Alignment and Alignment Change in the Indo-European Family (Oxford Studies in Diachronic and Historical Linguistics) 9780198857907, 019885790X

Citation preview

Alignment and Alignment Change in the Indo-European Family

O X F O R D ST U D I E S I N D I A C H R O N I C A N D H IST O R I C A L L I N GU IST I C S General editors Adam Ledgeway and Ian Roberts, University of Cambridge Advisory editors Cynthia L. Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Oxford RECENTLY PUBLISHED IN THE SERIES 43 Syntactic Features and the Limits of Syntactic Change Edited by Jóhannes Gı´sli Jónsson and Thórhallur Eythórsson 44 Romance Object Clitics Microvariation and Linguistic Change Diego Pescarini 45 The Diachrony of Differential Object Marking in Romanian Virginia Hill and Alexandru Mardale 46 Noun-Based Constructions in the History of Portuguese and Spanish Patrı´cia Amaral and Manuel Delicado Cantero 47 Syntactic Change in French Sam Wolfe 48 Periphrasis and Inflexion in Diachrony A View from Romance Edited by Adam Ledgeway, John Charles Smith, and Nigel Vincent 49 Functional Heads Across Time Syntactic Reanalysis and Change Edited by Barbara Egedi and Veronika Hegedűs 50 Alignment and Alignment Change in the Indo-European Family Edited by Eystein Dahl For a complete list of titles published and in preparation for the series, see pp. 366–86

Alignment and Alignment Change in the Indo-European Family Edited by

E Y ST EI N D A H L

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 Eystein Dahl 2022 © the chapters their several contributors 2022 The moral rights of the authors have been asserted 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: 2022935971 ISBN 978–0–19–885790–7 DOI: 10.1093/oso/9780198857907.001.0001 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 Series preface List of figures List of tables List of abbreviations The contributors

1. Alignment and alignment change in the Indo-European family and beyond Eystein Dahl 2. Alignment in Proto-Indo-European Eystein Dahl 3. Split alignment, mixed alignment, and the spread of accusative morphosyntax in some archaic Indo-European languages Paola Cotticelli and Eystein Dahl 4. The origin of ergative case markers: The case of Hittite revisited Silvia Luraghi and Guglielmo Inglese With an appendix by Petra Goedegebuure

vii viii ix xi xiv

1 24

64 123

5. Passives and anticausatives in Vedic Sanskrit: Synchronic and diachronic perspectives Hans Henrich Hock

166

6. Non-nominative arguments, active impersonals, and control in Latin Michela Cennamo and Claudia Fabrizio

188

7. Infinitives and subjecthood between Latin and Old Italian Claudia Fabrizio

221

8. Alignment changes with Spanish experiential verbs Chantal Melis

246

9. Armenian morphosyntactic alignment in diachrony Robin Meyer

277

vi

CONTENTS

10. Areal and diachronic trends in argument flagging across Slavic Ilja A. Serzˇant, Bjo¨rn Wiemer, Eleni Buzˇarovska, Martina Ivanová, Maxim Makartsev, Stefan Savic´, Dmitri Sitchinava, Karolı´na Skwarska, and Mladen Uhlik

300

References Index

328 361

Series preface Modern diachronic linguistics has important contacts with other subdisciplines, notably first-language acquisition, learnability theory, computational linguistics, sociolinguistics and the traditional philological study of texts. It is now recognized in the wider field that diachronic linguistics can make a novel contribution to linguistic theory, to historical linguistics and arguably to cognitive science more widely. This series provides a forum for work in both diachronic and historical linguistics, including work on change in grammar, sound, and meaning within and across languages; synchronic studies of languages in the past; and descriptive histories of one or more languages. It is intended to reflect and encourage the links between these subjects and fields such as those mentioned above. The goal of the series is to publish high-quality monographs and collections of papers in diachronic linguistics generally, i.e. studies focussing on change in linguistic structure, and/or change in grammars, which are also intended to make a contribution to linguistic theory, by developing and adopting a current theoretical model, by raising wider questions concerning the nature of language change or by developing theoretical connections with other areas of linguistics and cognitive science as listed above. There is no bias towards a particular language or language family, or towards a particular theoretical framework; work in all theoretical frameworks, and work based on the descriptive tradition of language typology, as well as quantitatively based work using theoretical ideas, also feature in the series. Adam Ledgeway and Ian Roberts University of Cambridge

List of figures 2.1. Singular and plural agreement with neuter plural subjects in Ancient Greek

40

2.2. Distribution of neuter nouns in core argument functions

48

2.3. Ranges (divided into 10% segments) of the probabilities of the different tenses and word classes of alignment probabilities for Proto-Indo-European, contrasted to grammatical hierarchies of categories observed in languages

52

2.4. Relative inherent animacy and preferred morphosyntactic alignment

54

2.5. Ergative → accusative alignment change and the animacy hierarchy

55

3.1. Actor–Undergoer Hierarchy

70

5.1. The development of vidyáte

175

5.2. Kulikov’s speculative scenario (2011a: 198) (with oval and circle added by author)

183

5.3. Alternative speculative scenario

186

10.1. Ratios of (flagging) alternation

307

10.2. Transitivity prominence of different Slavic languages

310

10.3. Transitivity degree of Slavic languages

311

10.4. Number of morphological cases in languages of Europe

314

10.5. Predicative possession marking in Europe

315

10.6. Nominative marking of subject-like argument across Slavic

320

10.7. Similarity dendrogram of Slavic languages

323

List of tables 1.1. Different alignment types

3

1.2. Three approaches to S, A, P

13

2.1. Nominative and accusative case endings in PIE

29

2.2. The nominative and accusative forms of 1st- and 2nd-person pronouns

30

2.3. Split case marking in PIE

30

2.4. Two complementary agreement patterns in PIE

35

2.5. Compatibility of nouns/pronouns with core argument functions in PIE

37

2.6. Singular and plural verb agreement in Homeric Greek

39

2.7. Agreement patterns of neuter plural subjects through the history of Greek

39

2.8. Proto-Indo-European as a split-agreement language

44

2.9. Distribution of neuter nouns in the core argument functions

47

2.10. Expected distribution of neuter nouns in core argument functions

48

2.11. Attractions and repulsions of neuters and core argument functions

49

2.12. The pre-PIE voice system according to Pooth and Orqueda (2021)

57

2.13. Outline of the PIE gender–number system according to Melchert (2014)

60

3.1. Some recurring alignment types

65

3.2. The periods of the Latin language (adapted from Cuzzolin and Haverling 2009: 20)

73

3.3. The nominal case system in Latin

74

3.4. Stages of Ancient Greek

80

3.5. The Greek nominal case system

81

3.6. Chronological stages of Vedic

88

3.7. The Vedic nominal case system

89

3.8. Chronological stages of Hittite

95

3.9. The Hittite case system

95

3.10. Chronological stages of Armenian

104

3.11. The Armenian case system

104

3.12. Chronological stages of Lithuanian

108

3.13. The Lithuanian nominal case system

109

3.14. Lithuanian one- and two-place verbs with non-nominative case marking of HRA (adapted from Wiemer and Bjarnadóttir 2014: 333)

113

x

LIST OF TABLES

3.15. Accusativization across Indo-European languages (preliminary version)

116

3.16. Accusativization/subjectization across Indo-European languages

118

3.17. Accusativization/subjectization across Indo-European languages

119

4.1. Coding of A, S, and O with common and neuter nouns

127

4.2. Secured nouns with -ant- as derivational morpheme (adapted from Goedegebuure 2018)

140

4.3. Distribution of derivational and ergative -ant- (adapted from Goedegebuure 2018)

141

4.4. Alignment in Proto-Anatolian

146

6.1. The passive–impersonal continuum

192

6.2. Synthetic and analytic passives

193

6.3. Personal~impersonal alternation with divalent verbs (NOM–ACC)

205

6.4. Personal~impersonal alternation with divalent verbs (NOM–OBL)

206

6.5. Personal~impersonal alternation with monovalent verbs (NOM (–DAT))

206

7.1. Aktionsart, macrorole, and infinitive subjects in Latin

226

7.2. Intransitive finite predicates with infinitives as subjects in Latin

227

7.3. Constraints on activity predicates with infinitives as subjects in Latin

227

7.4. Infinitive subjects instantiating active-inactive alignment in Latin

230

7.5. The behaviour of subject infinitives in Old Italian

234

7.6. Experiencer constructions with infinitive as stimulus in Latin and Old Italian

239

7.7. Differential marking of infinitival subjects in Old Italian

241

8.1. Competition between MIDDLE and DATIVE

269

8.2. Emotion of EGO favours DATIVE

270

9.1. 1.SG.IND forms of sirem ‘to love’ and hełum ‘to pour; to flow’

278

9.2. Declension of the noun am ‘year’, and the personal pronouns of the 1.SG and 3.SG

279

9.3. Summary of constituent marking and morphosyntactic alignment

284

9.4. Analogical shift explanation of morphosyntactic alignment in Classical Armenian (Stempel 1983)

289

9.5. Diachronic development of the Armenian PRS as seen in the 1.SG.ACT of grem ‘I write’

295

10.1. Structure of the database

305

10.2. The number of object-marking patterns across all 46 verb meanings in different Slavic lgs

318

10.3. Case frames and verbs for ‘leave sth to sb’

323

10.4. Dialect continua that have formed modern East Slavic languages

325

List of abbreviations 1 2 3 A A ABL ABS ACC ACT ADV AGR AOR Arm. ART AUX BC BO BS C CE CL COMP COND CONJ CONN CTH CVB DAT DEF DEM DET DIR DO DU EMPH ERG F FUT

first person second person third person bivalent subject/Actor macrorole bivalent subject ablative absolutive accusative active adverb agreement aorist Armenian article auxiliary Before Christ bivalent object bivalent subject common gender Christian Era clitic object marker complementizer conditional conjunction connective Catalogue des textes hittites converb dative [case] definiteness marker demonstrative determiner direct direct object dual emphatic particle ergative feminine future

xii

LIST OF ABBREVIATIONS

GEN GER gr. hit. IMP IMPERS IND INDF INF INS INTJ INTERR intr. IPF IPFV KBo KUB LOC LOG M MArm. MEA MH MID MP MS MWA N NACC NEG NH NOM NONFUT NONPST NP NS O O OBJ OBL OH OPT OS P P

genitive gerundive Greek Hittite imperative impersonal indicative indefinite infinitive instrumental interjection interrogative intransitive imperfect imperfective Keilschrifttexte aus Boghazko¨y Keilschrifturkunden aus Boghazko¨y locative logophore masculine Middle Armenian Modern Eastern Armenian Middle Hittite middle mediopassive Middle script/monovalent subject Modern Western Armenian neuter [gender] nota accusativi negation New Hittite nominative [case] nonfuture nonpast noun phrase New script bivalent object (= P) object [marking] object marker = nota accusativi (NACC) oblique Old Hittite optative Old script bivalent object (= O) past

LIST OF ABBREVIATIONS PART PASS PFV PL PLPF PN POSS POST PP PP PPP PRF PROG PROH PRON PRP PRS PRV PSTP PST PTCL PTCP Pth. QUOT REFL REL RRG S SAP s.v. sb. SBJ SG SPLV sth. SU SUP TNS TODP TSA U VOC

participle passive perfective plural pluperfect proper noun possessive posteriority; temporal suffix ‘since, after’; spatial case ‘behind’ prepositional phrase past/perfect participle past passive participle perfect progressive prohibitive marker pronoun preposition present preverb postposition past particle participle Parthian quotative particle reflexive relative Role and Reference Grammar monovalent subject speech act participant sub voce somebody subjunctive singular superlative something subject supine tense today’s past tense tense-sensitive alignment Undergoer macrorole vocative

xiii

The contributors Eleni Buzˇarovska is Professor of Linguistics in the English Department at the University of Cyril & Methodius (Skopje, North Macedonia). Her research encompasses issues from both theoretical and applied linguistics with a special focus on the typology of Balkan languages and contact phenomena within the Balkan context. Her special interest in the field of SLA is the role of Macedonian as L1 in the acquisition of English as a foreign language. Apart from co-authoring several books, she has published a number of articles in international journals and edited volumes on language change and on comparison of various syntax–semantic features in South Slavic languages. Michela Cennamo is Professor of Linguistics at the University of Naples ‘Federico II’. She is the author of numerous articles on the grammar and diachrony of transitivity, voice, argument structure, and alignment in Latin and Italo-Romance, with special focus on split intransitivity, auxiliary selection, passives, impersonals, and reflexives—also from a panRomance perspective. In her research she has paid particular attention to the Latin evidence for the early stages of these domains and their subsequent history in Italo-Romance. Her research interests also cover grammaticalization, valency, and the lexicon–syntax interface. Her recent publications include Argument Structure in Flux (2013, Benjamins, co-edited with Jóhanna Barðdal and Elly van Gelderen); Contrastive Studies in Verbal Valency (Benjamins, 2017, co-edited with Lars Hellan and Andrej Malchukov); Historical Syntax 2015 (Benjamins, 2020, co-edited with Claudia Fabrizio). Paola Cotticelli is Professor of Historical and Comparative Linguistics at the University of Verona. Since 2007, she has been the Director of the Department Research Centre for Studies in Metalanguage and History of Ideas. She participated in and directed many research programmes in the history of ideas, linguistic concepts, and grammarians. In 2007, she published Lessico di Linguistica (Edizioni dell’Orso, Alessandria). Her other research fields include Hittitology, historical linguistics and ancient languages, and graphematics. Eystein Dahl is a Research Associate in the Department of Language and Culture at UiT— The Arctic University of Norway. He has previously held positions at the University of Bergen, and at Goethe University in Frankfurt. His primary research interests lie in the interface between comparative philology and diachronic typology, and much of his recent work focuses on tense/aspect semantics, alignment typology, and morphosyntactic change in Vedic Sanskrit, Latin, Ancient Greek, and Hittite. Claudia Fabrizio is Associate Professor of Linguistics at the University ‘Gabriele D’Annunzio’ of Chieti/Pescara. She is the author of numerous articles on the morphosyntactic typology and alignment of Latin. Her research interests also include Italian lexicography and the history of linguistics. Her recent publications include Historical Syntax 2015 (Benjamins, 2020, co-edited with Michela Cennamo).

THE CONTRIBUTORS

xv

Hans Henrich Hock , Professor Emeritus, University of Illinois, works on Indo-European comparative–historical linguistics with main focus on Sanskrit. Major publications include Principles of Historical Linguistics (2021); Language History, Language Change, and Language Relationship (co-author, 2019); Studies in Sanskrit Syntax (ed., 1991); The Languages and Linguistics of South Asia: A Comprehensive Guide (co-editor, 2016). Guglielmo Inglese obtained his Ph.D. in linguistics from the University of Pavia and University of Bergamo in 2019, and then worked as a postdoctoral researcher first at the University of Pavia and then at KU Leuven. He is currently Senior Assistant Professor at the University of Turin, where he teaches general and historical linguistics. His main research interests include Indo-European and historical linguistics and linguistic typology. Martina Ivanová is Professor of Slovak Linguistics at the University of Prešov and the researcher at Ľ. Štúr Institute of Linguistics, Slovak Academy of Sciences, Bratislava (Slovakia). Her research focuses on Slovak grammar, morphematics, corpus and cognitive linguistics, and teaching Slovak as a second language. She has published monographs on modal verbs and modality, cognitive grammar, verb valency, morphematics, and derivational morphology. She is the co-author of five monolingual dictionaries of Slovak language (valency dictionary, dictionary of root morphemes, dictionary of word formation means). Currently, she is also the editor of a monolingual dictionary of contemporary Slovak, Slovnı´k súcˇasného slovenského jazyka (since 2018). Silvia Luraghi is Professor of Linguistics at the University of Pavia. She has published extensively on Indo-European linguistics with a focus on Hittite, Ancient Greek, and Latin. Her research interests cover historical linguistics, language typology, and cognitive linguistics. She is the editor of the Journal of Historical Linguistics. Maxim Makartsev is a research fellow at the Institute of Slavistics, Carl von Ossietzky University (Oldenburg) and a senior research fellow in the Institute of Slavic Studies, Russian Academy of Sciences (Moscow). His work is focused on grammatical semantics, verbal morphology, and morphosyntax in Balkan languages (especially Balkan Slavic). He also studies language contact and change in Slavic dialects in Albania. Chantal Melis is Professor of Hispanic Linguistics at the Universidad Nacional Autónoma de México. Her research activities focus on several aspects of the syntax of Spanish, both diachronic and synchronic, examined from a functional–typological perspective. She has published articles dealing with topics such as differential argument marking, nonnominative subjects, voice phenomena, event structure, and verbal aspect. She is currently working on light verb constructions. Robin Meyer completed his doctoral thesis on Iranian–Armenian language contact at the University of Oxford in 2017. After three years as the Diebold Research Assistant in Comparative Philology and Lector for Latin and Ancient Greek at Oxford, he is now Assistant Professor in Historical Linguistics at the Université de Lausanne. His research focuses on historical language contact, diachronic syntax, and the typology of language change.

xvi

THE CONTRIBUTORS

Stefan Savic´ obtained his Ph.D. at Rhodes University (Makhanda, South Africa) in 2021 with a thesis on tense and aspect in Xhosa. He is currently involved on a project examining the linguistic convergence areas in Europe. Work in progress includes typological and diachronic studies of future temporal reference and modality in the Slavic languages. Ilja A. Serzˇant is Professor of Slavic linguistics at the University of Potsdam. Previously, he has held professor and postdoctoral positions at Konstanz, Mainz, Leipzig, and Kiel universities. His research covers linguistic typology, historical linguistics, areal, and contact linguistics with a special focus on perfects, partitives, antipassives, differential argument marking systems, non-canonical argument coding, indexing, and efficiency. Dmitri Sitchinava is Senior Researcher at the Russian Language Institute, Russian Academy of Sciences, and Assistant Professor at the Higher School of Economics, Moscow. He studies Slavic languages, Old East Slavic (including birchbark letters), corpus linguistics, and typology of verbal tense and aspect. Karolı´na Skwarska is Head of the Department of Slavonic Linguistics and Lexicography, Institute of Slavonic Studies, Czech Academy of Sciences, Prague, Czech Republic. Her research interests are in the morphology, syntax, and semantics of Slavonic languages (especially Russian, Polish, and Slovenian), the semantics of case (genitives and accusatives, animacy), bilingual lexicography, and the valency of verbs, adjectives, and substantives. Her current projects are a multilingual valency dictionary and a comparative study of emotion and motion verbs in Slavonic languages. Mladen Uhlik is a researcher at the Fran Ramovš Institute of the Slovenian Language and assistant professor at the Department of Slavistics, Faculty of Arts, University of Ljubljana. His research deals with contrastive Slavic linguistics, specifically, comparison of Slovene with other South Slavic languages and Russian. He is a member of the editorial board of Encyclopedia of Slavic Languages and Linguistics published by Brill (2021–). At the Department of Slavistics, Faculty of Arts, University of Ljubljana, he teaches courses related to Russian morphology and syntax. Bjo¨rn Wiemer received his Ph.D. in Slavic and general linguistics in 1996 (Hamburg University). He worked as research assistant at the chair of Slavic Languages at Constance University (1996–2003). Subsequent to his postdoctoral thesis, which was devoted to grammaticalization (2002, habilitation/venia legendi in Slavic and Baltic linguistics) he continued doing research and teaching at Constance University until 2007, when he was appointed to the chair of Slavic Linguistics at Mainz University. His main topics of interest are aspect and other verbal categories, voice related phenomena, evidentiality and modality, clausal complementation, also from a diachronic perspective and in non-standard varieties, language contact, and areal linguistics. He has contributed to all of these domains with publications both on synchronic and diachronic issues.

1 Alignment and alignment change in the Indo-European family and beyond Eystein Dahl

1.1 Introduction This book explores alignment typology and change in some Indo-European languages. Alignment typology has been the subject of growing interest in recent years, both in synchronic and diachronic perspective (Harris and Campbell 1995, Donohue and Wichmann 2008, Yanagida and Whitman 2009, Whitman and Yanagida 2012, Coon 2013, Hasselbach 2013, Coghill 2016, Dahl and Stron´ski 2016, Gildea and Zuñiga 2016, Polinsky 2016, Aldridge 2017, Butt and Deo 2017, Haig 2017, Cristofaro and Zuñiga 2018, Zuñiga 2018). However, few existing studies deal with variation and change in the alignment typology across languages belonging to one linguistic family and the present volume aims to fill this gap partially, exploring alignment change in several Indo-European languages, including Vedic Sanskrit, Latin, Ancient Greek, Hittite, Armenian, Lithuanian, Old Italian, Spanish, and various Slavonic languages. Alignment in a strict sense comprises basic alignment patterns expressing core arguments of intransitive, transitive, and ditransitive predicates, but in the present context the term has a slightly extended sense, including both basic alignment patterns and various types of valency-decreasing constructions, most notably passives, anticausatives/middles, antipassives, impersonals/transimpersonals, and A- or Plability. As will be discussed in somewhat more detail below, there is an intimate diachronic relationship between some of these construction types and certain types of basic alignment patterns, a fact suggesting that valency-decreasing constructions and basic alignment patterns represent two dimensions of the same area of grammar (see Zúñiga and Kittila¨’s (2019) discussion of so-called symmetrical voice systems found in some Austronesian languages). Moreover, many of the branches/languages belonging to the Indo-European linguistic family show both head- and dependent-marking strategies of argument realization (see e.g. Nichols 1986, 1992) early on and generalize one of these strategies in the course of their recorded history, a fact with intriguing implications for diachronic alignment typology. The contributions in this volume, which we discuss more carefully in Section 1.4, present complementary perspectives on overlapping and analogous Eystein Dahl, Alignment and alignment change in the Indo-European family and beyond. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl, Oxford University Press. © Eystein Dahl (2022). DOI: 10.1093/oso/9780198857907.003.0001

2

EYSTEIN DAHL

patterns of variation and change within the broader field of alignment typology as presently defined. At the beginning of their attested tradition, most branches of the Indo-European linguistic family show a predominantly nominative–accusative organization of their basic alignment pattern. However, in the course of time many of the branches and individual languages under consideration underwent significant changes in their alignment system. There is some evidence that Proto-Indo-European (PIE) at an early stage had a basic alignment pattern with a different alignment type, which according to some scholars was ergative (see e.g. Willi 2017: ch. 9), while others believe it was a so-called stative–active or semantic alignment system (Pooth and Orqueda 2021). These and related issues are addressed in Chapter 2 of this volume. Taken together, these observations suggest that alignment is a diachronically unstable typological dimension in the Indo-European language family. This is somewhat surprising since large-scale typological studies like Nichols (1992) suggest that languages generally tend to show diachronic stability in their alignment patterns both from a genetic and from a geographic perspective. Harris (1990), on the other hand, examines various patterns of alignment change that readily lend themselves to analyses in terms of extension or analogy. Several of the examples provided by Harris (1990) involve Kartvelian languages, this linguistic family thus constituting a case analogous to Indo-European. These two linguistic families seemingly contrast with others like, e.g. Dravidian or Semitic, where alignment change appears to be less frequent.¹ Since basic alignment in general must be considered a core part of the grammar of a language, alignment change represents a surprising diachronic phenomenon. It is reasonable to assume that alignment change, when it occurs, often is an epiphenomenon of changes affecting other parts of the grammar of the language in question, typically involving one or more of its valency-affecting constructions, often involving the emergence of new aspect categories. In such cases, construction-specific developments through a series of reanalyses result in a major typological change in the basic alignment system, as is the case in Indo-Aryan (Dahl 2016, Dahl 2021a, Butt and Deo 2017). This is in line with the more general observation that morphosyntactic change rarely has a catastrophic, macroscale character, but rather tends to involve discrete small-scale changes (Roberts 2010, Traugott and Trousdale 2010). Many cases of alignment change appear to involve changes in detransitivizing morphosyntax, and it is therefore reasonable to assume that this area of grammar is particularly relevant for understanding the causes of alignment change. It is tempting to suggest that this may be somehow

¹ See e.g. Lehmann 1998 on Tamil, Krishnamurti 2003 on Dravidian more generally; Hasselbach 2013 shows that Semitic languages generally show nominative–accusative alignment, most likely inherited from Proto-Semitic; note, however, that Coghill 2016 shows that alignment change occurs in late stages of Aramaic.

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

3

connected with the fact that the main difference between the major alignment types ergative, accusative, semantic, and tripartite concerns how they classify subjects of intransitive clauses. These considerations suggest that intransitivity, detransitivization, and alignment change are intimately interconnected linguistic phenomena, together constituting a potentially very fruitful field of research.

1.2 Alignment typology There is considerable cross-linguistic variation as to how verbal arguments are morphosyntactically expressed and alignment represents one central parameter of typological variation. Basic alignment patterns express the arguments of intransitive or one-place, transitive or two-place, and ditransitive or three-place predicates. Other terms also used in subsequent chapters are monovalent, bivalent, and trivalent verbs. Here, we shall restrict our attention to one-place, or monovalent and two-place or, bivalent verbs. In alignment typology, the first argument of two-place predicates is represented by the shorthand A, the second argument of two-place predicates by P (or O), and the single argument of one-place predicates by S. Different constellations may hold between the three argument types, representing one of the following alignment types: • nominative–accusative or simply accusative alignment, where A and S have the same expression, while P is expressed in a different manner; • ergative–absolutive or simply ergative alignment, where P and S have the same expression, while A is expressed in a different manner; • split intransitive or semantic alignment, where S in some cases has the same expression as A, for which we use the shorthand SA , and in other cases receives the same marking as P, for which we use the shorthand SP ; • neutral alignment, where A, S, and P have the same expression; • tripartite alignment, where A, S, and P each have a unique expression; • double-oblique alignment, where A and P have the same expression, while S has a different one. A schematic representation of the various alignment types is given in Table 1.1. Table 1.1 Different alignment types Accusative

Ergative

Neutral

Tripartite

Double Oblique

Semantic

A=S≠P

A≠S=P

A=S=P

A≠S≠P

A=P≠S

A≠P, A = SA , P = SP

4

EYSTEIN DAHL

Among the alignment types outlined in Table 1.1, accusative and ergative have a more privileged status in linguistic theory than the others, partly because these two alignment patterns are more common across the world’s languages and partly because they are more widely studied than the other alignment types. In recent work, Creissels (2018) outlines a prototype-based model for ergative and accusative alignment, which he relabels P-unmarked and A-unmarked alignment. He notes (2018: 72) that ergative/P-unmarked systems have the following characteristics: • A tends to be expressed with a special case form or adposition and P with an unmarked case form; • if there is verb agreement, the verb agrees with P; • the coding of P is the default, obligatory type of argument coding, which also characteristically codes S. Accusative/A-unmarked systems, on the other hand, have the opposite characteristics (Creissels 2018: 73): • P is expressed with a special case form or adposition and A with an unmarked case form; • if there is verb agreement, the verb agrees with A; • the coding of A is the default, obligatory type of argument coding, which also characteristically codes S. At present, we primarily focus on morphological coding since various patterns of case marking/flagging and verb agreement/indexation constitute the main dimensions of typological variation in the languages under discussion. Although it is often the case that languages are characterized globally as accusative, ergative, etc., the alignment types defined above characteristically apply to the level of individual constructions rather than the level of overall language typology. Indeed, most of the languages belonging to the Indo-European family and, in particular, the various languages discussed in this volume all show a clear propensity towards nominative–accusative alignment, which, however, interacts in different ways with other alignment types. There are two main ways in which alignment types can interact, which we define as split alignment and mixed alignment. Along the lines of McGregor (2009, 2010), we use the term ‘split alignment’ to refer to cases where different alignment systems appear in different, well-defined lexical or grammatical contexts. A case in point concerns many of the present-day Indo-Aryan languages, which have an aspect-based split between accusative and ergative alignment, ergative alignment exclusively occurring with perfective forms of the verbal paradigm, and accusative alignment occurring with imperfective

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

5

forms of the verbal paradigm, as illustrated by the examples from Hindi in (1) and (2).² (1) a. laṛkā kitāb paṛh rahā hai boy:NOM book:F.ABS read PROG.M.SG be:PRS.3SG ‘The boy is reading the book.’ b. laṛkā sōn rahā hai boy:NOM sleep PROG.M.SG be:PRS.3SG ‘The boy is sleeping.’ c. ādmī laṛke=ko mar rahā hai man:NOM boy=ACC hit PROG.M.SG be:PRS.3SG ‘The man is hitting the boy.’

[Hindi]

[Hindi]

[Hindi]

These examples unambiguously show that imperfective verb forms have accusative alignment, A and S showing identical marking, in these cases laṛkā ‘boy’ and P showing a different case-marking pattern, the accusative form laṛke=ko.³ (2) a. laṛke=ne kitāb paṛhī boy=ERG book:NOM read:PST.PFV.M.SG [Hindi] ‘The boy has read the book.’ b. laṛkā sōyā boy:NOM sleep:PST.PFV.M.SG ‘The boy slept.’ [Hindi] c. tūphān=ne šīšā toḍ diyā storm=ERG glass:NOM break give:PST.PFV.M.SG ‘The storm broke the glass.’ [Hindi] (adapted from Mohanan 1994: 75) d. tūphān tham gayā storm:NOM stop go:PST.PFV.M.SG ‘The storm stopped.’ [Hindi] e. āj tūphān dekhā tum=ne today storm:NOM see:PST.PFV.M.SG 2SG/PL=ERG ‘You saw the storm today.’ [Hindi] Examples like those cited in (2) show that perfective verb forms in Hindi select an A argument with the clitic ergative marker -ne, and an S and P argument without this marker, thus instantiating ergative case marking.⁴ ² Here and in the following, Hindi is used as a shorthand for Hindi/Urdu for convenience. ³ Note, however, that examples (1a) and (1c) show that Hindi has differential object marking (DOM) in the imperfective, animate nouns like laṛkā ‘boy’ receiving explicit accusative case marking, and inanimate nouns like kitāb ‘book’ appearing in the unmarked form when appearing in P function. Thus, Hindi appears to have a noun–class-based split between accusative and neutral alignment in the imperfective domain, accusative alignment being restricted to animate nouns, inanimate nouns showing neutral alignment. ⁴ Even though Hindi is usually represented as a textbook example of an aspect-based split-ergative language, matters turn out to be less straightforward on closer examination. While examples like those

6

EYSTEIN DAHL

The data from Hindi illustrate how two or more alignment types may interact in different grammatical or lexical contexts, constituting split-alignment systems. On the other hand, we use the term mixed alignment where different alignment types interact in the same lexical or grammatical context. Most notably, mixed alignment obtains in cases where dependent marking and head marking instantiate different alignment types. The Indo-Aryan language Nepali provides an example of this, showing ergative alignment in the nominal morphology and accusative alignment in the verbal agreement patterns (Bickel 2011: 400). Consider, by way of illustration, the examples in (3). (3) a. ma gaẽ I:NOM go:PST.1SG ‘I went.’ [Nepali] (adapted from Bickel 2011: 401) b. mai=le timro ghar dekhẽ I=ERG your house:NOM.SG see:PST.1SG ‘I saw your house.’ [Nepali] (adapted from Bickel 2011: 401) c. hasan=le gāṛi chalauncha Hassan=ERG car:NOM drive:NONPST.3SG ‘Hassan drives cars.’ [Nepali] (adapted from Butt and Poudel 2007) d. hasan gāṛi chalauncha Hassan:NOM car:NOM drive:NONPST.3SG ‘Hassan drives cars.’ [Nepali] (adapted from Butt and Poudel 2007) e. hasan=le nacyo Hassan=ERG dance:PST.M.3SG ‘Hassan danced.’ [Nepali] (adapted from Butt and Poudel 2007) f. hasan nacyo Hassan:NOM dance:PST.M.3SG ‘Hassan danced.’ [Nepali] (adapted from Butt and Poudel 2007) cited in (2ab) are indicative of clear-cut ergative case marking, examples like (2c) and (2e) seem to be rather restricted, since the P argument strongly tends to be marked with the accusative clitic =ko, as illustrated by examples like the following: ādmī=ne laṛke=ko dekhā man-ERG boy=ACC see:PST.PFV.M.SG ‘The man looked at the boy’ [Hindi] bādalō-ne sūraj=ko gher liyā clouds=ERG sun=ACC surround take:PFV.M.SG ‘The clouds have surrounded the sun’ [Hindi]

hai be:PRS (adapted from Mohanan 1994: 75)

This constellation seems to reflect influence from the DOM in the accusative alignment pattern in the imperfective domain of the verb. Indeed, the tendency of using the marked =ko forms to express the P argument appears to be almost exceptionless with animate referents, a fact indicating that this particular noun type shows tripartite rather than ergative case marking in the perfective domain. Moreover, the fact, noted by Mohanan (1994: 75), that the ergative marker does not appear with inanimate nouns creates further complications. However, a more extensive exploration of these intriguing issues will have to be undertaken elsewhere.

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

7

Examples (3a) and (3b) clearly show that case marking follows an ergative pattern, while verb agreement follows a nominative pattern. Nepali also shows alternation between ergative and nominative case with unergative predicates both in the non-past and the past tenses without any change in verb agreement, as illustrated by the examples in (3c)–(3f ).⁵ It is tempting to hypothesize that these two phenomena are related. At any rate, examples like those cited in (3) show that case marking/flagging and verb agreement/indexation are independent dimensions of morphosyntax, that may, but need not converge on the same alignment type. Along the lines of scholars like Donohue (2008), ergative and accusative alignment may be defined as syntactic patterns, since they assign case marking and/or agreement to the argument types irrespective of semantic factors. In semantic alignment systems, on the other hand, the realization of the various argument types depends to a smaller or greater extent on the lexical semantic properties of the predicate. Semantic alignment often represents a special case of either accusative or ergative alignment, where accusative or ergative case marking is extended to the S argument of certain verbs. Two slightly different types of semantic alignment are often distinguished: fluid-S and split-S. Restricting the present focus to flagging/case marking strategies, fluid-S alignment obtains when the argument of certain one-place predicates may alternately select the same case marking as A or P, as illustrated in (4). (4) a. ram=ne royā Ram=ERG cried:PST.PFV.M.SG ‘Ram cried (intentionally).’ [Hindi] b. ram royā Ram:NOM cried:PST.PFV.M.SG ‘Ram cried.’ [Hindi]

(adapted from Butt 2017: 817)

(adapted from Butt 2017: 817)

c. ʔa: k’lú:k’lu:w 1SG.NOM cough ‘I coughed (intentionally).’ [Central Pomo] (adapted from Arkadiev 2008: 111) d. ṱo: k’lú:k’lu:w 1SG.ACC cough ‘I coughed (accidentally).’ [Central Pomo] (adapted from Arkadiev 2008: 111) Butt (2017: 816) notes that unergative one-place verbs generally select an ergativemarked subject in Hindi, as illustrated in (4a). However, they may alternately take a nominative-marked subject, as illustrated in (4b). Along the lines of Ahmed’s ⁵ It should be noted that in Nepali, this alternation only marginally occurs with one-place unergative predicates in the past tense, where the ergative case is predominant. Two-place agentive predicates exclusively select the ergative case. According to Butt and Poudel 2007, the alternation between the ergative and nominative with non-past forms of agentive two-place predicates often conveys a distinction between stage-level and individual-level predication.

8

EYSTEIN DAHL

(2010) analysis, semantic factors such as animacy, agency, dynamicity, or telicity determine whether a given one-place predicate behaves as unergatively or unaccusatively. In other words, the alternation in (4a) and (4b) is semantically, not syntactically determined, the ergative implying that the action is initiated by the subject and the nominative expressing that the action may have been due to an external cause (Butt 2017: 816). As regards the examples in (4c) and (4d), on the other hand, Arkadiev (2008: 111) notes that the alternation between the nominative and accusative in Central Pomo has a similar effect, the nominative implying that the subject coughed on purpose, whereas the accusative implies that the coughing happened spontaneously. In both these cases, semantic notions play a central role in argument realization, thus suggesting the conclusion that these constructions show fluid-S alignment. In a split-S system, the case marking of the argument is lexically determined, and does not alternate, as illustrated in (5). (5) a. Jon erori da Jon:ABS fall:PFV TNS:root ‘Jon has fallen.’ [Basque] (adapted from Berro and Etxepare 2017: 792) b. Jonek dantzatu du Jon:ERG dance:PFV TNS:root ‘John has danced.’ [Basque] (adapted from Berro and Etxepare 2017: 792) c. vaikas dainuoja child:NOM.SG sing:PRS.3 ‘The child is singing.’

[Lithuanian]

d. skauda kelį hurt:PRS.3 knee:ACC.SG ‘(My) knee hurts.’

[Lithuanian]

According to Berro and Etxepare (2017), Basque represents an ergative language where the expression of S depends on the lexical semantic features of the predicate. Specifically, change-of-state verbs, verbs denoting a spontaneous event, and a subset of aspectual predicates select absolutive case marking, as illustrated in (5a), whereas one-place verbs denoting volitional acts, emission, and involuntary bodily processes typically select ergative case marking, as illustrated in (5b). Lithuanian, on the other hand, has predominantly accusative alignment, as illustrated by the nominative-marked S argument in example (5c), but several experiential predicates select an S argument with accusative case marking, as illustrated in (5d). The examples given in (4) and (5) illustrate that fluid-S and split-S alignment patterns involve non-canonically case-marked S arguments, in that the predicates in (4b), (4d), (5b), and (5d) do not select the expected absolutive or nominative

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

9

case marking, respectively.⁶ While Fluid-S and split-S alignment represent salient types of semantic alignment, it seemingly makes sense to include other types of non-canonical argument realization patterns under the broader notion of semantic alignment. Consider the examples from Latin in (6). (6) a. Em, nunc tu mihi places PTCL now 2SG.NOM.SG 1SG.DAT please:PRS.2SG ‘Well, now you are pleasing me.’ [Latin] (Plaut. Merc. 911) b. quoius me nunc facti pudet which:GEN.SG 1SG.ACC now deed:GEN.SG be.ashamed:PRS.3SG ‘Of which deed I am now ashamed.’ [Latin] (Plaut. Bacch. 1016) Example (6a) illustrates that the verb placeo ‘please, be pleasing’ consistently selects a nominative–dative case frame instead of the expected nominative– accusative case frame, a fact reminiscent of the non-canonical marking of certain S arguments in split-S patterns. Example (6b), on the other hand, illustrates that the verb pudeo ‘be ashamed’ tends to select an accusative–genitive case frame instead of the expected, canonical argument realization pattern. Also, the verb appears in the default third singular form pudet a characteristic feature of impersonal forms in Latin rather than the first singular form pudeo. As noted above, causative–transitive verbs imply an asymmetric relation between the first and second participant, and other types of two-place predicates often have similar implications. It should be observed, however, that the experiential verbs placeo ‘please’ and pudeo ‘be ashamed’ illustrate two distinct ways of representing the relationship between the experiencer and the stimulus. The first verb selects the stimulus as the first participant, which receives the nominative case-marking characteristic of A arguments, while the second verb selects the experiencer as its first argument, which, however, has non-canonical accusative case marking. Bossong (1998) distinguishes between direct experiential constructions, that is, constructions where the experiencer is the first argument, and inverse constructions, that is, constructions where the stimulus is the first argument. The Latin data in (6) provide an intriguing illustration of how a predominantly nominative–accusative language may single out both types of experiential predicates vis-à-vis other types of two-place predicates. A growing body of evidence suggests that semantic alignment patterns may arise from so-called transimpersonal constructions (e.g. Malchukov 2008). Transimpersonal constructions are transitive impersonal constructions, that is, impersonal constructions with an implicit agentive argument. Restricting our attention to the Latin verb pudeo ‘be ashamed’, we may note that it is marginally attested as an intransitive personal verb and even as a causative–transitive verb with something like a dummy subject, as illustrated in (7). ⁶ Dixon (1994: 187) notes that ‘A split-S system (…), for instance, may arise through “grammaticalization” of a fluid-S system’.

10

EYSTEIN DAHL

(7) a. ita nunc pudeo atque ita nunc paveo thus now be.ashamed:PRS.1SG and thus now be.afraid:PRS.2SG ‘I am now so ashamed and I am now so afraid.’ [Latin] (Plaut. Cas. 877) b. nil-ne te pudet? nothing:NOM.SG-PTCL 2SG.ACC make.ashamed:PRS.3SG ‘Doesn’t anything make you ashamed?’ [Latin] (Plaut. As. 933) c. Nimio id quod pudet facilius by.far that:NOM.SG REL.NOM/ACC.SG make.ashamed:PRS.3SG easier fertur quam illud quod be.carried:PRS.3SG than that:NOM.SG REL.NOM/ACC.SG piget. annoy:PRS.3SG ‘That which makes (one) ashamed is far easier to endure than that which makes (one) annoyed.’ [Latin] (Plaut. Ps. 281) These examples illustrate that the verb pudeo has labile behaviour, being compatible with intransitive readings, as in (7a), and transitive readings, as in (7b) and (7c).⁷ In the present context, the transitive readings clearly show that the Latin verb pudeo had a transimpersonal use. The examples in (6) and (7) are found in Early Latin sources, Classical Latin only has the construction in (6b) with the verb pudeo. Hence, Latin appears to instantiate an unambiguous instance of the development from transimpersonal to semantic alignment.⁸ We now turn to the three last types of basic alignment: tripartite, double oblique, and neutral alignment.⁹ Both tripartite and double-oblique patterns are comparably rare across languages. Classical Armenian seemingly represents an instance of tripartite alignment, as illustrated in (8) (see Meyer 2016; Chapter 9 in this volume, for discussion). (8) a. ?ocˇʿ z-gir-n z-ayn icʿē NEG NACC-writing:ACC.SG-ART NACC-DEM.ACC.SG be:PRS.SBJ.3SG əntʿeṙcʿeal jer read:PTCP 2PL.GEN ‘Should you not have read this writing?’ (NT, Mk. 12:10 [Classical Armenian], adapted from Meyer 2016) b. or ustekʿ ustekʿ ekeal haseal ēin REL:NOM.SG/PL from.all.over come:PTCP arrive:PTCP be:PST.3PL ⁷ Example (7c) illustrates that the verb pigeo ‘be annoyed, annoy’, which also tends to select the accusative–genitive pattern illustrated in (6b), behaves in a manner analogous to pudeo, indicating that the behavioural pattern is not restricted to one lexeme. ⁸ It should be noted that an admittedly small group of experiential predicates show a behaviour analogous to pudeo. These are pigeo ‘be offended’, taedeo ‘regret’, paeniteo ‘repent’, and misereo ‘pity’ as well as their derivatives and compounds. ⁹ For ease of exposition, the following discussion only considers case-marking patterns, leaving out agreement.

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

11

i tʿikuns … in aid:ACC.PL ‘[W]ho had arrived from all over in aid …’ (Agatʿangełos 21.1 [Classical Armenian], adapted from Meyer 2016) The examples in (8) show that the Classical Armenian predicated verbal adjective in -eal may occur with a genitive-marked A (jer in (8a)), an accusative-marked P (z-gir-n z-ayn in (8a)), and a nominative-marked S (or in (8b)), instantiating tripartite alignment (Meyer 2016). As regards double-oblique alignment, Rushani, a variety of the Pamir language Shugni, provides an instance of this rather rare alignment pattern, which is restricted to the past tense, as illustrated in (9).¹⁰ (9) a. cˇi mum kitōb xˇēycˇ? who:OBL this:OBL.SG book:SG read:PRF ‘Who has read this book?’ [Rushani] (adapted from Payne 1980: 156) b. cˇāy yat? who:NOM come:PST ‘Who came?’ [Rushani] (adapted from Payne 1980: 156) c. tā virōd cˇi wunt? your brother who:OBL see:PST ‘Whom did your brother see?’ [Rushani] (adapted from Payne 1980: 156) These examples illustrate that the interrogative pronoun has the same form in A and P function (cˇi), and a different form in S function (cˇāy). Both tripartite and double-oblique alignment tend to be regarded as inherently unstable, transitory alignment types, representing stages in the development from ergative to accusative or from accusative to ergative alignment.¹¹ Moreover, both of these alignment types typically appear in languages with an ergative– accusative split, as an extension of one of the case-marking patterns from one domain to another.¹² Nevertheless, certain languages with tripartite alignment show considerable stability, as seems to be the case in the Pamir language Yazgulyami (Meyer 2016).¹³ ¹⁰ Like other languages of the Shugni or Shugni-Rushani branch of the North Pamir languages, Rushani has case-marking distinctions only in the pronouns, while bare noun forms are unmarked for case, accompanying demonstratives/articles providing case (Payne 1980, Edelman and Dodykhudoeva 2009). ¹¹ See e.g. Payne (1980) and Bubenik (2016). ¹² See e.g. Harris (1990), Harris and Campbell (1995). The Hindi data discussed above illustrate the rise of a tripartite pattern through extension, appearing to bear witness of a situation under development in the perfective domain where A is marked by the ergative clitic =ne, P is marked by the accusative clitic =ko, and S is unmarked, the use of =ko probably reflecting influence from the nominative– accusative pattern in the imperfective. As regards the rise of double-oblique systems, Payne (1980, see also Harris 1990, Harris and Campbell 1995) cites Rushani as a language where the extension of the use of the oblique case to express the P argument has spread from the accusatively aligned imperfective system to the (originally) ergatively aligned perfective/past tense system, yielding a double-oblique alignment pattern. ¹³ I am unaware of any examples in the literature of languages with a diachronically stable doubleoblique system.

