LANGUAGE AND LAW Edited by Peter Tiersma and Lawrence M. Solan

THE OXFORD HANDBOOK OF LANGUAGE EVOLUTION Edited by Maggie Tallerman and Kathleen Gibson

THE OXFORD HANDBOOK OF LEXICOGRAPHY Edited by Philip Durkin

THE OXFORD HANDBOOK OF LINGUISTIC ANALYSIS Second edition Edited by Bernd Heine and Heiko Narrog

THE OXFORD HANDBOOK OF LINGUISTIC FIELDWORK Edited by Nicholas Thieberger

THE OXFORD HANDBOOK OF LINGUISTIC INTERFACES Edited by Gillian Ramchand and Charles Reiss

THE OXFORD HANDBOOK OF LINGUISTIC MINIMALISM Edited by Cedric Boeckx

THE OXFORD HANDBOOK OF LINGUISTIC TYPOLOGY Edited by Jae Jung Song

THE OXFORD HANDBOOK OF MODALITY AND MOOD Edited by Jan Nuyts and Johan van der Auwera

THE OXFORD HANDBOOK OF NAMES AND NAMING Edited by Carole Hough

THE OXFORD HANDBOOK OF SOCIOLINGUISTICS Edited by Robert Bayley, Richard Cameron, and Ceil Lucas

THE OXFORD HANDBOOK OF TENSE AND ASPECT Edited by Robert I. Binnick

THE OXFORD HANDBOOK OF THE WORD Edited by John R. Taylor

THE OXFORD HANDBOOK OF TRANSLATION STUDIES Edited by Kirsten Malmkjaer and Kevin Windle