12

EYSTEIN DAHL

Before concluding the discussion of the various alignment types, a brief note should be added regarding neutral alignment. This alignment pattern is found in modern Romance languages like Italian, or Germanic languages like Norwegian. Consider the examples in (10) and (11). (10) a. se un ospite ama la vita notturna if a guest:SG love:PRS.3SG the life nightly ‘If a guest enjoys nightlife.’

[Italian]

b. é arrivato un ospite be:PRS.3SG arrive:PRF.PTCP a guest:SG ‘A guest has arrived.’

[Italian]

c. ho visto un ospite particolare have:PRS.1SG see:PRF.PTCP a guest:SG special:SG ‘I have seen a special guest.’

[Italian]

(11) a. Mann-en mater hund-en man:SG-DEF feed:PRS dog:SG-DEF ‘The man feeds the dog.’

[Norwegian]

b. Mann-en løper man:SG-DEF run:PRS ‘The man runs.’

[Norwegian]

c. Hund-en biter mann-en dog:SG-DEF bite:PRS man:SG-DEF ‘The dog bites the man.’

[Norwegian]

These examples illustrate that nouns have an identical form in all pertinent argument functions in Italian and Norwegian. These two languages instantiate two ways in which neutral alignment may arise. In general, the immutable Italian noun forms derive from Latin accusative forms (e.g. ospite < Lat. hospitem acc. sg. of hospes ‘host, guest’). The immutable noun forms in Norwegian, on the other hand, arose in Middle Norwegian through a phonologically determined merger of the Old Norse nominative and accusative forms (see e.g. Mørck 2005). At this point, a brief clarification of the status of the analytical notions A, S, and P/O, distinguished above is in order. Haspelmath (2011) notes that these typological terms are often taken for granted as unspecified primitive notions which, however, are used with (at least) three different meanings in the extant literature, which he labels Dixonian, Comrian, and Bickelian. Under Haspelmath’s (2011) analysis, these definitions differ along three dimensions: their nature (syntactic functions or generalized roles); scope (subclass of verbs or all verbs); and purpose (comparative and descriptive or descriptive only). Table 1.2 provides an overview of the implications of the different approaches. In his discussion of these three approaches, Haspelmath (2011) concludes that the Comrian approach, where A and P/O are defined as syntactic functions

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

13

Table 1.2 Three approaches to S, A, P

Nature: Scope: Purpose:

Dixonian

Comrian

Bickelian

syntactic functions (only subclass of verbs) comparative + descriptive

syntactic functions only subclass of verbs comparative

generalized roles all verbs comparative + descriptive

Source: adapted from Haspelmath (2011: 539).

associated with core transitive verbs (‘kill’, ‘break’, etc.) and employed as comparative concepts, is preferable to the Dixonian and Bickelian approaches. In the present context, we may note that a Comrian definition of A, P, and S has been explicitly adopted in at least two of the contributions to this volume (Chapters 2 and 3). Given that these terms are defined as comparative concepts, terms such as bivalent subject/first argument of bivalent predicates, bivalent object/second argument of bivalent predicates and monovalent subject/sole argument of monovalent predicates will be used in language-specific descriptions in this chapter as well as the two others already mentioned. On this background, basic alignment pattern(s) can be defined as the alignment pattern(s) shown by core transitive verbs in the unmarked voice in a given language. Another important parameter of variation is what Haspelmath (2015) labels ‘transitivity prominence’, which is understood as a scalar notion intended to capture to what extent different types of binary predicates adopt the morphosyntactic properties of the basic alignment pattern(s). This essentially quantitatively defined notion plays an important role in the contribution of Serzˇant et al. to this volume (Chapter 10). After having defined some basic theoretical and terminological notions in this section, we proceed with a brief outline of some diachronic dimensions of alignment change in the following section.

1.3 Alignment change in Indo-European and beyond Along the lines of the recent state-of-the-art survey of Zúñiga (2018), one can distinguish three mechanisms of alignment change, namely analogical extensions of alignment patterns, reanalysis of low-transitivity constructions, and borrowing/contact (see also Harris 1985, 1990, Plank 1985, 1995, Harris and Campbell 1995). While clear-cut cases of borrowing seem to be comparably rare, examples of alignment change via reanalysis and extension are more common. One can distinguish two main types of reanalysis that impact alignment systems. First, there is broad agreement that languages with ergative basic alignment can develop (split-)accusative patterns by a process where an antipassive or other type of voice construction is reanalysed as an unmarked transitive construction. A

14

EYSTEIN DAHL

recent study by Katarzyna Janic and Charlotte Hemmings (2021) shows that this has happened in Inuktitut, a language of the Eskimo–Aleut family, and in certain Western Austronesian languages, where the so-called Actor voice is generalized as an unmarked voice marker. On their analysis, these developments are motivated by a process of functional markedness reversal (Janic and Hemmings 2021). Interestingly, there seem to be no uncontroversial instances of the opposite development, ergative alignment arising from passives in languages with predominantly accusative alignment. In other cases, alignment change arises as a by-product of the grammaticalization of newly emerging tense/aspect morphosyntax, as discussed in some detail in Creissels (2018). This is, for instance, the case in Indo-Aryan, where a predicated P-oriented resultative verbal adjective develops into an anterior past tense with ergative alignment (see Dahl 2021a for discussion). The grammaticalization of a new tense/aspect category in Indo-Aryan was motivated by the fact that the aspectual distinctions associated with the inherited synthetic aspect stems were gradually lost and replaced by temporal remoteness and evidentiality distinctions (Dahl 2015). The inherited alignment system of Old Indo-Aryan had a predominantly nominative–accusative character, that developed a split-ergative character through the rise of an analytic construction with ergative alignment (see Dahl 2016, 2021a for discussion). The examples in (12) illustrate that the predicated verbal adjective in -tá- had a resultative function in Early Vedic, the oldest attested stage of Indo-Aryan (12a), an anterior function in Pāli, a later stage of Indo-Aryan (12b), and is a general past tense in Epic Sanskrit (12c). (12)

a. ásti sómo ayám ˘˙ sutáḥ be:3SG.PRS soma:NOM DEM.NOM extract:PPP.NOM.SG.M pı´banty asya marútaḥ / drink:PRS.3PL 3SG.GEN Maruts:NOM ‘This soma is pressed. The Maruts drink of it.’ (RV VIII 94.4 after Jamison 1990: 5 [Early Vedic]) b. Udenena upāsakena […] vihāro karapito hoti. Udena:INS layman:INS monastery.NOM.SG let.build:PPP copula:PRS.3SG ‘The layman Udena has had a monastery built […]’ (Mahavagga 111:5 after Peterson 1998:93 [Pāli]) c. ´sruto mayā ´sabdo hear:PPP.M.NOM.SG 1SG.INS sound:M.NOM.SG ‘I heard a sound.’ (Rāmāyana 2.58.13, adapted from Bynon 2005: 11 [Epic Sanskrit])

The development from resultative via anterior to (perfective) past tense is a well-known grammaticalization path and therefore unsurprising. An important point worth noting here is that example (12a) reflects the use of the predicated verbal adjective in a tense/aspect system with two other semantically specific

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

15

perfective/anterior categories, the aorist indicative and perfect indicative,¹⁴ while both (12b) and (12c) reflect stages of the language where no other semantically specific aspect categories exist, the predicated past participle being in complementary distribution with a simple preterite category.¹⁵ Thus, the rise of split ergativity in Indo-Aryan seems to be a direct consequence of the decline of inherited aspect categories in this branch. Here, the resultant ergative pattern is consistently ergative regarding both case marking/flagging and verb agreement/indexation. Extension can impact alignment systems on different levels. It was noted above in connection with so-called mixed-alignment patterns in Nepali that case marking/flagging and verb agreement/indexation are independent morphosyntactic dimensions, that often, though not always converge on the same alignment type. Broadly speaking, one may distinguish two distinct but analogous constraints on alignment change via pattern extension (Dahl 2021a). One has global scope, and roughly corresponds to Creissels’s (2018) Obligatory Coding Principle, ‘according to which all verbal predicative constructions in a language must include a nominal term showing a particular type of coding that can be viewed as the default type of argument coding in question’ (Creissels 2018: 59–60). Following Dahl (2021a), this principle may be interpreted as a tendency to generalize a particular alignment pattern across constructions. For example, as discussed in Meyer’s contribution to this volume, Armenian instantiates a clearcut case of a language where a split-alignment pattern is replaced by a consistently nominative–accusative pattern (Chapter 9). Moreover, another analogous principle seems to operate on a more fine-grained level, levelling out mismatches of the type represented by mixed alignment, where case marking/flagging and verb agreement/indexation instantiate different alignment types. This may be illustrated by the development of verb agreement with neuter plural subjects in Ancient Greek. As discussed at length in Chapter 2 of this volume, Ancient Greek had inherited an alignment system where neuter plural subjects generally did not trigger verb agreement. However, in Homeric Greek there is an incipient tendency to generalize plural agreement marking with neuter plural nouns/pronouns as well. Given that neuter nouns show neutral case marking, this development, motivated by the principle of generalization, leads to a situation where neuter nouns show mixed alignment with neutral case-marking and accusative agreement. However, unlike in other related languages like Vedic Sanskrit or Latin, where the mixed pattern prevails, it becomes obsolete at an early stage in Ancient Greek, neuter subjects showing consistent singular agreement throughout the Classical period. This is arguably a case where a mixed neutral–accusative alignment pattern is ousted by a consistently neutral pattern (see also Dahl 2021a, 2021b).

¹⁴ See Dahl (2010) for a survey of the Early Vedic tense/aspect system. ¹⁵ See Oberlies (2001: 228–243) for discussion of the past tense system of Pāli and Oberlies (2003: 143–170) for a survey of the uses of the tense categories in Epic Sanskrit.

16

EYSTEIN DAHL

Works such as Harris (1985, 1990) and Plank (1985, 1995) show that extension plays an important role in the development of case-marking systems. A much-discussed case is the so-called extended accusative in Late Latin (Plank 1985, Cennamo 2009, 2011). Like many Indo-European languages, Latin has a nominative–accusative basic alignment pattern and appears to have a high transitivity prominence in the sense of Haspelmath (2015). In Late Latin, however, accusative forms tend to show up in monovalent and bivalent subject function, as illustrated by the examples in (13). (13) a. fit orationem become:PRS.3SG prayer:ACC.SG ‘A prayer is said/takes place.’ [Late Latin] (Per. Aeth. 24.3 after Rovai 2014: 127) b. si iumentum morbum renalem temptavit if beast.of.burden:ACC.SG illness:ACC.SG renal:ACC.SG affect:PRF.3SG ‘If kidney disease has befallen the beast of burden.’ [Late Latin] (Mul. Chir. 55 adapted from Cennamo 2009: 325) Cennamo (2011) argues that the spread of accusative to monovalent subject function of unaccusative predicates (13a) and, subsequently, to subject function of unergative predicates and then to bivalent subject function, as in (13b), originates from impersonal constructions with an expressed object argument, as illustrated by the examples in (14). (14) a. Pessumis me modis despicatur domi worst:ABL.PL 1SG.ACC manner:ABL.PL despise:PRS.MP.3SG home:LOC ‘I am despised in the worst manner at home (lit.: It despises me (…)).’ [Early Latin] (Plaut. Cas. 186, see Cennamo 2011: 178)) b. Faciatur, si tibi videtur, et triclinia make:PRS.SBJ.MP.3SG if 2SG.DAT.SG see:PRS.MP.3SG CONJ table:ACC.PL ‘It should be arranged for dining tables, if it seems appropriate to you.’ [Silver Latin/Vulgar Latin] (Petr. Sat. 71, see Cennamo 2011: 178) Under this analysis, the accusative-marked object argument in impersonal constructions of the kind illustrated in (14) were reanalysed as unaccusative subject arguments, a change causing the accusative to be generalized as a marker of the monovalent subject function and then, by extension, to bivalent subject function (see Cennamo 2009, 2011 for elaborate discussion). From a diachronic typological perspective, this would instantiate a development from an accusative via a semantic to a neutral alignment system through the gradual extension of a P marker via SP and SA to A function. The insight that marginal constructions may furnish a point of departure for changes in basic alignment implies that individual languages may maintain remnants of archaic morphosyntax that in turn may serve as a cause of further innovations, also within the realm of alignment typology.

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

17

Having briefly addressed some recurrent mechanisms of alignment change, we now turn to a brief outline of the chapters in this volume.

1.4 The contributions in this volume Chapter 2 ‘A lignment in Proto-Indo-European’ by Eystein Dahl provides a comparative–historical reconstruction of the PIE alignment system based on comparative evidence from Anatolian, Indo-Iranian, Greek, and Italic, and a brief account of the most central hypotheses concerning the prehistory of the reconstructed system. Comparative data from these branches suggest that PIE had an NP-based split between non-neuter and neuter nouns and pronouns, nonneuter pronouns showing consistently nominative–accusative alignment, while neuter nouns display a behaviour reminiscent of semantic alignment. First, neuter nouns/pronouns only appear in monovalent subject and bivalent object functions. Second, neuter nouns/pronouns do not have distinct forms for nominative and accusative. Third, nouns and pronouns of this type do not trigger verb agreement when used in monovalent subject function. It is argued that these properties show that neuter nouns/pronouns show consistently different behaviour from non-neuters, a behaviour that involves neutral case marking, neutral alignment, and semantically motivated distribution over the core argument functions. The restriction of neuters to monovalent subject and bivalent subject function precludes an analysis in terms of neutral alignment, since this would presuppose that the non-distinct case forms of neuter nouns were also used in bivalent argument function in PIE, which does not seem to have been the case. Given this, it seems preferable to derive the PIE alignment system from a semantically aligned system, along the lines of Pooth and Orqueda (2021), since this hypothesis has a stronger explanatory power than the competing ergative hypothesis, as formulated by Willi (2017). In Chapter 3, ‘Split alignment, mixed alignment, and the spread of accusative morphosyntax in some archaic Indo-European languages’, Paola Cotticelli and Eystein Dahl explore the interaction between nominative–accusative alignment and other alignment types, including various patterns of non-canonical argument marking in Latin (Italic), Ancient Greek (Greek), Vedic Sanskrit (Indo-Iranian), Hittite (Anatolian), Classical Armenian (Armenian), and Lithuanian (Baltic). Taking the reconstruction model outlined in Chapter 2 as their frame of reference, the authors explore the various developments that have led to the various discrepancies between the languages under consideration. While the languages under consideration have analogous typological properties such as a predominantly nominative–accusative alignment, they show important differences on a more detailed level of analysis. A case in point concerns the morphosyntactic properties of neuter nouns, which show intriguing discrepancies across the

18

EYSTEIN DAHL

various languages. Another dimension of Cotticelli and Dahl’s contribution concerns whether and to what extent these languages show evidence of the same notion of morphosyntactic subjecthood, finding that Latin and Lithuanian have a more clear-cut, consistent, and complex notion of subjecthood than the others. Intriguingly, they are also the languages with most uncontroversial instances of non-canonical subject marking. An important finding of Cotticelli and Dahl’s analysis is that the variation shown by the languages is suggestive of a diachronic cline where a noun–class-based split between accusative and neutral alignment, reflected both in case marking and agreement, is replaced by a noun–class-based split between accusative and mixed neutral–accusative alignment. The former pattern is found in Hittite and Ancient Greek, the latter pattern is characteristic of Vedic Sanskrit and Latin. The last stage of this development is seen in Armenian and Lithuanian, which show consistently nominative–accusative alignment in the inherited part of their finite verbal systems. These observations lead the authors to postulate a diachronic drift towards consistently accusative morphosyntax in the Indo-European languages, instantiated in a series of small-scale changes affecting the inherited head-marking and dependent-marking structures. This process, which they label ‘accusativization’, did not impact Anatolian, which is the topic of Chapter 4. In Chapter 4 ‘The origin of ergative case markers: The case of Hittite revisited’ Silvia Luraghi and Guglielmo Inglese reassess the controversial issue of the rise of split ergativity in Hittite. Hittite has two classes of nouns: common and neuter, which differ fundamentally in their syntactic behaviour. While bare common gender nouns formally distinguish between nominative and accusative case and occur in any type of core argument function, bare neuter nouns have a syncretic form for these two cases and, significantly, are restricted to S and P functions. However, Hittite has a marker -anza /-ants/ which is variously interpreted as a derivational suffix with an individualizing function or as an ergative marker. The authors explore the origin and development of this suffix, which appears to represent a clear-cut example of a change from derivation to inflection. Through a careful examination of the various functions of the suffix -anza-, the chapter makes a convincing case for the claim that this suffix, in spite of its relative wide set of uses, has a unitary set of derived meanings ultimately arising from an original possessive meaning. As regards its development as an ergative marker, however, the authors assume with Goedegebuure 2018 that it is directly related to the use of the suffix as a means of individualization, an assumption supported by the fact that neuter nouns often indicate a low degree of individualization. The authors show that diachronic data from Hittite provide strong support in favour of the hypothesis that the inflectional, ergative use of the suffix -anza- gradually developed from its use as a derivational individualizing suffix in the oldest attested stage of the language. Comparative data from the closely related languages Luwian and Lycian suggest that the suffix had an individualizing function in Proto-Anatolian as well,

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

19

a fact supporting the hypothesis that the ergative use of the suffix -anza- is not inherited but represents a secondary development specific to Hittite. The chapter also contains an appendix by Petra Goedegebuure. In Chapter 5 ‘Passives and anticausatives in Vedic Sanskrit: Synchronic and diachronic perspectives’ Hans Henrich Hock reassesses the vexed question concerning the relationship between passives with suffix accent like ṛc-yá-te ‘be recited, praised’ from arc- ‘recite’ and a formally closely related group showing alternating accent and often considered as anticausatives, e.g. múc-ya-te /muc-yáte ‘be/become released’ from moc- ‘release’. There is general agreement that neither of these formations is inherited from PIE, where there was no dedicated passive or anticausative construction, the middle voice covering both. The development of a distinct passive formation with -yá- suffix is at least of Proto-Indo-Iranian age, as indicated by the fact that Iranian has analogous constructions. Hock explores the relationship between these two Vedic constructions, firmly establishing that accent does not provide a reliable cue for distinguishing between passive and anticausative functions. Absence of agent NPs, on the other hand, does not provide a reliable criterion for distinguishing between passives and anticausatives in Vedic, since even generally accepted passive constructions only have an explicit agent NP in about 11 per cent of their occurrences, including a number of somewhat uncertain cases. Converb control constitutes a more reliable parameter, since the S of anticausatives but not the promoted P of passives may show subject coreference with converbs. In spite of the seemingly close formal relationship between the two constructions, Hock suggests that they may in fact derive from two PIE present stem formations, one with zero grade root and suffix accentuation and another with full grade root and root accentuation. The Vedic data explored in this chapter provide a clear illustration of how a passive formation may develop into an anticausative formation. Chapter 6 ‘Non-nominative arguments, active impersonals and control in Latin’ by Michela Cennamo and Claudia Fabrizio provides a systematic analysis of different types of impersonal constructions in Latin. Within the Indo-European family, Latin has a particularly rich array of impersonal/non-canonical argument realization patterns, which partly show consistent case marking with specific groups of lexemes and partly alternate between different case-marking patterns. Cennamo and Fabrizio contrast the range of uses associated with impersonal constructions in the active voice with personal active and mediopassive forms, making a strong case for the claim that these construction types are part of a complex system of transitivity marking in Latin, where the notion of control plays a central role. Their analysis is based on a definition of control as a scalar or spectral notion, that mirrors the relative direct responsibility a given discourse participant has over the situation denoted by the verbal predicate. This rather complex notion includes or interacts with a variety of transitivity features such as agentivity, volitionality, individuation of participants, and the aspectual properties of

20

EYSTEIN DAHL

the predicate, according to Cennamo and Fabrizio. They convincingly show how certain verbs showing multiple argument realization options illustrate the ways in which different constructional patterns give rise to subtle semantic distinctions, often additionally enriched by contextual factors. Based on previous work by Cennamo (1997, 2005, 2016), they show how the various constructions instantiate different points on the so-called Passive-Impersonal Continuum, ranging from agent-defocusing/suppression via stativization, subjectization of a non-agent and topicalization of a non-agent to affectedness of the surface subjects. An important observation is that many of the constructions involved in the complex system of transitivity marking in Latin resemble constructions found in languages displaying semantic alignment. On Cennamo and Fabrizio’s analysis, the alignment system of Latin may be analysed as a syntactically based nominative–accusative pattern that interacts with various semantically oriented alignment patterns. In Chapter 7 ‘Infinitives and subjecthood between Latin and Old Italian’, Claudia Fabrizio examines the syntax of infinitives in Latin and the transition to Old Italian. In Latin, the use of infinitives in subject function is restricted to unaccusative one-place predicates denoting non-agentive states and achievements, as well as passive forms of activity predicates. In itself, the fact that infinitives appear as undergoers of intransitive and transitive predicates, that is, in SP and P function, and not as actors of intransitive and transitive predicates, that is, in SA and A function, is remarkable. However, as Fabrizio points out, adverbal infinitives in SP and P function show no difference in morphology, thus manifesting a behaviour apparently diverging from the predominant nominative–accusative alignment pattern of Latin. In this respect, infinitives show a marked contrast to deverbal action nouns, which has case inflection and are compatible with all argument functions. Given that adverbal infinitives in Latin only express the undergoer role, they represent a restricted type of semantically oriented alignment in the language. In Old Italian, on the other hand, infinitives are found in all argument functions, a fact suggestive of neutral alignment on Fabrizio’s approach. The chapter convincingly claims that experiencer verbs represented a plausible bridging context, since infinitives frequently appear as the stimulus argument of such verbs already in Latin. This assumption is strongly supported by the fact that a majority of the infinitives in subject function of two-place predicates in Old Italian have experiencer objects. Interestingly, Old Italian also developed a pattern where different PPs involving an infinitive are used to single out the various core argument functions, a remarkable example of the continuous realignment process in Romance languages. In Chapter 8 ‘A lignment changes with Spanish experiental verbs’, Chantal Melis explores the argument realization patterns of experiencer predicates in Spanish from a diachronic perspective. Experiential verbs are known to show considerable variation both within and across languages, a fact Melis takes to reflect the nature

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

21

of experiencer arguments, which appear somewhat ambiguous or vague when contrasted with more prototypically agentive or patientive arguments. The present study focuses upon a subclass of such predicates, emotional causatives, which characteristically have stimulus subjects. Given that predicates of this type show a somewhat unexpected or inverse constellation of their arguments, they often show diachronically unstable behaviour regarding argument realization, being particularly prone to adopting valency-affecting strategies that may contribute to singling out the experiencer argument. Through the history of Spanish, three argument realization patterns are employed with varying degrees of frequency, a middle construction inherited from Late Latin, a dative alignment pattern competing with the middle construction from the fifteenth century onwards, and a more recent splitintransitive pattern arising via a transimpersonal structure. Quantitative data from different stages of Spanish clearly show how the dative gradually invades the realm of the middle construction, being the preferred choice of many individual predicates in present-day Spanish. Melis shows that there is a clear propensity towards using the dative with first-person experiencers as opposed to third-person experiencers with most of the verbs under scrutiny, even if not all predicates behave alike in this respect. Interestingly, first-person experiencers also constitute the focal point where the admittedly rare transimpersonal construction of Modern Spanish appears, and characteristically involves cases where some unbounded discourse element furnishes the stimulus argument. Thus, Spanish experiential verbs constitutes a fruitful field for exploring how competing strategies contribute to forming a highly dynamic area of alignment typology that may ultimately be the starting point of more radical development patterns within this part of grammar. Chapter 9 ‘Armenian morphosyntactic alignment in diachrony’ by Robin Meyer examines the development of a periphrastic perfect in Classical Armenian, which represents an ergative/tripartite construction in an otherwise consistently nominative–accusative system. Thus, the Classical Armenian alignment system represents a clear-cut example of aspect-based split alignment. However, the tripartite alignment associated with the periphrastic perfect was gradually replaced by a nominative–accusative pattern, predominant in the synthetic tenses of the verb. A central problem in Armenian linguistics that has not been previously resolved concerns the origin of the periphrastic tenses. Having discussed the merits and shortcomings of previous proposals, Meyer makes a case for the claim that the Classical Armenian periphrastic perfect arose through the contact relationship between Armenian and West Iranian languages, most notably Parthian. During the time when the Armenian kingdom was subject to Iranian rule extended language contact took place, as shown by the considerable number of Iranian loanwords in Armenian. According to Meyer, this contact situation may be safely assumed to have been sufficiently intense as to have influenced Armenian syntax as well. On this background, he concludes that it is necessary to consider the possibility that the Armenian periphrastic perfect originated as a syntactic loan from Iranian.

22

EYSTEIN DAHL

This assumption is among other things supported by the fact that West Middle Iranian languages also show tense/aspect-based split alignment, with accusative alignment in the present system and ergative alignment in the periphrastic tenses. Moreover, through close comparison of Armenian and Parthian periphrastic constructions, Meyer convincingly demonstrates that the ergative pattern of the Armenian perfect arose through pattern replication, imitating the corresponding construction(s) in Iranian. The Armenian construction subsequently developed a tripartite alignment pattern, which in turn was replaced by an accusative pattern that became predominant in the later stages of Classical Armenian, yielding a consistently nominative–accusative alignment system across the board, which is still found in Modern Armenian. The last contribution of this volume is ‘Areal and diachronic trends in argument flagging across Slavic’ by Ilja A. Serzˇant, Bjo¨rn Wiemer, Eleni Buzˇarovska, Martina Ivanová, Maxim Makartsev, Stefan Savic´, Dmitri Sitchinava, Karolı´na Skwarska, and Mladen Uhlik. They examine the areal and diachronic patterns of case marking of core arguments across ten modern Slavic languages (Bulgarian, Macedonian, Serbian, Slovenian of the South Slavic sub-branch, Czech, Slovak, Polish of the West Slavic sub-branch, and Belarusian, Russian and Ukrainian from the East Slavic sub-branch), comparing them with Old Church Slavonic, the oldest attested Slavic language, which they employ together with data from other ancient Indo-European languages as a proxy for Proto-Slavic. The study mainly having an explorative character, its main aim is to arrive at a better understanding of the mechanisms causing divergent case-marking patterns in genetically closely related languages. In this respect, the Slavic languages provide a close to ideal case for study, given that their development into separate sub-branches is comparatively recent. The authors delimit three factors that have contributed to the relative homogeneity in case marking across modern Slavic languages, namely shared morphosyntactic inheritance, a historically relatively unitary areal dispersion, and common innovations resulting from typological pressure. The two first factors favour the comparably high degree of homogeneity in morphosyntax and lexicon across the modern Slavic languages, including their core argument-marking patterns. As regards typologically motivated factors, the authors point to the fact that the inherited Proto-Slavic alignment pattern with phonologically overt nominative and accusative case markers in most noun classes has been replaced by a system with morphologically unmarked subjects and objects with differential marking based on animacy in modern Slavonic languages. Nevertheless, there are some intriguing differences between modern Slavic languages, which the authors take to reflect their different geographic locations. The exploration of the diachronic and areal patterns in the development of the case marking across the Slavic languages is based on a database created by the authors, comprising translations of 46 verb meanings into each of the languages, disambiguated through different discourse contexts. Drawing on Haspelmath’s (2015) scalar notion of

ALIGNMENT IN INDO-EUROPE AN AND BEYOND

23

transitivity prominence, their sample mainly includes verbs located somewhere towards the middle of the scale, where variation can be expected. Among other findings, the chapter identifies some important tendencies across modern Slavic languages. For one thing, there is a marked reduction across the board as regards the number of available case frame for individual verbs. In some languages, e.g. Polish, Macedonian, and Bulgarian, this has led to a stronger degree of rigid and uniform government in case marking. Other languages, such as Slovak and Russian, display a less rigid character in this respect. In the case of Slovak, this appears to reflect an ongoing alignment change, in which the basic transitive pattern is generalized. As regards transitivity prominence, there is a clear-cut trend across the Slavic-speaking areas dividing them into a North-East Slavic area, including Russian, Belarusian, Ukrainian, and Polish, and a South-West Slavic area, including the remaining languages, a split that can be explained as a reflection of a macroareal tendency towards more morphological case marking in the eastern part of Eurasia and less or no cases in its western and South-Western parts.

Latin sources Plaut. As.: T. Maccius Plautus, Asinaria, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Bacch.: T. Maccius Plautus, Bacchides, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Cas.: T. Maccius Plautus, Casina, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Merc.: T. Maccius Plautus, Mercator, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Ps.: T. Maccius Plautus, Pseudolus, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905.

2 Alignment in Proto-Indo-European Eystein Dahl

The purpose of this chapter is threefold. First, it provides a critical assessment of the comparative evidence that forms the basis for the reconstruction of the Proto-Indo-European (PIE) alignment system, arguing that the resulting system shows a fundamental split between nominative–accusative and semantic alignment. Second, it attempts to evaluate this reconstructed system against competing hypotheses in recent contributions to the field (Carling and Cathcart 2021, Keydana 2018, Pooth and Orqueda 2021, Willi 2018). Third, it has the overarching aim to present readers with modest or little background knowledge of IndoEuropean studies in general, and alignment patterns in Indo-European languages more specifically, with an introduction to this intriguing but controversial field of research. The chapter is organized as follows. Section 2.1 is its main empirical part, giving a comparative survey of the primary data from Hittite, Vedic Sanskrit, Homeric Greek, and Early Latin that serve as basis for a comparative–historical reconstruction of the (late) PIE alignment system, on which the discussion in Chapter 3 is also based. Section 2.2 explores three hypotheses concerning the prehistory of the PIE alignment system, evaluating them in light of the analysis of the data given in Section 2.1. Section 2.3 contains a summary and conclusion. Theoretically and terminologically, the present chapter is in line with Chapter 3 in this volume. It should be noted from the outset that there has long been some controversy regarding the PIE alignment system. Broadly speaking, one can distinguish three different hypotheses. According to what may be regarded as a mainstream view, PIE had consistently nominative–accusative alignment. This assumption is rooted in the neogrammarian school of the late nineteenth and early twentieth centuries, representing the outcome of the pioneering comparative–historical syntactic work of Berthold Delbru¨ck and Karl Brugmann (Brugmann and Delbru¨ck 1893, 1897, 1900), which is still adhered to in state-of-the-art handbooks within the field (e.g. Keydana 2018). According to another hypothesis, PIE had what is now commonly labelled ergative–absolutive alignment, an assumption originating from the work of the Dutch scholar Uhlenbeck (1901, 1907) and reflected in refined form in Willi (2018). A third line of thought assumes that PIE was a language with semantic alignment, a hypothesis first developed by Soviet scholars from the 1940s onwards (Willi 2018: 521ff. for discussion and references) and developed further in more Eystein Dahl, Alignment in Proto-Indo-European. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl. Oxford University Press. © Eystein Dahl (2022). DOI: 10.1093/oso/9780198857907.003.0002

ALIGNMENT IN PROTO-INDO-EUROPE AN

25

recent work (see, e.g. Pooth and Orqueda 2021 with references). At this point, a note of clarification is in order. The traditional reconstruction of PIE effectively reflects a relatively advanced stage of development of the proto-language shortly before its ramification into the various branches to which the historically attested languages belong. Moreover, it is the result of applying the method of comparative reconstruction to the available data from the oldest stages of the daughter languages. The reconstruction of PIE as ergatively or semantically aligned, on the other hand, reflects an earlier or more remote stage of the proto-language, arrived at via the method of internal reconstruction, which primarily takes inconsistent or aberrant features in the reconstructed system and/or in individual languages or branches as input. This stage of the proto-language shall be referred to as pre-PIE. Moreover, there is broad agreement that the Anatolian branch was the first to split off from common PIE, non-Anatolian PIE in the present context being referred to as Core Indo-European (Core IE), in line with Melchert (2014). The evidence on which the comparative reconstruction in Section 2.1 is based has a three-fold character, involving case marking/flagging, verb agreement/indexation, and distributional restrictions upon the combination of certain types of nouns and certain classes of verbs. The present work is based on the working hypothesis that these three classes of phenomena constitute independent dimensions of morphosyntax that may but need not converge. To keep the presentation and discussion of primary evidence as informative and coherent as possible, I have chosen to draw primarily on data from the Anatolian language Hittite, the Indo-Iranian language Vedic Sanskrit, the variety of Greek known as Homeric or Epic Greek, and Early Latin from the Italic branch. These branches have been selected because they comprise the oldest attested languages within the Indo-European family that at the same time have a reasonably rich corpus of sources. As will be clear from the following outline, the languages under consideration show several important similarities as well as some striking differences within the realm of alignment typology. Thus, they provide a fruitful and nuanced point of departure for a comparative reconstruction of the PIE alignment system.

2.1 Comparative–historical outline of the PIE alignment system The present section provides a comparative–historical outline of the alignment patterns of Hittite, Vedic Sanskrit, Homeric Greek, and Early Latin, which forms the basis of a reconstruction of the PIE alignment system. This outline begins with a couple of observations I take to be uncontroversial, namely that most the oldest attested Indo-European branches show a combination of case marking/flagging and agreement/indexation in argument realization and that these two morphosyntactic dimensions tend to converge on a predominantly nominative–accusative alignment system. This is illustrated by the examples from Hittite, Vedic Sanskrit, and Homeric Greek given in (1).

26

EYSTEIN DAHL

(1) a. aliya-n=kan aliyanzina-š apēl=pát aliya.animal-ACC.SG=PTC aliyanzina-NOM.SG DEM.GEN.SG=PTC mīya-š kuen-zi miya-NOM.SG strike:PRS-3SG ‘The aliyanzina, its own miya will strike the Aliya-animal.’ (KUB 30.36 ii 11–12; CTH 401.1.A [Hittite]) b. bṛ́haspáti-r hán-ty amı´tra-m Bṛhaspati-NOM.SG kill:PRS-3SG enemy-ACC.SG ‘Bṛhaspati kills the enemy.’ (RV VI 73.3; [Vedic Sanskrit]) c. tò-n dè kten-eî égkheï phaı´dimo-s he-ACC.SG PTC kill:FUT-3SG spear:DAT glorious-NOM.SG Héktōr Hektor:NOM ‘Him shall glorious Hektor kill with (his) spear.’ (Hom. Il. 15.65 [Homeric Greek]) d. hospe-s necavi-t hospite-m captu-m manu host-NOM.SG kill:PRF-3SG guest-ACC.SG take:PPP-ACC.SG hand:ABL.SG ‘A host killed a guest, taken by (his) hand.’ (Pl. Most. 479 [Early Latin]) These examples illustrate the detailed parallelism in alignment holding between the early Indo-European languages. Specifically, we observe that the subject argument of two-place verbs is marked by the morpheme -š in Hittite, -r in Vedic Sanskrit, and -s in Homeric Greek and Early Latin,¹ and that the object argument is marked by a nasal morpheme -n in Hittite and Homeric Greek and -m in Vedic Sanskrit and Early Latin.² Finally, there is the third singular verb agreement marker -zi in Hittite, -ti in Vedic Sanskrit, -eî in Homeric Greek, and -t in Early Latin.³ The examples in (2) illustrate that one-place verbs behave in an analogous manner in these languages. (2) a. ta LUGAL-waš LÚ hinkulaš paizzi CONN king-GEN.SG offerant:NOM.SG go:PRS.3SG ‘The king’s offerant goes.’ (KUB 2.7 ii 9, CTH 626.Tg04.I.1.B [Hittite]) b. út sū́ryo jyótiṣā devá eti // up sun:NOM.SG light:INS god:NOM.SG go:PRS.3SG ‘The sun, the god, goes up with his light.’ (RV IV 13.1d after Jamison and Brereton 2014: 575) ¹ Note that the -r in Vedic Sanskrit is a sandhi variant of a morpheme that in other contexts surfaces as -s, and that the long vowel in the final syllable of the nominal stem type reflected in the Homeric Greek name Héktōr usually is taken to reflect a secondary development from *-or-s, a process known as Szemerényi’s law. ² Note that the bilabial nasal -m in Vedic Sanskrit and Early Latin is older, the dental nasal in Hittite and Greek representing secondary developments. ³ The ending -ti in Vedic Sanskrit directly reflects the original PIE third singular ending of so-called athematic verbs, which regularly develops into -zi in Hittite. The ending -eî in Homeric Greek reflects the original PIE ending of so-called thematic verbs. The Early Latin third singular ending -t arose by way of neutralization of the PIE primary and secondary endings *-ti and *-t.

ALIGNMENT IN PROTO-INDO-EUROPE AN

27

c. en d’ arkhòs ébē polúmētis in PTCL leader:NOM.SG go:AOR.3SG of.many.counsels:NOM.SG Odusseús Odysseus:NOM ‘And Odysseus of many counsels went on board as leader.’ (Hom. Il. 1.311) d. at etiam cubat cuculus CONJ still sleep:PRS.3SG cuckoo:NOM.SG ‘And the cuckoo is still asleep.’ (Pl. As. 923) Taken together, the examples in (1) and (2) suffice to show that at least some types of nouns show consistent nominative–accusative alignment in the archaic Indo-European languages, employing the same case and agreement markers, a fact that allows us to reconstruct this pattern for PIE.⁴ Thus, the basic alignment pattern of PIE may be reconstructed as a nominative–accusative constellation (A=S≠P). However, equally uncontroversial is the fact that the formal case distinctions illustrated in (1) and (2), where the subject argument of one- and two-place verbs have nominative case marking, while the object argument has accusative case marking, are not applicable or reconstructable across the board. So-called neuter nouns and pronouns do not distinguish nominative and accusative case forms but employ the same form in all pertinent argument functions, as illustrated by the examples from Vedic Sanskrit and Homeric Greek in (3) and (4), respectively. (3) a. námo dādhāra pṛthivī́m utá dyā́m reverence:NOM/ACC.SG uphold:PRF.3SG earth:ACC CONJ heaven:ACC ‘Reverence upholds heaven and earth’. (RV VI 51.8 after Jamison and Brereton 2014: 846) ́ b. rājā cid ebhyo náma ´ıt king:NOM.SG even they:DAT.PL reverence:NOM/ACC.SG PTCL kṛṇoti make:PRS.3SG ‘Even the king makes his bow to them.’ (RV X 34.8 after Jamison and Brereton 2014: 1431) c. idáṃ pitṛ́bhyo námo astv this:NOM/ACC.SG father:DAT.PL homage:NOM/ACC.SG be:PRS.IMP.3SG adyá today ‘Let this homage here today be for the forefathers.’ (RV X 15.2 after Jamison and Brereton 2014: 1393) ⁴ Here too, attention should be drawn to the fact that the final syllable of the Vedic forms sū́r yo and devá in (2b) represent sandhi variants of -as.

28

EYSTEIN DAHL

(4) a. ándras mèn kteı´nousi, pólin dé te man:M.ACC.PL PTCL slay:PRS.3PL city:F.ACC PTCL CONJ pûr amathúnei fire:N.NOM/ACC.SG destroy:PRS.3SG ‘They slay the men and fire destroys the city.’ (Hom. Il. 9.594) b. mega d’ ´ıakhe thespidaès big:NOM/ACC.SG PTCL roar:IPF.3SG portentous:NOM/ACC.SG pûr fire:NOM/ACC.SG ‘And the portentous fire roared loudly.’ (Hom. Il. 23.216) c. Hē ́ phaistos dè katésbese thespidaès Hephaistos:NOM PTCL put.out:AOR.3SG portentous:NOM/ACC pûr fire:NOM/ACC.SG ‘Hephaistos put out the portentous fire.’ (Hom. Il. 21.381) The examples in (3) and (4) illustrate that Early Vedic neuter nouns like námas ‘homage, reverence’⁵ and Homeric Greek neuter nouns like pûr ‘fire’ have the same form in all of the core argument functions. As regards case marking, then, nouns of this type instantiate a neutral alignment pattern. Accordingly, we may reconstruct a noun–class-based split-alignment pattern for PIE, in which some nouns and pronouns show nominative–accusative case marking, while others show neutral case marking. Similar considerations apply to Hittite and Early Latin. At this point, a few remarks seem appropriate. First, an important difference between Anatolian languages like Hittite, on the one hand, and Indo-Iranian, Greek, and Italic, on the other, concerns the fact that the latter three branches have a distinction between masculine and feminine gender shared with other branches of Indo-European, which does not appear in Anatolian as reflected in Hittite. This distinction pervades the nominal, adjectival, and pronominal system but it is controversial whether a three-way gender distinction ever existed in the prehistory of Anatolian. This is one among several grammatical features that figure as arguments for the assumption that the Anatolian branch was the first to split off from PIE. Under this analysis, the distinction between masculine and feminine gender is a feature of core-IE. At present, it suffices to note that both masculine and feminine nouns and pronouns have distinct nominative and accusative case forms, so our conclusion, that PIE had an NP-based split accusative–neutral alignment system remains valid. A second point concerns the morphological case markers for nominative and accusative employed by the different noun types. Neuter nouns have an alternation between a zero ending and an ending *-m in the singular, which appear

⁵ Note that the forms námo and náma are sandhi variants of námas.

ALIGNMENT IN PROTO-INDO-EUROPE AN

29

with so-called athematic and thematic stems, respectively.⁶ In the plural, we find reflexes of an ending *-h2 across the board. Masculine and feminine nouns show the accusative singular ending *-m across the board, and alternate between a nominative ending in *-s found with most noun stem types and a zero ending appearing with noun stems ending in *-h2 , which characteristically but not necessarily have feminine gender. In the plural, masculine and feminine nouns have the ending *-es in the nominative and the ending *-ns in the accusative. Adjectives follow essentially the same pattern. The pronominal system shows an analogous distribution of endings, except for the fact that the predominant neuter singular ending for nominative/accusative is *-d and that the masculine nominative singular ending in some cases is zero. Table 2.1 contains an overview of the nominative and accusative case endings of nouns and demonstrative/relative/interrogative pronouns in the singular and plural.⁷ Table 2.1 Nominative and accusative case endings in PIE Masculine Singular Nouns and adjectives Nominative *-s

Feminine

Neuter

Plural

Singular

Plural

*-es

*-s/*-Ø

*-es

Accusative Pronouns Nominative

*-m

*-ns

*-m

*-ns

*-s/*-Ø

*-i/*-es

*-Ø

*-es

Accusative

*-m

*-ns

*-m

*-ns

Singular

Plural

*-m/*-Ø

*-h2

*-d/*-m

*-h2

At this point, it should be noted that first- and second-person pronouns essentially show the same distinction between nominative and accusative in the languages under consideration here. However, given that the exact reconstruction of the individual personal pronouns is rendered difficult because of the considerable language-specific differences between them, I have chosen to give the languagespecific forms in Table 2.2 It should be noted that the data given in Table 2.2 yields a considerably simplified picture of the situation in the individual languages, which show a great deal of variation in this morphosyntactic realm. Table 2.2 shows that personal pronouns systematically distinguish between nominative and accusative case in the languages under consideration, except for Early Latin, where we find syncretic forms in the plural. Given that the pronominal stems underlying the forms given in Table 2.2 largely seem to reflect common inheritance, we may safely conclude that this feature should ⁶ Thematic stems end in the so-called thematic vowel *-o-, athematic stems lack this vowel. ⁷ It should be noted that comparative evidence suggests that PIE had a third number category, the dual, in addition to singular and plural. For ease of exposition, it has been omitted here.

30

EYSTEIN DAHL

Table 2.2 The nominative and accusative forms of 1st- and 2nd-person pronouns Hittite 1st-person pronouns singular plural Nominative ūk wēš Accusative ammuk anzaš

Vedic Sanskrit

Homeric Greek

Early Latin

singular plural ahám vayám asmā́n mā́m

singular plural egṓ hēmeîs emé hēméas

singular plural ego nōs mē nōs

2nd-person pronouns Nominative zik šumeš tvám Accusative tuk šumaš tvā́m

yūyám sú yuṣmā́n sé

humeîs tū huméas tē

vōs vōs

be reconstructed for PIE. Table 2.3 provides a schematic representation of this case-marking split. Table 2.3 Split case marking in PIE

The above considerations show that PIE had a split case-marking system, with neuter nouns and pronouns showing neutral case marking and other nouns and pronouns showing nominative–accusative case marking. Another important dimension of morphosyntactic alignment is verb agreement. Comparative evidence shows that most of the inflectional categories of the PIE verbal system had a predominantly accusatively oriented agreement pattern, that is, finite verb forms show agreement in person and number with their subject argument.⁸ Put in slightly different terms, verb agreement or indexation operates on an S/A basis across the languages of our survey. Examples from Hittite, Vedic Sanskrit, Homeric Greek, and Early Latin are given in (5)–(8). (5) a. kappuwanteš=pat=mu=kan antuḫšeš number:PART.NOM.PL.C=PTCL=1SG.DAT=PTCL people:C.NOM.PL isparter escape:PST.3PL ‘Only few people escaped from me.’ (KUB 19.37 iii 25, CTH 61) ⁸ Note that the term ‘subject’ in the present context is taken as a convenient descriptive cover term for the sole argument function of monovalent verbs (S) and the first or highest-ranked argument function of bivalent and trivalent verbs (A). The question as to whether it is possible to delineate a notion of subject with identical or analogous morphosyntactic properties across these and other Indo-European languages is discussed in Chapter 3.

ALIGNMENT IN PROTO-INDO-EUROPE AN

31

b. vā́caṃ parjányajinvitāṃ prá maṇḍū́kā speech:ACC.SG quickened.by.parjanya:ACC.SG PRV frog:M.NOM.PL avādiṣuḥ speak:AOR.3PL ‘The frogs have spoken forth a speech quickened by Parjanya.’ (RV VII 103.1 after Jamison and Brereton 2014: 1013) c. tákha gár se kataktanéousin Akhaioì pántes quickly for 2SG.ACC slay:FUT.3PL Achaean:NOM.PL all:NOM.PL ephormēthéntes rush.on:AOR.PASS.PART.NOM.PL ‘For soon will all the Achaeans rush onto you and kill you.’ (Hom. Il. 6.409–410) d. di illum omnes perdant god:NOM.PL 3SG.M.ACC all:NOM.PL destroy:PRS.SBJ.3SG ‘May all the gods destroy him.’ (Plaut. Men. 596) The examples in (5) illustrate that nominative plural forms of common gender nouns in Hittite (antuḫšeš ‘men, people’) and masculine gender nouns in Vedic Sanskrit (maṇḍū́kās ‘frogs’),⁹ Homeric Greek (Akhaioì ‘Achaeans’), and Early Latin (di omnes ‘all gods’ have accompanying verb forms showing agreement in person and number (isparter ‘(they) escaped’, avādiṣuḥ ‘(they) spoke’, kataktanéousin ‘(they) slayed’, perdant ‘may (they) destroy’). In connection with the examples in (1) we saw that nominative singular forms of the same types of nouns co-occur with third singular verb forms. These observations allow us to conclude that common and masculine gender nouns (and pronouns),¹⁰ as well as feminine nouns (and pronouns)¹¹ triggered verb agreement in PIE. The examples in (6) and (7) illustrate that second-person singular and plural pronouns behave in an identical manner. (6) a. zīk=ma=šmaš=kan anda handāsi 2SG.NOM=CONN=they:DAT=PTCL PRV join:PRS.2SG ‘But you join them.’ (KBo 4.14 ii 75, CTH 123) b. sunvánti sómān pı´basi tvám eṣām press:PRS.3PL soma:ACC.PL drink:PRS.2SG 2SG.NOM DEM.PRON.GEN ‘They press soma-drinks, you drink of them.’ (RV X 28.3, see Jamison and Brereton 2014: 1419)

⁹ The actually attested form maṇḍū́kā is a sandhi variant, the final -s being elided because of the initial a- in the following word. ¹⁰ For reasons of space, I refrain from illustrating the behaviour of masculine gender pronouns, which show exactly the same behaviour in this respect as masculine nouns. ¹¹ Feminine gender nouns and pronouns show the same behaviour as masculine nouns regarding verb agreement in the archaic Indo-European languages with a three-way gender distinction.

32

EYSTEIN DAHL c. sù dè taût’ anemṓlia 2SG.NOM PTCL DEM.PRON.NOM/ACC.PL windy:NOM/ACC.PL bázeis speak:PRS.2SG ‘But you speak these empty [words].’ (Hom. Il. 4.355) d. quoi tu abiens offers REL.DAT.SG. 2SG.NOM depart:PRS.PTCP.NOM.SG offer:PRS.2SG morbum distress:ACC.SG ‘On whom you bestow sorrow by departing.’ (Plaut. As. 593)

Here, we may observe that nominative forms of second-person singular pronouns (zīk, tvám, su, tu) consistently trigger second-person singular agreement on the verb (handāsi ‘you join’, pı´ basi ‘you drink’, bazeis ‘you speak’, offers ‘you offer, bestow’). (7) a. mān šumeš kuwatqa idālu iyatteni if 2PL.NOM in.any.way evil:NOM/ACC.SG do:PRS.2PL ‘If you in any way commit (any) evil (act).’ (KBo 5.3 iv 31, CTH 42) . b. yūyáṃ dhattha rā́jānam ˘ ´sruṣṭimántam 2PL.NOM establish:PRS.2PL king:ACC.SG hearing.attentively:ACC.SG ‘You establish a king with attentive hearing.’ (RV V 54.14 after Jamison and Brereton 2014: 733) ken ē humeîs Troı´ēn c. eis hó until REL.NOM/ACC.SG PTCL or 2PL.NOM Troy:ACC.SG eúpurgon hélēte well.walled:ACC.SG take:AOR.SBJ.2PL ‘Until you either capture well-walled Troy (…).’ (Hom. Il. 7.71) d. quid vos agitis? INTERR.NOM/ACC.SG 2PL.NOM do:PRS.2PL ‘What are you doing?’ (Plaut. Cas. 358) These examples illustrate that nominative forms of second-person plural pronouns (šumeš, yūyám, humeîs, vos) trigger corresponding agreement marking on the accompanying verb (iyatteni ‘you commit’, dhattha ‘you establish’, hélēte ‘you will capture’, agitis ‘you are doing’).¹² Thus, we conclude that first- and secondperson pronouns show accusative alignment in case as well as agreement, just like common or masculine/feminine gender nouns and pronouns. A last point that should be made is that comparative evidence shows that PIE was a pro-drop language, as illustrated by the examples in (8), where the second-plural verb forms parkueššteni ‘you are pure’, rājatha ‘you rule/shine’ and pheúgete ‘you flee’ are not ¹² It should be noted that first-person pronouns behave in an identical manner.

ALIGNMENT IN PROTO-INDO-EUROPE AN

33

accompanied by any explicit subject pronoun. These examples also serve to illustrate that the agreement patterns illustrated in examples (5)–(7), which mostly involved two-place constructions, also apply to one-place predicates. D (8) a. takku parkueššteni šumel LAMMA=KUNU if be.pure:PRS.2PL 2PL.GEN tutelary.deity=2PL.GEN ‘If you are pure (i.e. found innocent), (it is) your tutelary deity.’ (KUB 13.4 iv 32, CTH 264) b. vı´ párvateṣu rājatha apart mountain:LOC.PL rule/shine:PRS.2PL ‘You rule [/shine] throughout the mountains.’ (RV VIII 7.1 after Jamison and Brereton 2014: 1042) c. póse pheúgete phôta idoûsai whither flee:PRS.2PL man:ACC.SG see:AOR.PART.NOM.PL.F ‘Whither shall you flee having seen a man?’ (Hom. Od. 6.199) d. quid denique agitis INTERR.NOM/ACC.SG then do:PRS.2PL ‘What did you do then?’ (Plaut. Bacch. 294)

Thus, we may preliminarily conclude that the predominant verb agreement pattern in PIE also instantiated nominative–accusative alignment. However, so far we have not reviewed any examples involving neuter nouns, which, again, turn out to display somewhat idiosyncratic behaviour, at least in some of the languages under consideration. First, there is significant discrepancy in the comparative evidence as regards the agreement patterns of plural neuter nouns in subject function. In Vedic Sanskrit, neuter plurals in subject function generally trigger plural verb agreement, just like other nouns and pronouns, as illustrated in (9a). In Hittite, neuter plurals generally do not trigger plural verb agreement but appear with singular forms of the verb, as demonstrated in (9b). In Homeric Greek, plural forms of neuter nouns alternately select plural and singular forms of the verb when used in subject function, as shown in (9c) and (9d), respectively. Finally, example (9e) illustrates that Early Latin behaves like Vedic Sanskrit, showing consistent plural verb agreement with neuter plural nouns and pronouns. (9) a. svarvı´de bhúvanāni prathanta sun.finder:DAT.SG being:N.NOM/ACC.PL spread.out:PRS.INJ.3PL ‘The worlds spread out for the finder of the sun.’ (RV IX 94.2 after Jamison and Brereton 2014: 1332) b. nu=wa=kan apē=ya uddār and=QUOT=PTCL DEM.NOM/ACC.PL=and word:NOM/ACC.PL QATAMMA lagāru thus fall:IMP.3SG ‘And thus these words shall fall.’ (KBo 2.3 iii 21–22, CTH 404.1.II.A)

34

EYSTEIN DAHL c. hoîa theáōn lepta te kaì such:NOM/ACC.PL goddess:GEN.PL delicate:NOM/ACC.PL CONJ CONJ kharı´enta kaì aglaà érga beautiful:NOM/ACC.PL CONJ beautiful:NOM/ACC.PL work:NOM/ACC.PL pélontai be:PRS.3PL ‘Such as are the handiworks of goddesses, delicate and beautiful and glorious.’ (Hom. Od. 10.222–223) d. oú toi téknon emòn dédotai NEG 2SG.DAT child:VOC my:VOC give:PRS.MID.3SG polemḗïa érga of.war:NOM/ACC.PL work:NOM/ACC.PL ‘Not to you, my child, works of war are given.’ (Hom. Il., 5.428) e. folia nunc cadunt leaf:NOM/ACC.PL.N now fall:PRS.3PL ‘Now leaves are falling.’

(Pl. Men. 375)

Examples like these illustrate that the agreement patterns of neuter plural nouns do not yield a unitary picture that may serve as a basis for comparative reconstruction. This raises the question as to which of the two patterns is older. At present, we preliminarily conclude that plural forms of neuter nouns trigger plural verb agreement in some languages and singular verb agreement in others. Given that the third singular form of the verb also is the form of choice in impersonal constructions, one way of analysing the pattern with singular verb agreement as involving lack of agreement, which might in turn be understood as an instantiation of neutral alignment.¹³ We thus seem to have two complementary agreement patterns in the languages under scrutiny, which show competition in Homeric Greek: one consistently nominative–accusative and one split between nominative–accusative and neutral. These are given in Table 2.4. Note that the two scenarios given in Table 2.4 have rather different implications for the alignment pattern of neuter nouns in PIE. If one assumes that PIE nouns and pronouns had nominative–accusative agreement across the board, neuter nouns had mixed alignment, instantiating neutral case marking, as established above, and nominative–accusative agreement. If, on the other hand, PIE nouns and pronouns had split agreement, neuter nouns and pronouns would show consistently neutral alignment. We shall return to this problem in Section 2.1.1.

¹³ This suggestion requires some clarification. In languages with an alignment system combining case marking/flagging and agreement/indexation like the Indo-European languages under consideration, subject agreement is an integral part of the predominant alignment pattern. The lack of subject agreement in specific morphosyntactic contexts involves a neutralization of semantic dimensions that are grammatically relevant in other morphosyntactic contexts, such as number in the present case.

ALIGNMENT IN PROTO-INDO-EUROPE AN

35

Table 2.4 Two complementary agreement patterns in PIE

Another intriguing problem concerns the distribution of different types of nouns and pronouns with different types of predicates. Hitherto, we have tacitly assumed that neuter nouns and pronouns essentially show the same distribution as common gender or masculine/feminine gender nouns and pronouns. This assumption is in need of some adjustment. First, we may note that there is a broad tendency across the languages under scrutiny that neuter gender nouns only to a limited extent have animate or human referents, common gender or masculine/feminine nouns representing the main expression of animate and human referents.¹⁴ Another, related observation is that neuter gender nouns are (marginally) attested in subject function of two-place predicates in Vedic Sanskrit and Homeric Greek. Consider the examples in (10). (10) a. yád ánti yác ca dūraké REL.NOM/ACC near REL.NOM/ACC CONJ little.distance:LOC bhayáṃ vindáti mā́m ihá peril:N.NOM/ACC find:PRS.3SG 1SG.ACC here ‘What peril nearby and what at a little distance finds me here.’ (RV IX 67.21 after Jamison and Brereton 2014: 1297) b. toùs dè khlōròn déos hē ́ irei DEM.ACC PTCL pale:N.NOM/ACC.SG fear:N.NOM/ACC.SG seize:IPF.3SG (Hom. Il. 7.479) ‘Then pale fear seized them.’ In Hittite, on the other hand, neuter nouns such as ēšḫar ‘blood’ only appear in subject function of unaccusative one-place verbs and in object function of two-place verbs in their bare form, being supplemented by a form with the suffix -ant- in subject function of two-place verbs, as illustrated in (11) (see, however, the discussion in Section 2.2.1.2 and Chapter 4 in this volume).

¹⁴ To my knowledge, there exists no systematic study of nominal gender and types of reference in the older Indo-European languages.

36 (11)

EYSTEIN DAHL a. nu apāt ēšḫar kuṷapi paizz[i] CONN DEM.N.NOM/ACC.SG blood:N.NOM/ACC.SG when flow:PRS.3SG ‘And when the blood flows.’ (KUB 14.3 iv 52, CTH 181) b. nu ēšḫar akkuškandu CONN blood:N.NOM/ACC.SG drink:IMP.3PL ‘Now, they shall drink blood!’ (KBo 20.73 iv 10, CTH 458.1.1.A ) E c. nu ešḫananza linkiaz ḫalinduwa CONN blood-ant:NOM.SG.C perjury-ant:NOM.SG.C palace É.DINGIR.MÉŠ lē ēpzi temples PROH seize:PRS.3SG ‘Blood (and) perjury shall not seize the palace (and) the temples.’ (KUB 30.34 iv 7, CTH 400.1)

Examples (11a) and (11b) illustrate the use of the bare neuter noun ēšḫar ‘blood’ in subject function of the unaccusative monovalent verb pai- ‘go’ and in object function of the bivalent verb aku- ‘drink’. In example (11c) we observe that the form ešḫananza appears in subject function of the bivalent verb ēp- ‘take, seize’. The status of such forms and of the function of the suffix -ant- in this morphosyntactic context is controversial. Here, we follow Goedegebuure’s (2018) analysis, that -ant- was a derivational suffix serving to derive secondary common gender nouns from neuter and common gender nouns in the oldest attested stage of the language, which then developed into a derivational marker of agentivity primarily selecting neuter nouns at its base in the middle stage of the language, eventually becoming an ergative marker in the later language (see Chapter 4 in this volume). On this analysis, only common gender nouns could appear in subject function of bivalent and trivalent verbs in the oldest attested stage of Hittite. while neuter gender nouns were restricted to subject function of monovalent predicates and object function of bivalent verbs.¹⁵ Again, we are faced with two somewhat different scenarios for PIE, one according to which neuter nouns are perfectly compatible with all three core argument functions, and another according to which such nouns are restricted to subject function of monovalent verbs and object function of bivalent verbs. In a broader perspective, these two scenarios have important repercussions for the reconstruction of the PIE alignment system when compared with the behaviour of other nouns and pronouns, which are compatible with all of the core argument functions, as shown above. The first scenario would involve a situation where all nouns and pronouns are compatible with the full range of argument functions, their distribution only being constrained by lexical semantic ¹⁵ As regards the use of neuter nouns in subject function of monovalent verbs, they generally seem to appear with unaccusative predicates only.

ALIGNMENT IN PROTO-INDO-EUROPE AN

37

Table 2.5 Compatibility of nouns/pronouns with core argument functions in PIE

features, while the other scenario would involve a situation where neuter nouns and pronouns are subject to morphosyntactic constraints that do not apply to other nouns and pronouns. The two alternatives are schematically represented in Table 2.5.¹⁶ Taken together, the properties summarized in Tables 2.3–2.5 demonstrate that neuter nouns show a rather disunited behaviour across the languages under scrutiny, a fact rendering a reconstruction of their status in PIE challenging. To clarify this issue, two sets of problems need to be addressed. On the one hand, we must ascertain which of the agreement patterns is older, and, on the other hand, it is necessary to clarify whether the morphosyntactic restrictions neuter nouns show in Hittite represents an archaism or an innovation. These issues will be dealt with in Sections 2.1.1 and 2.1.2, respectively.

2.1.1 The agreement patterns of neuter nouns in PIE Above we saw that plural forms of neuter nouns and pronouns do not show a unitary agreement pattern across the three branches under consideration. In Indo-Iranian, as reflected in Vedic Sanskrit and Italic, as instantiated by Early Latin, nouns of this class show plural agreement marking on the verb, a feature shared with other branches such as Germanic as represented by Gothic, illustrated in (12). ¹⁶ Note that BS is shorthand for Bivalent Subject, MS for Monovalent Subject, and BO for Bivalent Object.

38

EYSTEIN DAHL

(12) ni indrobnaina izwara hairtona nih NEG be.troubled:PRS.SBJ.3PL your:N.PL heart:NOM.PL NEG.CONJ faurhtjaina be.afraid:PRS.SBJ.3PL ‘And your hearts shall not be troubled, nor shall they be afraid.’ (John 14:27 [Gothic]) The fact that the pattern with plural verb marking is found across several branches is a strong prima facie indication that it may be old, and that the pattern instantiated in Hittite and Greek, illustrated in (9) is secondary. A hypothesis along such lines may seem to find some support in diachronic evidence from Greek. As shown in (9cd), Homeric Greek shows an alternation between the two agreement patterns, as also shown by the examples in (13). (13)

a. kaì tôn mèn hamartêi doúrata makrà CONJ the:GEN.PL PTCL at.once spear:NOM/ACC.PL long:NOM/ACC.PL ek kheirôn ḗïxan from hand:GEN.PL spring:AOR.3PL ‘And the long spears sprang from the hands of both at once.’ (Hom. Il. 5.656–657 [Homeric Greek]) b. kanákhize dè doúrata púrgōn ring:IPF.3SG PTCL beam:NOM/ACC.PL tower:GEN.PL ballómen’ throw:PRS.MID.PART.NOM/ACC.PL ‘The beams of the towers rang as they were thrown.’ (Hom. Il. 12.36–37 [Homeric Greek])

Since Homeric Greek shows a systematic alternation between the singular and plural agreement pattern, it provides a convenient point of departure for exploring the diachronic relationship between these two constructions. Although many aspects of the origin and development of the two Homeric epics remains disputed, we shall assume at present that they essentially represent the last stage of a long bardic tradition the roots of which date back at least to the mid-second millennium BCE. Among the many controversies in Homeric studies is the question whether both epics can be ascribed to a single author, but most specialists seem to agree that both the Iliad and the Odyssey in the form that we know them can be dated to about the eighth century BCE (see, e.g. Kirk 1985, Janko 2011, West 2011a). We noted above that neuter plural subjects show plural verb agreement in Vedic Sanskrit, Latin, and Gothic. Interestingly, however, this construction constitutes slightly under one-quarter of the total number of examples of neuter plural subjects in Homeric Greek, as shown by the data in Table 2.6.¹⁷ ¹⁷ The data given in Table 2.6 were collected through a targeted search in the Perseus under PhiloLogic database.

ALIGNMENT IN PROTO-INDO-EUROPE AN

39

Table 2.6 Singular and plural verb agreement in Homeric Greek

Iliad Odyssey Total

Singular

Plural

Total

106 62 168

37 25 62

143 87 230

These numbers show that there is a slight difference between the Iliad and the Odyssey regarding the frequency of neuter plural subjects and it is therefore important to clarify whether this difference is statistically significant. A chi-square test of the raw frequencies in Table 2.6 yields a p-value of .748, a χ2 value of 0.103 and one degree of freedom (χ2 (1) = 0.103), indicating that the difference between the two texts is not statistically significant and that we may treat the two works as one, unitary corpus (see also Dahl 2021b).¹⁸ The data in Table 2.6 show that Homeric Greek has a clear preference for the construction with neuter plural subjects and singular verb agreement. It is dubious whether this fact alone can shed any light upon the question as to which of the two agreement patterns is older. The evidence provided by Mycenaean, which represents an older variety of Ancient Greek than Homeric, does not suffice to make any firm conclusions as to the relative frequency of the two constructions. In other, later varieties of Ancient Greek, the plural agreement pattern becomes obsolete, the singular pattern being the only option for neuter plural subjects. Consider, by way of illustration, the diachronic data given in Table 2.7.¹⁹,²⁰ Table 2.7 Agreement patterns of neuter plural subjects through the history of Greek

Singular agreement Plural agreement

Archaic Homer

Hesiod

Classical Herodotus

Thucydides

Hellenistic Polybius

168 62

25 7

38 3

50 0

34 2

¹⁸ These results were obtained by using the chisqu.test() function in the standard package of R (R Core Team 2019). The data and R script are available at https://github.com/eystdahl/AlignmentPIE. ¹⁹, The data furnishing the basis for the frequencies in Table 2.7 are drawn from the following sources in the Perseus under PhiloLogic database (http://perseus.uchicago.edu/): The Iliad and The Odyssey (Homer), the Theogony, Works and Days and Shield of Heracles (Hesiod), Book 1 and 2 of Histories (Herodotus), Book 1 and 2 of The Peloponnesian War (Thucydides) and the Histories (Polybius). ²⁰ For a schematic survey of the chronological stages of Ancient Greek, see Chapter 3 in this volume.

40

EYSTEIN DAHL

Table 2.7 shows that while the plural agreement pattern is robust in the Archaic Period as reflected in the Homeric Corpus and Hesiod, this construction virtually disappears in Classical Greek, as represented by Herodotus and Thucydides, a situation also holding in early Hellenistic Greek (Koiné), as represented by Polybius.²¹ A chi-square test of the data in Table 2.8 yielded a p-value of 7.125e-06, a χ2 value of 29.198 and one degree of freedom (χ2 (4) = 29.198), so that it is unlikely that the diachronic distribution of the two construction types is due to chance. A schematic representation of the relative frequency of the two constructions through the history of Ancient Greek is given in Figure 2.1 The data given in Table 2.7 and Figure 2.1 strongly indicate that the pattern with singular agreement replaced the pattern with plural agreement in the Classical Period. Mutatis mutandis, one might assume that the agreement pattern shown by Vedic Sanskrit, Latin, and Gothic is the original one, and that a process analogous to the one we observe in historically attested Ancient Greek took place at some stage in the prehistory of Hittite, resulting in an identical situation. There are, however, at least two sets of evidence that speak against a scenario along such lines. First, although the plural agreement pattern is robust in Vedic Sanskrit, there are three examples in the Rigveda where neuter plural subjects select a singular form of the verb, one of which is given in (14a).²² One could argue that the few instances of this pattern in Vedic Sanskrit represent sporadic cases of the same

Relative frequency (%)

100

Singular Plural

80 60 40 20 0 Homer

Hesiod

Herodotus Periods

Thucydides

Polybius

Fig. 2.1 Singular and plural agreement with neuter plural subjects in Ancient Greek Source: adapted from Dahl (2021a: 424).

²¹ At this point, it should be noted that certain texts from the early Hellenistic period, notably the Septuagint, show a significant attraction towards the plural agreement pattern. However, given that the Septuagint is a translated text, where the source languages, Biblical Hebrew and Aramaic, have complex patterns of verb agreement, even involving agreement in gender, secondary influence from the source languages cannot be excluded. A more thorough discussion based on data from a broader set of Hellenistic sources is far beyond the scope of the present work and will have to be undertaken elsewhere. ²² See Delbru¨ck (1888: 83).

ALIGNMENT IN PROTO-INDO-EUROPE AN

41

innovative construction that prevailed in Ancient Greek and, by hypothesis, Hittite. However, an explanation along such lines would also have to account for the fact that the closely related Old Iranian language Gatha Avestan seemingly shows a clear preference for the singular agreement pattern when plural neuter nouns and pronouns appear in subject function, as illustrated in (14b). (14)

te ápi devéṣv a. sárvā tā́ all:N.NOM/ACC.PL this:N.NOM/ACC.PL 2SG.GEN PTCL god:M.LOC.PL astu be:PRS.IMP.3SG ‘Let all this of yours be also among the gods.’ (RV I 162.14 after Jamison and Brereton 2014: 345; [Early Vedic]) ̥ b. mazdā saxv arә¯ mairištō wise.one:NOM verse:N.NOM/ACC.PL remembering:SPLV yā zī vāuuǝrǝzōi pairī ciϑīt̰ REL:N.NOM/ACC.PL PTCL make:PRF.MID.3SG ADV PTCL ‘The Wise One best remembers the verses, which have been made around (here).’ (Y. 29.4; [Gatha Avestan])

These examples illustrate that neuter plural NPs such as sárvā tā́ ‘all these (things)’ and yā ‘which’ co-occur with singular verb forms, sporadically in Early Vedic and more regularly in Gatha Avestan.²³ Taken together, these observations suggest that the construction with neuter plural subjects and singular verbs was available and possibly more productive in Proto-Indo-Iranian, the common ancestor of Vedic Sanskrit and Gatha Avestan. Under this analysis, the sporadic cases of this construction in Vedic Sanskrit represent archaic relics, reflecting a pattern that had become almost entirely replaced by the pattern with plural verb agreement. The hypothesis that the plural agreement pattern has replaced the singular agreement pattern in Early Vedic may find some support in the somewhat more recent variety of Avestan commonly labelled Young Avestan, where both constructions appear, as illustrated in (15). In this variety, however, the plural agreement pattern seemingly has almost entirely replaced the singular agreement pattern.²⁴ ²³ At this point, it must be emphasized that the actually attested examples of this construction in Gatha Avestan are rather few, only around five cases being cited in the pertinent philological literature (see, e.g. West 2011b: 9). Moreover, all of the Gatha Avestan examples occur in relative clauses, and this could be of some importance, since agreement patterns in relative clauses do not necessarily coincide with agreement patterns in matrix clauses, as noted in Dahl (2021a: 432, fn. 10). For present purposes, we note that there are no examples of the plural agreement pattern in the Gatha Avestan texts and take this as an archaic feature of Gatha Avestan. However, I hasten to add that the evidence provided by this stage of the Avestan language can hardly be regarded as abundant, so that its significance in absolute terms remains somewhat weak. ²⁴ This observation is based on the fact that standard grammars and handbooks of Avestan only mention two examples of this construction in Young Avestan, the one cited in (15a) given in Skjærvø (2009: 103) and one given in Reichelt (1909: 301) originating from a citation in a Pahlavi translation of Vendīdād 2.40. It is, of course, not excluded that a comprehensive examination of the entire Young

42 (15)

EYSTEIN DAHL a. vǝrǝziiatąm=ca iδa vohu be.produced:PRS.IMP.3SG=CONJ here good:N.NOM/ACC.PL vāstriia pasture:N.NOM/ACC.PL ‘And good pastures shall be produced here.’ (Vīsp. 15.1 [Young Avestan]) b. kuua tā dāϑra INTERROG.ADV the:N.NOM/ACC.PL offering:N.NOM/ACC.PL paiti.haṇjasǝnti go:PRS.3PL ‘Where do the offerings go?’ (Vend. 19.27 [Young Avestan])

These examples illustrate that neuter plural nouns such as vāstriia ‘pastures’ and dāϑra ‘offerings’ appear with singular verb forms such as vǝrǝziiatąm, albeit very rarely, or plural verb forms in Young Avestan. Thus, the situation in this variety of Avestan seems to be identical as in Early Vedic, suggesting that Old Avestan reflects the Proto-Indo-Iranian system, whereas both Young Avestan and Early Vedic present us with parallel innovations in this morphosyntactic realm. As per Cotticelli and Dahl (Chapter 3 in this volume), this development would represent a type of morphosyntactic change contributing to a higher degree of consistency in morphosyntactic alignment, which they label generalization in line with Dahl (2021a). More specifically, the development postulated for Indo-Iranian would involve what Cotticelli and Dahl (this volume) label ‘accusativization’, that is, the diachronic increase of accusative features. In Anatolian, as reflected in Hittite, on the other hand, neuter plural subjects mostly show singular verb agreement, as illustrated in (9b). However, here too, there is some interesting diachronic variation.²⁵ Specifically, in New Hittite there are a couple of examples of neuter plural subjects appearing with a third plural verb form, as illustrated in (16). (16)

a. nu=mu KUR.KUR.MEŠ hūmanda menaḫḫanda CONN=1SG.ACC/DAT lands all:NOM/ACC.PL against kururiaḫḫir make.war:PST.3PL ‘And all the lands made war against me.’ (KBo 10.2 i 25, CTH 4.II.A [NH] after Gertz 1982: 79)

Avestan corpus might yield some more examples of the construction with singular agreement, but such an endeavour is far beyond the scope of the present work. ²⁵ See Chapter 3 in this volume, for a schematic survey of the chronological stages of Hittite.

ALIGNMENT IN PROTO-INDO-EUROPE AN

43

b. nu=šši=kan išḫaḫru [par]ā PA4 ḪI.A -uš mān CONN=3SG.DAT=PTCL tear:NOM/ACC.PL forth canals like aršanzi flow:PRS.3PL ‘And his tears flow forth like canals.’ (KUB 33.87 i 29–30, CTH 345.I.2.A [NH] after Gertz 1982: 130) An interesting feature of these examples is that they appear in Late Hittite texts, example (16a) appearing in a Late Hittite copy of an Old Hittite text (Gertz 1982). A careful examination of Gertz’s (1982) comprehensive study of the nominative– accusative neuter plural in Anatolian suggested that the two examples cited in (16) are isolated in Hittite.²⁶ Since the available examples belong to the later chronological layer of Hittite, it is reasonable to conclude that here too, plural verb agreement with neuter plural subjects represents a secondary development. The two other Old Anatolian languages, Luwian and Palaic,²⁷ also show singular verb agreement with neuter plural subjects, a property that clearly represents common inheritance in Anatolian and from a comparative perspective may be regarded as an archaism shared by Ancient Greek, Indo-Iranian, and Anatolian. Under the present analysis, then, singular verb agreement was the predominant and possibly the only available alignment pattern with neuter plural subjects in three of the earliest attested branches of Indo-European, Anatolian, IndoIranian, and Greek. However, this begs the question as to how this hypothesis can accommodate the diachronic data from Ancient Greek discussed previously in this section. One way of accounting for the fact that we find both agreement patterns in Homeric Greek is to assume that it reflects an incipient tendency to use plural verb forms that gained some ground early in the history of Ancient Greek but in the end did not become grammaticalized and disappeared. This is unlike Indo-Iranian, and, by hypothesis, Italic, Germanic, and other branches, where the plural pattern prevailed with neuter plural subjects. In Anatolian, on the other hand, this pattern seemingly arose late and is marginal in the latest attested stage of Hittite. According to this hypothesis, all of the pertinent branches show signs of a development from singular to plural verb agreement with neuter plural subjects, which, as already suggested for Indo-Iranian, reflects a general tendency across the Indo-European family to generalize accusative morphosyntax, that is, accusativization.²⁸ From the previous discussion, the impression arises that the development of plural verb agreement with neuter plural subjects represents independent innovations in the different branches. While this clearly is ²⁶ This impression was corroborated by a scrutiny of the data in Prins (1997). ²⁷ See Melchert (2003, 2008) for observations on the Luwian and Palaic evidence. ²⁸ See Chapter 3 in this volume.

44

EYSTEIN DAHL

Table 2.8 Proto-Indo-European as a split-agreement language

the case in Anatolian, evidence from Indo-Iranian and Greek is less compelling in this respect. It therefore seems most parsimonious to assume that plural verb agreement with neuter plural subjects was incipient in Core IE. From the overall perspective of the present work, the fact that we can reconstruct singular verb agreement for neuter plural subjects in PIE has two important corollaries. First, this means that neuter gender nouns/pronouns had neutral agreement in PIE, and second, this in turn implies that PIE had split nominative–accusative and neutral agreement. Table 2.8 provides a schematic representation of the split agreement system of PIE. Taken together with the above observations, summarized in Table 2.3, we may therefore tentatively conclude that comparative data suggest that PIE had a consistent noun–class-based split between nominative–accusative and neutral alignment, which involved both case marking/flagging and verb agreement/indexation. However, as will be clear from the discussion in the next section, this reconstruction runs into another difficulty, namely the fact that there is evidence that neuter nouns did not occur freely in all the core argument functions in PIE.

2.1.2 Morphosyntactic restrictions of neuter nouns in PIE In Section 2.1 we noted that there is some divergence between the languages that form the basis of this comparative study as regards the ability of neuter nouns/pronouns to appear in first argument function of bivalent and trivalent verbs. While we find unequivocal examples of this in Vedic Sanskrit, Homeric Greek, and Early Latin (see (18)), examples from Hittite are scarce and seemingly non-existent in the earliest stage of the language. A systematic scrutiny of volumes 1–9 of Jaan Puhvel’s Hittite Etymological Dictionary (HED) yielded only three possible examples of bare neuter nouns occurring in first argument function of a bivalent verb, one from Middle Hittite and two from New Hittite. These are given in (17).

ALIGNMENT IN PROTO-INDO-EUROPE AN (17)

45

a. nu=mu=ššan šēr aššul natta CONN=1SG.ACC/DAT=CONN PTCL good:NOM/ACC.SG NEG išduwari become.apparent:PRS.MP.3SG ‘Good does not become apparent to me.’ (KUB 30.10 Rs. 19 CTH 373.A [MH]) b. antuḫšātar=ma=wa=naš population:NOM/ACC.SG=CONN=QUOT=1PL.ACC/DAT arantallin[zi be.insurgent:PRS.3PL ‘The population is insurgent against us.’ (KUB 6.41 i 25 CTH 68.D [NH]) GIŠ c. eyan maḫḫan uktūri iyatniyan eya-tree:NOM/ACC as ever grow:PTC.NOM/ACC.N nu ḫurpaštanuš arḫa ŪL išḫuwai CONN leaf:ACC.PL PRV NEG throw:3SG.PRS ‘A s the eya-tree is ever verdant and does not shed its leaves (…).’ (KUB 29 1 iv 18 CTH 414.1A [NH])

At this point, a critical reader would object that the examples in (17) provide rather weak evidence for the use of bare neuter nouns in first argument function of bivalent predicates in Middle and Late Hittite. First, the enclitic pronouns mu ‘1sg’ and naš ‘1pl’ in (17a) and (17b), respectively, both are syncretic for accusative and dative case. Therefore, one may harbour some doubt as to the argument status of these pronouns and, consequently, to the status of the bare neuter nouns such as aššul ‘(something) good, favour’ in (17a) or antuḫšātar ‘population’ as first arguments of bivalent predicates rather than sole argument of monovalent predicates. Example (17c) admittedly provides somewhat more compelling evidence for neuter nouns in bivalent subject function, in this case GIŠ eyan ‘eya-tree’, given that the noun ḫurpaštanuš ‘leaves’ is a plural common gender noun with accusative case marking. While this appears to be a bona fide case of a bare neuter noun in first argument function of a bivalent verb, it should be emphasized that it is attested in a late stage of the language, thus indicating that, to the extent that such nouns could appear in first argument function, this was a secondary development with an extremely low text frequency.²⁹ Having established that the employment of bare neuter nouns in first argument function of bivalent verbs represents a secondary and rather late development in Hittite, we now go on to explore the main problem of this part of the chapter, whether neuter nouns could appear in bivalent subject position in PIE. Given ²⁹ It is tempting to suggest that cases like those cited in (17a) and (17b) represented a bridging context, from the basis of which cases like (17c) arose. However, this possibility will have to be explored elsewhere.

46

EYSTEIN DAHL

that Old Hittite most likely did not have this feature, the research question essentially is whether the Old Hittite situation can plausibly be assumed for PIE or not. The feature in question shows up in Vedic Sanskrit and Homeric Greek, as illustrated by examples (3a) and (4a), as well as in Early Latin, as illustrated in example (18). (18) Quia mihi misero cerebrum excutiunt CONJ 1SG.DAT poor:DAT.SG brain:NOM/ACC knock.out:PRS.3PL tua dicta, soror POSS.2SG.NOM/ACC.PL word:NOM/ACC.PL sister:VOC ‘Because your words are knocking out the brain of poor me, sister.’ (Plaut. Aul. 151–152 [Early Latin]) Here, the neuter plural noun phrase tua dicta ‘your words’ is the first argument of the bivalent verb excutiunt ‘are knocking out’, the second argument of which is cerebrum ‘brain’. This example is a prima facie indication that Early Latin behaves like Early Vedic and Homeric Greek and unlike Hittite in this respect. It remains to be explored, however, whether this in fact is the case. From a more general perspective, the fact that three of the four languages under discussion allow neuter nouns as bivalent subjects provides strong prima facie evidence for the hypothesis that Early Vedic, Homeric Greek, and Latin preserve the PIE situation, while the situation in Old Hittite is innovative. Alternately, one could assume that Old Hittite is archaic in this respect, while the syntactic flexibility of neuter nouns represents common or (partially) independent, parallel innovation in the other languages, that probably was incipient in Core IE. One immediate objection against the former hypothesis is that it violates Occam’s razor, since it would seem to presuppose the loss of a syntactic option for which there is no positive evidence and the later re-emergence of the same syntactic option. While this is not an optimal situation, this objection cannot be regarded as a decisive counterargument, since we find an analogous situation in Ancient Greek, where the plural agreement pattern turns up again in the post-classical language (see Section 2.1.1, n. 21). However, if it is correct that Hittite lost the possibility of employing bare neuter nouns in bivalent subject function, it would be reasonable to expect the three other languages to show approximately the same degree of permissiveness regarding neuter nouns in this function. If, on the other hand, this is not the case, this would imply that the situation in Hittite is more archaic, the other languages representing different stages of innovation. To explore this issue, representative samples of neuter nouns in core argument functions were collected from Early Vedic, Homeric Greek, and Early Latin.³⁰ The ³⁰ Here, a brief note on the samples is in order. In the case of Early Vedic and Homeric Greek, the samples were collected through a targeted search for nominative/accusative forms of neuter nouns in the VedaWeb database (https://vedaweb.uni-koeln.de/) and the Chicago Homer database

ALIGNMENT IN PROTO-INDO-EUROPE AN

47

Table 2.9 Distribution of neuter nouns in the core argument functions

Bivalent subject Monovalent subject Bivalent object

Vedic Sanskrit

Homeric Greek

Early Latin

5 21 129 N = 155

8 56 116 N = 180

1 46 109 N =156

distribution of neuter nouns over the three core functions in the samples from these three languages is given in Table 2.9. Again, I employ the terms bivalent subject, monovalent subject, and bivalent object for the core argument functions in description on language-specific level, reserving the terms A, S, and P for the argument functions associated with prototypical transitive verbs such as ‘kill’, ‘break’, ‘cut’ (Haspelmath 2011, Dahl 2021a: 424). These raw numbers indicate that there is some variation between the three languages under consideration regarding their respective permissiveness of neuter nouns in bivalent subject function. Specifically, Homeric Greek appears to be more permissive in this respect than Vedic Sanskrit and Early Latin. Figure 2.2 provides a visualization of the relative frequencies of neuter nouns in the three core argument functions. Figure 2.2 illustrates that the three languages show divergent proportional distributions of neuter nouns in core argument function. For present purposes, however, it is essential to establish whether the differences in distribution regarding neuter nouns in bivalent subject function are statistically significant. First, we employ the chi-square test to see whether the raw distributions in Table 2.9 are due to chance and find that it yields a p-value of 0.0003742, a χ2 value of 20.634, and four degrees of freedom (χ2 (4) = 20.634), indicating that the distributional differences between the three languages is significant, assuming the conventional significance level of 0.05.³¹ Moreover, the dataset yields a Cramer’s V value of 0.145, implying a rather small effect size.³² Since the chi-square test assesses the distribution as a whole, this does not necessarily imply that the different distributions of neuter nouns in first argument (https://homer.library.northwestern.edu/), respectively. The Early Vedic corpus from which the sample was collected comprises book II of the Rigveda, while the Homeric Greek corpus comprises the first three books of the Iliad. The Early Latin sample was collected through manual searches in the Plautine texts of the PHI database of Classical Latin Texts (https://latin.packhum.org/), and comprises the following comedies of Plautus: Aulularia, Casina, and Menaechmi. The examples are available at https://github.com/eystdahl/AlignmentPIE. ³¹ These results were obtained by using the chisqu.test() function in the standard package of R (R Core Team 2019). The full R script is available at https://github.com/eystdahl/AlignmentPIE ³² The Cramér’s V value was obtained by using the assocstats() function in the additional R package vcd (Meyer et al. 2017).

48

EYSTEIN DAHL 100

Relative frequency

80

BS MS BO

60 40

20

0 Vedic Sanskrit

Homeric Greek

Early Latin

Fig. 2.2 Distribution of neuter nouns in core argument functions BS = Bivalent Subject; MS = Monovalent Subject; BO = Bivalent Object

Table 2.10 Expected distribution of neuter nouns in core argument functions

Bivalent subject Monovalent subject Bivalent object

Vedic Sanskrit

Homeric Greek

Early Latin

4.4 (5) 38.8 (21) 111.8 (129)

5.1 (8) 45.1 (56) 129.8 (116)

4.4 (1) 39.1 (46) 112.5 (109)

function of bivalent verbs are statistically significant. A first step to clarify this is to calculate the expected frequencies for each of the cells, i.e. the occurrences of neuter nouns in each of the argument functions in each of the languages one would expect if there were no difference between the languages in this respect. These are given in Table 2.10; the observed frequencies are given in parentheses for convenience.³³ The data in Table 2.10 show that there is divergence between the expected and observed frequencies across the board. It remains to be established, however, to what extent these differences are significant. To this aim, we apply the Fisher Exact ³³ The expected value of a given cell is obtained by multiplying the row sum with the column sum and dividing the product with the total sum. Here and below, I used the chisqu.test()$expected function in the standard package of R (R Core Team 2019) to obtain the expected values. The full R script is available at https://github.com/eystdahl/AlignmentPIE

ALIGNMENT IN PROTO-INDO-EUROPE AN

49

Table 2.11 Attractions and repulsions of neuters and core argument functions

Bivalent subject Monovalent subject Bivalent object

Vedic Sanskrit

Homeric Greek

Early Latin

[+]0.4675 [-]2.858e-05 [+]9.72e-05

[+]0.09328 [+]0.01273 [-]0.002964

[-]0.03375 [+]9.72e-05 [-]0.2593

Test to each of the cells.³⁴ The results are given in Table 2.11;³⁵ the notations [+] and [−] indicate that the observed value is greater than or smaller than the expected value. Along the lines of Janda et al. (2013: 89ff.), observed values greater than the expected values are referred to as ‘attractions’, while observed values smaller than the expected values are referred to as ‘repulsions’. Assuming 0.05 as the level of statistical significance, the results reported in Table 2.11 show that the deviations between the observed and expected values are statistically significant in six out of 12 cases. For present purposes, the most relevant finding is that Homeric Greek and Vedic Sanskrit show a slight but insignificant attraction towards employing neuter nouns in bivalent subject function while Early Latin shows significantly greater restrictiveness than the two other languages in this very same respect.³⁶ These observations show that Early Latin resembles Hittite rather than Vedic Sanskrit and Homeric Greek regarding the use of neuter nouns and pronouns in bivalent subject function. This, in turn, supports the hypothesis that the situation in Old Hittite, where neuter nouns do not appear in this argument function, is more archaic. An important corollary of this analysis is that the distribution of neuter nouns in bivalent subject function found in Early Latin is more conservative than those found in Homeric Greek and Vedic Sanskrit. These considerations invite the conclusion that PIE had a distributional split between neuter and non-neuter nouns, where neuter nouns/pronouns were restricted to monovalent subject and bivalent object function, whereas nonneuter nouns/pronouns appeared in all three argument functions, albeit with a special case form in the bivalent object function. Indeed, the incipient change ³⁴ Janda et al. (2013: 13) note that ‘[t]he Fisher Test takes into account the overall distribution of values in our table and, when applied to each cell, it can tell us the probability that each value could deviate even more from the expected value. This means that lower probabilities reflect stronger attractions and repulsions.’ ³⁵ The results reported on in Table 2.11 were obtained by using the fisher.test() function in the standard package of R (R Core Team 2019), specifying ‘alternative = “greater”’ when the observed value is greater than the expected value and ‘alternative = “less”’ when it is smaller than the expected value. The full R script is available at https://github.com/eystdahl/AlignmentPIE ³⁶ For the sake of completeness, it should be noted that Vedic Sanskrit and Homeric Greek diverge in significant respects with regard to their relative attraction/repulsion towards employing neuter nouns in monovalent subject and bivalent object function.

50

EYSTEIN DAHL

reflected in Early Latin, which by hypothesis appears in more developed form in Homeric Greek and Vedic Sanskrit may be regarded as a case of generalization or accusativization as defined by Cotticelli and Dahl (Chapter 3 in this volume), where neuter nouns gradually spread to bivalent subject function. To sum up. Under the present analysis, we reconstruct an alignment system for PIE which shows a consistent split between neuter and non-neuter nouns and pronouns on three morphosyntactic levels, case marking, verb agreement and what one can loosely refer to as core argument distribution. Comparative evidence from Hittite, Vedic Sanskrit, Homeric Greek, and Early Latin suggests that the original system showed convergence on all three levels and that this convergence was gradually replaced by other patterns. Interestingly, changes in this realm seem to affect each morphosyntactic dimension independently. This is, for instance, suggested by the fact that Homeric Greek shows conservatism in verb agreement but not regarding core argument distribution, while Early Latin shows innovation in verb agreement and conservatism in core argument distribution. Overall, case marking represents the most stable morphosyntactic dimension, being maintained in all the languages considered here (see also Chapter 3 in this volume). I hope to have shown that the present analysis allows for reconstructing a consistent splitalignment system for PIE. In the next section, we shall briefly review some recent hypotheses concerning the prehistory of the PIE alignment system.

2.2 Alignment in PIE and pre-PIE This section briefly outlines some recent contributions to a long-standing debate on the prehistory of the PIE alignment system. As noted in the introductory section, there are three different approaches to this question, the accusative hypothesis (Section 2.2.1), the ergative hypothesis (Section 2.2.2), and the semantic hypothesis (Section 2.3.3). Section 2.3.4 explores the background for the occurrence of singular agreement with neuter plural subjects in PIE.

2.2.1 The accusative hypothesis According to what may be regarded as the standard hypothesis, PIE had an accusatively oriented alignment system. This hypothesis was established in the late nineteenth and early twentieth centuries by the pioneering comparative syntactic work of scholars like Karl Brugmann and, especially, Berthold Delbru¨ck (see, e.g. Brugmann and Delbru¨ck 1893, 1897, 1900). At this point, it should be noted that this generation of scholars did not have any knowledge of Hittite or the Anatolian branch of Indo-European more generally, since Hittite was first identified as an Indo-European language in 1902 (Knudtzon 1902: 41–53), followed by

ALIGNMENT IN PROTO-INDO-EUROPE AN

51

more systematic studies during the First World War (Hrozný 1915, 1917, 1919). However, given that Hittite, like all other archaic Indo-European languages, also shows (some degree of ) accusative alignment, it seems reasonable to conclude, with Keydana (2018: 2197) that ‘PIE was a nominative–accusative language. As all attested old IE languages are of this type, hypothesizing any other syntactic type would be highly implausible.’ Observations along these lines find strong empirical support in recently published work by Gerd Carling and Chundra Cathcart (2021), who base their reconstruction of the PIE alignment system on a Bayesian phylogenetic model. The dataset employed in their study comprises 126 Indo-European languages drawn from the DiACL database (Carling 2017). As noted by the authors, ‘[t]he data consist of comparative concepts, i.e. abstractions of grammar properties in attested languages, which are not connected to any morphological material that can be reconstructed. In this aspect, the model complements models of reconstruction, which used morphological reconstruction in combination with syntactic meaning reconstruction and diachronic typology’ (Carling and Cathcart 2021: 383). Furthermore, they emphasize that the application of the computational phylogenetic method to a linguistic family generates insights into the predominant dynamics of change characteristic of that family, as well as the stability of a given type of categorical feature in a phylogeny, i.e. its phylogenetic stability. The probability of a given feature at the nodes in the phylogenetic tree is calculated on features present in the language-specific data, thus precluding, as the authors rightly note, the reconstruction of features and states that are unattested in the phylum (Carling and Cathcart 2021: 384). In spite of the considerable variation in alignment found across the Indo-European languages, their study results in a reconstruction of PIE where nominative–accusative alignment features turn out to constitute the features that can be reconstructed with the highest probability for PIE. An important and very interesting finding of their work is that there is a clear tendency that different features within a given dimension show a hierarchical organization, which generally show a high degree of correspondence with independently established marking hierarchies found across languages. For example, while there is a clear tendency that the probability of nominative–accusative alignment outranks the probability of other alignment types across pronouns, nouns, and tenses, this alignment types shows a higher probability with pronouns than with nouns, with present progressive than with simple past, and so on. A schematic representation is given in Figure 2.3. These findings appear to correlate in a general way with the claims made in the present chapter, although it should be emphasized that Carling and Cathcart’s (2021) analysis does not seem to capture the crucial differences between neuter and non-neuter nouns identified as a central feature here. This is partly because their analysis is based on a Comrian-style definition of the core argument functions A/S/P, implying that they mainly take the default marking patterns of prototypically transitive and intransitive predicates into account. From

52

EYSTEIN DAHL Probability range 1−0.9 0.9−0.8 0.8−0.7 0.7−0.6 0.6−0.5 0.5−0.4 0.4−0.3 0.3−0.2 0.2−0.1 0.1−0 More frequent

Pronouns Present progressive

NEUTR

NOM

Simple past

NOM

ERG

NEUTR TRI

Nouns Present progressive Simple past

NEUTR

NOM NOM

NEUTR

TRI ERG

TRI

Less frequent Lower probability

Higher probability

Fig. 2.3 Ranges (divided into 10% segments) of the probabilities of the different tenses and word classes of alignment probabilities for Proto-Indo-European, contrasted to grammatical hierarchies of categories observed in languages NOM = nominative–accusative, NEUTR = neutral marking, ERG = ergative, TRI = tripartite Source: From Carling and Cathcart (2021: 381). Reprinted with kind permission from John Benjamins Publishing Company, Amsterdam/Philadelphia. [www.benjamins.com]

this perspective, Carling and Cathcart’s (2021) methodology complements the comparative–historical-oriented approach pursued in this chapter, arriving at partially but not completely overlapping results.

2.2.2 The ergative hypothesis In the brief outline of the computational phylogenetic study by Carling and Cathcart (2021) given in the previous section, it was noted that the reconstruction of features and states that are unattested in the phylum is precluded in their model. Moreover, while the two approaches arrive at grossly similar results, an important question arising from the present investigation concerns the fact that PIE is reconstructed as a language with noun–class-based split alignment. From a general typological perspective, this problem does not seem to be very severe, since splitalignment systems are widespread and may to some extent be regarded as quite normal (see, e.g. Moravcik 1978 or Dahl 2021a for discussion and references). In a comparative–historical context, on the other hand, irregular and/or non-unitary features tend to be of special interest, since they often represent relics of a previous linguistic stage, in which they represented regularities, but it remains controversial whether this principle can be applied to syntactic reconstruction (see, e.g. Viti 2014 for critical discussion). In the present section we review a recent proposal based on internal reconstruction by Andreas Willi (Willi 2018), according

ALIGNMENT IN PROTO-INDO-EUROPE AN

53

to whom the nominative–accusative system found in (late) PIE developed from an ergative alignment system in pre-PIE. The idea that the late PIE system is secondarily derived from an erstwhile ergative system goes back to a seminal, very short paper by Christianus Cornelius Uhlenbeck (1901). From the observation that neuter nouns such as *iugóm ‘yoke’ ̑ (ved. yujám, gr. zugón) and *médh u ‘honey’ (ved. mádhu, gr. méth u) have identical forms for the nominative and accusative across the board, he inferred that the categories in question originally had neither nominative nor accusative function but a more general function which in turn developed into nominative and accusative (Uhlenbeck 1901: 170). On his analysis, this assumption finds support in the fact that thematic neuter nouns like *iugóm ‘yoke’ share the marker *-m with accusative ̑ singular forms of non-neuter nouns such as *uĺ̥ kȗ om ‘wolf ’ (ved. vŕ̥kam, gr. lúkon), ̑ which have a distinct nominative form *uĺ̥ kȗ os (ved. vŕ̥kas, gr. lúkos) with the ̑ characteristic nominative case marker *-s. Arguing that the suffix *-m cannot originally have been an accusative case marker, given that it is used as a nominative marker with thematic neuter nouns, he suggests (Uhlenbeck 1901: 170) that at some remote prehistoric stage PIE did not have nominative–accusative case marking but rather a type of marking where there was an agentive case (Activus), restricted to first argument function of bi- and trivalent verbs (A) and a nonagentive case (Passivus), denoting the other two core argument functions (S/P).³⁷ This is, in essence, the first formulation of the hypothesis that pre-PIE had an ergatively oriented alignment pattern. Various scholars have explored the explanatory potential of the ergative hypothesis since its first appearance c.120 years ago. For a thorough and clarifying discussion of previous literature concerned with this question, see Willi (2018: ch. 9). The original proposal and its corollaries have been subject to various types of criticism, the perhaps most central of which is that the distribution of case markers implied by Uhlenbeck’s (1901) model is the opposite of what one would expect from Silverstein’s (1976) well-known animacy scale. Specifically, as illustrated by the examples above, pre-PIE as conceived of by Uhlenbeck would seem to represent an alignment system with a marked ergative case marker *-s (e.g. *uĺ̥ kȗ os) and an unmarked (allomorphic) absolutive marker ̑ *-Ø (e.g. *médh u). Moreover, the presumed ergative case marker *-s appears with non-neuter nouns, a class that generally tends to include a greater share of referents with potential for appearing in A function than neuter nouns, in particular animate referents. Taken together, these two facts suggest that pre-PIE had a splitalignment system, where non-neuter nouns had ergative alignment whereas neuter nouns had neutral alignment. From Silverstein’s (1976) implicational hierarchy,

³⁷ It should be noted that Willi (2018) operates with a Dixonian definition of the analytical terms A, S, and O/P (see Chapter 1).

54

EYSTEIN DAHL

however, one would expect neuter nouns, which are less prone than non-neuters to denote animate referents, to show ergative marking as well. Consider the schematic representation in Figure 2.4 1st/2nd > 3rd/demonstratives > proper nouns > human > animate > inanimate Accusative Ergative

Fig. 2.4 Relative inherent animacy and preferred morphosyntactic alignment Source: adapted from Dahl & Stroński (2016: 6).

Figure 2.4 illustrates two important corollaries of the animacy hierarchy. First, higher-ranked types of NPs tend to have accusative alignment, while lower-ranked ones are more prone to have ergative alignment. Second, if a given NP has ergative alignment, all nominals to its right have ergative alignment. The ergative hypothesis, at least in its original form, would result in a reconstruction that would represent something of a typological anomaly. This, in turn, would run counter to the general principle that reconstructed stages of a language should be typologically plausible. However, Willi (2018: 511) rightly observes that the animacy hierarchy obtains in cases where a language shows split ergativity. In languages with ergative morphosyntax across the board or split ergativity based on tense/aspect/mood (TAM) distinctions, on the other hand, the animacy hierarchy does not play an equally pivotal role (Willi 2018: 514). This observation provides a simple and elegant solution to the problem raised above, namely that pre-PIE was a language with fully ergative alignment in the nominal and pronominal system (Willi 2018: 515). However, even textbook examples of languages with ergative case marking, such as Dyirbal, show a split between ergative marking with nouns and accusative marking with pronouns (Dixon 1994: 13–15, Dahl 2021a: 419–420). From this perspective, this version of the ergative hypothesis may turn out to be somewhat too radical and perhaps less typologically plausible than one could wish.³⁸ Nevertheless, it has a clear advantage over the original version of the hypothesis, reducing the typological implausibility considerably. Moreover, Willi’s (2018) model has the additional advantage that it provides an explicit diachronic scenario ³⁸ Indeed, one might be tempted to suggest a slightly altered version of the ergative hypothesis, according to which pre-PIE was a language with split accusative–ergative case marking, where nouns followed an ergative–absolutive pattern, while pronouns had a nominative–accusative inflection.

ALIGNMENT IN PROTO-INDO-EUROPE AN

55

of the development from the pre-PIE ergative case marking system to the PIE split accusative–neutral case-marking system. By this analysis, the original, consistently ergative case system of pre-PIE gradually develops into an accusative system, by extending the ergative case marker to the intransitive subject first among the pronouns, then among animate nouns, and then among inanimate nouns. However, in the case of inanimate nouns, a split arises between inanimate nouns that readily occur in first argument function of bi- and trivalent predicates, which develop distinct nominative and accusative forms, and inanimate nouns that do not readily appear in this function, which have syncretic forms for the nominative and accusative (Willi 2018: 515). This diachronic scenario is schematically represented in Figure 2.5 (adapted from Willi 2018: 515). Stage

Pronouns

Animate nouns Inanimate nouns

I

Aerg | SPabs

Aerg | SPabs

Aerg | SPabs

II

ASnom | Pacc

Aerg | SPabs

Aerg | SPabs

III

ASnom | Pacc

ASnom | Pacc

Aerg | SPabs

IV

ASnom | Pacc

ASnom | Pacc

ASnom | Pacc or ASPnom-acc

= ergative system

= accusative system

Fig. 2.5 Ergative → accusative alignment change and the animacy hierarchy Source: adapted from Willi (2017: 515).

These observations show that Willi’s (2018) diachronic scenario straightforwardly accounts for how the PIE split accusative–neutral case-marking system arose.³⁹ A further advantage of this set of assumptions is that it may also be able to explain the distributional restrictions on neuter nouns, which by hypothesis could mirror the fact that the neuter noun class arose from a lexically restricted subclass of inanimate nouns consisting of nouns that tended not to appear in first argument function of bivalent verbs. Specifically, the rise of a morphosyntactic distinction between neuter and non-neuter nouns and pronouns might be assumed to involve not only a morphological split in case marking but also a distributional split between nouns compatible with bivalent subject function, i.e. non-neuter nouns, and nouns incompatible with bivalent subject function, i.e. neuter nouns, rising from the strengthening of an erstwhile semantically motivated distributional restriction. ³⁹ At this point, it should be emphasized that Willi’s (2018) model provides a considerably more complex explanation, also involving a number of specific mechanisms of change based on a subtle analysis of the pre-PIE verbal system and its development to PIE. For reasons of space, only his analysis of case marking is taken into consideration here, since it seems most relevant for the issues dealt with in this chapter.

56

EYSTEIN DAHL

This brief outline of the ergative hypothesis shows that it provides an explanatory model that plausibly accounts for the rise of split accusative–neutral alignment in PIE, including two of the three peculiarities of neuter nouns, namely their neutral case-marking patterns and their exclusion from bivalent subject function. The background for the lack of agreement with neuter nouns will be addressed in Section 2.2.4. The next section examines some aspects of the hypothesis that pre-PIE had semantic alignment.

2.2.3 The semantic hypothesis According to another line of research, pre-PIE had a semantically oriented alignment system. As noted previously in this chapter, this hypothesis originates from publications from the 1940s, 1950s, and 1960s by Soviet scholars and became more generally known to western scholars from the early 1970s onwards. Important early contributions to this tradition include Gamkrelidze and Ivanov (1984) and Lehmann (1989a; 1989b). I refer to Willi (2018: 520– 532) for comprehensive, critical discussion of the various contributions to this field of research. In the present context, we shall briefly review the most relevant aspects of a more recent version of the hypothesis that pre-PIE had semantic alignment, as formulated in works such as Pooth et al. (2019) and Pooth and Orqueda (2021). Pooth and Orqueda (2021: 120) note that the accusative and ergative hypotheses are both based on the assumption that most verbs were lexically classified as inherently mono-, bi- or trivalent, a broadly accepted assumption in Indo-European linguistics. They argue that this view is unsatisfactory, since it is unclear how it can deal with labile verb forms, a phenomenon that seems to be particularly closely linked to middle voice, as suggested by data from Early Vedic and Old Hittite (Pooth and Orqueda 2021: 120). They draw the conclusion that ‘we should consider giving up the conception that the PIE verbal lexicon was dominated by valency and argument structures in the same way as in younger IE languages. We contend that the PIE verbal lexicon was rather dominated by agency and prototypical transitivity in semantic terms’ (Pooth and Orqueda 2021: 120). An important corollary of this analysis is that the pre-PIE voice system was organized as a privative opposition between a semantically specific proto-active voice and a semantically rather general proto-middle voice, which did not represent a valency-decreasing category in a strict sense but rather a more general detransitivizing category with valency-decreasing potential among other readings. This is illustrated in Table 2.12 (adapted from Pooth and Orqueda 2021: 120–121).

ALIGNMENT IN PROTO-INDO-EUROPE AN

57

Table 2.12 The pre-PIE voice system according to Pooth and Orqueda (2021) Proto-active

Proto-middle

*g ȗh én-m slay:1sg.active (agentive) I slew (it) (intentionally)

*g ȗh n-h2 e slay:1sg:detransitive I slew it non-intentionally I was slain (by s.o.) I slew myself I slew it for my own benefit, for myself I slew it (and it is/was my own) I made it slain (indirect causation) *mólh2 -h2 e slay:1sg:detransitive(:stative-habitual) I am grinding it I grind non-intentionally I am grinded (by s.o.) I grind myself I grind it for my own benefit, for myself I grind it (and it is my own) I make it grinded (indirect causation) →I grind it

*mélh2 -mi grind:1sg:activeprogressive I am grinding it

This model captures the idea that pre-PIE proto-active forms like *g ȗhénm or *mélh2 mi had a specifically agentive meaning, whereas proto-middle forms such as *g ȗhnh2 é or *mólh2 h2 e expressed various types of non-agentive semantics, determined by the specific context. The authors note that this type of voice system ‘matches those found in languages with semantic alignment’ (Pooth and Orqueda 2021: 121). Having established this, they go on to reconstruct the PIE case system. First, they suggest that neuter nouns were incompatible with the first argument function of bi- and trivalent verbs in PIE, based on the behaviour of bare neuter nouns in Hittite (Pooth and Orqueda 2021). This essentially implies that neuter nouns like *dóru ‘wood’ (ved. dā́ru, gr. dóru, hit. taru) were originally only able to express the monovalent subject function and the bivalent object function, their use in bivalent subject function representing a secondary development (Pooth and Orqueda 2021: 123 with references). They conclude that this implies a change in alignment type, at least as far as neuter nouns are concerned (Pooth and Orqueda 2021: 123). Second, they argue that the suffix *-m, which appears as an accusative marker with non-neuters and a nominative–accusative marker with neuters in the daughter languages, originally was a dative–allative case marker (Pooth and Orqueda 2021: 124). On their analysis, antipassive constructions containing proto-middle verb forms provided a context where neuter nouns such as *dóru could surface as a derived, agentive S argument with an oblique P-like argument like *h2 nér-m in the dative-allative case. Consider, by way of illustration, the reconstructed examples in (19) (adapted from Pooth and Orqueda 2021: 124, 142).

58 (19)

EYSTEIN DAHL a. *dóru *h2 nér-m *h1 ar-ó wood man-DAT/ALL hit-PROTOMIDDLE.3SG ‘wood (lance) hit at man’ b. *h2 nér *h1 ék̂ uo-m *g ȗh én-o ̑ man horse-DAT/ALL slay-PROTOMIDDLE.3SG ‘man is able/can/is expected to/may slay at horse’ c. *h2 nér-s *h1 ék̂ uo-m *g ȗh én-o ̑ man-AGENTIVE horse-DAT/ALL slay-PROTOMIDDLE.3SG ‘man is/was (willingly) hunting at horse/is was willing to slay horse, man is/was/will be slaying horse, hunting at horse’

Pooth and Orqueda (2021: 124) note that the construction in (19a), with a protoneuter noun in derived S function, may be expected to have been comparatively rare with proto-neuters in derived S function, a restriction motivated by the fact that such nouns were generally barred from A function. The construction in (19b), on the other hand, would have been more frequent, contrasting with the proto-active construction, characteristically showing a variety of contextually determined modal meanings, as indicated in the translation. Finally, Pooth and Orqueda (2021: 142) define a second variety of this antipassive-like construction, where the derived S argument is marked by the agentive suffix *-s, which they take to have had a range of more emphatically volitional set of possible interpretations, illustrated in (19c). This constructional variant would show closer competition with the proto-active construction and ultimately develop into a full-fledged active construction, according to Pooth and Orqueda (2021). This hypothesis implies that the antipassive construction of the type illustrated in (19) provided a bridging context for the reanalysis of the dative-allative marker *-m as an accusative marker. A third observation made by Pooth and Orqueda (2021: 124ff.) is that the agentive core case marker, *-s, strongly resembles the genitive–possessive marker *-s and the ablative-genitive marker *es/*-os. They rightly note that this type of formal overlap is reported on in languages showing ergative alignment, such as Inuktitut (Eskimo–Aleut), and seems to be less common in languages with accusative alignment (Pooth and Orqueda 2021: 24).⁴⁰ On their analysis, the alternating use of the agentive marker, illustrated in (19b) and (19c), would represent so-called optional ergative marking, a type of system where the presence or absence of an ergative/agentive marker is determined by semantic and/or pragmatic factors, for

⁴⁰ In his cross-linguistic study of agent markers, Palancar (2002: 229) found that 11 out of 65 ergative markers in his sample showed syncretism with a possessive marker. See also McGregor (2017: 449–452), who notes that instrumental, genitive, oblique, and ablative markers are most frequently suggested as sources for ergative case markers.

ALIGNMENT IN PROTO-INDO-EUROPE AN

59

instance the presence or absence of agentive features (McGregor 2009, 2010).⁴¹ An intriguing corollary of this part of the proposal is that it provides an elegant explanation of the relationship between nominative singular forms like *ph2 tḗr ‘father’ (ved. pitā́, gr. patḗr ) or *h1 ék̂ uo-s ‘horse’ (ved. ás´vas, gr. hı´ppos, lat. equus) vis̑ à-vis corresponding vocative singular forms like *ph2 ́ter (ved. pı´tar, gr. páter) or *h1 ék̂ ue (ved. ás´va, gr. hı´ppe, lat. equue). Under the present analysis, these forms ̑ would originate from the agentive forms *ph2 tér-s (> *ph2 térr > *ph2 tḗr by Szemerényi’s law) and *h1 ék̂ uo-s, and the non-agentive forms *ph2 ́ter and *h1 ék̂ ue, ̑ ̑ respectively, the non-agentive forms being reinterpreted as vocatives. It also provides a straightforward way to account for the allomorphs *-s/*-Ø in the feminine singular, which would reflect the same alternation/optional case-marking pattern. From this brief outline it should be clear that the semantic hypothesis has considerable explanatory potential, even if, as in the case of the ergative hypothesis as formulated by Willi (2018), we have only addressed some of the many intriguing perspectives it raises. From the perspective of the present chapter, the undoubtedly most important dimension of the semantic hypothesis is that it straightforwardly and explicitly accounts for the distributional restrictions of neuter nouns in PIE as reflecting a paradigmatic gap associated with such nouns in pre-PIE. In this respect, the semantic hypothesis has an indubitable advantage over the two other hypotheses discussed earlier, since it provides a typologically well-founded explanation for the case marking and restricted distribution of neuter nouns. In the following section, we shall address the third aberrant dimension of this class of nouns, the fact that neuter plural nouns in subject function appeared with singular verb forms in PIE.

2.2.4 Singular agreement with neuter plural subjects In Section 2.1, we saw that three of the most archaic, reasonably well-attested branches of Indo-European, namely Anatolian, Indo-Iranian, and Greek, provide comparative evidence favouring the assumption that neuter plural subjects did not trigger verb agreement in PIE, instead co-occurring with singular verb forms. Under this assumption, the languages where neuter plural subjects consistently or partly appear with plural verb forms, are innovative in this respect. In this section, we shall briefly review some assumptions that may contribute to explaining the lack of plural verb agreement with neuter plural subjects. The original function of the neuter plural nominative–accusative suffix, which in current Indo-European linguistics is standardly reconstructed as *-h2 ⁴¹ It should be noted that the term ‘optional’ in this context does not imply totally free variation; rather, it denotes a particular subtype of differential marking, with alternation between a phonologically overt morpheme and zero.

60

EYSTEIN DAHL

(see Table 2.1), remains somewhat disputed (see, e.g. Schmidt 1889, Neu 1969: 240, Eichner 1985, Melchert 2011a, Dahl 2021b). First, the suffix *h2 is also attested as an alternative plural ending with non-neuter nouns across the Indo-European family, as suggested by pairs of alternate plural forms such as alpēš ~ alpa from alpa- ‘cloud’ (m.) in Hittite, Gatha Avestan maṣ̌ iiā̊ŋhō ~ maṣ̌ iiā from maṣ̌ iia- ‘mortal, man’, kúkloi ~ kúkla from kúklos ‘wheel, circle’ in Greek and locī ~ loca from locus ‘place’ (m.) in Latin, where the forms in -a#/-ā# reflect the suffix *-h2 . Second, certain derivational feminine gender stems are formed with an identical suffix, as illustrated by examples such as *h1 ék̂ uo-h2 ‘mare’ (ved. ás´vā, lat. equa) to *h1 ék̂ uȏ ̑ ‘horse’, suggesting that the suffix *-h2 - did not necessarily encode plural number.⁴² In a pioneering study, Johannes Schmidt suggested that these forms were originally feminine singular forms with a collective value, a hypothesis largely based on the co-occurrence of neuter plural subjects with singular verb forms (1889: 8–9). The hypothesis that the plural function of this morpheme derives from an earlier collective function is generally accepted in Indo-European linguistics, but it remains disputed how this function relates to the feminine singular function, illustrated by *h1 ék̂ uo-h2 ‘mare’. This topic has engendered a vast literature, I refer to the papers ̑ in Neri and Schuhmann (2014) for a representative survey of different approaches and a rich bibliography. The following outline builds on Melchert’s (2011a, 2014) account. According to Melchert (2014: 257), PIE had a number system comprising four grammatical categories: singular, dual, count plural, and ‘collective’, which he regards as a set plural. By his analysis, a central difference between animate (~ non-neuter) and inanimate (~ neuter) nouns is that animate nouns were compatible with all the number categories, while inanimate nouns were incompatible with count plural as a grammatical category, which, however, does not imply that such nouns were inherently uncountable (Melchert 2014: 257, fn. 3). Table 2.13 gives a schematic representation of the PIE gender–number system according to Melchert (2014).

Table 2.13 Outline of the PIE gender–number system according to Melchert (2014)

⁴² See, e.g. Kim (2014) for discussion and references. He notes that examples like these are rather rare, most examples of this particular type ‘can be shown to continue substantivized feminine thematic adjectives in *-eh2 - or rather *-e-h2 -’ (2014: 119 with references).

ALIGNMENT IN PROTO-INDO-EUROPE AN

61

The defective number paradigm of inanimate/neuter nouns reflects the fact that gender and number were closely interrelated in PIE, animate/non-neuter nouns tending to be strongly associated with referential individuation (Melchert 2014: 258 with references). Furthermore, the PIE inflectional collective plural suffix *-h2 developed from a pre-PIE derivational suffix ‘that formed secondary exocentric concrete nouns referring to entities viewed as [+bounded, + internal structure] (i.e. non-mass and consisting of matching parts) such as *uorbh eh2 ‘enclosure’ ̑ (attested in the Hittite plurale tantum warpa)’ (Melchert 2014: 258). The same suffix was employed to form plural forms of mass nouns, as illustrated by examples such as PIE *uédōr ‘(bodies of ) water’ (hit. witār, gr. húdōr) from pre-PIE *uédoȓ ̑ h2 and PIE *h1 néh3 mōn ‘(set of ) names’ (got. namo, ved. nā́māni) from pre-PIE *h1 néh3 mon-h2 (see also Kim 2014: 115). The fact that the relationship between sg. *uédor and pl. *uédōr was no longer morphologically transparent in PIE strongly ̑ ̑ suggests that the plural function of the suffix *-h2 must be of pre-PIE age (Melchert 2011b: 396). Since these plurals are very archaic, their co-occurrence with singular verb forms does not suffice to show that they have developed from previous singular forms. On the other hand, in some languages different types of nouns show different agreement patterns in accordance with their position in the animacy hierarchy (see Section 2.2.2). For example, in the Western Malayo-Polynesian language Muna, plural pronouns and human plural nouns trigger plural verb agreement, while plural forms of inanimate nouns show singular verb agreement and nonhuman animates show alternation between these two patterns (Corbett 2000: 71). Thus, the co-occurrence of neuter plural subjects with third singular verb forms may be taken to reflect an archaic verb agreement rule, according to which inanimate nouns do not trigger verb agreement but select the unmarked, default verb form (see also Melchert 2011a: 396). However, while these observations provide a motivation for the lack of agreement with neuter plural subjects, the relationship between the plural marker *-h2 and the formally identical marker, later characteristic of feminine nouns and adjectives, remains mysterious. According to Melchert (2011a: 397), both originate from a derivational marker *-h2 , by which concrete nouns were derived from action nouns. These derived nouns could either be understood as [+bounded, internal structure] or as [+bounded, +internal structure]. In the first case, they were categorized as grammatically animate, selecting the (accusative) marker *-m when being used in bivalent object function. In the second case, they were classified as inanimate collective pluralia tantum, from which the suffix spread as a general collective plural marker (Melchert 2011a: 397–398). In this section, we have reviewed Melchert’s (2011a, 2014) account of the functions and the origin of the collective marker *-h2 . In the context of the present chapter, the most important dimension of this model is that it provides a convincing explanation for the fact that neuter plural subjects co-occurred with singular verb forms in PIE. Specifically, we have seen, that certain languages, like Muna,

62

EYSTEIN DAHL

display an animacy-based split agreement pattern, where inanimate nouns do not trigger verb agreement, and assume that a similar agreement rule was operational in pre-PIE and PIE as well. Here too, PIE neuter nouns seem to have maintained a morphosyntactic feature of great antiquity.

2.3 Conclusion Based on a comparison of data from Hittite, Vedic Sanskrit, Homeric Greek, and Early Latin, this chapter has argued that PIE had an NP-based split-alignment system, where non-neuter nouns showed consistently nominative–accusative alignment in case marking, verb agreement, and distribution over the core argument functions, whereas neuter nouns do not distinguish nominative and accusative case in their paradigm, do not trigger verb agreement and only occur in monovalent subject function and bivalent object function. We also briefly considered three hypotheses regarding the prehistory of the PIE alignment system, noting that the peculiar properties of neuter nouns are most easily accounted for as remnants from a prehistoric stage where pre-PIE had a semantic alignment system (Pooth and Orqueda 2021). The hypothesis that the pre-PIE alignment system was of the semantic type is perfectly compatible with the assumption that inanimate/neuter nouns were unable to trigger verb agreement, appearing with the default third singular form of the verb (Melchert 2011a, 2014). Under this analysis, neuter nouns have maintained some very archaic morphosyntactic features, which are gradually replaced with nominative–accusative features in the daughter languages, as discussed in Chapter 3 in this volume.

Text editions consulted Hom. Il.: Homerus, Ilias. Recensuit Martin L. West, Mu¨nchen: Saur, 1998–2000 (2.vol) Hom. Od.: Homerus, Odyssea. Recognovit P. von der Mu¨hll, Mu¨nchen: Saur, 2005 [1945] Plaut. As.: T. Maccius Plautus, Asinaria, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Bacch.: T. Maccius Plautus, Bacchides, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Cas.: T. Maccius Plautus, Casina, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Men.: T. Maccius Plautus, Menaechmi, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904.

ALIGNMENT IN PROTO-INDO-EUROPE AN

63

Plaut. Most.: T. Maccius Plautus, Mostellaria, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905. RV – Rigveda, Theodor Aufrecht (ed.) Die Hymnen des Rigveda. 2. Auflage. Bonn: A. Marcus, 1877 (2. vol.) Vend.: Vendīdād, in Karl F. Geldner (ed.) Avesta. Die heiligen Bu¨cher der Parsen. III Vendidad, Stuttgart: Kohlhammer, 1895. Vīsp.: Vīspered, in Karl F. Geldner (ed.) Avesta. Die heiligen Bu¨cher der Parsen. II Vispered und Khorde Avesta, Stuttgart: Kohlhammer, 1889. Y: Yasna, in Karl F. Geldner (ed.) Avesta. Die heiligen Bu¨cher der Parsen. Erster Teil: Yasna, Stuttgart: Kohlhammer, 1886.

3 Split alignment, mixed alignment, and the spread of accusative morphosyntax in some archaic Indo-European languages Paola Cotticelli and Eystein Dahl

This chapter explores how nominative–accusative alignment interacts with other alignment types, notably neutral alignment, semantic, i.e. Split-S or Fluid-S alignment and tripartite alignment including various patterns of non-canonical argument realization in Latin, Classical Greek, Vedic Sanskrit, Hittite, Classical Armenian, and Lithuanian. These languages represent different branches of the Indo-European linguistic family, and all show basic alignment patterns with a predominantly nominative–accusative orientation. However, in a more detailed perspective these languages show complex patterns of mixed alignment and nonnominative coding which contribute to considerable variation between them at a micro level. While the data we present here are representative rather than exhaustive, they suffice to illustrate how analogous processes of morphosyntactic change may have rather different results across languages belonging to the same linguistic family. The reconstructed alignment system of PIE outlined in Chapter 2 serves as a background for the analysis of the language-specific data dealt with in this chapter. An important aim of this chapter is to highlight the various developments that have contributed to shaping the alignment systems of each of the languages under consideration. In recent years, there has been a surge of interest in oblique or non-canonical subjects in Indo-European languages (see, e.g. Barðdal et al. 2012, 2020). An important set of controversial issues concerns whether and to what extent constructions of this type are inherited or not and, more generally, whether the morphosyntactic tests for subjecthood can be reconstructed to some prehistoric linguistic stage. These and related questions form a thematic thread through the present chapter. The chapter is organized as follows. Section 3.1 outlines the main theoretical assumptions on which the following discussion is based. Section 3.2 presents data from the various languages under consideration, with Section 3.2.1 dealing with Latin, Section 3.2.2 with Greek, Section 3.2.3 with Vedic Sanskrit, Section 3.2.4 with Hittite, Section 3.2.5 with Armenian, and Section 3.2.6 with Lithuanian. Paola Cotticelli and Eystein Dahl, Split alignment, mixed alignment, and the spread of accusative morphosyntax in some archaic Indo-European languages. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl. Oxford University Press. © Paola Cotticelli and Eystein Dahl (2022). DOI: 10.1093/oso/9780198857907.003.0003

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

65

Section 3.3 provides a discussion of the main findings of the previous section. Section 3.4 contains the summary and conclusion.

3.1 Alignment and argument marking: general considerations Alignment is a cover term for the ways in which core arguments of predicates can be morphosyntactically expressed across languages. As outlined in the introductory chapter, it is common to distinguish three analytical primitives, namely A, representing the first argument of two-argument clauses, P, the second argument of two-argument clauses, and S, the sole argument of one-argument clauses. Some more or less frequently recurring alignment types are schematically represented in Table 3.1. These definitions reflect that accusative alignment involves identical coding of A and S, while P has different coding, that ergative alignment obtains when S and P have identical expression, distinct from A, that neutral alignment makes no formal distinction between A, S, and P, that tripartite alignment has different expressions for A, S, and P and that double-oblique alignment obtains when A and P have the same coding, S having a different expression. At this point, we emphasize that accusative, ergative, and neutral patterns are considerably more frequently attested than tripartite and double-oblique ones. A fifth alignment type is semantic alignment, which comprises several subtypes, most notably split-S and fluid-S alignment.¹ As discussed in the introductory chapter, split-S alignment describes a situation where the encoding of S depends on the lexical semantic properties of the verbal predicate, selecting the same encoding as A with highly agentive predicates and that of P with highly patientive predicates. Fluid-S alignment, on the other hand, obtains when a group of predicates alternately select A- or P-encoding, depending on whether an agentive or non-agentive meaning is intended. Consider the examples in (1). Table 3.1 Some recurring alignment types Accusative

Ergative

Neutral

Tripartite

Double Oblique

A=S≠P

A≠S=P

A=S=P

A≠S≠P

A=P≠S

¹ At this point it should be noted that Dixon (1994: 28–35) employs the term ‘semantically based marking’ in a somewhat different manner from the term ‘semantic alignment’ as used in the present context, namely, to describe systems where morphosyntactic marking is directly determined by semantic roles rather than the general syntactic functions A, S, and P. As examples of this kind of system, he mentions Folopa, which belongs to the Teberan family of Papua New Guinea, and the Tibeto-Burman language Manipuri.

66

PAOL A COTTICELLI AND EYSTEIN DAHL

(1) a. vaĩkas šóka child:NOM.SG dance:PRS.3 ‘The child dances.’ [Lithuanian] b. skauda pilvą hurt:PRS.3 stomach:ACC.SG ‘(The/my) stomach hurts.’ [Lithuanian] c. (as) vuizˇnas 1SG.ERG fall:AOR.1SG.ERG ‘I fell down, on purpose.’ ([Tsova-Tush/Batsbi] adapted from Holisky 1987: 105) d. (so) vozˇensO 1SG.NOM fall:AOR.1SG.NOM ‘I fell down, by accident.’ ([Tsova-Tush/Batsbi] adapted from Holisky 1987: 105)² Examples (1a) and (1b) illustrate that in Lithuanian, which shows a predominantly accusative-based alignment system, many one-place verbs show the default nominative case-marking pattern, while some verbs of this type show accusative case marking, instantiating a lexically rather restricted pattern of split-S alignment. The examples in (1c) and (1d), on the other hand, show that Tsova-Tush/Batsbi, which belongs to the Nakh branch of the Northeast Caucasian languages, allows for both ergative and absolutive case marking and verbal cross-referencing with one and the same verb, thus instantiating a fluid-S pattern.³ As also discussed by Cennamo and Fabrizio (Chapter 6 in this volume), semantic alignment patterns often form subsystems in predominantly nominative–accusative or ergative-alignment systems (see also Rovai 2007a, 2012, Cennamo 2011). At present, we are mainly concerned with the formal morphosyntactic encoding of arguments, which involves two parameters, case marking or flagging and verb agreement or indexation. While there is a tendency across languages that these two morphosyntactic dimensions converge, cases where they instantiate different alignment types are not uncommon, and are usually referred to as mixed alignment, a term we adopt in the present work. As we argue in Sections 3.2.1 and 3.2.3, neuter nouns in Latin and Vedic Sanskrit instantiate mixed alignment, with neutral case marking and nominative–accusative agreement. As argued in Chapter 1, mixed alignment should be kept strictly apart from phenomena covered by the term ‘split alignment’. Split alignment involves systematic alternations between different alignment patterns in different morphosyntactic environments. Two common types of common splits may be distinguished. Some languages show tense-/aspect-based split alignment, as illustrated by the examples ² Note that capital O is used by Holisky (1987) to indicate final reduced vowels with unclear phonetic quality. ³ Note that in the tradition of Caucasian linguistics, the case used to mark S/P is called nominative, as reflected in the glossing of example (1d).

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

67

from Hindi in Chapter 1, where perfective and anterior forms show ergative alignment, while accusative alignment is found elsewhere. Another common type of split involves different types of nouns/pronouns and their relative inherent animacy, as reflected in Silverstein’s animacy hierarchy (see Chapter 2 in this volume for brief discussion). These two types of split tend to show complementary distribution across languages, a fact Dahl (2021a) takes to reflect that they develop along different diachronic paths.⁴ Both aspect- and person-based splits may involve flagging or indexation only, thus interacting in sometimes rather complex ways with mixed alignment. Consider, by way of illustration, the situation in Vedic Sanskrit, described in more detail in Section 3.2.3, where we find an NP-based split between two classes of nouns/pronouns. An important difference between these two classes of nouns and pronouns is that one, here preferentially labelled ‘non-neuter’,⁵ shows a systematic morphological distinction between nominative and accusative case forms, while the other, conventionally labelled ‘neuter’, has syncretic forms for these two cases. The examples in (2) and (3) illustrate this difference as reflected in the behaviour of masculine nouns such as páti- ‘lord, husband’ and neuter nouns such as havı´s- ‘oblation, offering’. (2) a. pátir vı´´svasya bhū́mano vy àkhyad lord:NOM.SG all:GEN.SG creation:GEN.SG PRV survey:AOR.3SG ródasī ubhé world.half:ACC.DU both:ACC.DU ‘A s lord of all creation, he has surveyed both world-halves.’ (RV IX 101.7 after Jamison and Brereton 2014: 1348 [Vedic]) b. dīrghā́yur asyā yáḥ pátir long.lived:NOM.SG 3SG.F.GEN.SG REL.M.ACC.SG husband:NOM.SG jī́vāti ´sarádaḥ ´satám live:PRS.SBJ.3SG autumn:ACC.PL hundred ‘Who as her husband will live, long-lived, through a hundred autumns.’ (RV X 85.39 after Jamison and Brereton 2014: 1524 [Vedic]) ́ c. jāyā pátiṃ vahati vagnúnā wife:NOM.SG husband:ACC.SG carry:PRS.3SG uproar:INS.SG ‘The wife conveys the husband amid the uproar.’ (RV X 32.3 after Jamison and Brereton 2014: 1427 [Vedic])

⁴ At this point, it should be emphasized that aspect-based and person-based splits in some cases occur in the same alignment system. This is, for instance, the case in Punjabi, where an aspect-based alignment split interacts with a person-based split, as discussed in Dahl (2021a). ⁵ At present, we employ ‘non-neuter’ as a cover term for ‘masculine/feminine’, since this distinction plays a subordinate role in the present context, as masculine and feminine nouns and pronouns have analogous morphosyntactic properties distinguishing them from neuter nouns. See Section 3.2.3 and Chapter 2 for discussion.

68

PAOL A COTTICELLI AND EYSTEIN DAHL

(3) a. ā́ asya havı´s tanvàḥ kā́mam PRV 3SG.M.GEN oblation:NOM/ACC.SG body:GEN.SG desire:ACC.SG r̥dhyāḥ satisfy:AOR.OPT.3SG ‘Might the oblation bring to fulfillment the desire of his body.’ (RV III 50.1 after Jamison and Brereton 2014: 533 [Vedic]) . b. yásya idám ápyam ˘ REL.M.SG.GEN here.and.now watery:NOM/ACC.SG havı´ḥ priyáṃ devéṣu gácchati oblation:NOM/ACC.SG dear:NOM/ACC.SG god:LOC.PL go:PRS.3SG ‘[W]hose own dear watery oblation goes here and now to the gods.’ (RV X 86.12 after Jamison and Brereton 2014: 1528 [Vedic]) c. agnı´r havı´ḥ ´samitā́ sūdayāti Agni:NOM oblation:NOM/ACC.SG ´samitar:NOM.SG sweeten:PRS.SBJ.3SG ‘Agni, the S´amitar, will sweeten the offering.’ (RV III 4.10 after Jamison and Brereton 2014: 474 [Vedic]) The examples in (2) illustrate that masculine nouns such as páti- ‘lord, husband’ have a nominative form pátis which appears in A and S function, and an accusative form pátim that appears in P function,⁶ while the examples cited in (3) show that neuter nouns like havı´s- employ the same form in all of these functions. As will be shown in Section 3.2.3, first- and second-person pronouns also show nominative–accusative alignment in Vedic. Thus, Vedic has an NP-based split between nominative–accusative and neutral alignment, neuter nouns and pronouns instantiating neutral alignment, while other nouns and pronouns show nominative–accusative alignment. As far as Vedic is concerned, this split only concerns case marking/flagging, since neuter nouns/pronouns show the same agreement patterns as non-neuter nouns/pronouns in this language. As noted previously, this implies that neuters instantiate mixed neutral/nominative–accusative alignment, with neutral case marking and nominative–accusative agreement.⁷ The previous discussion implicitly assumes that the terms A, S, and P represent syntactic functions rather than generalized semantic roles. In the following, we take a Comrian approach to these terms, assuming that they are primarily suited to serve comparative purposes (see Chapter 1 above). One reason for this is that they appear to be too coarse-grained to be useful in distinguishing between different verb classes and individual verbs in a language-specific perspective. As regards the scope of these terms, it seems reasonable to restrict the application of A and P to constructions appearing with bivalent verbs with a clear-cut transitive–causative meaning involving a change of state, such as ‘kill’, ‘break’, ‘cut’, while S might be ⁶ Note that the actually attested forms pátiṃ in (1c) and havíḥ in (2b–c) reflect sandhi variants of the underlying forms pátim and havís, respectively. ⁷ See the discussion of nominative–accusative and neutral agreement patterns in Chapter 2.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

69

defined on the basis of constructions selected by monovalent verbs denoting an uncontrolled change of state such as ‘die’, ‘rust’, ‘get lost’, ‘grow’, along the lines proposed by Haspelmath (2011). An indubitable advantage of this approach is that it firmly roots the typological terms A, S, and P in classes of predicates which tend to show unitary behaviour both within and across languages, in the sense that they characteristically select the default, unmarked argument realization pattern or patterns in a given language, and therefore constitute an ideal basis for cross-linguistic comparison. On the other hand, it accommodates the intuition that a comparative concept cannot fruitfully be employed as a descriptive device. Consider, by way of illustration, the examples in (4). (4) a. ngatu n̪an̪a kalkaŋa 1SG.ERG 3SG:ACC hit:PST ‘I hit her’ (adapted from Blake 1977: 11 [Wangkumara]) b. cui ego iam linguam praecidam REL.DAT.SG. 1SG.NOM ADV tongue:ACC.SG cut.off:PRS.SBJ.1SG ‘I shall immediately cut off her tongue.’ (Plaut. Aul. 189 [Latin]) c. si aliud nil sit, tui me, if other:NOM/ACC.SG nothing be:PRS.SBJ.3SG 2SG.GEN 1SG.ACC uxor, pudet wife:VOC.SG be.ashamed:PRS.3SG ‘If there were nothing else, I would be ashamed of you, (my) wife.’ (Plaut. As. 933 [Latin]) There is a clear sense in which the ergative-marked argument ngatu ‘I’ in the example from Wangkumara in (4a) can be fruitfully compared with the nominative-marked argument ego ‘I’ in the Latin example in (4b), in that they represent different morphosyntactic instantiations of the comparative category A. On the other hand, there is a clear sense in which the nominative-marked argument ego in (4b) and the accusative-marked argument me ‘me’ in the Latin example in (4c) can be fruitfully compared, both representing instantiations of the category subject in Latin (see Section 3.2.1). However, it is not clear to us that these two comparative dimensions can be meaningfully collapsed into one, either by characterizing all the pertinent arguments in (4) as instantiations of A or, for that matter, to define them as instantiations of a general or universal category of subject, without reducing the analytical power of these concepts significantly. In this chapter, we therefore rely on a two-tiered framework, involving, on one hand, a cross-linguistic dimension employing A, S, and P as comparative concepts on interlanguage level and a language-specific dimension employing notions such as subject and object on intra-language level, on the other. At this juncture, a brief reflection on the notions of subject and object is necessary. While these and similar terms are sometimes used without any further

70

PAOL A COTTICELLI AND EYSTEIN DAHL

qualifications, their theoretical status is all but controversial. Unfortunately, a review of the different definitions of these terms available in the literature and their respective merits and shortcomings is far beyond the scope of this chapter. Here, we adopt a Role and Reference Grammar-style (RRG) monostratal, functional, lexicalist approach, assuming that the meaning of verbs and other predicative elements condition argument functions and that predicates and arguments, as well as adjuncts, are mapped directly into the clause structure, labelled the Layered Structure of the Clause in RRG.⁸ Another feature of RRG which is crucial in the following is the postulation of two generalized semantic roles or ‘macroroles’, namely Actor and Undergoer. These notions are defined in terms of ‘generalizations across classes of specific argument positions in logical structure’ (Van Valin and LaPolla 1997: 142), which in turn are arrived at via predicate decomposition. With bivalent predicates, Actor and Undergoer are taken to represent opposite poles of a continuum, as it were, the Actor–Undergoer hierarchy, illustrated in Figure 3.1. ACTOR

Arg. of DO [

UNDERGOER

1sg arg. of

1st arg. of

2nd arg. of

Arg. of

do’(x, …

pred’ (x,y)

pred’ (x,y)

pred’ (x)

= increasing markedness of realization of argument as macrorole]

Fig. 3.1 Actor–Undergoer Hierarchy Source: after Van Valin (2005: 61). Note that the logical notations DO and do’ reflect the difference between a prototypical agent and less prototypical causers.

The Actor–Undergoer hierarchy reflects the insight that there is a correlation between the Actor and the most agent-like argument available and between the Undergoer and the most patient-like argument. This in turn highlights the central role played by the basic opposition between the cardinal arguments of core bivalent predicates, that is, predicates with meanings such as ‘kill’, ‘break’, ‘cut’ in argument realization. As regards monovalent predicates, the Actor–Undergoer hierarchy allows for distinguishing between one-place predicates with an Actor argument, corresponding to unergative verbs, and one-place predicates with an Undergoer argument, corresponding to unaccusative verbs. Taking the macroroles as our point of departure, we may now proceed to the formulation of an operative definition of notions like subject and object. Drawing on the discussion in Van Valin (2005: 89–107), we define the notion of subject in terms of generalized privileged syntactic argument, which in turn is anchored in the notion of restricted neutralization of argument functions. In the RRG ⁸ See Van Valin (2005) and Van Valin and LaPolla (1997) for more elaborate discussion of these questions.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

71

framework developed in Van Valin (2005), restricted neutralization of argument functions is understood as a construction-specific phenomenon, which singles out particular constellations of the macroroles. Consider the examples from Latin in (5). (5) a. Quia ego hanc amo et haec Because 1SG.NOM 3SG.F.ACC love:PRS.1SG CONJ 3SG.F.NOM me amat 1SG.ACC love:3SG.PRS ‘Because I love her and she loves me.’ (Pl. As. 631 [Latin]) b. nam cum pugnabant maxume, ego tum fugiebam CONJ when fight:IPF.3PL mightily:SPLV 1SG.NOM then flee:IPF.1SG maxume mightily:SPLV ‘For when they were fighting most mightily, then I was fleeing most mightily.’ (Plaut. Amph. 199 [Latin]) c. tum pol ego perii misera then by.Pollux 1SG.NOM be.ruined:PRF.1SG miserable:F.SG.NOM ‘Then, by Pollux, I am ruined.’ (Plaut. Merc. 510) d. ego (…) cum patrono disertissimo 1SG.NOM PRP advocate:ABL.SG eloquent:SPLV.ABL.SG comparer compare:PRS.1SG.MP ‘I am being compared with a most eloquent defence lawyer.’ (Cic. Quinct. 1.2 [Latin]) These examples illustrate that the Actor argument of bivalent verbs such as amo ‘love’ in (5a), the Actor argument of unergative monovalent verbs like fugio ‘flee’ in (5b), the Undergoer argument of unaccusative monovalent verbs like pereo ‘to be ruined’ in (5c) and the Undergoer argument of mediopassive forms of bivalent verbs like comparo ‘compare’ in (5d) behave identically in that they all trigger verb agreement in Latin. On the other hand, this neutralization does not include the Undergoer argument of bivalent verbs, as shown by example (5a), where the verb forms amo ‘love’ and amat ‘loves’ appear with the accusative forms hanc ‘her’ and me ‘me’, respectively, neither of which triggers verb agreement. Given that verb agreement involves neutralization of some but not all the core argument types, it is said to be restricted, picking out a subset of them, which together represent the privileged syntactic argument associated with verb agreement in Latin. As should be clear from the examples in (5), this amounts to saying that Latin has an agreement pattern instantiating nominative-accusative alignment, operating on an A/S basis. This illustrates that the approach developed here, where the notions A, S, and P are regarded as parameters for cross-linguistic comparison, while notions such as subject or object are language-specific, is

72

PAOL A COTTICELLI AND EYSTEIN DAHL

descriptively and analytically adequate. At this point, we would like to emphasize that the existence of one construction showing restricted neutralization in a given language is a necessary, but not sufficient condition for establishing that the notion of subject plays a role in that language within this framework. Van Valin (2005: 99) notes that many languages show the same privileged syntactic argument across constructions, and this observation underlies the definition of subject as generalized privileged syntactic argument. From a different perspective, Falk (2006: 12–16) makes a principled distinction between so-called same-subject and different-subject languages, based on the association of a range of subject properties across languages. The languages we are dealing with in this chapter all seem to show a high degree of internal consistency in this respect, thus representing same-subject languages. However, as we demonstrate below, this does not mean that the same set of constructions are equally relevant for delimiting the generalized privileged syntactic argument across the languages under consideration. Unfortunately, a full elaboration of these issues cannot be pursued here, but we trust that the evidence provided is sufficient to convince the reader, and hope to come back to this problem on a later occasion. Note that the RRG framework only postulates one syntactic function, namely the generalized privileged syntactic argument, and recognizes no syntactic functions corresponding to direct and/or indirect object. For present purposes it suffices to note that the traditional notion of direct object is essentially identical with the Undergoer argument of two-place predicates, which, in the languages under consideration here, also represents their non-privileged argument. In this work, we distinguish three core arguments, the first or most Actor-like argument of bivalent predicates, the second and most Undergoer-like argument of bivalent predicates, and the sole argument of monovalent predicates, which we sometimes refer to as ‘monovalent subject’ for convenience.⁹ Before concluding this section, a couple of further points need some elaboration. As noted in Chapter 1, previous work on alignment change has identified three mechanisms of change within this realm, namely extension of case-marking and/or agreement patterns, reanalysis of low-transitivity constructions, and borrowing. It was noted above that case marking and agreement represent independent dimensions of grammar, and that they may reflect different patterns, resulting in mixed alignment, but that they strongly tend to converge on one alignment type. Creissels’s (2018: 73) Obligatory Coding Principle represents an analogous tendency, the gradual extension of a particular type of case and/or agreement pattern to an increasing number of verbal predicative constructions in a language. Fol-

⁹ Since the first argument of bivalent predicates has essentially identical properties as the first argument of trivalent predicates, and the second argument of these two classes of predicates also do not show differences we deem relevant for present purposes, we only distinguish between monovalent and bivalent predicates for convenience in what follows.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

73

lowing Dahl (2021a), we name these two diachronic constraints ‘harmonization’ and ‘generalization’, respectively, and assume that they are pivotal as conditioning factors on analogical processes on a language-specific macro level. Drawing on Harris (1990) and Harris and Campbell (1995), we also assume that constraints with more restricted scope may have important impact on analogical processes, as formulated in the Complementarity Principle. In some cases, such local processes may be expected to operate independently of or even counteract more general patterns of change. We argue that this may turn out to be the case with so-called extended accusative patterns (Plank 1985), which is attested in several of the object languages of this study.

3.2 Alignment and argument marking in Indo-European This section comprises the main empirical part of our chapter and is structured as follows: Section 3.2.1 deals with Latin; Section 3.2.2 is devoted to Greek; Vedic Sanskrit is the topic of Section 3.2.3; Hittite is explored in Section 3.2.4; Section 3.2.5 discusses Armenian; Section 3.2.6 focuses on Lithuanian. Each of the sections gives a brief overview of the chronological stages of the language and a survey of its case system.

3.2.1 Latin Latin belongs to the Italic branch of Indo-European, being related to Faliscan, Oscan, Umbrian, and a number of other more or less fragmentarily attested languages. Table 3.2 outlines the diachronic stages of Latin.¹⁰ Table 3.2 The periods of the Latin language (adapted from Cuzzolin and Haverling 2009: 20) Archaic Latin Early Latin Classical Latin (‘Golden Age Latin’) Postclassical Latin (‘Silver Latin’) Late Latin

7th century BCE–c.240 BCE c.240 BCE–c.90 BCE c.90 BCE–14 CE 14 CE–c.200 CE c.200 CE–c.600 CE

Latin has a nominal system distinguishing six case categories and two numbers, as outlined in Table 3.3. ¹⁰ See Weiss (2009: 23–24) for a slightly different periodization of Latin.

74

PAOL A COTTICELLI AND EYSTEIN DAHL Table 3.3 The nominal case system in Latin

Nominative Vocative Accusative Genitive Dative Ablative

equus ‘horse’ (m.)

villa ‘house’ (f.)

bellum ‘war’ (n.)

Singular

Plural

Singular

Plural

Singular

Plural

equus eque equum equī equō equō

equī equī equōs equōrum equīs equīs

villa villa villam villae villae villā

villae villae villas villārum villīs villīs

bellum bellum bellum bellī bellō bellō

bella bella bella bellōrum bellīs bellīs

First, we may observe that there is a split between accusative and neutral alignment in the Latin nominal and pronominal system, masculine and feminine nouns (and pronouns) showing accusative alignment with distinct nominative and accusative case forms, and neuter nouns (and pronouns) showing neutral alignment, having syncretic forms for these two case categories. As argued in Chapter 2, this is an inherited feature of the Latin alignment system. As regards verbal agreement/indexation, on the other hand, Latin shows a predominantly nominative-accusative pattern. For instance, the examples in (6)–(8) illustrate that masculine, feminine, and neuter nouns all trigger plural number agreement on finite verbs. (6)

a. legem consules conscripserunt law:ACC.SG consul:NOM.PL pass:PRF.3PL ‘The consuls passed a law.’

(Cic. Att. 4.1.6)

b. sed postridie Sesti pueri venerunt but the.day.after Sestus:GEN.SG slave:NOM.PL come:PRF.3PL ‘But the following day Sestus’ slaves came.’ (Cic. Att. 3.17.1) (7)

a. quem etiam tuae superiores REL.ACC.SG now.also your:NOM.PL previous:NOM.PL litterae confirmarunt letter:NOM.PL strengthen:PRF.3PL ‘Which now also your previous letters have strengthened.’ (Cic. Fam. 5.21.3) b. tabernae mihi duae corruerunt shop:NOM.PL 1SG.DAT two:NOM.PL collapse:PRF.3PL ‘Two shops of mine have collapsed/two shops have collapsed for me.’ (Cic. Att. 14.9.1)

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX (8)

75

a. neque doctissimorum (…) praecepta te fugiunt NEG.CONJ learned:SPLV.GEN.PL precept:NOM.PL 2SG.ACC flee:PRS.3PL ‘And the precepts of the most learned men do not escape you.’ (Cic. Fam. 4.1.1) b. cum (…) tecta ardebunt when house:NOM.PL burn:FUT.3PL ‘When houses will burn.’ (Cic. Cat. 1.29)

Taken together, these considerations suggest that the Latin agreement system is organized on a consistently nominative–accusative basis and that there is a strong correlation between nominative case marking and verbal agreement. The fact that all noun types appear in all core argument functions and trigger verb agreement when appearing in subject function of monovalent and bivalent verbs shows that nominative–accusative alignment is predominant in Latin, warranting the view that neuter nouns have syncretic forms for the nominative and accusative and that the mixed alignment shown by this subclass of nouns is neutralized and accommodated to the general alignment pattern operating in the language. From a more general perspective, this implies that neuter nouns show mixed alignment, case morphology following a neutral pattern and verb agreement instantiating a nominative–accusative pattern. As we argue later, this reflects the fact that the Latin alignment system is in a relatively advanced stage of development vis-à-vis some of the other languages examined in this study (see also Chapter 2). We would also like to draw attention to the fact that the first argument of two-place predicates and the single argument of one-place predicates are consistently singled out as the privileged syntactic argument in several constructions in Latin, including being the (often unexpressed) addressee of second-person directives, as illustrated in (9a–b) and being omitted in so-called subject control constructions,¹¹ as illustrated in (9c–d). (9)

a. tu condicionem hanc accipe 2SG.NOM proposal:ACC.SG DEM.ACC.SG accept:PRS.IMP.2SG ‘You, accept this proposal’ (Plaut. Aul. 237) b. strenue curre in Piraeum quickly run:PRS.IMP.2SG PRP Piraeus:ACC ‘Run quickly to Piraeus.’

(Plaut. Trin. 1102–1103)

c. qui cudere soles REL.M.NOM.SG forge:PRS.INF use.to:PRS.2SG plumbeos nummos leaden:ACC.PL coins:ACC.PL ‘You, who are in the habit of forging leaden coins.’

(Plaut. Most. 892)

¹¹ We take control verbs to be more or less coextensive with what Pinkster (2015: 210) labels ‘auxiliary verbs’, that is, verbs obligatorily selecting a complement infinitive, the subject argument of which is co-referent with the subject of the matrix verb and, moreover, left unexpressed.

76

PAOL A COTTICELLI AND EYSTEIN DAHL d. anus hic solet cubitare old.woman:NOM.SG here use.to:PRS.3SG lie.down:PRS.INF ‘The old woman usually lies down here.’ (Plaut. Curc. 76)

These examples illustrate that both these constructions select the first argument of two-place verbs like accipio ‘accept’ and cudo ‘coin’ and the single argument of one-place verbs like curro ‘run’ and cubito ‘lie down’ as their privileged syntactic argument. Other morphosyntactic constructions picking out these two core arguments as syntactically privileged include the present participle, object raising and conjunction reduction (see Fabrizio forthcoming, Barðdal et al. 2020 for discussion).¹² At this point, we also note that Latin is a pro-drop language. In view of the above observations, it may seem paradoxical that Latin also seems to be more permissive with regard to non-canonical marking of core arguments than many other languages in this study. Restricting the present focus to the syntactically privileged argument in Latin, there are at least two types of nonnominative marking which may seem to represent alternative expressions of the syntactically privileged argument, namely accusative and dative (see, e.g. Fabrizio forthcoming, Barðdal et al. 2020). The examples in (10) illustrate each of these types. (10) a. quoius me nunc facti pudet REL.GEN.SG 1SG.ACC now deed:GEN.SG be.ashamed:PRS.3SG ‘Which deed I am now ashamed of.’ (Plaut. Bacch. 1016) b. adulescentes, haec si vobis lex young.man:VOC.PL DEM.NOM.SG if 2PL.DAT law:NOM.SG placet please:PRS.3SG ‘Young men, if this law pleases you (…)’ (Plaut. Merc. 1025) These examples illustrate that certain experiencer verbs, such as pudeo ‘be ashamed’ in (10a) regularly select an accusative-marked experiencer argument in Latin (me), whereas others, like placeo ‘please’ (10b) obligatorily select a dativemarked experiencer argument (vobis). The stimulus argument, on the other hand, is expressed by the genitive case (quoius facti) in (10a) and by the nominative case (haec lex) in (10b). We refer to Cennamo and Fabrizio’s contribution to the present volume (Chapter 6) for a thorough exploration of the complex patterns of argument alternation and non-canonical argument marking in Latin. At present, it should be noted that the case frame represented in (10a) is restricted to a closed, ¹² Additional constructions which are regularly mentioned among the putative subject tests in Latin are the property of being the antecedent in clause-bound and long-distance reflexivization (see Fabrizio forthcoming, Barðdal et al. 2020 for discussion and examples). Given that these two morphosyntactic patterns do not single out the first argument of bivalent verbs and the sole argument of monovalent verbs but are also, albeit marginally, compatible with the second argument of bivalent verbs as well as adjunct agent phrases, they are excluded from the present discussion.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

77

well-defined group of experiential predicates,¹³ whereas the one illustrated in (10b) seems to be considerably more inclusive. When faced with non-canonical argument realizations such as the ones illustrated in (10), the question arises as to which of the arguments is the syntactically privileged one. Fabrizio (forthcoming) explores the behavioural patterns of accusative-marked and dative-marked experiencer arguments, observing that accusative-marked experiencers like me in (10a) are singled out as the privileged syntactic argument across a considerable number of constructions in Latin,¹⁴ whereas dative-marked experiencers like vobis in (10b) only fall within this category with regard to two constructions.¹⁵ From this perspective, one is tempted to conclude that accusative-marked experiencers fall within the category of generalized syntactically privileged argument, representing bona fide non-canonical subjects, whereas dative-marked experiencers do not. From Fabrizio’s (forthcoming) discussion of the pertinent data it appears that it is the nominative-marked stimulus argument that has privileged syntactic status with verbs such as placeo (see, however, Barðdal et al. 2020 for a different view). The above considerations constitute strong evidence in favour of the assumption that Latin had a class of clear-cut accusative-marked subjects. Moreover, at least as far as Latin goes, there is only very scarce evidence for an inherited dative subject construction. This is, however, not to deny that there are significant semantic correspondences between predicates occurring with dative-marked nominal and pronominal constituents in Latin and in other archaic Indo-European languages, as demonstrated by Barðdal et al. (2012). As regards the Latin accusative subject construction, on the other hand, some further considerations are in order. First, mediopassive verb forms sporadically co-occur with an argument-like accusative-marked noun or pronoun, as illustrated in (11). (11) a. incerte errat animus; praeterpropter doubtfully wander:PRS.3SG mind:NOM.SG more.or.less vitam vivitur life:ACC.SG live:PRS.3SG.MID ‘The mind wanders doubtfully; one lives life more or less.’ (Enn. Sc. 241 V5 /202 Jocelyn, after Goldberg and Manuwald 2018: 88) b. pessumis me modis despicatur domi worst:ABL.PL 1SG.ACC way:ABL.PL despise:PRS.3SG.MP home:LOC ‘I am despised in the worst ways at home.’ (Plaut. Cas. 185, Cennamo 2011: 178) ¹³ These predicates are pudeo ‘be ashamed’, pigeo ‘be offended’, taedeo ‘regret’, paeniteo ‘repent’, and misereo ‘pity’ as well as their derivatives and compounds. ¹⁴ These include being the addressee of second-person imperatives, the target of the present participle, the omitted argument in control constructions, conjunction reduction, and raising to object, as well as reflexivization, which, however, does not pick out the same set of arguments as the other constructions. ¹⁵ These are conjunction reduction and reflexivization.

78

PAOL A COTTICELLI AND EYSTEIN DAHL c vixi annos bis centum; live:PRF.1SG year:ACC.PL twice hundred nunc tertia vivitur aetas now third:NOM.SG live:PRS.3SG.MID age:NOM.SG ‘I have lived through hundred years twice; now the third age is being lived.’ (Ov. Met. 12.187–188)

Examples (11a–b) illustrate that mediopassive, i.e. (derived) monovalent verb forms such as vivitur ‘be lived’ and despicatur ‘be despised’ occasionally appear with an accusative-marked core argument, like vitam ‘life’ in (11a) and me ‘me’ in (11b). Interestingly, however, example (11c) shows that the form vivitur may also appear with a nominative-marked core argument, such as tertia aetas ‘third age’.¹⁶ Second, certain nouns seemingly vacillate between masculine and neuter, as shown by the examples in (12a-c). (12) a. iam tibi (…) crassus corius already 2SG.DAT thick:NOM.SG skin:NOM.SG redditu=st give.back:PPP.NOM.SG=be:PRS.3SG ‘The thick skin has already been returned to you.’ (Plaut. Fr. 122 L; Cennamo 2009: 313) b. cum legeres, si unam peccavisses when read:IPF.SBJ.2SG if one:ACC.SG miss:PLPF.SBJ.2SG syllabam fieret corium tam syllable:ACC become:IPF.SBJ.3SG skin:NOM/ACC.SG so maculosum quam est nutricis spotted:NOM/ACC.SG as be:PRS.3SG nurse:GEN.SG pallium gown:NOM/ACC.SG ‘If you, when you are reading, should miss one syllable, (your) skin would be as spotted as a nurse’s gown is.’ (Plaut. Bacch. 434, see also Cennamo 2009: 313) c. heri in tergo meo tris facile yesterday in back:ABL.SG my:ABL.SG three easily corios contrivisti bubulos skin:ACC.PL wear.out:PRF.2SG of.oxen:ACC.PL ‘Yesterday you easily wore out three ox hides (i.e. whips of oxhide) on my back.’ (Plaut. Poen. 138–139 after De Melo 2012: 31) ¹⁶ A critical reader might object that the accusative form vitam in example (11a) could also represent a so-called internal accusative, denoting a cognate object. However, as illustrated by (11c), cognate objects are marked by the accusative in active voice constructions like vixi annos bis centum and with the nominative in mediopassive voice constructions like tertia vivitur aetas. Pinkster (2015: 1214– 2115) notes that ‘[p]assivization is generally speaking a good proof for subjecthood’ in Latin.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

79

Here, we see that the noun under scrutiny shows alternating appearance in core argument function of (basic and derived) monotransitive verb forms. It appears in the masculine nominative singular form corius in (12a), and in the neuter nominative/accusative singular form corium in (12b), which, formally speaking, could also represent an accusative form of a masculine noun. Given the data in (11), where clear-cut accusative forms appear with (derived) monotransitive verb forms, it seems reasonable to regard the form corium as a bona fide accusative form of a masculine noun corius. This interpretation is also suggested by the example cited in (12c), where a masculine accusative plural form of the same noun is attested.¹⁷ A number of scholars (e.g. Rovai 2007a, 2012, Cennamo 2011) regard examples like those cited in (11) and (12) as reflecting a semantically aligned subsystem in Early Latin. We adopt this approach here and note that the data considered above indicate that this system was of the fluid-S type. Tikkanen’s (2011) survey of the Sabellic case system does not mention any examples of analogous case-marking patterns in the other attested Italic languages, a fact that might easily reflect their scarce attestation. We therefore postpone an assessment of the origin of the Latin accusative subject construction until we can draw on comparative evidence from the other Indo-European languages under scrutiny. Finally, a case could be made for the claim that Latin has at least one productive verbal construction that instantiates an ergative-like alignment pattern, or, put otherwise, picks out the Undergoer argument of bivalent predicates and the only argument of monovalent predicates as their privileged syntactic argument. This is the perfect participle, as illustrated in (13). (13) a. ego autem a patre eram 1SG.NOM CONJ PRP father:ABL.SG be.IPF.1SG deductus ad Scaevolam introduce:PRF.PTCP.NOM.SG PRP Scaevola:ACC ‘I had already been introduced by my father to Scaevola.’ (Cic. Amic. 1) b. postriduo natus sum ego the.day.after be.born:PRF.PTCP.NOM.SG be:PRS.1SG 1SG.NOM ‘The next day I was born.’ (Plaut. Mil. 1082) c. quas senatus sine ulla varietate REL.ACC.PL senate:NOM.SG without any:ABL.SG variety:ABL.SG est secutus be:3SG follow:PRF.PTCP.NOM.SG ‘Which the senate has followed without any exception.’ (Cic. Cat. 3.6)

¹⁷ For the sake of completeness, it should be mentioned that masculine forms of this noun only appear in Early Latin; in later stages of the language, the noun is consistently neuter.

80

PAOL A COTTICELLI AND EYSTEIN DAHL

Here, we observe that the perfect participle forms deductus ‘introduced’ is oriented towards the undergoer argument of the base verb deducō ‘introduce’, whereas natus ‘born’ selects an undergoer-like argument of the base verb nascor ‘be born’. As will be discussed in the following sections, the Latin perfect participle originates from an originally P-oriented resultative verbal adjective, which was restricted to Undergoer arguments of bivalent and monovalent predicates. In Latin, it is included in the system of voice oppositions, being, among other things, employed in analytic mediopassive constructions, as illustrated by the examples in (13). It should also be noted that examples like the one cited in (13b) mainly appear with so-called deponent verbs in Latin, that is, verbs exclusively or mainly showing mediopassive forms. Interestingly, example (13c) illustrates that some deponent verbs, such as sequor ‘follow’ denote a situation with two participants, and with such verbs, the passive perfect participle picks out the first argument, a fact indicating that this construction shows unrestricted neutralization of the three core argument functions. This in turn indicates that the perfect participle has a neutral rather than ergative-like orientation in Latin. In any case, this construction is rather isolated in the predominantly accusatively oriented alignment system of Latin, being restricted to valency-reducing functions. In this section we have seen that the Latin alignment system shows a predominantly accusative-based structure and that a restricted class of predicates select non-canonical subjects with accusative case marking. In Early Latin, there is some evidence of fluid-S marking, involving an alternation between nominative and accusative case marking. It was also briefly noted that the Latin perfect participle has an ergative-like orientation, which, however, plays a marginal role in the overall alignment system.

3.2.2 Ancient Greek Greek is generally taken to represent an independent branch of the Indo-European family.¹⁸ Ancient Greek is a cover term for the oldest stages of Greek, delimiting a period from the fifteenth century BCE to the fourth century CE. Table 3.4 outlines the diachronic stages of Ancient Greek. Table 3.4 Stages of Ancient Greek Mycenaean Archaic Greek Classical Greek Hellenistic Greek

15th century BCE–12th century BCE 8th century BCE– c.480 BCE c.480 BCE–323 BCE 323 BCE–4th century CE

¹⁸ See, however, the discussion of the affiliation of Armenian at the beginning of Section 3.2.5.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

81

The Ancient Greek nominal system distinguishes five case categories and three numbers, as outlined in Table 3.5.

Table 3.5 The Greek nominal case system khṓrā ‘land’ (f.)

híppos ‘horse’ (m.)

Nom Voc Acc Gen Dat

dôron ‘gift’ (n.)

Singular Dual

Plural

Singular Dual

Plural

Singular Dual

Plural

hippos híppe híppon híppou híppōi

híppoi híppoi híppous híppōn híppois

khṓrā khṓrā khṓrān khṓrās khṓrāi

khṓrai khṓrai khṓrās khṓrōn khṓrais

dṓron dṓron dṓron dṓrou dṓrōi

dṓra dṓra dṓra dṓōn dṓrois

híppō híppō híppō híppoin híppoin

khṓrā khṓrā khṓrā khṓrain khṓrain

dṓrō dṓrō dṓrō dṓroin dṓroin

As indicated by the data in Table 3.5, the Ancient Greek case system shows a split between accusative and neutral alignment, neutral case marking being associated with neuter nouns and pronouns, as also shown to be the case in Latin in the previous section. In the following, we mainly deal with data from Homeric Greek. The examples in (14) and (15) illustrate the use of masculine and feminine nouns in core argument functions in Ancient Greek. (14) a. hoppóte koûroi erı´sseian perì múthōn whenever boy:NOM.PL strive:AOR.OPT.3PL PRV word:GEN.PL ‘Whenever the young men contended about words (…)’ (Hom. Il. 15.284) b. koûroi mèn krētêras epestépsanto potoîo boy:NOM.PL PTCL bowl:ACC.PL fill:AOR.3PL.MID drink:GEN.SG ‘The servants filled the bowls with drink.’ (Hom. Il. 1.470) c. zōoùs ek potamoîo duṓdeka léksato living:ACC.PL PRP river:GEN.SG twelve choose:AOR.3SG.MID koúrous boy:ACC.PL ‘He chose twelve young men alive from the river.’ (Hom. Il. 21.27) (15) a. hṑs éphato klaı´ous’ epì thus speak:AOR.3SG.MID crying:NOM.SG PRV PTCL dè stenákhonto gunaîkes PTCL groan:IPF.3PL.MID woman:NOM.PL ‘Thus she spoke, crying. The women groaned in answer.’ (Hom. Il. 19.301)

82

PAOL A COTTICELLI AND EYSTEIN DAHL b. toùs d’ állous moi étikton this:ACC.PL PTCL other:ACC.PL 1SG.DAT bring.forth:IPF.3PL enì megároisi gunaîkes PRP palace:DAT.PL woman:NOM.PL ‘The others the women in the palace brought forth for me.’ (Hom. Il. 24.497) c. dṓsō d’ heptà gunaîkas amúmona érga give:FUT.1SG PTCL seven woman:ACC.PL noble:ACC.PL work:ACC.PL iduìas skilled:ACC.PL ‘And I shall give seven women skilled in noble handwork.’ (Hom. Il. 9.128)

These examples illustrate that plural forms of masculine and feminine nouns systematically trigger plural agreement on the verb. Moreover, they suffice to show that verb agreement follows a nominative–accusative pattern with masculine and feminine nouns and pronouns in Ancient Greek. This is the case with first- and second-person pronouns as well. In other words, verb agreement selects the first argument of bivalent verbs and the only argument of monovalent verbs as its syntactically privileged argument with masculine and feminine nouns and pronouns as well as speech act participant pronouns. Ancient Greek, like Latin, is a pro-drop language. Neuter nouns and pronouns have syncretic forms for the nominative and accusative in Ancient Greek, a distinction that plays an important role elsewhere in the alignment system of Ancient Greek. Consider, by way of illustration, the examples in (16), where the noun pûr ‘fire’ appears in all core argument functions. (16) a. méga d’ ´ıakhe thespidaès pûr great:NOM.SG PTCL shriek:IPF.3SG god.kindled:NOM.SG fire:NOM.SG ‘And the great god-kindled fire roared.’ (Hom. Il. 23.216) b. ēu΅te pûr aı΅dēlon epiphlégei as fire:NOM.SG annihilating:NOM.SG burn.down:PRS.3SG áspeton húlēn immense:ACC.SG forest:ACC.SG ‘A s an annihilating fire burns down an immense forest.’ (Hom. Il. 2.455) c. pûr dè Menoitiádēs daîen fire:ACC.SG PTCL son.of.Menoetius:NOM.SG kindle:IPF.3SG mega isótheos fṓs great:ACC.SG godlike:NOM.SG man:NOM.SG ‘And the son of Menoetius kindled a great fire, the godlike man.’ (Hom. Il. 9.211)

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

83

These examples suffice to illustrate the neutral alignment pattern of neuter nouns and pronouns in Ancient Greek. As regards verb agreement, we note that plural forms of neuter nouns generally do not consistently trigger plural verb agreement, unlike masculine and feminine nouns. Consider the examples in (17). (17) a. tòn d’ oú pote kúmata leı´pei 3SG.ACC PTCL NEG ever waves:NOM.PL leave:3SG.PRS ‘The waves never leave it.’ (Hom. Il. 2.396) ̀ b. kaì dē doûra sésēpe neôn and PTCL plank:NOM.PL rot:PRF.3SG ship:GEN.PL kaì spárta léluntai rope:NOM.PL loosen:PRF.3PL.MID CONJ ‘The planks of the ship are rotten, and the ropes are untied.’ (Hom. Il. 2.135) These examples illustrate that plural forms of neuter nouns in subject function (kúmata, doûra, spárta) alternately co-occur with singular (leı´pei, sésēpe) and plural verb (léluntai) forms in Homeric Greek. As noted by Dahl (2021b; Chapter 2 in this volume), examples of plural neuter nouns with plural verb forms amount to slightly less than one-third of the total number of plural neuter nouns in subject function in the Homeric epics. As argued by Dahl (Chapter 2), the pattern where plural neuter nouns co-occur with singular verb forms seems to be more archaic, reflecting a stage where these noun forms had a collective function (see also Melchert 2011a; 2014 for discussion). Within the synchronic alignment system of Ancient Greek, however, their co-occurrence with the paradigmatically unmarked third singular form of the verb may be regarded as an instantiation of neutral agreement. In a synchronic perspective, a case could be made for the claim that neuter nouns generally show a consistently neutral alignment pattern in Ancient Greek, and we therefore preliminarily conclude that there is a nounbased split between accusative and neutral alignment in this language. According to this approach, examples like spárta léluntai in (17b), where plural forms of neuter nouns trigger are suggestive of an incipient alignment change from a consistently neutral alignment pattern to a mixed pattern of the type we have seen in Latin. We conclude that Greek neuter nouns differ from other noun classes in that they tend not to trigger verb agreement when used with plural morphology. As discussed in Dahl (2021b, Chapter 2 in this volume), the alternation between singular and plural number agreement with plural forms of neuter nouns is characteristic for Archaic Greek. In Classical Greek, the singular agreement pattern prevails. On the assumption that the singular agreement pattern represents a default choice where the absence of plural agreement marking amounts to neutral alignment, the minority pattern with plural neuter nouns/pronouns and plural verb forms represents a mixed neutral/accusative pattern. From this perspective, the development from Archaic to Classical Greek may be reinterpreted as a change where

84

PAOL A COTTICELLI AND EYSTEIN DAHL

the competition between a mixed neutral/accusative and a consistently neutral alignment pattern is resolved via harmonization, the tendency for constructions to converge on one alignment type (see also Dahl 2021a, and Chapter 1 in this volume). In this specific case, then, harmonization outranks generalization, as it were, the change resulting in a noun–class-based split between masculine and feminine nouns and pronouns showing consistent accusative alignment, and neuter nouns and pronouns showing consistent neutral alignment. We now turn to a brief outline of some other morphosyntactic behaviour patterns that are relevant for delimiting the generalized privileged syntactic argument in Ancient Greek. The example in (18a) illustrate that only the first argument of bivalent predicates and the sole argument of monovalent predicates can be addressees of second-person directives. The examples in (18b–c) show that control predicates pick out the very same core arguments. (18) a. all’ ´ıthi mḗ m’ eréthize but go:PRS.IMP.2SG PROH 1SG.ACC provoke:PRS.IMP.2SG ‘But go! Do not provoke me!’ (Hom. Il. 1.32) b. Kúklops, ouk ar’ émelles Kyklops:VOC.SG NEG PTCL be.destined:IPF.2SG análkidos andròs hetaı´rous feeble:GEN.SG man:GEN.SG companion:ACC.PL édmenai en spêi glamphurôi kraterêphi eat:PRS.INF PRP cavern:DAT.SG hollow:DAT.SG brutal:DAT.SG bı´ēphi force:DAT.SG ‘Kyklops, not a feeble man’s companions were you destined to eat in the hollow cave with brutal force.’ (Hom. Od. 9.475–476) c. kaì gàr egṓ pot’ émellon en CONJ for 1SG.NOM once be.destined:IPF.1SG PRP andrásin ólbios eînai man:DAT.PL rich:NOM.SG be:PRS.INF ‘For once I was also destined to be a rich man among men.’ (Hom. Od. 18.138) Example (18a) illustrates that second-person imperative forms like ´ıthi ‘go!’ and eréthize ‘provoke!’ pick out the first argument of bivalent verbs and the sole argument of monovalent verbs. Examples (18b–c) show that control infinitives like the ones associated with the verb méllō ‘be destined’ show gapping of the very same core arguments, the undergoer argument of bivalent predicates being explicitly expressed, like hetaı´rous ‘companions’ in (18b). Some predicates select potential non-canonically case-marked subjects in Ancient Greek, some examples of which are given in (19).

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX (19) a. kaì Zēnì tôn sôn, (…), CONJ Zeus:DAT ART:GEN.PL 2SG.POSS.GEN.PL mélei pónōn be.an.object.of.care:PRS3SG trouble:GEN.PL ‘And Zeus cares about your troubles.’

85

(Eur. Her. 717)

b. soı´ te gàr paı´dōn tı´ deî 2SG.DAT.SG PTCL PTCL child:GEN:PL why need:PRS.3SG ‘But why do you need (more) children?’ (Eur. Med. 565) These examples illustrate that a small number of bivalent predicates select a dative–genitive case frame in Classical Greek, illustrated by the dative forms Zēnì ‘Zeus’ and soı´ ‘you’ and the genitive forms sôn pônōn ‘your troubles’ and paı´dōn ‘children’. Attention should also be paid to the fact that the verb usually occurs in the unmarked, third singular form in such cases. Conti (2010) identifies four predicates that select this case frame, namely mélō ‘be an object of care’, metamélō ‘repent’, deō (déomai/déuomai) ‘need’, and elleı´pō ‘lack’. Some of the pertinent predicates show alternation between dative and accusative marking of one of the arguments, as illustrated by example (19c), where we find the accusative form me ‘me’ corresponding to soı´ in (19b). (19) c. tês túkhēs me deî mónon ART:GEN.SG fortune:GEN.SG 1SG.ACC need:PRS.3SG only ‘Now I only need (good) luck.’ (Eur. Hel. 1424) Drawing on the model outlined in Cole et al. (1980), Conti (2010) explores the behaviour of the two arguments of this group of predicates, concluding that the classical syntactic behavioural properties are more or less equally distributed among them (Conti 2010: 268).¹⁹ Interestingly, she notes that only elision with complement infinitives and agreement in number and person represent behavioural properties that are exclusive of subjects in Ancient Greek (Conti and Luraghi 2014: 454), leading her to the conclusion that the subject function is syntactically somewhat underspecified in this language (Conti 2010: 269). This situation is quite different from that seen in Latin, as should be clear from the discussion in Section 3.2.1. From a diachronic point of view, it should be noted that constructions like those illustrated in (19) do not occur in the Homeric texts. For example, the verb mélō ‘be an object of care’ consistently selects a nominative–dative case frame, as illustrated in (20a). Example (20b) shows that the nominative-marked argument is the target of the imperative, not the dative-marked one referring to the second person.

¹⁹ These include conjunction reduction, binding of reflexive pronouns, and possible ellipsis in raising to object or AcI clauses involving a series of infinitives.

86

PAOL A COTTICELLI AND EYSTEIN DAHL

(20) a. pólemos d’ ándressi melē ́ sei paˆsi war:NOM.SG PTCL man:DAT.PL be.an.object.of.care:FUT.3SG all:DAT.PL málista d’ emoı´ most PTCL 1SG.DAT ‘And warfare shall concern all men, and most of all me.’ (Hom. Il. 6.492–493) b. soì d’ autôi melétō 2SG.DAT PTCL REFL.DAT.SG be.an.object.of.care:PRS.IMP.3SG hekatēbóle phaı´dimos Héktōr far.darting:VOC.SG glorious:NOM.SG Hector:NOM.SG ‘Let glorious Hector be an object of care for yourself, O far-darting one (i.e. Apollo).’ (Hom. Il. 15.231) We take data like these as strong prima facie evidence that the dative–genitive case frame illustrated in (19) represents an inner-Greek innovation, an assumption that is corroborated by the fact that no parallel constructions exist in other, related languages.²⁰ It possibly arose via influence from another construction with genitive-marked experiencer, illustrated in (21). (21) a. autôn gàr sphetárēisin atasthalı´ēisin ólonto DEM.GEN.PL PTCL own:DAT.PL wickedness:DAT.PL perish:AOR.3PL.MID ‘They (viz. Odysseus’ comrades) perished, indeed, through their own blind folly.’ (Hom. Od. 1.7–9 after Conti and Luraghi 2014: 455) b. en khṓrai épipton hekatérōn PRP place:DAT fall:IPF.3PL both:GEN.PL ‘Men of both sides fell in their position.’ (Xen. Hell. 42.20 after Conti and Luraghi 2014: 458) According to Conti and Luraghi (2014: 457), genitive-marked subject arguments generally tend to appear with monovalent predicates with a patient- or theme-like argument, as illustrated by the examples in (21). An interesting feature of these and similar examples in Ancient Greek is that the genitive-marked arguments in the plural (autôn, hekatérōn) co-occur with plural verb forms (ólonto, épipton). Furthermore, they observe that the exceptions to this general rule are comparably few and typically involve existential sentences, where also nominative-marked arguments may show lack of agreement, a fact taken to reflect that genitive-marked arguments have been grammaticalized as non-canonical subjects in this particular function (Conti and Luraghi 2014: 455–456). The general ability to trigger verb agreement is restricted to the use of the genitive as marker of subjects of unaccusative verbs, as shown by examples like those cited in (19), where genitive plural forms co-occur with unmarked, third singular verb forms. ²⁰ See Conti and Luraghi (2014: 62) for discussion and an alternative interpretation of the Ancient Greek data.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

87

We may now summarize the main points made in this section. Ancient Greek has a case system with a split between accusative and neutral case marking, neuter nouns and pronouns consistently showing syncretic forms for the nominative and accusative case, while other nouns and pronouns have distinct nominative and accusative case forms. As shown in Section 3.2, Ancient Greek shares this feature with Latin. Verb agreement likewise generally follows an accusative pattern, identifying the first argument of bivalent verbs and the single argument of (simple and derived) monovalent verbs with masculine and feminine nouns and pronouns as well as with first- and second-person pronouns. The agreement patterns of neuter nouns and pronouns presents a somewhat more complex picture. In Archaic Greek, plural neuter forms alternately co-occur with singular and plural verb forms, which we interpreted as an alternation between neutral and accusative agreement. In Classical Greek, plural neuters do not show this alternation, cooccurring with singular forms only (see Chapter 2 for discussion). We have also seen that there are some intriguing examples of potential non-canonical argument realization patterns in Ancient Greek, one involving a dative–genitive case frame and another involving genitive-marked unaccusative subjects. The former of these was argued to represent an inner-Greek innovation, while the latter may be archaic, as will be discussed later on. Another important observation concerns the status of classical behavioural subject properties such as control of reflexives or conjunction reduction, which are not restricted to the first argument of twoplace verbs and the argument of one-place verbs in Ancient Greek. On the other hand, control predicates and the possibility of being the addressee of directives, both of which identify these two core arguments as syntactically privileged, seem to represent relatively uncontroversial subjecthood tests in Ancient Greek. From a more general perspective, subjecthood represents a somewhat less developed phenomenon in Ancient Greek than in Latin, given that there appear to be fewer constructions that unambiguously converge on the same subset of core arguments as their syntactically privileged argument.

3.2.3 Vedic Sanskrit In the present context, Vedic refers to the language of the so-called Vedas and their commentaries. These reflect the oldest attested stage of the Indo-Aryan subbranch of Indo-Iranian. As the historical context of the Vedic texts remains opaque, it is difficult to establish an absolute chronology for these texts. It is likely that the earliest extant text, the Rigveda, had attained the form known to us somewhere around 1200 BCE, parts of it originating from the first half of the second millennium BCE. The latest Vedic texts are probably not much younger than 600 BCE. Our absolute chronological framework thus covers about six centuries, from approximately

88

PAOL A COTTICELLI AND EYSTEIN DAHL Table 3.6 Chronological stages of Vedic Early Vedic: Early Middle Vedic: Middle Vedic: Late Middle Vedic: Late Vedic ≈ Classical Sanskrit

The language of the Rigveda (RV ) The language of the mantra parts of the Atharvaveda and of the Yajurveda The language characterizing the oldest Vedic prose texts of the Yajurveda The language of the younger Vedic prose texts belonging to the Early Brāhmaṇas The language of the youngest Vedic prose texts belonging to the Late Brāhmaṇas and Early Upaniṣads

1200 to 600 BCE. One may distinguish at least five distinct chronological stages of Vedic. These are summarized in Table 3.6 (see, e.g. Witzel 1989, 1995).²¹ Here, we mainly deal with data from Early Vedic. The Vedic nominal system distinguishes six case categories and three numbers, as outlined in Table 3.7. Like Latin and Greek, Vedic has three noun classes: masculine, feminine, and neuter. As shown in Table 3.7, masculine and feminine nouns have distinct forms for nominative and accusative as do masculine/feminine as well as first- and second-person pronouns, while neuters do not distinguish between these two case categories. At the beginning of its attested tradition, Vedic Sanskrit has a predominantly nominative–accusative alignment orientation in the main aspect stems of the verb system, the Present, Aorist, and Perfect.²² The examples in (22) illustrate that masculine nouns like devá- have a nominative singular case-marker -s appearing with monovalent verbs, as in (22a) and in the first argument function of bivalent verbs, as in (22b).²³ This ending is distinct from the accusative singular case-marker -m, which is the default marker of the second argument of bivalent verbs. (22) a. út sū́ryo jyótiṣā devá eti up sun:NOM.SG light:INS.SG god:NOM.SG go:PRS.3SG ‘The sun, the god, goes up with his light.’ (RV IV 13.1d after Jamison and Brereton 2014: 575) ²¹ Kulikov (2013) provides a slightly different chronological framework. ²² See Dahl (2010) for a discussion of the aspectual semantics of the Present, Aorist, and Perfect stems in Early Vedic. ²³ More precisely, devá and devó are two sandhi variants of the nominative singular form devás. Likewise, the forms śarádo, mā́sā and dyā́va in (22c) reflect the nominative plural forms śarádas, mā́sās and dyā́vas.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

89

b. ūrdhváṃ ketúṃ savitā́ devó as´ret upright:ACC.SG banner:ACC.SG Savitar:NOM god:NOM erect:AOR.3SG ‘Divine Savitar has braced his beacon upright.’ (RV IV 14.2a after Jamison and Brereton 2014: 577) c. ná yáṃ járanti ´sarádo ná mā́sā NEG REL.ACC.SG age:PRS.3PL autumn:NOM.PL NEG month:NOM.PL ná dyā́va ´ındram avakars´áyanti NEG day:NOM.PL Indra:ACC make.lean:PRS.3PL ‘Indra, whom the autumns do not age, nor do the months, nor days make lean’ (RV VI 24.7 after Jamison and Brereton 2014: 807) These examples also illustrate that nominative-marked subjects trigger verb agreement, the singular form deváḥ ‘god’ agreeing with the singular verb forms eti ‘goes’ and as´ret ‘has braced’, and the plural forms ´sarádas ‘autumns’, mā́sās ‘months’ and dyā́vas agreeing with the plural verb forms járanti ‘age’ and avakars´áyanti ‘make lean’. In (23) we can observe plural forms of neuter nouns with identical endings in all three core argument functions, the sole argument of monovalent verbs and the first argument of bivalent verbs co-occurring with plural verb forms. Table 3.7 The Vedic nominal case system áśva- ‘horse’ (m.)

Nominative Vocative Accusative Genitive Dative Instrumental Locative Ablative

yugá- ‘yoke’ (n.)

Singular Dual

Plural

Singular Dual

áśvas aśva áśvam áśvasya áśvāya áśvena áśve áśvād

áśvās áśvās áśvān áśvānām áśvebhyas áśvebhis áśveṣu áśvebhyas

yugám yugám yugám yugásya yugā́ya yugéna yugé yugā́d

áśvā áśvā áśvā áśvayos áśvābhyām áśvābhyām áśvayos áśvābhyām

devī- ‘goddess’

Nominative Vocative Accusative Genitive Dative Instrumental Locative Ablative

Singular Dual

Plural

devī́ dévī ī́m devyā́s devyái devyā́ devyā́m devyā́s

s̄ ́ s dévīs ī́s devīnā́m devī́bhyas devī́bhis devī́ṣu devī́bhyas

devī́ dévī ī́ devyós devī́bhyām devī́bhyām devyós devī́bhyām

yugā́ yugā́ yugā́ yugáyos yugā́bhyām yugā́bhyām yugáyos yugā́bhyām

Plural yugā́/yugā́ni yugā́/yugā́ni yugā́/yugā́ni yugā́nām yugébhyas yugébhis yugéṣu yugébhyas

90

PAOL A COTTICELLI AND EYSTEIN DAHL

(23) a. ā́d ´ıt prā́pas´yan bhúvanāni vı´´svā after that see:IPF.3PL being:NOM/ACC.PL all:NOM/ACC.PL ‘Only after that did all living beings see in front of (them).’ (RV X 88.11 after Jamison and Brereton 2014: 1534) b. vı´´svāny atasā́ vánāni ná arvā́g all:NOM/ACC.PL shrubs:NOM/ACC.PL tree:NOM/ACC.PL NEG close ´ındraṃ pratimā́nāni debhuḥ Indra:ACC equivalent:NOM/ACC.PL deceive:PRF.3PL ‘A ll the bushes and trees have not deceived Indra as near-equivalents (of soma).’ (RV X 89.5 after Jamison and Brereton 2014: 1536) c. yā́ni indrāgnī cakráthur REL.NOM/ACC.PL Indra.and.Agni:VOC do:PRF.2DU vīryā́ṇi manly.deed:NOM/ACC.PL ‘O Indra and Agni, which manly deeds you two have done (…)’ (RV I 108.5 after Jamison and Brereton 2014: 255) Macdonell (1916: 289) notes that exceptions to these general agreement rules are rare, and we conclude that Vedic generally shows consistent nominative– accusative agreement on the verb (see also Delbru¨ck 1888: 83, Chapter 2 in this volume). From the overall perspective of this chapter, this implies that neuter nouns show mixed alignment in Vedic, instantiating neutral case marking and accusative agreement.²⁴ It should also be noted that Vedic is a pro-drop language. The above data show that verb agreement picks out the first argument of bivalent verbs and the only argument of monovalent verbs as syntactically privileged. Other constructions identifying the same two arguments as syntactically privileged include being addressee of directives, as illustrated by the examples in (24).²⁵ (24) a. ´sivás tvaṣṭar ihā́ gahi kind:NOM.SG Tvaṣṭar:VOC here come:AOR.IMP.2SG ‘Tvaṣṭar, come here as the kind one.’ (RV V 5.9 after Jamison and Brereton 2014: 668) b. jāmı´m ájāmiṃ prá mṛṇīhi ´sátrūn kin:ACC.SG non.kin:ACC.SG PRV crush:PRS.IMP.2SG rival:ACC.SG ‘Kin or non-kin, pulverize the rivals.’ (RV IV 4.5 after Jamison and Brereton 2014: 564) ²⁴ Delbru¨ck (1888: 83) identifies three clear-cut (‘sichere’) examples of plural neuter nouns accompanied by singular verb forms in the Rigveda (RV I 81.3, I 162.14 and VII 21.6) and only one example of a singular noun with a plural verb form in all of the Vedic texts he had at his disposal (MS I 4.14), so both types are marginal in Vedic. We refer to Chapter 2 in this volume for a discussion of the implications of this fact. ²⁵ Other relevant constructions include reflexivization, control over converbs, word order, and certain deverbal nominalizations (Hock 1990, 1991). It should be noted that there are several competing strategies as regards reflexivization in Vedic (see Orqueda 2019 for a survey).

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

91

Example (24a) illustrates that the sole argument of monovalent verbs like gam‘come’ is the addressee of second-person imperatives such as gahi ‘come’. In (24b), we observe that second-person imperative forms of bivalent predicates like mari ‘crush’ address the first argument. In previous sections, we have seen that control infinitives pick out the same set of core arguments as verb agreement and imperative reference. In Vedic, the situation is somewhat different, as illustrated by the examples in (25).²⁶ (25) a. ná asyā vas´mi vimúcaṃ ná āvṛ́taṃ NEG DEM.ABL.SG wish:PRS.1SG release:INF NEG turn.back:INF púnar again ‘I do not desire to be released from it[/her], nor to turn back here again.’ (RV V 46.1 after Jamison and Brereton 2014: 721) b. ´sakéma tvā samı´dham be.able:AOR.OPT.1PL 2SG.ACC kindle:INF ‘Might we be able to kindle you.’ (RV I 94.3 after Jamison and Brereton 2014: 230) These examples illustrate that there is no restricted neutralization of the three core arguments with control infinitives in Early Vedic. Specifically, while the infinitive form āvŕ̥tam ‘turn back’ in (25a) shows that the argument of monotransitive predicates can be coreferent with the subject of the matrix clause, and the form samı´dham ‘kindle’ in (25b) illustrates that this is the case with the first argument of bivalent predicates as well, the form vimúcam ‘release’ in (25a) demonstrates that the undergoer argument of bivalent predicates also has this property. Examples like these clearly show that infinitives are underspecified with regard to voice in Vedic (Keydana 2013: 83). These considerations suggest that control infinitives do not pick out a syntactically privileged argument in Early Vedic, and therefore do not constitute a morphosyntactic environment relevant for determining subjecthood in this language. Given the observations in the previous sections, this could be interpreted as an indication that the notion of subjecthood is less developed in Early Vedic than in Ancient Greek and Latin. As regards non-canonical subject constructions in Vedic, evidence is somewhat scanty. The examples in (26a–b) illustrate that subject-like experiencer arguments of one-place verbs have accusative forms in two cases in the Rigveda. Examples (24c–d) show that the genitive has an analogous use. (26) a. strı´yaṃ dṛṣṭvā́ya kitaváṃ tatāpa woman:ACC.SG see:ABS player:ACC.SG burn:PRF.3SG ‘Having seen his wife, the gambler is in pain.’ (RV X 34.11a after Hock 1991: 127) ²⁶ See Keydana (2013) for a thorough exploration of the infinitives in Early Vedic.

92

PAOL A COTTICELLI AND EYSTEIN DAHL b. ná mā taman ná ´sraman NEG 1SG.ACC be.exhausted:AOR.SBJV.3SG NEG be.weary:AOR.SBJV.3SG ná utá tandrat NEG CONJ grow.fatigued:3SG.AOR.SBJV ‘I will not become exhausted, not become weary and not grow fatigued.’ (RV II 30.7ab after Dahl 2014) c. ápāyy asya ándhaso mádāya drink:AOR.3SG.PASS DEM.GEN.SG soma.juice:GEN.SG exhilaration:DAT.SG ‘The drinking of this soma juice for exhilaration has begun.’ (RV II 19.1)²⁷ d. yád vaı´ púruṣasya āmáyati if PTCL man:GEN.SG be.afflicted.with.pain:PRS.3SG ‘If a man is afflicted with illness (…).’ (MSP 1.8.9, Hock 1990: 129)

Examples (26a–b) illustrate that monovalent predicates occasionally co-occur with accusative-marked nouns, like kitavám ‘gambler’ in (26a), or pronouns, like mā ‘me’, in (26b), which seemingly express their single argument. An analysis along these lines finds some support in the fact that the noun kitavám in (26a) is coreferent with the subject of the converb dṛṣṭvā́ya ‘having seen’, a property characteristically but not exclusively associated with subjects in Vedic (see Hock 1990, 1991 for discussion). Analogous constructions are found in Latin (Section 3.2.1), Hittite (Section 3.2.4), and in the closely related Old Iranian language Avestan (Danesi 2014), so that a case could be made for the claim that this construction is inherited in Vedic.²⁸ The examples in (26c–d) show that genitive-marked nouns seemingly also appear in argument function of monovalent predicates. However, there is a slight but important difference between the two constructions instantiated here. In (26c), the genitive has a clear-cut partitive function, expressing that a part of the soma juice has been consumed and alternating with the nominative.²⁹ Given that the form ápāyi ‘be drunk’ has a passive-like detransitivizing function, this example may be analysed as a special instantiation of the pervasive use of the partitive genitive with ingestion verbs in Vedic, characteristically surfacing as object marking alternating with the accusative (Dahl 2009, 2014).³⁰ Example (26d), on the other hand, illustrates a somewhat different use of the genitive in argument marking, namely, to express the only argument of certain monovalent experiential predicates (see Hock 1990: 129 for examples). With a verb such as āmaya- ‘be afflicted with illness’, the genitive seems to be the preferred or exclusive argument realization option. Hock (1991) provides convincing evidence that ²⁷ Jamison and Brereton (2014: 428): ‘The drinking of this plant for exhilaration has begun’. ²⁸ It should be noted, however, that all the examples involving accusative subjects in Avestan cited in Danesi (2014) stem from the more recent part of the corpus. ²⁹ Note that the same verb form is used in RV I 175.1 with a neuter noun (máhas ‘might’). ³⁰ As discussed in Dahl (2014), the alternating use of accusative and genitive as object markers found in Vedic has almost exact parallels in Homeric Greek, a fact suggesting that this alternation pattern is inherited.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

93

genitive-marked possessor NPs show subject-like behaviour in Sanskrit but the pertinent data from the earliest stages of the language are rather scarce. Thus, we remain agnostic regarding the subject status of non-canonically case-marked arguments like those illustrated in (26) for Vedic. It should be noted, however, that the use of the genitive to express the sole argument of one-place experiential predicates has a striking parallel in Homeric Greek, and it is therefore reasonable to assume that the construction is inherited. Although several constructions display a nominative–accusative orientation in their selection of syntactically privileged arguments in Vedic, there are also certain constructions that instantiate an ergative-like orientation. These include the socalled past passive participle (PPP), passive aorist, and future passive participle or gerund. Here, we give a brief outline of the PPP, which is a P-oriented resultative verbal adjective in Vedic. Consider, by way of illustration, the examples in (27), where the PPP appears as the verbal head of its respective clause. (27) a. ayáṃ vāṃ mitrāvaruṇā nṛ́bhiḥ DEM.NOM.SG 2DU.DAT Mitra.and.Varuṇa:VOC man:INS.PL sutáḥ press:PPP.NOM.SG sóma ā́ pītáye sutáḥ // soma:NOM.SG PRV drink:INF press:PPP.NOM.SG ‘Here it is, pressed by men for you two, Mitra and Varuṇa—the soma pressed for drinking.’ (RV I 137.3 after Jamison and Brereton 2014: 309) b. kvà ṛtám pūrvyáṃ where righteousness:NOM/ACC.SG earlier:NOM/ACC.SG gatáṃ go:PPP.NOM/ACC.SG ‘Where has the earlier righteousness gone?’ (RV I 105.4)³¹ c. hatā́ ´ındrasya ´sátravaḥ sárve slay:PPP.NOM.PL Indra:GEN enemy:NOM.PL all:NOM.PL ‘Slain were all the rivals of Indra/by Indra.’ (RV X 155.4)³² These examples show that the PPP has a different agreement pattern from the verb forms we have encountered earlier, in that it aligns the sole argument of

³¹ See Jamison and Brereton’s (2014: 251) translation: ‘Where has my earlier “truth” gone?’. ³² See, however, Jamison and Brereton’s (2014: 1639): ‘Slain were all the rivals of Indra’.

94

PAOL A COTTICELLI AND EYSTEIN DAHL

monovalent verbs and the second argument of bivalent verbs. In all cases, these arguments show the nominative case marking typical of subjects, in (27a) the actor argument of the verbal predicate has instrumental case marking and in (27c) it has genitive marking. It is uncontroversial that the ergative constructions in Hindi and other New Indo-Aryan languages ultimately derive from the Vedic PPP but the status of the source construction is less clear (see Butt and Deo 2017 for discussion). According to one analysis, the source construction was a predicated P-oriented resultative which first developed into a passive construction and then into an inverse construction eventually becoming an ergative construction in Pāli and other Middle Indo-Aryan languages (see, e.g. Dahl 2016). On this approach, example (27a) represents a bona fide passive construction with an instrumental-marked agent. According to a slightly different analysis, the source construction was anticausative and originally restricted to unaccusative intransitives, as in (27b), spreading to transitive predicates via unergative intransitives, representing an indirect evidential marker. On this analysis, the agentive argument would reflect a possessor raised out of its NP, the instrumental marking of the agent in (27a) representing an innovation (see, e.g. Bynon 2005). This approach would seem to imply that example (27c) should be given precedence, since the genitive-marked noun ´ındrasya may either have an agentive or a possessive reading, since the fact that the enemies of Indra were slain by him is a well-established part of Vedic world knowledge. We remain agnostic as to whether the development of the PPP through the various stages of Indo-Aryan is most satisfactorily explained by the passiveto-ergative scenario or the anticausative-to-ergative scenario. At any rate, there appears to have been an emerging alignment split already at an early stage of the Vedic language between the predominant nominative–accusative system that pervaded most of the nominal and verbal system, and the ergative or proto-ergative alignment system associated with the PPP and the future passive participle. In this section we have seen that Vedic has a basic alignment split between consistently nominative–accusative alignment with masculine and feminine nouns, and mixed alignment with neuter nouns. Vedic only shows very few examples of neuter plural nouns with singular verb forms, which most likely represent isolated relics of an Indo-Iranian system where neuter nouns had consistently neutral agreement, as shown by comparative evidence from Old Avestan (see the discussion in Chapter 2). We have also seen that control infinitives do not represent a morphosyntactic environment relevant for determining subjecthood in Vedic. As regards non-canonical subject constructions, there are only a few isolated examples in Vedic, some of which are mirrored by analogous categories in other languages, a point that will be resumed later on.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

95

3.2.4 Hittite Hittite belongs to the Anatolian branch of Indo-European and was the main administrative language of the Ḫatti Kingdom during parts of the second millennium BCE. Table 3.8 gives an overview of the chronological stages of the Hittite language. Table 3.8 Chronological stages of Hittite Old Hittite Middle Hittite Late Hittite

c.1650–1450 BCE c.1450–1380 BCE c.1350–1200 BCE

Hittite nouns distinguish seven case categories and two number categories. Table 3.9 gives an overview of the Hittite case system. Table 3.9 The Hittite case system antuḫša- ‘human being’, laḫḫa- ‘campaign’ (c.) šaḫḫan- ‘tax, obligation’, genu- ‘knee’ (n.)

Nominative Vocative Accusative Genitive Dative–Locative Ablative Instrumentala Allativeb

Singular Plural

Singular Plural

antuḫšaš antuḫšeš antuḫša antuḫšeš antuḫšan antuḫšuš antuḫšaš antuḫšaš antuḫši antuḫšaš antuḫšaz antuḫšet laḫḫa

šaḫḫan šaḫḫani šaḫḫan šaḫḫani šaḫḫan šaḫḫani šaḫḫanaš šaḫḫanaš šaḫḫani šaḫḫanaš šaḫḫanaz šaḫḫanit genuwa

a

The instrumental merges with the ablative in Middle Hittite, and is completely obsolete in New Hittite (Hoffner and Melchert 2008: 268). b

The allative case primarily occurs in Old Hittite, has no plural form, and is restricted to nouns denoting inanimate objects (Hoffner and Melchert 2008: 262).

Hittite has two noun classes: common and neuter, a central difference being that common nouns distinguish nominative and accusative case forms in the singular and plural, while neuter nouns are generally taken to have syncretic forms for these two case categories, as illustrated in Table 3.9. Common nouns show consistently nominative–accusative alignment with finite verb forms, also triggering verb agreement, as illustrated by the examples in (28).

96

PAOL A COTTICELLI AND EYSTEIN DAHL

(28) a. a]lpašš=a arḫa paizzi (…) a]lpašš=a cloud:NOM=CONJ away go:3SG.PRS cloud:NOM=CONJ ṷezzi come:3SG.PRS ‘And the cloud goes away and the cloud comes.’ (KUB 33.21+ iv 17–18, The Storm-god of the Queen Ašmunikkal, CTH 326.A, NS) b. nu alpaš GIG-an ŪL taraḫzi CONN cloud:NOM illness:ACC NEG overcome:3SG.PRS ‘And the cloud does not overcome the illness.’ (KUB 17.8 + iv 9, Incantation and Myth, CTH 457.1.A, NS) c. mān uwarka[ntan] antuḫšan uwanzi if fat:ACC.SG man:ACC.SG see:PRS.3PL n=an=kan kunanzi CONN=3SG.ACC=PTCL kill:PRS.3PL ‘If they see a fat man they kill him.’ (KBo 3.60 ii 3–4, Cannibal Text, CTH 17.1, OH/NS) d.

LÚ.MEŠ

ḫalliyarēš PĀNI GIŠ.d INANNA.ḪI.A ḫallari_singer:NOM.PL in.front lyre:ACC.PL paršnanzi squat:PRS.3PL ‘The ḫallari-singers squat in front of the lyres.’ (KUB 11.13 ii 15, AN.TAḪ.ŠUMSAR Festival: days 18–19, CTH 613.1.A, LNS)

m e. nu=mu Nuwanzaš GAL.GEŠTIN CONN=1SG.ACC Nuwaranza:NOM wine.chief:NOM.SG EN.MEŠ=ya ḫūmanteš INA URU Tegaramma mēnaḫḫanda lord:NOM.PL=CONJ all:NOM.PL in city.of.Tegaramma to uēr nu=mu anda wemiēr come:PST.3PL CONN=1SG.ACC PRV find:PST.3PL ‘Nuwanza the Chief of the Wine and all the lords came to (meet) me in T. and found me there.’ (KBo 4.4 + iii 20–22, Annals of Muršili II, CTH 61.II.5.B, NH/NS)

These examples illustrate that common nouns select nominative case when appearing in argument function with monovalent (28a, e) and in first argument function of bivalent (28b, e) verbs, and accusative case in second argument (28c) function of bivalent verbs. We can also observe that plural forms of common nouns trigger plural verb agreement in monovalent subject function (28d) and in bivalent first argument function (28e). Mutatis mutandis, common pronouns and firstand second-person pronouns show the same behaviour. We also note that Hittite is a pro-drop language.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

97

Hittite neuter nouns and pronouns, on the other hand, show a different behaviour from common nouns and pronouns, as illustrated by the examples in (29). First, bare neuter nouns are used as the sole argument of a restricted group of monovalent predicates (29a) and as the second argument of bivalent predicates (29b). Neuter nouns in first argument function of bivalent predicates consistently show a special form in -ant- + -s (29c) (see Chapter 4 in this volume for discussion of this form). Second, plural forms of neuter nouns do not trigger plural verb agreement (29d) (see also Chapter 2 in this volume). (29) a. nu apāt ēšḫar kuṷapi paizz[i] CONN DEM.NOM/ACC.SG blood:NOM/ACC.SG when flow:PRS.3SG ‘And when this blood flows.’ (KUB 14.3 iv 52, Letter of Tawagalawa, CTH 181, NS) b. nu ēšḫar akkuškandu CONN blood:ACC.SG drink:IMP.3PL ‘Now, let them drink blood!’ (KBo 20.73+ KBo 17.54 iv 10, Incantation Ritual, CTH 458.1.1.A, MS) URU c. nu KUR Ḫatti=ya apāš išḫananza arḫa CONN land:ACC Hatti=CONJ DEM.NOM.SG blood:NOM.SG away namma zinni[t] further finish:PST.3SG ‘That blood(shed) further finished off the land of Hatti too.’ (KUB 14.14+ rev. 11–12 after Goedegebuure 2015, Plague Prayer of Muršili II, CTH 378.1.A, NH/NS)

d. nu=wa=kan apē=ya uddār QATAMMA CONN=PTCL=PTCL DEM.NOM.PL=CONJ word:NOM.PL thus lagāru fall:IMP.3SG ‘And thus these words shall fall.’ (KBo 2.3+ iii 21–22, Ritual of Maštigga, CTH 404.1.II.A, NS) These data are intriguing in several respects. As regards agreement, Hittite nouns and pronouns show a split analogous to the one found in Ancient Greek, where masculines and feminines as well as speech act participant pronouns display a consistent nominative–accusative pattern, while neuters alternate between a nominative–accusative and a neutral pattern, as argued above. An important difference is that plural forms of neuter nouns and pronouns consistently appear with singular verb forms in Hittite, as illustrated by example (29d). We return to the implications of this fact shortly. Another interesting problem concerns case marking. There is widespread agreement in the scholarly tradition that data like those cited in (28) and (29) show that Hittite has a noun–class-based alignment split, common nouns having nominative–accusative alignment and neuter

98

PAOL A COTTICELLI AND EYSTEIN DAHL

nouns instantiating ergative–absolutive alignment (Laroche 1962, Garrett 1990, Rizza 2009, 2010, Melchert 2011b, Goedegebuure 2012, 2015, 2018, Chapter 4 in this volume). Goedegebuure (2018) and Luraghi and Inglese (Chapter 4 in this volume) make a strong case for the claim that this split represents a secondary development, which essentially involves the grammaticalization of an originally derivational marker to a case marker (see also Dahl 2021a, Chapter 4 in this volume). In recent work, Dahl (2021a) draws attention to the fact that neuter nouns do not freely occur with intransitive predicates but seem to be restricted to a subclass of unaccusative verbs, namely verbs that do not entail a sentient and volitional subject argument. Consider the examples in (30). (30) a. udder=ma=kan kue KAxU-az parā word:NOM.PL=PTCL=PTCL REL.NOM.PL mouth-ABL PRV iyattari come:PRS.3SG ‘Words which come forth from the mouth.’ (KUB 27.29+ ii 17–18, Ritual of Allaituraḫi, CTH 780.II.Tf06.A, NS) b. welluwaš anda welkuwan lē huwāi meadow:LOC.PL PRV grass:NOM.SG NEG grow:PRS.3SG ‘In the meadows grass shall not grow.’ (KBo 6.34+ ii 40–41, Military Oath, CTH 427.A, NS) c. utnē [kuit k]uit=pat araiš land:NOM.SG INDEF.NOM.SG=PTCL rise:PST.3SG ‘Whatever land made an uprising.’ (KBo 3.22 obv. 11–12, Anitta Text, CTH 1.A, OH/OS) These examples are intended to illustrate that neuter nouns in monovalent argument function primarily occur with predicates that do not entail animacy or volitionality (28a–b) but that they in some rare cases appear in contexts implying an agentive reading (28c). Given that neuter nouns generally denote inanimate referents, this observation is in itself not very surprising. For our purposes, however, this indicates that the distribution of the two noun classes in Hittite with intransitive predicates is determined by their relative agentivity, a fact suggestive of semantically oriented alignment. This observation is in line with the fact that plural forms of neuter nouns and pronouns in monovalent subject function cooccur with singular verb forms. Specifically, given that neuter nouns and pronouns are generally restricted to two core argument functions in Hittite, namely that of monovalent predicates with an undergoer-like argument and the second argument of bivalent predicates, the fact that neuter plural forms in monovalent subject function appear with singular verb forms indicates that they are treated on a par with the undergoer argument of bivalent predicates. We take these properties to be characteristic of semantic alignment and conclude that the emergent split ergative system in Hittite interacts with a noun–class-based semantic alignment and that

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

99

the rise of ergativity with neuter nouns is intimately connected with their inability to appear in first argument function of bivalent predicates and their limited ability to appear with monovalent predicates implying an animate or volitional argument (see Chapter 4 in this volume). In this connection, attention should be drawn to the fact that the distinction between two classes of monovalent predicates plays a role in other areas of morphosyntax in Hittite. One such area concerns enclitic subject pronouns, which are restricted to monovalent predicates with an undergoer-like argument, as illustrated by the examples in (31). (31) a. šu=uš tameššir š=e akir CONN=ACC.3PL oppress:PST.3PL CONN=NOM.3PL die:PST.3PL ‘And they oppressed them. And they died.’ (KBo 22.2 vs. 12–13, CTH 3.1.A [OH]) b. EGIR-an=ma=aš irmaliyattat=pat then=PTCL=NOM3SG become.ill:PST.3SG.MP =PTCL ‘Then he became sick.’ (KBo 3.4 i 5–6, CTH 61.1A [NH]) These examples show that third-person enclitic subject pronouns like e ‘they’ and aš ‘he/she’ appear with unaccusative predicates throughout the attested history of the language (see Garrett 1996 for discussion). As we shall see below, the distinction between the two classes of monovalent predicates also manifests itself in auxiliary selection. Given this alignment split in the Hittite nominal and pronominal system, one would expect there to be relatively few constructions in the language that select syntactically privileged arguments that can be generalized across all noun and pronoun classes. Restricting our focus to common nouns and pronouns, and to firstand second-person pronouns, we note that the property of being the addressee of second-person imperative forms is characteristic of monovalent arguments and first arguments of bivalent verbs, as shown in (32).³³ (32) a. ammukk=a=za=pa anda ēp 1SG.ACC=CONN=REFL=PTCL ADV take:IMP.2SG lē=mu genzuwaiši PROH=1SG.ACC spare:PRS.2SG ‘Include me! Do not spare me!’ (KBo 3.7 iii 29, CTH 321.A) ³³ To our knowledge, no systematic investigation into the subject properties of Hittite exists and, consequently, the status of subjecthood as a syntactically relevant dimension and its foundations in Hittite remain unexplored. Here, we only offer a few observations of particular relevance for the present chapter but hasten to add that a full examination of these matters would go far beyond the scope of this chapter, and will have to be undertaken elsewhere.

100

PAOL A COTTICELLI AND EYSTEIN DAHL b. INA KUR ḪUR.ŠAG Zippaslā ēš in land Zippaslā.mountains sit:IMP.2SG ‘Settle down in the land of the Zippaslā-mountains.’ (KUB 14.1 vs. 16, CTH 147)

Another morphosyntactic context which serves to establish subjecthood in other, related languages is control infinitives. The examples in (33) suffice to illustrate that control infinitives do not show restricted neutralization of the core argument functions in Hittite. (33) a. nu 1-aš 1-aš INA É.DINGIR-LIM šarā šēšūwanzi CONN one:NOM.SG one:NOM.SG in temple up sleep:INF lē=pat karštari PROH=PTCL neglect:PRS.3SG.MID ‘Let him not neglect—each in his turn (lit. one by one)—to spend the night up in the temple.’ (KUB 13.4 iii 5–6, CTH 364.A after Hoffner and Melchert 2008: 336) b. maḫḫan=ma=za ḫalkueššar ḫandāuwanzi zinnāi when=CONN=REFL material:NOM/ACC.SG prepare:INF finish:PRS.3SG ‘When he finishes preparing the materials (for the festival) (…).’ (KUB 27.59 i 23–24, CTH 691.1 after Hoffner and Melchert 2008: 335) c. nu māḫḫan ŠA GALHI.A waršiyaš memiyanieš CONN when of cups soothing:GEN.SG word:NOM.PL ḫurlili memiyawanzi zinnandari Hurrian speak:INF finish:PRS.3PL ‘When they finish speaking in Hurrian the words of soothing the cups (lit. when the words of soothing the cups are finished to be spoken).’ (KUB 29.8 obv. i 1–2, CTH 777.Tf10.2A after Lyutikova and Sideltsev 2021: 3) These examples illustrate that the matrix verb of control infinitives in Hittite alternately show coreference with monovalent arguments as in the case of šēšūwanzi ‘sleep, spend the night’ in (33a), with the first argument of bivalent predicates, as shown by ḫandāuwanzi ‘prepare’ in (33b), or with the second argument of bivalent predicates, as shown by memiyawanzi ‘speak’ in (33c). This behaviour is motivated by the fact that Hittite infinitive forms are underspecified for voice (Hoffner and Melchert 2008: 332). Therefore, control infinitives do not constitute a morphosyntactic environment relevant for defining subjecthood in Hittite. As regards non-canonical subject constructions, we note that some experiential verbs select an experiencer argument in the accusative or dative–locative, as illustrated in (34) (Luraghi 2010, Dardano 2018).

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

101

(34) a. m]ān antuḫšan SAG.DU-SÚ ištara[k-zi ] if human.being:ACC.SG head-POSS.3SG be(come).ill:PRS.3SG n=an naššu apēniššan išta[r]akzi CONN=3SG.ACC or similarly be(come) ill:PRS.3SG ‘If a man is ill in his head, or he is ill in a similar manner.’ (KUB 8.36 ii 12–13, Tablet catalogue, CTH 279.3, NS) f b. [k]inun=a kāša Gaššuliyawiyaš tuēl GÉMETUM now=CONN here Gaššuliyawiya:NOM you:GEN servant [n]=an tamaštat ištarkiat GIG-anza be(come).ill:PST.3SG CONN=3SG.ACC sickness:NOM oppress:PST.3SG ‘Now Gaššuliyawiya here, your servant, has become ill, [a]nd the sickness has oppressed her.’ (KBo 4.6 obv. 24–25, Muršili’s Prayer to Lelwani for the Recovery of Gaššulijawija, CTH 380.1.A, NH/NS)

c. nu=šši šakuwaš peran É-ri=kán anda CONN=3SG.DAT eyes:GEN.PL before house:LOC.SG=PTCL PSTP katkattiyazi tremble:PRS.3SG ‘And in the house he suffers trembling before his eyes.’ (KUB 29.9 i 5–6, Lunar eclipse, CTH 532.II.5, NS) In a recent discussion of non-canonical subject constructions in Hittite, Dardano (2018) notes that the verbs selecting this type of argument realization pattern tend to show subject alternation. For example, the verb ištark- ‘be, become ill’ alternately selects an experiencer argument in the accusative, as in (34a), or the nominative, as in (34b). Example (34c) illustrates that experiencer arguments in a few cases appear in the dative–locative. While most of the examples discussed in the pertinent research literature involve experiencer predicates, attention should be drawn to the fact that certain unaccusative verbs in the singular sporadically select subject argument in the plural in Late Hittite, as illustrated in (35) (Hoffner and Melchert 2008: 241). Note that the nouns tedanuš ‘hair’ and ḫēwēš ‘rain’ are common gender. (35) a. tedanuš ārašzi hair:NOM/ACC.PL flow:PRS.3SG ‘Hair flows’ (KUB 41.8+ ii 3–4, Ritual for the Purification of the House, CTH 446.C, LNS) b. ḫēwēš [kī]ša BURU.HI.A SIG5-anta rain:NOM/ACC.PL become:PRS.3SG.MP crops thrive:PRS.3PL.MP ‘Rains will occur, crops will thrive.’ (KUB 8.1(+) iii 8–9, Lunar eclipse, CTH 532.II.1.A, NS) Data like these indicate that plural forms of common nouns no longer trigger plural agreement on the verb, at least not with unaccusative intransitive verbs. It is tempting to view this fact in connection with another development in Late

102

PAOL A COTTICELLI AND EYSTEIN DAHL

Hittite. As noted by Hoffner and Melchert (2008: 70, 80), late Hittite witnesses a development where nominative and accusative plural forms are used interchangeably, the distinction between them eventually being lost and the accusative ending becoming the default ending of the merged nominative–accusative plural. These considerations suggest that data of the type illustrated in (35) may be indicative of a situation where accusative forms are used in non-canonical subject constructions with certain non-experiential predicate types. This possibility will have to be more thoroughly explored elsewhere. Before concluding this section, attention should be drawn to the fact that Hittite has a resultative P-oriented verbal adjective which shows an ergatively based orientation, selecting monovalent subjects and second arguments of bivalent predicates as syntactically privileged core arguments. This may be illustrated by the different orientations of appant- ‘taken’ from ēp- ‘take, seize’ and piyant- ‘given’ from pai‘give’, on one hand, and pānt- ‘gone’ from pāi- ‘go’ and ḫuwant- ‘running’ from ḫuwai- ‘run’, on the other (Hoffner and Melchert 2008: 339).³⁴ Among other things, these verbal adjectives are used in periphrastic perfect and stative constructions in Hittite. This is a further morphosyntactic area where semantic alignment plays a role in Hittite, as illustrated by the following examples. (36) a. karū irmalianza ēšta already become.ill:PART.C.NOM.SG be:PST.3SG ‘(He) had already become ill.’ (KBo 4.12 vs. 22, CTH 87) b. nu aruwan ḫarkanzi CONN prostrate:PART.N.NOM/ACC.SG have:PRS.3PL ‘They have bowed down.’ (IBoT 3.140, 4, CTH 631.5) c. nu=mu ištamaššan kuit ḫarker CONN=1SG.ACC hear:PART.N.NOM/ACC.SG CONJ have:PST.3PL ‘Because they had heard about me (…).’ (KBo 5.8 i 24, CTH 61.II.7a, NH after Hoffner and Melchert 2008: 311) Here, we observe that periphrastic perfect forms of unaccusative verbs like irmaliya- ‘be(come) ill’ are formed by means of the participle in the nominative case, agreeing with the subject in gender and number, and the auxiliary verb ēš‘be’, while corresponding forms of unergative verbs like aruwai- ‘prostrate oneself, bow down’ show the neutral nominative–accusative singular form of the participle and the auxiliary ḫar(k)- ‘hold, have’. This is yet another area in Hittite grammar that seems to be indicative of semantic alignment. ³⁴ The last example illustrates that the verbal adjective does not always have a resultative meaning but may show a processually oriented meaning.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

103

Summarizing the main points of this section, we observe that Hittite has split alignment in the nominal system between common nouns, which show consistent nominative–accusative patterning in case and agreement, and neuter nouns, which have consistent neutral alignment in case as well as agreement. Moreover, we have seen that bare neuter nouns seem to be restricted to second argument function of bivalent predicates and to subject function of monovalent predicates with an undergoer-like argument. Common nouns and pronouns, on the other hand, appear in all core argument functions. These facts are indicative of semantic alignment. This somewhat defective behaviour provided favourable conditions for the development of the newly emerging ergative marker, as discussed in detail by Goedegebuure (2018) and in Chapter 4 in this volume. The assumption that the distribution of Hittite nouns is based on semantic alignment finds support in the fact that other grammatical phenomena, such as enclitic subject pronouns and auxiliary selection with the periphrastic perfect, are sensitive to the distinction between unergative and unaccusative predicates. Related observations pertain to verb agreement, which serves to pick out a consistent syntactically privileged argument for common nouns and pronouns as well as speech act participant pronouns but not in the case of neuters. We have also seen that, while the property of being the addressee of imperatives constitutes a bona fide subject test in Hittite, control infinitives do not, since they allow for all three core arguments to be coreferent with the subject of the matrix verb.

3.2.5 Classical Armenian For the purposes of this chapter, Armenian is regarded as an independent branch of Indo-European, historically spoken in the Armenian highlands.³⁵ Table 3.10 gives an overview of the main chronological stages of Armenian.³⁶ Here, we mainly deal with Old Armenian/Classical Armenian. Unlike the other languages dealt with in this chapter, Armenian has lost morphological noun–class distinctions, partly because of word-final syllable loss/apocope.³⁷ Nouns and pronouns formally distinguish seven case categories and two numbers, as shown in Table 3.11. ³⁵ It is controversial to what extent Armenian represents a separate, independent branch of the Indo-European family. Some scholars, like Holst (2009), claim that Armenian belongs to a closely related group of Balkan-Indo-European languages, which also includes Greek, Albanian, Phrygian, and (Ancient) Macedonian. Other scholars, like Clackson (1994), assume that Proto-Armenian and Proto-Greek were geographically proximate but represent two distinct branches of Indo-European. ³⁶ We emphasize that the chronological framework presented here is intended as a broad diachronic survey. For a more detailed discussion of the various diachronic and diatopic varieties of Armenian with a rich bibliography, see Weitenberg (2017) and Sayeed and Vaux (2017). ³⁷ The loss of genus as a nominal and pronominal category in Classical Armenian seemingly is in line with a more general areal tendency, since languages such as Hurrian and Urartian as well as Kartvelian languages do not show this type of distinction either (see Holst 2009: 102 with references).

104

PAOL A COTTICELLI AND EYSTEIN DAHL Table 3.10 Chronological stages of Armenian Old Armenian/Classical Armenian Middle Armenian/Cilician Armenian ‘Civil Armenian’ Modern Armenian

Standard Western Armenian

Standard Eastern Armenian

5th century CE 11th–15th century CE 14th/17th–18th century CE 19th century CE onwards

Table 3.11 The Armenian case system am ‘year’

Nominative Accusative Genitive Dative Locative Ablative Instrumental

Singular

Plural

am am ami ami ami amē amaw

amk‘ ams amac‘ amac‘ ams amac‘ amawk‘

Classical Armenian generally has a consistently nominative–accusative alignment system regarding both case marking and verb agreement in its two main aspect stems, the Present and Aorist.³⁸ As illustrated by the data in Table 3.11, however, nouns show distinct nominative and accusative forms in the plural only, having syncretic forms for these two case categories in the singular. In the pronominal system, we find the same pattern of syncretism with the demonstrative pronouns, while first- and second-person pronouns show distinct nominative and accusative forms in the singular as well as in the plural. In addition, Classical Armenian shows differential object marking, employing the prefix z-, the so-called nota accusativi, when the second argument of bivalent predicates is definite or marked as given.³⁹ Finally, Classical Armenian is a pro-drop language. The examples in (37) illustrate a number of characteristic properties of the nominative–accusative alignment system shown in the Classical Armenian present and aorist system. ³⁸ See Vogt (1930), Jensen (1959: 114–117) and Klein (2017), for a discussion of the semantics of the Present and Aorist stems in Classical Armenian. ³⁹ See Mu¨th (2014) for a recent study of differential object marking in Classical Armenian.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

105

(37) a. Ew Mariam z=amenayn z=bans z=aysosik CONJ Mary:NOM NACC=all NACC=word:ACC.PL NACC=DEM.ACC.PL pahēr keep:IPF.3SG ‘And Mary kept all these words.’ (Lk. 2:19) b. Ew darjan hoviwk‘=n, (…) ew CONJ return:AOR.3PL shepherd:NOM.PL=DEF CONJ ōrhnēin z=Astowac praise:IPF.3PL NACC=God:ACC.SG ‘And the shepherds returned (…) and were praising God.’ (Lk. 2:20) c. Ew gayin ar̄ na, berēin andamaloyc CONJ come:IPF.3PL PRP 3SG.ACC carry:IPF.3PL paralysed:ACC.SG mi barjeal i ˇc‘oric‘ one:ACC.SG lift:PTCP PRP four:ABL.PL ‘And they came to him, they were carrying a paralysed (man), lifted by four (men).’ (Mk. 2:3) These examples illustrate that nouns (Mariam, hoviwk‘-n) trigger number agreement on the verb (pahēr, darjan, ōrhnēin) when used in bivalent first argument function and monovalent subject function. We also observe that nouns in second argument function of bivalent predicates select the nota accusativi prefix z- when the referent has already been introduced in the discourse (z-amenayn z-bans) or has unique reference (z-Astowac). The examples in (37) suffice to show that verb agreement singles out the monovalent subject and the first argument of bivalent verbs as the syntactically privileged argument. Ko¨lligan (2013) suggests that conjunction reduction and reflexive control also select these two arguments as their privileged syntactic argument. The examples in (38) illustrate the patterns of reflexive control.⁴⁰ (38) a. gnesc‘en iwreanc‘ kerakowrs buy:AOR.SBJ.3PL REFL.DAT.PL food:ACC.PL ‘[They] may buy food for themselves.’ (Matth. 14:15, adapted from Klein 2017: 1112–1113) b. xelamowt linēr i srti knowing:NOM/ACC.SG become:IPF.3SG PRP heart:LOC.SG iwrowm REFL.POSS.DAT.SG ‘And [she] became wise in her heart.’ (Lk. 2:19)

⁴⁰ At this point, we would like to point out that Classical Armenian has a class of infinitives, which are, however, voice neutral, as was shown to be the case in Vedic and Hittite (see, e.g. Vogt 1935, Stempel 1983, Vanséveren 1997). From this perspective, infinitives show unrestricted neutralization of the core arguments, and consequently, control infinitives do not represent a subject test in Classical Armenian.

106

PAOL A COTTICELLI AND EYSTEIN DAHL

These examples illustrate that the reflexive pronoun iwr ‘self ’ is characteristically controlled by the first argument of bivalent predicates like gnesc‘en ‘[they] buy’ and by the monovalent argument, as in xelamowt linēr ‘[she] became wise’. There does not appear to be any clear-cut examples of non-canonical subject constructions with finite forms of the present and aorist systems in Classical Armenian. In other words, lexically determined non-canonical subjects are not, to our knowledge, attested in the language, and the present and aorist show a uniformly nominative–accusative alignment pattern. Interestingly, however, Classical Armenian also has a P-oriented resultative verbal adjective in -eal, which formally belongs to the aorist stem and selects the monovalent subject and the bivalent second argument as its syntactically privileged argument. This is shown by forms such as bereal ‘brought, borne’ from berem ‘bring, bear’, where the verbal adjective targets the second argument, and kac‘eal ‘stood’ from kam ‘stand’. The verbal adjective is productively employed in an analytic anterior or perfect construction with the (expressed or unexpressed) copula em ‘be’ as auxiliary. Unlike the finite forms of the present and aorist system, the analytic perfect construction instantiates a tripartite alignment pattern, as illustrated by the examples in (39a–b) (see also Meyer 2016, 2017, Chapter 9 this volume). Note also that the participle is employed in a periphrastic passive construction, as illustrated in (39c). (39) a. ?ocˇ‘ z=gir=n z=ayn ic‘ē NEG NACC=writing:ACC.SG=DEF NACC=DEM.ACC.SG be:PRS.SBJ.3SG ǝnt‘eṙc‘eal jer read:PTCP 2PL.GEN ‘Have you perhaps not read this writing.’ (Mk. 12:10 adapted from Meyer 2017: 112) b. zi omank‘ i doc‘anē i her̄astanē for some:NOM.PL PRP DEM.3.ABL.PL PRP distant.land:ABLSG ekeal en come:PTCP be:PRS.3PL ‘For some of them have come from far away.’ (Mk. 8:3) c. … erkri ew mardkan, or i nmanē earth:GEN.SG CONJ mankind:GEN.SG REL:NOM.SG/PL PRP he:ABL.SG en stełceal. be:PRS:3PL create:PTCP ‘[O]f earth and mankind, which were created by him’ (Agatʿangełos 52.4 adapted from Meyer 2016) These examples illustrate that the verbal adjective in -eal shows tripartite marking of core arguments. In (39a), the verbal adjective ǝnt‘eṙc‘eal from the bivalent verb ǝnt‘er̄nowm ‘read’ selects a genitive-marked first argument (jer ‘you’) and an accusative-marked second argument (z=gir=n z=ayn ‘this writing’). In

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

107

(39b), the verbal adjective ekeal from the monovalent verb gam ‘come’ selects a nominative-marked argument (omank‘ ‘some’). Meyer (2017) notes that the tripartite alignment pattern of the periphrastic perfect is rather well established in Classical Armenian, suggesting that it originates from an ergative alignment pattern in Proto-Armenian (see also Meyer in this volume—Chapter 9). Example (39c) represents a clear-cut passive, the prepositional phrase i nmanē ‘by him’ expressing the (demoted) agent. On the other hand, there are some examples that deviate from this pattern, as illustrated by the examples in (40). (40) a. k‘o ekeal ēr 2SG.GEN come:PTCP be:IPF.3SG ‘You had come.’ (Gen. 27.33, adapted from Ko¨lligan 2013: 77) b. es tesil teseal z=hreštak Astowcoy 1SG.NOM dream:ACC see:PTCP NACC=angel:ACC God:GEN.SG ‘I saw a vision, the angel of the Lord.’ (Agat‘angelos 250.8, adapted from Ko¨lligan 2020: 357) Assuming that the tripartite alignment pattern illustrated in (39) was the predominant one with the periphrastic perfect in Classical Armenian, examples like those cited in (40a–b) can be argued to have non-canonical argument marking.⁴¹ Example (40a) shows that perfect forms of monovalent predicates like ekeal from the monovalent verb gam ‘come’ sometimes select a genitive-marked argument (Ko¨lligan 2013: 77 for more examples of this construction). The fact that the same predicate appears with a nominative-marked argument in (39b) indicates that there is monovalent subject alternation between the nominative and genitive with some verbs at least. It is tempting to analyse this as an instance of incipient alignment change via extension of the genitive from the bivalent perfect of the type illustrated in (39a). Under this analysis, the extension of the genitive to monovalent subject function might be taken to reflect a change from tripartite to accusative alignment, reflecting a so-called marked-nominative structure, differing from the case-marking pattern found with finite forms of the present and aorist.⁴² Analogous considerations might seem to apply to example (40b), where the first argument (es ‘I’) of the verbal adjective teseal from the bivalent predicate tesanem ‘see’ shows nominative case. Assuming that the Classical Armenian perfect shows tripartite alignment, this too could be interpreted as a non-canonical pattern of argument marking. As suggested by Ko¨lligan (2020: 357), however, this pattern ‘probably foreshadows the development in the later stages of the lan-

⁴¹ Meyer (2017) employs the term ‘non-standard’ for these patterns. ⁴² According to Meyer’s (2017) study, the pattern with genitive-marked monovalent subjects was already in decline in the Classical Armenian of the fifth century CE.

108

PAOL A COTTICELLI AND EYSTEIN DAHL

guage which shows nominative–accusative alignment also in the perfect’.⁴³ The pattern with monotransitive arguments in the genitive declines in later stages of the language, reflecting a gradual change from tripartite to consistently accusative alignment, converging on the case-marking patterns found with finite forms of the present and aorist stem. If the present analysis is on the right track, one might hypothesize that there was a competition between the marked-nominative pattern employing the genitive and the unmarked nominative–accusative pattern, with the unmarked pattern eventually prevailing, instantiating a case of generalization in the sense defined above. In this section we have seen that Classical Armenian has a split between consistently nominative–accusative alignment with finite forms belonging to the Present and Aorist system and tripartite alignment associated with the verbal adjective in -eal. There does not seem to be any clear-cut examples of non-canonical subject constructions in the present/aorist system but the periphrastic perfect shows some patterns that arguably could be regarded as non-canonical. However, these cases rather appear to reflect competing innovative patterns, a marked-nominative one and an unmarked, general one, both of which arise as alternatives to the established tripartite alignment pattern of the periphrastic perfect.

3.2.6 Lithuanian Lithuanian belongs to the Baltic branch of the Indo-European language family. It is mainly spoken in the Eastern Baltic region, even though a sizable minority is spread across other parts of Eastern Europe and other continents. The oldest extant remains of Lithuanian date to the sixteenth century CE, so that our chronological framework comprises approximately 500 years. Table 3.12 outlines the diachronic stages of Lithuanian. Table 3.12 Chronological stages of Lithuanian Old Lithuanian Modern Lithuanian

16th–17th century CE 18th century CE onwards

At present, we are mainly but not exclusively dealing with data from Modern Lithuanian. The Old Lithuanian noun system distinguished ten case categories

⁴³ According to Meyer’s (2017) study, there was a distinct increase in the use of the nominative to express the first argument of bivalent predicates in the periphrastic perfect construction in the Classical Armenian of the fifth century CE.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

109

and three numbers, as illustrated in Table 3.13. This inventory has become somewhat reduced in Modern Lithuanian. The boldfaced case categories are only found in Old Lithuanian. Table 3.13 The Lithuanian nominal case system miškas ‘forest’

nominative vocative accusative genitive dative locative/inessive instrumental illative adessive allative

Singular

Dual

Plural

miškas miške mišką miško miškui miške mišku miškañ miškíep(i) miškóp(i)

miškù miškai miškù miškai miškù miškus miškų miškám miškams miškuose miškam ˜ miškais miškuosna miškuosemp miškump

Lithuanian has two noun–class categories, masculine and feminine, both of which show a systematic morphological distinction between nominative and accusative case, except in the dual, as illustrated in Table 3.13.⁴⁴ Thus, the Lithuanian nominal system is organized on a consistently nominative–accusative basis. This is also shown by the fact that nominative forms of both classes appear in subject function of bivalent and monovalent verbs, accusative forms occurring in object function of bivalent verbs, as illustrated by the examples in (41).⁴⁵ (41) a. Seminãrijos gyvẽnimas, tiesà, gerókai jàs seminary:GEN.SG life:NOM.SG indeed practically 3PL.ACC apgrióve˙ destroy:PST.3SG/PL ‘Indeed, seminary life practically destroyed them.’ (Viñcas Mykoláitis-Pùtinas, Altorių šeše˙ly) b. Mìršta zˇmóne˙s kãro laukuosè die:PRS.3SG/PL people:NOM.PL war:GEN.SG field:LOC.SG baisiosè kancˇiosè horrible:LOC.PL.F agony:LOC.PL ‘People die on the fields of war in horrible agony.’ (Balys Sruoga, Dievų miškas) ⁴⁴ The adjectival and participle system has a special so-called neuter form, which appears in a restricted number of morphosyntactic contexts (Mathiassen 1996, Ambrazas 2006: 135–137). ⁴⁵ The examples are adapted from https://lrc.la.utexas.edu/eieol/litol.

110

PAOL A COTTICELLI AND EYSTEIN DAHL c. kur̃, bū́davo, àtvezˇa kãlinį ir̃ where use.to:PST.3SG/PL bring:PRS.3SG/PL prisoner:ACC.SG CONJ tuojau˜ nuzˇùdo immediately kill:PRS.3SG/PL ‘(Camps) where they used to bring a prisoner and immediately kill (him).’ (Balys Sruoga, Dievų mìškas)

These examples also illustrate that Lithuanian does not distinguish third-person singular and plural verb forms, a fact reminiscent of the behaviour of neuter nouns in Hittite and Ancient Greek. The lack of number distinction distinguishes thirdperson verb forms from first- and second-person verb forms. The examples in (42) illustrate that Lithuanian has distinct verb forms for the first- and second-person singular and plural. (42) a. aš pàzˇvelgiau 1SG.NOM take.a.glance:PST.1SG ‘I took a glance.’ (Vyautas Macˇernis, Trecˇióji vìzija) b. Tė́ve mū́sų, kurs esì dangujè Father:VOC.SG 1PL.GEN who:NOM.SG be:PRS.2SG heaven:LOC.SG ‘Our Father, who is in heaven.’ (Bishop Paltarokas, Katalikų tikybos katekizmos) c. kaip ir mẽs atléidzˇiame sávo kaltiniñkams like CONJ 1PL.NOM forgive:PRS.1PL REFL.GEN.SG offender:DAT.PL ‘Like we forgive our own offenders.’ (Bishop Paltarokas, Katalikų tikybos katekizmos) d. O jū̃s, kõ stóvite? oh 2PL.VOC INTERROG.ADV stand:PRS.2PL ‘O you, why are you standing?’ (Šatrijos Ragana, Živile) Taken together, the examples in (41) and (42) show that Lithuanian has consistently nominative–accusative alignment in its case and agreement system, with syncretic (zero) third-person singular and plural verb agreement. Note also that Lithuanian is a pro-drop language. The examples in (43) suffice to illustrate that the Old Lithuanian alignment system is essentially identical to the one found in Modern Lithuanian. (43) a. Apaˆßtałai pásiunte mókitinius ir apostle:NOM.PL send:PST.3SG/PL disciple:ACC.PL CONJ ąntwieczius sawús representative:ACC.PL REFL.POSS ‘The apostles sent the disciples and their own representatives.’ (Daukša, Postilie 246 after Senn 1957: 171)

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

111

b. Abrômas númire ir Pránaßai Abraham:NOM die:PST.3SG/PL CONJ Prophet:NOM.PL ‘Abraham is dead, and (so are) the Prophets.’ (Daukša, Postilie 127 after Senn 1957: 169) c. tu Samaritônas essi ir welinuwą 2SG.NOM Samaritan:NOM.SG be:PRS.2SG CONJ evil.spirit:ACC.SG turi have:PRS.2SG ‘You are a Samaritan and have an evil spirit.’ (Daukša, Postilie 127 after Senn 1957: 169) As regards non-canonical argument realization patterns, on the other hand, Lithuanian is rather permissive. In recent years, a growing body of research dealing with argument realization in Baltic has emerged, much of which focusses on non-canonical constructions in Lithuanian from a synchronic, diachronic, and comparative point of view (see, e.g. Piccini 2008, Barðdal et al. 2012, Holvoet 2013, Serzˇant 2013, Bjarnadóttir 2014a, 2014b, Serzˇant and Bjarnadóttir 2014, and the papers in Holvoet and Nau 2014, 2015, 2016). The examples in (44) illustrate some cases of non-nominative argument marking with one-place predicates. (44) a. mane pykina 1SG.ACC nauseate:PRS.3SG/PL ‘I feel nauseous.’

(after Bjarnadóttir 2014a: 32)

b. miškè áuga visókių mẽdzˇių wood:LOC.SG grow:PRS.3SG/PL all.kinds:GEN.PL tree:GEN.SG ‘Trees of any kind grow in the wood.’ (adapted from Senn) c. Mán buvo šálta 1SG.DAT be:PST.3SG/PL cold:ADV ‘I am cold (lit. to me is cold).’ (adapted from Bjarnadóttir 2014a: 24) These examples illustrate that certain one-place predicates may select a monovalent subject argument in the accusative (44a), genitive (44b), or dative (44c).⁴⁶ The examples in (45) illustrate more complex patterns. (45) a. vaikui pagailo gyvulio child:DAT.SG feel.sorry:PRS.3SG/PL animal:GEN.SG ‘The child feels sorry for the animal.’ (after Bjarnadóttir 2014a: 33) b. Mán skau˜da dañtį 1SG.DAT hurt:PRS.3SG/PL tooth:ACC.SG ‘My tooth hurts.’ (adapted from Magnani 2019: 188) ⁴⁶ It should be noted that the construction with the genitive, illustrated in (47b) alternates with the canonical nominative construction: vaikai áuga ‘the children grow’.

112

PAOL A COTTICELLI AND EYSTEIN DAHL c. Jam veidą beria spuogais: 3SG.DAT face:ACC.SG cover:PRS.3SG/PL pimple:INS.PL ‘His face becomes covered with pimples.’ (after Wiemer and Bjarnadóttir 2014: 321) d. Man skamba ir skamba ausyse 1SG.DAT sound:PRS.3SG/PL CONJ sound:PRS.3SG/PL ear:LOC.PL ‘(It) is ringing heavily in my ears (lit. to me (it) is sounding and sounding in the ears).’

In these examples, we find an experiential predicate with two non-nominative arguments, a pattern analogous to those found in Latin and Ancient Greek. As regards the dative-marked NPs in (45), a point of clarification is in order. While vaikui ‘child’ in (45a) represents a relatively clear-cut case of a dative-marked experiencer argument, as indicated by the verb pagaile˙ti ‘pity’, jam in (45c) rather appears to represent an external possessor, as noted by Wiemer and Bjarnadóttir (2014: 321). At this point, it should be noted that Lithuanian has two ways of expressing predicative possession, one involving an existential sentence with dative marking of the possessor, and another involving the transitive verb ture˙ti ‘have’. By Wiemer and Bjarnadóttir’s (2014) analysis, many cases where a dativemarked experiencer-like NP appears would represent instantiations of the former type of predicative possession construction. While the interpretation of the examples in (45b) and (45d) may seem less obvious, a case could clearly be made for the claim that they also represent external possessors. Bjarnadóttir (2014a: 83– 88) provides a thorough survey of the predicates selecting for non-nominative marking of their highest-ranked argument in Modern Lithuanian. Wiemer and Bjarnadóttir’s (2014) study reports the distribution of the various non-nominative patterns in Lithuanian given in Table 3.14.⁴⁷ The numbers given in Table 3.14 indicate a distinct preference for selecting accusative as the marker of the highest ranked argument, while the dative follows with some margin. In contrast, the other three non-nominative argument marking patterns have a rather marginal distribution. Bjarnadóttir (2014a: 33– 36) explores the syntactic behaviour of first arguments of bivalent predicates with dative case marking, showing that they behave like nominative-marked arguments with respect to reflexivization, raising to subject constructions, control of converbs but not control infinitives.⁴⁸ She provides the examples in (46), which illustrate the dative-marked pattern with reflexive control and raising to subject. ⁴⁷ The meticulous reader will notice that the numbers and percentages given in Table 3.14 differ slightly from those given in Wiemer and Bjarnadóttir (2014: 333, table 3.2). The discrepancy concerns the instrumental and locative, which in the latter are given as 3/1.5 per cent and 2/1 per cent respectively. However, the list in Bjarnadóttir (2014a: 83–88) provides two examples of predicates selecting instrumental case-marking and two selecting locative case marking, rendering the assumption of a small error in Wiemer and Bjarnadóttir’s table plausible. ⁴⁸ See Ambrazas (2006: 373–375) for a discussion of the functions of the infinitive in Lithuanian.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

113

Table 3.14 Lithuanian one- and two-place verbs with non-nominative case marking of HRA (adapted from Wiemer and Bjarnadóttir 2014: 333)

Accusative Dative Genitive Instrumental Locative Σ

N

%

109 77 9 2 3 200

54.5 38.5 4.5 1.0 1.5

(46) a. Katrei reike˙jo savo drauge˙s Katre˙:DAT need:PST.3 REFL.POSS friend:GEN ‘Katre˙ needed her friend.’ (adapted from Bjarnadóttir 2014a: 34) b. man prade˙jo skaude˙ti galvą 1SG.DAT begin:PST.3 hurt:INF head:ACC ‘My head began to hurt/It hurts me with regard to (my) head.’ (adapted from Bjarnadóttir 2014a: 35)⁴⁹ Before concluding this section, we would like to draw attention to the fact that Lithuanian has a P-oriented resultative verbal adjective in -t-. In the grammatical literature, this form is classified as past passive participle (PPP) (see, e.g. Mathiassen 1996: 162), reflecting the fact that it represents an essential element of the past passive forms of the paradigm of bivalent or trivalent predicates, as illustrated in (47a). Interestingly, however, the verbal adjective in -t- can also be formed from monovalent predicates, as shown in (47b), the resultant construction representing an impersonal passive with the so-called neuter or default form of the participle, which characteristically has an evidential meaning (Ambrazas 2006: 282). (47) a. Àš ne kar̃ tą esù tė́vo išbártas 1SG.NOM NEG time:ACC.SG be:PRS.1SG father:GEN.SG scold:PPP.NOM.SG ‘I have been scolded by father more than once.’ (after Ambrazas 2006: 252) b. Jõ jau˜ bùvo išė̃jęta 3SG.GEN already be:PST.3 go:PPP.DF ‘He was already gone (lit. By him was already gone).’ (Ambrazas 2006: 282) ⁴⁹ The interpretation of man ‘1SG.DAT’ as a first argument or subject in this example is motivated by the fact that the verb skaude˙ti ‘hurt’ seemingly only requires a dative-marked argument, other arguments seemingly being optional (Wiemer and Bjarnadóttir 2014: 332).

114

PAOL A COTTICELLI AND EYSTEIN DAHL

These examples illustrate that the P-oriented verbal adjective in -t- is fully incorporated in the verbal voice system as a passive participle in Lithuanian. In this section, we have briefly reviewed the most important features of the Lithuanian argument realization system. We have seen that Lithuanian has a consistently nominative–accusative alignment system involving both case and agreement, and that the language is permissive regarding non-canonical argument realization patterns, involving both alternation and consistent lexical selection of non-canonical case marking.

3.3 Continuity and innovation in alignment and argument marking In the previous section we explored some dimensions of the alignment systems of Latin, Ancient Greek, Vedic Sanskrit, Hittite, Classical Armenian, and Lithuanian. We have seen that, although they belong to the same linguistic family, they display considerable variation in this realm of grammar. In this section, we explore whether any interrelations between the various parameters exist, and outline an explanation. A first set of problems concerns the relationship between split alignment and mixed alignment. In Latin, Ancient Greek, Vedic, and Hittite we find a nominal and pronominal case-marking system with a split between accusative and neutral alignment, although the distribution of Hittite neuters is suggestive of a split between accusative and semantic alignment. In contrast, Classical Armenian and Lithuanian show unitary nominative–accusative alignment in their nouns and pronouns. As regards agreement, on the other hand, most of the languages display predominantly accusatively oriented verb agreement, except for Ancient Greek and Hittite, where we arguably find a split between accusative and neutral verb agreement, the latter being most clearly visible with neuter plural nouns, which, however, show alternation between neutral and accusative agreement in Homeric Greek. In Latin, Vedic, and to some extent Homeric Greek, neuter nouns and pronouns show mixed alignment with neutral case marking and accusative agreement, while they show consistently neutral alignment in Hittite and Classical Greek. Furthermore, while non-neuter nouns and pronouns are compatible with all three argument functions in all the languages with a noun–class-based alignment split, bare neuter nouns are restricted to second argument function of bivalent verbs and subject function of monovalent verbs in Hittite, but appear in all three argument functions in Latin, Ancient Greek, and Vedic. We would like to suggest that these differences reflect different stages in a generalization process, which progressively increases the number of accusative features and which we label ‘accusativization’ for convenience. In view of this, we take the restriction

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

115

of neuter nouns to second argument function of bivalent verbs and monovalent argument function in Hittite to be an archaic feature, in line with the claims set out in Chapter 2. Thus, an early stage in the accusativization process consists in a change which removes this restriction, which sets Hittite apart from the other languages showing a distinct neuter noun and pronoun class. In Hittite, the limits imposed by the restriction create favourable conditions for the grammaticalization of an ergative case marker with neuter nouns (see Goedegebuure 2018, Dahl 2021a, and Chapter 4 in the present volume for details). Along similar lines, we assume that the lack of number agreement with plural forms of neuter nouns, found in Hittite and Ancient Greek, but also as relics in Vedic, represents an archaic feature, as argued in Chapter 2. We therefore propose that the analogical spread of plural agreement to neuter plural subjects represents a second stage in the accusativization process. In Section 3.2.2, we saw that plural forms of neuter nouns alternately co-occur with singular and plural verb forms in Homeric Greek, and we deem it likely that this alternation reflects an incipient stage of this analogical extension of accusative agreement to neuter nouns and pronouns. The hypothesis that this change postdates the removal of the restriction of neuters in first argument function of bivalent predicates finds some support in the fact that neuter nouns appear in this function in Homeric Greek, as shown in Section 3.2.2 and Chapter 2. From this perspective, Vedic and Latin represent more advanced stages of accusativization, as it were, than Homeric Greek, where the innovative agreement pattern alternates with the archaic pattern, which is still predominant. Interestingly, the pattern with singular agreement on the verb is generalized in Classical Greek, indicating that the innovative pattern resulted in a failed change in this branch, despite its prevailing in Vedic and Latin. This may be interpreted as a case where harmonization as defined above prevails over generalization, that is, a consistently neutral alignment pattern is preferred over a mixed neutral–accusative one. A third stage in the accusativization process is the elimination of the noun– class-based split between neuter and non-neuter nouns, which sets Classical Armenian and Lithuanian with their consistent distinction between nominative and accusative case marking throughout the nominal and pronominal system, distinguishing them from the other languages under scrutiny. Table 3.15 gives a schematic representation of the different stages of accusativization in the languages in this study. The above observations suggest that we may discern a general drift towards accusativization across the Indo-European languages, which primarily affects the distribution of neuter nouns and the formal systems of case marking and verb agreement. This process involves analogical generalization of already established patterns. Another mechanism of alignment change involves grammaticalization of newly emerging morphological markers, which sometimes result in developments that run counter to accusativization. We have already mentioned the development

116

PAOL A COTTICELLI AND EYSTEIN DAHL Table 3.15 Accusativization across Indo-European languages (preliminary version)

of a new ergative case marker in Hittite as an example of this, resulting in a noun– class-based split ergative–accusative alignment system (Goedegebuure 2018, Dahl 2021a, Chapter 4 in the present volume). Here, new case morphology arises in response to the restricted syntactic distribution of neuter nouns, which is resolved via extension in the other languages under discussion (Dahl 2021a). Another development pattern involves the grammaticalization of P-oriented verbal adjectives into perfects, as seen in Indo-Aryan and Armenian. Dahl (2021a) claims that the grammaticalization of the Vedic verbal adjective in -tá- into a perfect with ergative alignment is accommodated by the fact that the main aspect stems of the verb gradually lose their aspectual distinctions, developing into tense categories with neutral aspectual value (Dahl 2015; Chapter 2). Meyer (2017, Chapter 9 in the present volume) makes a strong case for the claim that the Classical Armenian periphrastic perfect with tripartite alignment arose as a result of language contact with West Middle Iranian languages. Again, one observes that new alignment patterns can arise via grammaticalization, but it should be noted that developments of this kind are language-specific in nature, unlike accusativization, which seems to be a more general tendency in the Indo-European family. This is, among other things shown by the fact that the Vedic adjective in -tá- has exact parallels in Latin and Lithuanian, where this formation has been incorporated in the voice system, representing a mediopassive and a passive participle, respectively. Analogous observations apply to Classical Armenian, where there is a periphrastic passive construction employing the participle in -eal, as noted in Section 3.2.5.⁵⁰ ⁵⁰ Given that Latin, Armenian, and Lithuanian show advanced stages of accusativization, it is tempting to speculate that the development of the P-oriented verbal adjectives into voice categories is somehow correlated with a highly developed accusative morphosyntax. While this is certainly an attractive possibility, an assumption along such lines would have to specify under what conditions this specific type of grammaticalization might be expected to arise and whether such conditions might be plausibly assumed to have obtained in the immediate prehistory of these languages. An examination of these matters is far beyond the scope of the present chapter, and will have to be undertaken elsewhere.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

117

Another dimension of the data we have reviewed above concerns the notion of subject as generalized privileged syntactic argument. We have not aimed at determining the full range of constructions that contribute to determining the generalized privileged syntactic argument in each of the languages under consideration, since this would be impossible within the limits of a single chapter. Given that this topic is far better studied in the case of some of the languages than others, the picture we have given necessarily remains incomplete. Among the findings we deem important is the fact that control infinitives represent a reliable subject test in Latin, Greek, and Lithuanian, where we find restricted neutralization of monovalent subjects and bivalent first arguments with the infinitives, but not in Hittite, Vedic, and Armenian, where there appears to be unrestricted neutralization of the core arguments with infinitives. This difference represents an important typological parameter of variation, which we take to reflect diachronic differences between the two groups of languages under consideration, the languages where infinitives show unrestricted neutralization of core arguments being more archaic.⁵¹ This assumption is in line with the more general suggestion made in Conti (2010: 269), that the subject was a somewhat underdeveloped or underspecified category in Greek and other archaic Indo-European languages, which only gradually developed into the kind of fully fledged subject category one finds, for instance, in Latin. Following this hypothesis, which remains to be systematically explored, a process involving the development of syntactic subjecthood operated across the Indo-European languages, parallel to the accusativization process schematically represented in Table 3.16. This process may be labelled ‘subjectization’ for convenience and consists in the gradual development of subjecthood through an increase in the number of constructions picking out the monovalent subject and the bivalent first argument as their privileged syntactic argument. While both processes ultimately contribute to increasing the number of accusative features, it is not self-evident that they operate in tandem, a question that will have to be explored in future research. Since unrestricted neutralization with control infinitives is a parameter of variation relevant for the classification of languages as showing a higher or lesser degree of development in subjecthood/accusative morphosyntax, we include this in the updated Table 3.16. Finally, we have seen that all the languages under consideration show examples of non-nominative or non-canonical subjects. However, there is considerable dif⁵¹ At this point, we would like to emphasize that it is controversial whether there was an infinitive category in Proto-Indo-European. Many of the infinitive constructions found in the individual branches seem to be branch-specific or only appear in a small number of branches (Fortson 2010: 107–108). Infinitive forms often have the appearance of being branch-specific innovations, having arisen via independent grammaticalization of fossilized case forms of different verbal nouns and this may contribute to explaining their different behavioural patterns in the languages that form the empirical base for this study.

118

PAOL A COTTICELLI AND EYSTEIN DAHL Table 3.16 Accusativization/subjectization across Indo-European languages

ference among the languages as regards the extent to which such constructions are employed. For example, Latin and Lithuanian show a far richer range of nonnominative subject constructions than Hittite or Vedic. It is striking that Latin and Lithuanian, which are among the languages in our sample that show the highest degree of accusativization/subjectization, are also most permissive with regard to non-canonical subject constructions. This observation may at first seem quite paradoxical, since one might expect languages with a high degree of accusativization/subjectization to show uniformity in morphosyntactic alignment, that is, to be restrictive regarding non-canonical argument realization patterns. This is not necessarily the case, however, since even closely related languages with typologically similar properties may differ considerably in the extent to which they show non-canonical marking, as explored in a recent study by Haspelmath (2015). Based on these observations, we hypothesize that there is a correlation within the IndoEuropean family between the degree of accusativization/subjectization and the degree of acceptance of non-canonical subject marking. If this is correct, noncanonical subjects are licensed, as it were, by uniformly accusative morphosyntax and a well-defined notion of subjecthood, which we above all find in Latin and Lithuanian. Given that the evidence for non-canonical subject constructions in the other languages under scrutiny is sporadic and rather scarce, and, in some cases have a secondary character motivated by language-internal factors, we remain

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

119

Table 3.17 Accusativization/subjectization across Indo-European languages

sceptical as to whether this construction type can be reconstructed for Proto-IndoEuropean. Table 3.17 presents a final survey of the distribution of the properties thus far outlined. Table 3.17 summarizes the observations made in the above sections. While it provides a reasonably clear picture of the general development towards a greater number of accusativization/subjectization features, one immediately notes that Ancient Greek and Classical Armenian show some divergence. We have already dealt with the fact that Ancient Greek shows a diachronic preference for the pattern with neuter plural subjects and singular verb forms, which we analysed as a case where harmonization prevails over generalization. As regards the properties of having infinitives showing non-restricted neutralization and only sporadic cases of non-canonical subject marking in Classical Armenian, it is tempting to suggest that they are interrelated. Section 3.2.5 demonstrated that uncontroversial cases of non-canonical subject marking mainly appear in the periphrastic perfect, where they seem to be motivated by a general tendency to replace the tripartite alignment pattern associated with that construction with an accusative pattern. Apart

120

PAOL A COTTICELLI AND EYSTEIN DAHL

from this, no clear-cut examples of non-canonical subject constructions appear, a fact that within the present framework may be hypothesized to be a consequence of a somewhat underdeveloped notion of subjecthood in the language, reflected, among other things, in the fact that the infinitive shows unrestricted neutralization of core arguments. The fact that Classical Armenian does not show these features in spite of figuring among the languages with a high degree of morphosyntactic accusativity as regards case and agreement shows that accusativization and subjectization are independent processes that may but need not proceed in a parallel fashion.

3.4 Summary and conclusion In this chapter, we have explored the relationship between split alignment, mixed alignment, and the spread of accusative morphosyntax based on comparative evidence from Latin, Ancient Greek, Vedic Sanskrit, Hittite, Armenian, and Lithuanian. A central aim of ours has been to establish how these different factors interact in the development of languages within one linguistic family, concerning both the historically attested language-specific development patterns of each language and their function in a comparative–historical perspective. We hope to have made a convincing case for the claim that the various languages under consideration reflect different stages in a general development towards more consistently accusative morphosyntax and at the same time illustrate that this tendency towards accusativization does not necessarily result in consistently accusative alignment. Perhaps the most striking finding of this work is the amount of variation within the realm of alignment typology within one linguistic family both concerning synchronic and diachronic factors, a fact suggesting that this is an area of grammar which, at least in some language families, shows a strong propensity to diachronic change. Our discussion of the data has mostly had a qualitative character, and it remains a task for future research to explore the extent to which the main claims made here find support in quantitative data.

Text editions of Greek, Latin and Vedic texts Cic. Att.: M. Tullius Cicero, Epistulae ad Atticum, in Cicero’s Letters to Atticus, ed. D. R. Shackleton Bailey, Leipzig, Teubner, 1965–1968. Cic. Cat.: M. Tullius Cicero, In Catilinam, in M. Tulli Ciceronis Orationes Vol I, ed. Albert C. Clark, Oxford: Clarendon, 1905.

THE SPRE AD OF ACCUSATIVE MORPHOSYNTAX

121

Cic. Fam.: M. Tullius Cicero, Epistulae ad Familiares in M. Tulli Ciceronis Epistulae Vol I, ed. W. S. Watt, Oxford: Clarendon, 1982. Cic. Amic.: M. Tullius Cicero, Laelius de Amicitia, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, Part 4, vol. 3, ed. C. F. W. Mueller, 1890. Cic. Quinct.: M. Tullius Cicero, Pro Publio Quincto, in M. Tulli Ciceronis Orationes Vol IV, ed. Albert C. Clark, Oxford: Clarendon, 1909. Eur. Hel.: Euripides, Helena in Euripides Fabulae Vol III, ed. James Diggle, Oxford: Clarendon, 1994. Eur. Her.: Euripides, Heracleidae in Euripides Fabulae Vol II, ed. James Diggle, Oxford: Clarendon, 1981. Eur. Med.: Euripides, Medea in Euripides Fabulae Vol I, ed. James Diggle, Oxford: Clarendon, 1984. Hom. Il.: Homerus, Ilias. Recensuit Martin L. West, Mu¨nchen: Saur, 1998-2000 (2.vol) Hom. Od.: Homerus, Odyssea. Recognovit P. von der Mu¨hll, Mu¨nchen: Saur, 2005 [1945] Ov. Met.: P. Ovidius Naso, Metamorphoses, in Metamorphoses in Two Volumes, ed. F. J. Miller and G. P. Goold, London, Loeb Classical Library, 1977–1984. Plaut. Amph.: T. Maccius Plautus, Amphitruo, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. As.: T. Maccius Plautus, Asinaria, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Aul.: T. Maccius Plautus, Aulularia, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Bacch.: T. Maccius Plautus, Bacchides, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Cas.: T. Maccius Plautus, Casina, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Curc.: T. Maccius Plautus, Curculio, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Fr.: T. Maccius Plautus, Fragmenta, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905. Plaut. Merc.: T. Maccius Plautus, Mercator, in T. Macci Plauti Comoediae. Tomus I, ed. W. M. Lindsay, Oxford: Clarendon, 1904. Plaut. Mil.: T. Maccius Plautus, Miles Gloriosus, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905. Plaut. Most.: T. Maccius Plautus, Mostellaria, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905. Plaut. Poen.: T. Maccius Plautus, Poenulus, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905.

122

PAOL A COTTICELLI AND EYSTEIN DAHL

Plaut. Trin.: T. Maccius Plautus, Trinummus, in T. Macci Plauti Comoediae. Tomus II, ed. W. M. Lindsay, Oxford: Clarendon, 1905. RV – Rigveda, Theodor Aufrecht (ed.) Die Hymnen des Rigveda. 2. Auflage. Bonn: A. Marcus, 1877 (2. vol.)

4 The origin of ergative case markers The case of Hittite revisited Silvia Luraghi and Guglielmo Inglese With an appendix by Petra Goedegebuure

4.1 Introduction It has long been observed (Benveniste 1962, Laroche 1962) that in Hittite neuter nouns feature a dedicated form in -ants when functioning as the subject of a transitive verb. When they function as subjects of intransitive verbs, they show the nominative–accusative form common to other ancient Indo-European languages. In this way, Hittite marks a distinction between arguments in A-position on the one hand, and arguments in S and O position on the other. This state of affair has prompted an intense debate as to whether neuter ant-nouns should be interpreted as carrying a derivational -ant- suffix plus the common gender nominative ending -s, or as instantiating a system of NP split ergativity, with -ants behaving as a fully fledged ergative case marker. Over several decades, proponents of the derivational hypothesis have argued against the more recent inflectional hypothesis outlined in Garrett (1990a, 1990b), which rests on the assumption that the -ants suffix taken by neuters in A-position originated from an ablative case ending, and must be analysed as an inflectional morpheme at all language stages. Goedegebuure (2012, 2018) has cast new light on the issue, by taking a diachronic perspective on the Hittite data. Goedegebuure argues that both schools of thought are essentially correct, but that their observations apply to different stages of the language: the suffix started out as derivational in Old Hittite (OH), and only later on was reanalysed as an inflectional case marker. In this chapter, we build upon Goedegebuure’s findings and investigate further the origin of the Hittite ‘ergative’ -ant- and the change from derivation to inflection. To do so, we survey the occurrence of -ant- suffixed nouns in the oldest Hittite texts and set out to establish the original meaning of -ant-, also based on comparison with other -ant- formations in Hittite as well as with cognate suffixes in other ancient Indo-European languages. As we show, once properly understood, Hittite provides unique insights on the rise of gender-based split ergativity, and further enhances our understanding of the possible origins of Silvia Luraghi, Guglielmo Inglese, and Petra Goedegebuure, The origin of ergative case markers. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl. Oxford University Press. © Silvia Luraghi, Guglielmo Inglese, and Petra Goedegebuure (2022). DOI: 10.1093/oso/9780198857907.003.0004

124

SILVIA LURAGHI ET AL.

ergative case markers in the languages of the world. Indeed, as we will show in Section 4.5, the Hittite development follows a path that tends not to be included among the various grammaticalization paths identified in the literature (e.g. McGregor 2017). The assumption of the derivational origin of the ergative ending has further implications for the reconstruction of the original state of affairs concerning alignment in Hittite and in Proto-Anatolian. Indeed, as there is no evidence for nominative–accusative alignment of neuter nouns at any language stages, one should likely reconstruct an earlier gender-based active alignment system. In other words, Proto-Anatolian should be reconstructed as having two classes of nouns, an active class (corresponding to common gender) and an inactive one (corresponding to neuter). The latter could never occur in A function. Notably, this reconstruction is in accordance with well-known facts about the distribution of neuter and non-neuter nouns in Proto-Indo-European (PIE) (Meillet [1921] 1948, Gamkrelidze and Ivanov 1995). This chapter is organized as follows. In Section 4.2, we review previous scholarship on Hittite split ergativity, including arguments for and against the derivational and the inflectional hypotheses. In Section 4.3, we discuss the semantics of other formations displaying -ant-suffixation in Hittite and show how they relate to ‘ergative’ -ant- according to different scholars. We then illustrate our views on the semantics of the -ant- suffix and its development. Section 4.4 focuses on the innerHittite development of -ant- nouns over time. We show that the evidence from Hittite texts, once chronologically ordered, indicates that suffix was derivational in OH but became an inflectional ergative ending by New Hittite (NH) times.¹ After discussing evidence from other ancient Anatolian languages, we outline a tentative reconstruction of alignment in Proto-Anatolian, and hint to its implications for the reconstruction of PIE. We also suggest that the emergence of ergativity in Hittite and possibly in Lycian and Luwian should be taken as pointing towards areal convergence among Indo-European and non-Indo-European languages of ancient Anatolia. In Section 4.5 we argue that the development of Hittite -antsheds new interesting light on the dynamics of the processes that lead to the emergence of systems of split ergativity, and frame the development of Hittite -antwithin the diachronic typology of ergative case markers. Section 4.6 contains the conclusion.

¹ In this chapter, we follow the chronology established in the CHD for Hittite texts. It must be noted that the Hittite language is usually regarded as showing three distinct stages: O(ld) H(ittite), M(iddle) H(ittite), and N(ew) H(ittite). As scribal habits were often such that texts were partially updated when copied from originals of older stages, this distinction must be matched with the dating of the script, whereby a distinction is made among O(ld) S(cript), M(iddle) S(cript), and N(ew) S(cript). Only sources from OH/OS can be considered fully reliable for the earliest stage, while copies are not always reliable.

THE ORIGIN OF ERGATIVE CASE MARKERS

125

4.2 Hittite neuter ant-nouns: inflection or derivation? This section discusses the -ant- suffix. As this suffix is sometimes described as ‘animatizing’, a preliminary description of the Hittite gender system is in order to understand the relation between animacy and grammatical gender. We discuss this issue in Section 4.2.1. Then, we provide some more details on split ergativity in Hittite (Section 4.2.2), and briefly review previous scholarship on the nature of the suffix -ants, starting with derivational approaches (Section 4.2.3), and then proceeding to inflectional approaches (Section 4.2.4).

4.2.1 Gender in Hittite Hittite, as well as other Anatolian languages, features a two-way gender system with a common gender and a neuter. The interpretation of the Anatolian data with respect to the reconstruction of the PIE gender system has been a matter of disagreement, which need not concern us here: it only needs to be remarked that the name ‘common gender’ was originally meant to indicate conflation in a single gender of the PIE masculine and feminine. Later research has shown that Anatolian likely reflects the PIE situation, and that at an even earlier stage of Pre-PIE the two-way gender system was animacy-based, with a class of animate vs a class of inanimate nouns.² A reflex of this early distribution remains in the neuter of Hittite as well as of most other Indo-European languages, which contains almost exclusively inanimate nouns. Gender assignment with other gender(s), in turn, is largely arbitrary: indeed, the Hittite common gender contains animate as well as (numerous) inanimate nouns, in much the same way as the feminine and masculine gender of other Indo-European languages.³ Animacy is not the only feature that is relevant for Hittite neuter nouns. Besides being mostly inanimate, neuter nouns most often show a low degree of individuation: in this respect, the Hittite neuter shares the characteristic features shown by nouns that can be reconstructed as constituting the core of this gender in PIE (Ostrowski 1985, Matasovic´ 2004, Luraghi 2009b, 2011). As Goedegebuure (2018) points out, Hittite neuter nouns are often uncountable, referring to masses and collectivities. To these, abstract nouns also need to be added, as various suffixes that derive abstract nouns from other types of bases assign neuter gender. An example ² This reconstruction dates back to Brugmann (1891)—see Luraghi (2009a). For a survey of the decades-long discussion on the Anatolian gender system, see Ledo-Lemos (2003: 41–94) and Matasović (2004: 36–41). See also Chapter 2 in this volume. ³ On gender and animacy in Hittite, see further Starke (1977), Weitenberg (1987), (1995), and Zeilfelder (2001). Weitenberg speaks of a three-class gender system combining grammatical gender and animacy and including: Class I animate and grammatically common gender nouns, Class II inanimate and grammatically common gender nouns, Class III inanimate and grammatically neuter nouns (on animate neuter nouns, see further on in this section).

126

SILVIA LURAGHI ET AL.

is the suffix -tar, which forms abstract and collective nouns that can also refer to animates, as in ḫuidar ‘animal’, ‘cattle’, and antuḫšatar- ‘humanity’, ‘population’ (Kloekhorst 2008: 226; see example (7)). In such cases, low individuation can override animacy. As we will see in the next section, low individuation plays a role in the spread of the -ant- suffix as attested in Hittite texts.

4.2.2 Hittite split ergativity The existence of a system of split ergativity based on nominal gender is one of the most debated topics in Anatolian linguistics. Discussion on this issue was initiated by Laroche’s (1962) seminal paper devoted to Hittite -ant- formations. In his paper, Laroche pointed out a notable syntactic distribution of the suffix -ant-, which he described as derivational, whereby neuter nouns require to be moved into common gender by the suffix -ant-s/-ant-es (spelled -an-za/- an-te-eš) containing the common gender nominative endings -s (singular) and -es (plural) when they occur as the subject of a transitive verbs. Following this view, gender motion allows neuter nouns to behave as common gender nouns, which freely occur with their regular nominative case ending (see the discussion of example (8)).⁴ This pattern can be illustrated by comparing the behaviour of the common gender noun LUGAL-u-⁵ ‘king’ with the neuter noun paḫḫur ‘fire’ in (1) and (2). (1)

a. LUGAL-uš 3-ŠU aīš=šet king.NOM 3.times mouth(N).NOM/ACC=3SG.POSS.NOM/ACC.N ārri wash.PRS.3SG ‘The king washes his mouth three times.’ (KBo 17.1+ i 15) b. LUGAL-uš arḫa paizzi king.NOM away go.PRS.3SG ‘The king goes away.’

(KBo 20.10+ i 2)

c. ta LUGAL-un šuppiaḫḫi CONN king.ACC purify.PRS.3SG ‘And he purifies the king.’

(KBo 20.10+ ii 8)

⁴ Three occurrences of neuter nouns in A function are mentioned in the literature. In the case of ḫandaiš- (KBo 3.23 i 6; OH/NS), Melchert (2017a) argues that this is in fact a common gender noun. Slightly more compelling is the case of šāwar (KUB 30.34 iv 4–10; MH/NS), which however as pointed out in Melchert (2017a) occurs in a string of sentence with verb gapping, that is ‘there is no collocation of šāwar with a transitive verb and its direct object’. Finally the neuter noun partawar (KUB 17.10 ii 35ʹ–36ʹ, OH/MS) is argued to be syntactically a direct object in Goedegebuure (2018). ⁵ As commonly is done in the transcription from cuneiform script, capital letters indicate sumerograms and italics capitals indicate akkadograms.

THE ORIGIN OF ERGATIVE CASE MARKERS

127

(2) a. kēdani=ma paḫḫur urāni DEM.DAT=PTCL fire(N).NOM/ACC burn.PRS.3SG.MID ‘And near to this one burns a fire.’ (KUB 23.59 iii 9) b.

LÚ

MUḪALDIM=kan ḫašši paḫḫur warnuzzi cook=PTCL hearth.DAT fire(N).NOM/ACC make.burn.PRS.3SG ‘The cook lights the fire on the hearth.’ (KUB 11.35 v 16)

c. man=an paḫḫuwenanza arḫa warnuzi IRR=3SG.ACC fire.ant.NOM away make.burn.PRS.3SG ‘May the fire burn him completely.’ (KBo 32.14 ii 6–7) When occurring as subject, the common gender noun LUGAL-u- ‘king’ takes the nominative case ending -s, both with the transitive verb ārr- ‘wash’ in (1a) and the intransitive verb pai- ‘go’ in (1b). By contrast, when it occurs as direct object of the transitive verb šuppiaḫḫi, it occurs in the accusative case form LUGAL-un, as in (1c). The neuter noun paḫḫur displays a different distribution. The nominative/accusative neuter case form paḫḫur is used when the noun occurs either as the subject of the intransitive verb ūr- ‘burn’, as in (2a), or when it occurs as the object of the transitive causative counterpart warnu- ‘make burn’, as in (2b). Conversely, in (2c), the noun occurs as the subject of transitive warnu- ‘make burn’ and appears in the paḫḫuwen-anza form. As we discuss extensively in the next sections, it is debated whether this form should be analysed as common or neuter gender. The distribution of the two nouns can be summarized as in Table 4.1. This behaviour is not limited to Hittite and is also displayed by the Luwian antis and Lycian -ẽti suffixes. We return to Luwian and Lycian data and on their bearing for the reconstruction of split ergativity in the Anatolian language family in Section 4.4.3. Even though these facts are basically undisputed, common agreement on their interpretation is far from being reached, still fuelling a lively scholarly debate.⁶ Approaches to the subject can be broadly distinguished into those that interpret Table 4.1 Coding of A, S, and O with common and neuter nouns

A S P

LUGAL-u- ‘king (c.)’

paḫḫur ‘fire (n. or c.?)’

paḫḫur ‘fire’ (n.)

LUGAL-uš LUGAL-uš LUGAL-un

paḫḫuwen-anza — —

— paḫḫur paḫḫur

⁶ See Dardano (2010, 2013), Melchert (2011b), and Goedegebuure (2012, 2018) among most recent studies. A detailed discussion on the various positions on the topic in earlier scholarship can be found in Rizza (2010, 2014).

128

SILVIA LURAGHI ET AL.

Hittite -ant- as a derivational morpheme and those that treat -ants as an inflectional case ending. We review the pros and cons of the two approaches in Sections 4.2.3 and 4.2.4.

4.2.3 Derivational approaches Derivational approaches, also called lexicalistic approaches (Dardano 2013), set out from the assumption that the transitive subject position is only available to common gender nouns. As a consequence, neuter nouns must be transferred to common gender by means of the suffix -ant-, and accordingly receive the common gender nominative case ending -s when they occur as subjects of a transitive verb. Individual scholars differ as to the specific semantic contribution of the suffix. According to some, the suffix has a personifying (Benveniste 1962, Neu 1989, Dardano 2010) or animatizing function (Laroche 1962, Carruba 1992, Dardano 2013, Teffeteller 2015), or it can be a sort of singulative, with an individualizing function (Oettinger 2001, Zeilfelder 2001, Josephson 2004, Goedegebuure 2018). We will return to these accounts in Section 4.4.2 when discussing the origin of the -ant- suffix. In any case, all proponents of the derivational hypothesis agree in assuming a semantic motivation for its occurrence. Evidence adduced in favour of the derivational nature of -ant- as a common gender suffix includes agreement with common gender adjectives and determiners, e.g. kāš tuppianza ‘this (c) tablet’, along with the fact that -ant-nouns display case forms other than the nominative, as shown in (3). (3)

kāšata=šmaš=kan utniyandan lālūš dāḫḫ[un] INTJ=3PL.DAT=PTCL country.ant.GEN.PL tongue.ACC.PL take.PST.1SG ‘Look! I took from you the tongues of the lands.’ (KBo 17.1+ i 11’, OH/OS)

Example (3) contains the genitive plural form utniyandan from utniyant‘country’, ‘population’, an -ant- formation based on utne- ‘land’. This is the lexical item that shows the biggest number of non-nominative cases, and the nature of its formation has been discussed, as we will show in Sections 4.3 and 4.4.2. Moreover, a limited number of common gender nouns may optionally take the suffix -ant- when occurring in agent function. An example is ḫaššanza (spelled GUNNI-anza) from common gender ḫašša- ‘hearth’ in (4). (4)

tagā[nzipaš] earth.NOM 4 4

ḫuimpaš É.ŠÀ-nanza ḫ.NOM interior.ant.NOM ḫalḫaldumma[rianteš?] corner.ant.NOM.PL

GUNNI-anza hearth(c).ant.NOM

THE ORIGIN OF ERGATIVE CASE MARKERS

129

É

ḫilaš KÁḪI.A -eš arḫa tarnan[du] courtyard.NOM gate(PL).NOM.PL away release.IMP.3PL ‘[May] the ear[th], the ḫuimpa-, the interior, the hearth, the four corners, the courtyard, (and) the gates release (the evil things).’ (KUB 7.41 i 19–21, MH/MS?) Finally, in several occurrences -ant- forms function as subjects of intransitive verbs as KU.BABBAR-anza ‘silver’ in (5). (5)

KU.BABBAR-anza anda parna=šša pait silver(c).ant.NOM inside house.DIR=3SG.POSS.DIR go.PST.3SG ‘Silver entered his house.’ (KUB 17.4 8’, MH/NS)

Notably, the form KU.BABBAR-anza ‘silver’ in this passage refers to a mythological hero, thus supporting the animatization/personification theory. In the same text, this noun also occurs in the vocative. An interesting example that also points towards the derivational nature of the suffix is (6). (6)

kašš=a=za URU-az (var. URU-aš) DEM.NOM=CONJ=REFL city.ant.NOM parnanzašš=a(var. parnaš) UDU.A.LUM DÙ-ru house.ant.NOM=CONJ ram become.IMP.3SG.MID ‘Let both this town and house become a ram.’ (KUB 41.8 iv 30, MH/NS)

In example (6), two -ant- forms are coordinates, URU-az ‘town’ from the common gender noun ḫappiriya- (here spelled with the Sumerogram URU) and parnanza ‘house’ from the neuter noun per, and function as subject of an intransitive verb. The duplicate has the nominative forms URU-aš and parnaš in place of the -antforms, thus showing that the suffix is optional in this context, hence sharing a feature of derivation rather than inflection. The form parnaš is also interesting because it can be taken as the nominative of the common gender noun parnafeaturing thematization as a gender motion strategy partly interchangeable with -ant- suffixation, or as a free-standing genitive. We return to the interpretation of this form in Section 4.4.1.

4.2.4 Inflectional approaches Proponents of the inflectional approach argue that -ant- is not a derivational morpheme, and that the forms -ants and -antes should not be analysed as containing the suffix plus the common gender nominative endings (-ant-s and -ant-es). Rather, these forms should be analysed as inflectional case endings, constituting the ergative case in the paradigm of neuter nouns. This theory was first outlined by Garrett (1990b). In his view, Hittite attests to a split-ergative system, with common

130

SILVIA LURAGHI ET AL.

gender nouns inflecting according to a nominative–accusative pattern and neuter nouns following an ergative–absolutive pattern. Following this approach, the use of -ants is essentially driven by syntactic considerations. Specifically, proponents of the inflectional approach deny a connection of -ants with agentivity/animacy, and instead argue that the occurrence of the suffix only depends on grammatical gender. An often-cited piece of evidence in support of this interpretation is that neuter nouns that indicate animate referents, such as antušḫatar ‘population’, still require the -ant- ending when occurring as transitive subject, as in (7). (7)

antuḫšannanza URUḪI.A =ŠUNU population.ant.NOM city(PL)=3PL.POSS ‘The population reoccupied their towns.’

EGIR-pa eppir back take.PST.3PL (KBo 8.77 rev. 14)

On the other hand, the numerous common gender nouns that indicate inanimate entities appear in their nominative form even when functioning as subject of a transitive verb. Consider example (8). (8) parnanza=at tarnau ištarniyaš=at house.ant.NOM=3SG.ACC.N release.IMPV.3SG inner.NOM=3SG.ACC.N GIŠ annašnanza tarnau luttanza=at a.ant.NOM release.IMPV.3SG window.ant.NOM =3SG.ACC.N tarnau ištarniyaš=at ḫilaš release.IMPV.3SG inner.NOM=3SG.ACC.N courtyard.NOM tarnau release.IMPV.3SG ‘May the house release it, may the inner a. release it, may the window release it,… may the inner courtyard release it.’ (KUB 17.10 iv 10–11, OH/MS) The passage in (8) contains a series of transitive sentences each featuring the imperative tarnau ‘may release’, the accusative object pronoun =at ‘it’ and an inanimate subject. In the first three sentences, the subjects are neuter gender nouns and systematically occur in their -ant- form. By contrast, in the fourth sentence the common gender noun ḫila- ‘courtyard’ occurs in its base form with the nominative case. As discussed by Melchert (2011b), this distribution can hardly be explained by recurring to an ‘animatizing’ force of the suffix -ant-, as inanimate but common gender ḫila- does not require the suffix (see Melchert 2011b for discussion and more examples). Mismatches between animacy and grammatical gender neatly show how the distribution of -ants is mostly sensitive to the latter for nouns in A position, that is, it is triggered by grammatical rather than semantic factors.⁷ ⁷ Moreover, Goedegeburre (2012) also mentions that ant-nouns usually trigger neuter agreement with anaphoric pronouns, thus bearing further evidence that the suffix does not transfer the noun to a different gender class (but see Dardano 2013: 52 and Goedegebuure 2018 for counterarguments).

THE ORIGIN OF ERGATIVE CASE MARKERS

131

Example (8) is also instructive as to the patterns of agreement between -antnouns and adjectival modifiers. As we have already observed, common gender agreement of -ant- nouns with demonstratives is often taken as a piece of evidence in favour of the derivational nature of -ant-. The same agreement pattern also applies to adjectives: in (8) both common gender ḫila- and neuter ergative annašnanza trigger common gender agreement with the adjective ištarniyaš ‘inner’. However, this is not the entire picture, as occasionally neuter -ant- noun in A position trigger agreement with adjectives that are likewise suffixed in -ant-, as in example (9), in which ergative [utnēant]eš ‘land’ is modified by the ant-suffixed adjective araḫzenanteš. (9)

kinun=a araḫzenanteš [utnēant]eš ḫūmanteš KUR now=PTC bordering.ant.NOM.PL land(N).ant.NOM.PL all.NOM.PL land URU KÙ.BABBAR-TI [walḫ]anneškuwan dāer Hatti hit.IPFV.IPFV.SUP put.PST.3PL ‘But now all the bordering lands began to attack the land of Hatti.’ (KUB 24.3 ii 49’–50’, MH/NS)

The pattern whereby ant-nouns trigger agreement with ant-adjectives when they occur in transitive subject position is not random. As Goedegebuure (2018) points out, ant-adjectives modifying ant-nouns become increasingly common in Middle Hittite (MH) and NH texts. This can be taken as evidence that starting from MH -ants/-antes started behaving as a fully fledged ergative case, had become part of the paradigm of neuter nouns, hence it triggered agreement with adjectives in the same way as other case endings did. In this respect, it is interesting to observe that the older copy of example (9) features the unsuffixed form [araḫz]eniēš (KUB 24.4 + KUB 30.12 rev. 7–8, MH/MS). We will return to this point in Sections 4.3 and 4.4. Note that even by the time lexical adjectives start to take ergative agreement, quantifiers and demonstratives do not, as shown by the occurrence of ḫūmanteš ‘all’ in (9). The inflectional approach has proven highly influential (Melchert 2011b) and has been adopted in recent descriptions of the Hittite language (Hoffner and Melchert 2008: 72–73, Melchert 2017a). Most importantly, Garrett’s view has also been taken as the standard for Hittite in general literature on ergativity (see, e.g. Dixon 1994: 187–188, Legate 2014, McGregor 2017). Comparatively less attention has been paid to the way in which the pattern of distribution of -ant historically came about. According to Garrett (1990a), Hittite ergative -ants originated out of an earlier ablative marker. A variant of this scenario is sketched by Patri (2007), who argues that -ants is synchronically to be interpreted as an ablative case ending. However, the latter analysis has been shown to be seriously flawed (Goedegebuure 2012, Dardano 2013), hence we will not pursue it further here. Garrett (1990a: 277–278) maintains that the putative development of the ablative into an ergative marker took place in cases in which ablative NPs that

132

SILVIA LURAGHI ET AL.

expressed the cause/instrument of a two-place verb were reinterpreted as agentive subjects of the clause, in absence of a distinct overt subject. These are contexts such as (10). (10)

URU

Samuḫann=a… alwanzešnaza šer šunnista S.ACC=CONJ witchcraft.ABL/ERG above fill.PST.3SG ‘And he filled Samuha … with witchcraft’ > ‘And witchcraft filled Samuha.’ (StBoT 24 ii 78, from Garrett 1990a: 278)

However, as convincingly argued by Goedegebuure (2012: 290–293), Garrett’s analysis is seriously flawed by morphological considerations. As a matter of fact, the ablative ending -anza is only a morphophonologically conditioned allomorph of the regular ablative ending -az/ats/. This allomorph was originally restricted to -r/n- neuter nouns and only marginally expanded to other neuters (Hoffner and Melchert 2008: 77). Therefore, if the ergative originated out of the ablative case, it is not clear why the more common ending -az was not selected as source. Note further that the existence of ‘ergatives’ in -az cannot be taken as evidence of an ablative origin of the ergative case. Neuter nouns used as transitive subjects and taking the -az ending are extremely rare and mostly restricted to NH texts, and they likely result from the -ants ergative ending undergoing nasal loss before the affricate /ts/, which is a phenomenon basically operating in NH only (Kimball 1999: 317). Given its extremely limited distribution, the ablative allomorph -anza could hardly have provided the source for the ergative ending. To this, one can also add that the syntactic scenario envisaged by Garrett, as exemplified in (10), runs into problems. As a matter of fact, in OH/OS texts the ablative was virtually limited to the encoding of local relations, mostly source and, in some fixed expressions such as kunnaz ‘on the right’, GÙB-laz ‘on the left’, location. In one occurrence it indicated cause, a function further attested in OH/NS texts and likely of PIE origin (Hoffner and Melchert 2008: 264–266, Luraghi 2015: 610). Only later did the ablative start replacing the instrumental case in the encoding of instrument (Melchert 1977, Luraghi 1997: 12–13, Melchert and Oettinger 2009, Melchert 2016).

4.3 Semantics of the suffix -antBesides neuter nouns in A-position, Hittite attests to several formations that bear the suffix -ant- (Oettinger 2001, Josephson 2004, Hoffner and Melchert 2008: 55– 56, Dardano 2010), and scholars hold different views concerning the possible relations among them, as well as concerning the grouping of actual -ant- formations. Interestingly, even proponents of quite similar theories do not agree on how to assign a specific form to one of the groups that they set up. For this reason, we think that a correct assessment of the relationship between these -ant- formations

THE ORIGIN OF ERGATIVE CASE MARKERS

133

and ergative -ant is in order. Accordingly, in this section we survey all -ant- formations available in Hittite, and discuss the etymologies and the interpretations of their meaning proposed by different authors. Broadly speaking, -ant-formations can be grouped as follows:

A. Nominal formations: A.1. Season terms: ḫamešḫa- ‘spring-time in general’ vs ḫamesḫ-ant- ‘the (next/following) spring’ A.2. Neuter nouns: watar vs witenant- ‘water’ (including ‘ergative’ forms/forms with gender motion) A.3. Common nouns denoting human beings: ḫuḫḫa- = ḫuḫḫ-ant- ‘grandfather’ and šankunni- = šankunniy-ant- ‘priest’ (this class includes kinship terms and some names of functions such as ‘priest’) B. Adjectival formations B.1. Possessive denominal adjectives: peruna- ‘rock’ > perun-ant- ‘rocky’.⁸ These include: aku- ‘stone’ > aku-ant- ‘stony’;⁹ annanuzzi- ‘leather restreaming gear for animals, halter’ > annaniziant- ‘geared’;¹⁰ arman‘illness’ > irmanant- ‘ill’;¹¹ ešḫar ‘blod’ > isḫaškant- ‘bloody’;¹² kureššar‘headdress’ > kurešnant- ‘wearing a headdress’;¹³ nata- ‘reed’ > natant‘provided with a drinking reed’;¹⁴ išmeri- ‘bridle’ > išemri-ant- ‘bridled’;¹⁵ išuwan- ‘residue’ > išuwananza;¹⁶ gurzippant- ‘gorgeted, wearing a hauberk’, based on a borrowed Hurrian noun *gurzippi-.¹⁷ B.2. Deadjectival formations: ikuna- = ikun-ant- ‘cold’ (on these see especially Dardano 2010, who lists 18 supposedly deadjectival adjectives; for some of them no nominal base is attested in Hittite). C. Participles of transitive and intransitive verbs: akk-ant- ‘dead’ from ak- ‘die’. D. Use with numerals: e.g. GIŠ harpa=ma 1-anta ‘one set of woodpile(s)’ (KBo 17.3 iv 25)

⁸ KBo 19.76 i 62, NH/NS; KUB 14.16 iii 8–9, NH/NS (on different views about the status of the base stem see Kloekhorst 2008 s.v. vs HED). ⁹ N4 akuwanduš KASKALḪI.A -uš ‘stony roads, roads covered with pebbles’ (KUB 35.84 ii 4, NS). According to Kloekhorst (2008: 168), who follows Hoffner (1978: 245), the noun aku- should be taken as meaning ‘sea shell’, and the -ant- derivate as meaning ‘covered with shells’. ¹⁰ KBo 17.15 rev. 9, 11, OH/OS. ¹¹ According to HEG and Hoffner and Melchert (2008), contra HED s.v.: IGIḪI.A -waššanza GIGanza ‘eye-ailment/ill eye’ (KBo 5.9 i 16, NH/NS), SAG.DU-aššis IGIḪI.A -waššiš GIG-anteš ‘head and eye-ailments’ (KUB 5.6 i 47, NH/NS). ¹² KBo 25.147 obv. 8, 11, n. 3 i 24, n. 4 ii 7, all OH/OS. ¹³ KBo 19.144 i 6, NS, KBo 11.12 i 4, MH. ¹⁴ 2 GAL KÙ.BABBAR natānteš [TU]R-TIM 2 GAL GUSKIN natānteš TUR-TIM ‘two silver cups, straw-equipped, small; two gold cups, straw-equipped, small’ (KUB 31.53 i 2, NH/NS). ¹⁵ ANA awiti GUSKIN=kan išmeriyanti ‘on a bridled golden awiti’ (KUB 38.1 ii 8). ¹⁶ KBo 2.4 iii 27, iv 5, ii 23, KBo 23.95 vs 11, Bo 3481 iv 14. ¹⁷ Besides akkadographic GUR-ZI-IP (KBo 26.147 8, KBo 2.1. ii 21–23, KUB 38.6 i 27).

134

SILVIA LURAGHI ET AL.

As discussed by Hoffner and Melchert (2008: 55–56), these formations can be descriptively regarded as instantiating a single -ant- suffix. Disagreement exists however on the extent to which all these formations reflect one or multiple historical sources (Melchert 2000, Oettinger 2001, Dardano 2013). Let us start with the participle suffix -ant-, whose Indo-European origin is undisputed, as the suffix appears in participles across Indo-European languages. Fellner and Grestenberger (2018) proposed that participle formations in *-e/ontshould be traced back to possessive *-ent-, originally a denominal possessive suffix. Indeed, participles are adjectival formations that indicate a property of an entity in much the same way as denominal possessive adjectives.¹⁸ The close semantic affinity between possessive adjectives and participles is reflected in the assumption by some authors, especially Puhvel (see under the individual lemma in the HED), that some of the denominal formations in B.1 are not denominal possessive adjectives but rather participles of otherwise unattested denominal verbs. For example, Puhvel (HED s.v.) views kurešnant- ‘wearing a headdress’ (KBo 19.144 i 6, NS, KBo 11.12 i 4, MH) as the participle of unattested *kuresnai- ‘provide with headdress’ from the noun kureššar- ‘headdress’. He suggests similar explanations for annaniziant-, ishaškant-, natant- and išuwanant-. However, such an assumption is unwarranted, and there is no need to regard the possessive and the participial suffixes as two homophonous but unrelated suffixes: in fact suffixes that are used to build both possessive denominal adjectives and participles, i.e. verbal adjectives, are not uncommon in IE languages, as in the case of *-to-, attested both in e.g. Lat. amatus ‘loved’ from amo ‘to love’ and barbatus ‘bearded’ from barba ‘beard’ (Dardano 2010: 174). Following this approach, the adjectival possessive function displayed by formations in B.1 can be taken as being directly connected with participles, and is of PIE origin as well. An oftenquoted example is the adjective *bʰérǵʰ-ont-/bʰr̥ǵʰ-n̥t- ‘high, having height’ from the noun *bʰérǵʰ-/bʰr̥ǵʰ- ‘height’ (Melchert 2017b: 218). This formation is possibly the source of several other formations. On the other hand, possessive adjectives could have easily provided, already in PIE, the model for nouns in -ant- via substantivized adjectives, e.g. *gérh2 o-nt- ‘old’ but also ‘old man’ (Oettinger 2001: 303). Among Hittite -ant- formations that we have discussed in Section 4.2, Garrett (1990a, 1990b) views the words utneanza ‘population’ and tuekkanza ‘person’ as instantiating ‘real’ derivates, containing the possessive suffix -ant-. Recall that Garrett, as explained in Section 4.2.4, thinks that the suffix is an inflectional ending along the whole period covered by written attestations: for this reason, nouns ¹⁸ Hittite participles perform functions typically associated with adjectives cross-linguistically (see Rieken 2017): among adjectival properties, one can mention the fact that Hittite adjectives are most often employed attributively as NP modifiers, their position with respect to the modified NP largely resemble that of regular adjectives, and they can also be used with adverbial modifiers such as mekki ‘much’ (see Frantikova 2015 for details).

THE ORIGIN OF ERGATIVE CASE MARKERS

135

that display cases other than the ergative, as utneanza, or are common gender, as tuekkanza are at odds with his description. Hence the solution to consider them as containing a different suffix, and instantiating possessive derivates. Goedegebuure also distinguishes between possessive derivates and other instances connected with the ‘ergative’ function of the suffix. In particular, she singles out three items: wiyana- ‘wine’ lingai- ‘oath’, and utne ‘land’. The former two are common gender nouns, while utne is a neuter. This last word is the most attested in OH (see Table 4.2 and Appendix), and the one that shows the largest number of cases other than the nominative (except for nouns denoting time units, which Goedegebuure leaves out of consideration). According to Goedegebuure, the -ant- suffix with lingai- and utne can either have the individualizing meaning that she views as the origin of the ergative function, or it can mean ‘pertaining to’. Hence, we should reckon with two pairs of lexical items: i ii iii iv

linkiyant- = pertaining to oath = oath deity linkiyant- = inividuated > agentive > ergative oath/perjury utne = utniyant- ‘of the land’ > population utne = inividuated > agentive > ergative land/country

The only two occurrences of winiyant- are taken by Goedegebuure as instantiation the first type: winiyant- = pertaining to wine = ‘wine deity’. However, it looks doubtful that there is any point in maintaining this distinction. Clearly, meanings i and iii are metonymically derived from meanings ii and iv, and the metonymical reading is suggested by each specific context, rather than being a lexical feature of the nouns in question. Metonymy is defined as ‘a cognitive process in which one conceptual entity, the vehicle, provides mental access to another conceptual entity, the target, within the same idealized cognitive model’ (Radden and Ko¨vecses 1999: 21). In particular, the metonymy by which a land or country stands for its population is extremely common, and occurs even in contexts in which the target is not even clearly identifiable, as discussed in Croft (1993: 353). Passages considered by Goedegebuure as instantiating either meaning do not stand up to closer scrutiny. Consider examples (11) and (12). (11) nu=z=apa utniyanza ḫūmanza iški(š)=šmet CONN=REFL=PTCL land.ant.NOM all.NOM back(N).ACC=3SG.POSS.N anda URU Ḫattuša lagan ḫard[u] (…) in H. bend.PTCP.NOM/ACC.N have.IMP.3SG nu=šše=pa utniyanza ḫūmanza anda inaraḫḫi CONN=3SG.DAT=FOC land.ant.NOM all.NOM in strengthen.PRS.3SG ‘Let the entire land keep their back bend down towards Hattusa (The Labarna, the king, is vigorous) and the entire land acts vigorously(?) for him.’ (KUB 36.110 rev. 8–12, OH/OS)

136

SILVIA LURAGHI ET AL.

(12)

kuīt=wa REL.NOM.N=QUOT

KUR[-e]anza=ma=t[ta k]ūrur land(N).ant.NOM=PTCL=2SG.DAT hostility.ACC

ēpzi (…) [nu=w]ar=at ūk ḫūdāk take.PRS.3SG CONN=QUOT=3SG.ACC.N 1SG.NOM suddenly wa[lḫ]mi strike.PRS.1SG ‘Whatever land seizes hostility against y[ou, (as long as the army of his majesty is in battle, now, concerning the fact that I am near), I will fight it (i.e. the country) immediately!’ (KUB 14.1 obv. 25, MH/MS) Goedegebuure points out that for (11) CHD S sub šaḫeššar translates ‘entire land’, but in her opinion ‘entire population’ is a better translation. In this light, it is unclear why the meaning ‘country’ should be preferred to the meaning ‘population’ in (12). In the case of linkiant-, Goedegebuure claims that there are in fact two homonyms differing with regard to whether the noun has an individualizing meaning or not, as in (13) and (14). (13)

n=an linkiyanteš ēppir CONN=3SG.ACC oath.ant.NOM.PL take.PRS.3PL ‘(He took an oath before the gods and then transgressed the oath), the oath -deities seized him.’ (KBo 6.34 iii 15–17, MH/NS)

(14)

nu=tta uittu kēl ŠA SÍSKUR [l]inkiyanza CONN=2SG.ACC come.IMP.2SG DEM.GEN of ritual oath.ant.NOM ēpdu take.IMP.2SG ‘Let the oath of this ritual come and seize you!’ (KBo 11.72 ii39–40 MH?/NS)

In (14), reference is made to the oath of a specific ritual; hence the noun linkiyanza cannot refer to a deity, as linkianteš does in (13). Note however that even in this case the two meanings, being in complementary distribution, should better be viewed as context-dependent. It is not clear whether Goedegebuure assumes the existence of two homophonous but distinct suffixes. Note further the different interpretation of utne ‘population’ between Goedegebuure (< pertaining to the land) and Garrett (< who has the land). According to Melchert (2000) the use of the suffix with season names (A.1) and with numerals (D) are related through the concept of individuation. Melchert, following Goetze (1951: 469–470), argues that names of seasons show a distribution whereby unsuffixed forms refer to a generic time of year, whereas suffixed forms are used to systematically indicate an identifiable time reference,

THE ORIGIN OF ERGATIVE CASE MARKERS

137

e.g. ḫamešḫa- ‘spring-time’ as opposed to ḫamešḫa-ant- ‘the following spring’.¹⁹ Similarly, a ‘delimiting’ function of -ant- is clearly at place with numeral expressions in agreement with collective nouns (see Hoffner and Melchert 2008: 159–163 for details). According to Melchert (2000: 67) ‘Anatolian speakers eventually felt the need to provide count plurals to neuters, and the device most widely used was a -nt- suffix, attached to a modifying numeral.’ In practice, the suffix provides a plurative to otherwise uncountable neuter nouns (see also Josephson 2004: 99 and below). This is taken by Melchert (2000) and Oettinger (2001) as evidence to reconstruct a PIE *-e/ont- ‘individualizing suffix’, which in their view constitutes an extension of the nasal suffix *-e/on-. This suffix is in turn employed in a number of nominal ‘individualizing’ formations in e.g. Latin, e.g. Nero ‘the man (par excellence)’ kunant- ‘killed’ and pai- ‘go’ > pant- ‘gone’ (Hoffner and Melchert 2008: 339, Frotscher 2013, Dardano 2013, Rieken 2017, Inglese and Luraghi 2020). However, participles of intransitive atelic verbs are not resultative, and rather indicate atelic situations, as in e.g. arant- ‘standing’ from ar-ta(ri) ‘stand’. Notably, when based on activity verbs, the participle indicates an ongoing event, agentively carried out by the subject, as in e.g. iya- ‘march’ > iyant‘marching’, so that these participles can easily be substantivized into agent nouns, as in e.g. LÚ ḫuyant- ‘fugitive’ from ḫuwai- ‘run’ (on which see also Dardano 2013: 59 and Rieken 2017: 401). As for the form of the suffix, Josephson writes that ‘possibly the -t of -n-t- adds specificity to what has already been characterized by the individualizing suffix -n-. Cf. -to in Russian kto-to as an indefinite of the specific’ (2004: 106; this also refers to the suffix of the Indo-European participle). Goedegebuure (2018) takes a different stance, and views individuation as the basic semantic feature that triggers the use of ‘ergative’ -ant- with neuter nouns in group A.2, without speculating on its relation with other -ant- formations. Goedegebuure builds upon Rijkhoff ’s (2002) classification of nominals according

THE ORIGIN OF ERGATIVE CASE MARKERS

139

to the features of [±SHAPE] and [±HOMOGENEITY]. Homogeneous nouns are characterized by divisibility, since all of their parts have the same properties, and they can be further distinguished into collectives [+SHAPE], such as population, and masses [-SHAPE], such as water. In other words, homogeneous, hence uncountable nouns, have a low degree of individuation. Low individuation in turn correlates with low agency, to the effect that uncountable nouns are unable to occur in A position, unless they are individuated, and thus agentivized. Individuation is precisely the function of -ant-, as shown by the fact that in OH and MH texts lexemes that occur with -ant- largely fall within semantic classes of less countable nouns, such as liquids, e.g. wātar ‘water’, abstracts, e.g. aššul ‘well-being’, and collectives, e.g. antuḫšatar ‘population’. Individuation as an original function can also easily explain the use of -ant- with numerals when they agree with collectives, and as Goedegebuure points out, -ant-numerals do in fact co-occur with derived collective -ant-nouns. As discussed in Section 4.2.1, in Anatolian languages collectives and mass nouns predominantly belong to the neuter gender. Hence, the association of -ant- with the neuter, which later gave rise to the system of gender-based NP split ergativity, is at its onset only a by-product of the correlation between low individuation and neuter gender. Starting from individualization, the suffix might also be associated with agentivization, thus favouring an ergative reinterpretation of -ant- with neuter nouns in A position. It must be kept in mind that Goedegebuure (2018) keeps some -ant- formations separate from those that she considers as individualized, both in cases in which she sees pairs of homonyms (see above the discussion of udne/udnyant-) and in cases of specific lexemes. Oddly enough, time units are among the ones that she considers as not featuring the individualizing suffix, while Melchert argues that the suffix in this case is in fact individualizing (see note 7 above).

4.4 From Old to New Hittite: the rise of an inflectional category In this section, we illustrate Goedegebuure’s (2018) findings, and highlight some problems with her theory concerning the replacement of the derivational suffix -ant- with thematization once the former had been reanalysed as an inflectional ending (Section 4.4.1). We then propose our own interpretation of the semantic development of the suffix that underlies the change from derivation to inflection (Section 4.4.2). Finally, we discuss the implications of the Hittite data, along with parallel data from other Anatolian languages, for the reconstruction of alignment in Proto-Anatolian and possibly in PIE (Section 4.4.3).

4.4.1 Historical development of -ant- in the Hittite texts Goedegebuure (2018) analyses all occurrences of -ant- formation in chronologically ordered Hittite texts. The contrast between OH and MH texts on the one

140

SILVIA LURAGHI ET AL.

hand and NH texts on the other attests to a different status of the suffix in the two groups of texts. While in NH -ant- forms do in fact behave as fully fledged inflectional forms providing an ergative case for neuter nouns in subject function with transitive verbs, at the OH/MH stage derivational features were prominent. Table 4.2, which is adapted from table 1 in Goedegebuure (2018), shows the extent to which -ant- forms can be considered derivational at different language stages.²¹ Goedegebuure considers those that show one or more of the following features

Table 4.2 Secured nouns with -ant- as derivational morpheme (adapted from Goedegebuure 2018)

²¹ We decided to exclude a number of -ant- formations from the counts in Table 4.2. In the first place, the noun marnuwa(nt)- ‘marnuwa-beer’ is not included. Unlike the other nouns included in Table 4.2, marnuwa(nt)- is commonly held to be originally an -ant- formation, whose forms based on the stem marnuwa- are a secondary backformation (see HED and Kloekhorst 2008 s.v.). Given its dubious etymology, we thus think that this noun does not synchronically instantiate productive -ant- suffixation of the type illustrated in Table 4.2, and therefore we leave it out of consideration. In the second place, we did not include the nouns ḫuḫḫant-, šankunniyant-, and kaenant-. These nouns are derived from common gender ḫuḫḫa- ‘grandfather’, šankunni- ‘priest’, and kaena- ‘in-law’ respectively. Their -ant- forms remain derivational up to NH, but the function of the suffix with these nouns is difficult to assess, and they are synchronically lexicalized. Similar considerations hold for season terms, such as ḫamešḫantfrom ḫamešḫa- ‘spring’ that, as we discussed note 7 above, are also likely to constitute early cases of lexicalized ant-suffixed nouns.

THE ORIGIN OF ERGATIVE CASE MARKERS

141

as derivational nouns: (a) have cases other from the nominative, (b) are based on common gender nouns and/or (c) function as subjects of intransitive verbs.²² The data in Table 4.2 shows that derivational -ant- is only attested in OH and MH texts, and it entirely disappears in NH texts. As Goedegebuure (2018) points out, this contrasts with the distribution of ‘ergative’ -ant-. As shown in Table 4.3, ergative -ant- remains quite frequent even in NH texts, so that quantitative data fully supports the hypothesis of a demise of derivational -ant- in NH. Table 4.3 Distribution of derivational and ergative -ant- (adapted from Goedegebuure 2018)

To sum up, arguments that -ant- behaves as a derivational morpheme only hold if one looks at OH and MH texts. At this language stage, one finds -ant- common gender nouns as well as -ant- nouns as subjects of intransitive verbs, and -antnouns occur in cases other than the nominative. By contrast, these objections drop if applied to NH material, in which the behaviour of -ant- is better captured in terms of an inflectional ergative case marker. Indeed, in NH -ant- nouns in cases other than the nominative, as well as common gender ones and those subjects of intransitive verbs, disappear, and one then finds occurrences of adjectives displaying ergative agreement. In other words, once Garrett’s hypothesis of an ablative origin has been discarded, the development of Hittite -ant- can be understood as a change from derivation to inflection. In Goedegebuure’s view, a consequence of this change was the rise of thematization as a means of indicating individuation. Originally, the suffix -ant- was derivational and functioned as a gender motion suffix by assigning common gender to neuter nouns. Later on, the function of -ant- was taken over by thematization in -a-, to the effect that in NH texts -ant- derivatives were replaced by thematized common gender forms. According to Goedegebuure, this is an ongoing ²² Nouns that show cases other than the nominative are pedant-, utneant-, tapišant-, kištant-, kaenant-, ḫappešnant-, KÙ.BABBAR-ant-. To the nouns listed in Table 4.2, Goedegebuure (2018) also adds wetenanza (< watar- n. ‘water’), which occurs once in KBo 64.4 + KUB 31.86 ii 11ʹ (MH/NS, CTH 261) as subject of a verbs that does not feature a direct object, šara arnu-. This verb only occurs twice in the whole Hittite corpus and its meaning is not completely clear (see HED s.v.), notably, however, the verb arnu- ‘bring’ is transitive (it contains the causative suffix -nu-, see Kloekhorst s.v.). Hence, it is safe to assume either that the sentence in which wetenanza occurs as subject contains a null object recoverable from the preceding context, or that the verb must be taken as unergative. Notably, as argued by Garrett (1990b), subjects of unergative verbs behave as subjects of transitive verbs.

142

SILVIA LURAGHI ET AL.

process, and developed parallel to the reanalysis of -ant- as dedicated ending for neuter nouns in A function, which in turn led the suffix -ant- to be absorbed into the nominal paradigm, and become a case ending for the ergative of neuter nouns in NH, thus creating a split-ergative system. Notably, the onset of the shift of ant- from derivational suffix to inflectional ending can be detected as early as OH, whereas thematisizing -a- only occurs from MH/NS onwards. This distribution suggests that the development of ergative -ant- started before the extension of thematization, so that thematization was only secondarily co-opted to fill in the functional gap that was being left open from -ant-’s specialization as an ergative case ending. Goedegebuure provides data that, in her view, should support her interpretation of this development showing that some nouns that have -ant- forms in OH and/or in MH only have -a- forms in NH. However, it is not always clear what the data show in terms of replacement. For example, the MH form paḫḫuwanza is argued to have been replaced by the thematicized form paḫḫuwenaš in NH, in which paḫḫuwanza does not occur. Unfortunately, the reasoning is flawed, as the form paḫḫuwanza, as for example in (2c) discussed in Section 4.2.2, is regarded as containing an ergative ending, which had already been reanalysed as inflectional, while the form paḫḫuwenaš is regarded to be derivational. Why a fully fledged inflectional form that belonged to the paradigm of the word paḫḫur- should have been replaced by a derivational form is not at all clear. Notably, the form paḫḫuwenaš is also the genitive of the neuter noun paḫḫur, and it is interpreted as such by Garrett (1990b: 50), as well as by several other authors (Dardano 2006: 228, Kassian and Yakubovich 2007: 436, Shatskov 2011: 146–147, 152) and by the CHD s.v. Following this other approach, the genitive paḫḫuwenaš is a so-called free-standing genitive. Let us consider the occurrence in (15). (15)

mān antuḫš[an] ḫūwaḫḫ[urtin] paḫḫuwenaš ēpzi if man.ACC throat.ACC fire(N).GEN take.PRS.3SG ‘If the fire (lit. that of the fire) seizes the throat of a man.’ (KUB 8.36 iii2 NH/NS)

Example (15) is from a list of medical texts. In this occurrence, paḫḫuwenaš indicates the type of disease that someone might have when they feel soreness in their throat. Compare this occurrence with the free-standing genitive tayazilaš ‘the (compensation) for theft’ Laws §§94–95, frequently also in reference to a human participant and functioning as an agent noun: ‘he of the theft; the thief ’ Laws §73, (Hoffner and Melchert 2008: 256). As the examples show, free-standing genitives provide a means to refer to a specific agent, or to a specific inanimate referent, possibly even occurring in A function as in (15) if we accept the interpretation given here. Even when the free-standing genitive does not occur in A function, it remains a definite individuated entity (see further its use with the genitive of personal pronouns, as in ammel daḫḫi ‘I will take (what is) mine’ KUB 13.35 ii 3; from Hoffner

THE ORIGIN OF ERGATIVE CASE MARKERS

143

and Melchert (2008: 256)). We return to this important feature of free-standing genitives in Section 4.4.2. Similar considerations hold for example (6). According to Goedegebuure (2018), who follows Laroche (1971: 190) and shares the interpretation proposed in Tschekoff (1978: 228–232), the form parnaš in the duplicate is not the genitive of the base neuter noun per, but the nominative of another common gender derivate, parna- which contains the thematic vowel -a-. However, parnaš might also be a free-standing genitive, as suggested in Garrett (1990b: 50), followed by Kassian and Yakubovich (2007) and Shatskov (2011). Note further that Dardano (2013) discussing the meaning and the function of -ant- and -a- formations points out that thematization with gender motion and often animatizing function has parallels in other Indo-European languages, such as Ancient Greek and Sanskrit (see Lazzeroni 2002b; see further nepiš ‘sky (n)’ > nepiša-š ‘sky (c)’; Dardano 2013: 45). However, this parallel is not as compelling as it might look at first sight in consideration of the Hittite chronology (thematization as described in this section is not attested in OH according to Goedegebuure 2018). A more compelling case is constituted by the noun erman (or irman) ‘illness’, which shows -ant- forms at all language stages and -a- forms in MH and NH. This noun occurs 12 times in the -ant- form, arguably functioning as ergative and not as a derivate, and five times thematized. Of the latter occurrences, one is an accusative ([Ḫ]UL-un irman Bo 12.94:7, NH?/NS). Examples are (16) and (17). (16) idalauwanza k[ū]n ue[miškiddu] irmananza evil(N).ant.NOM DEM.ACC find.IPFV.IMP.3SG illness(N).ant.NOM uemiškid[du] ÚS-anza uemišk[iddu] find.IPFV.IMP.3SG death(N).ant.NOM find.IPFV.IMP.3SG ‘Let evil find this one, let illness find him, let death find him.’ (Bo 3428rev. 1–4, cited in Wilhelm 1994: 43) (17) GIG-zi=ma=mu ḪUL-lu GIG-aš=mu [mekki] ail.PRS.3SG=PTCL=1SG.ACC bad(N).ACC illness.NOM=1SG.ACC much tamaššan ḫarzi oppress.PTCP.NOM/ACC.N have.PRS.3SG ‘I’m seriously ill, and the illness has badly much oppressed me.’ (KUB 19.5 obv. 5–6, NH/NS) Note further that in the case of irman- (n.), ermanza (c. or erg.) and irmana- (c.) the -ant- form and thematization coexist in MH and NH: this gives more likelihood to Geodegebuure’s theory of a replacement, as one can view the -ants form as containing a grammaticalized inflectional case, and the -a- form as supplying the non-subject forms for a common gender noun, as it does in the case of the accusative occurrence mentioned above, perhaps along with a specific semantics.

144

SILVIA LURAGHI ET AL.

Summing up: the extent of the putative replacement seems to be at best much more limited than in Goedegebuure’s view.

4.4.2 The semantic development of -antGoedegebuure (2018) argues that the suffix added a semantic feature to the base noun that was not commonly found in neuter nouns, viz. individualization and contingently also animacy, hence being especially suitable to contexts in which the noun functioned as subject of a transitive verb. She does not try to connect this meaning of the suffix with the other meanings illustrated in Section 4.3, in fact she distinguishes between ‘individualizing’ (hence potentially ergative) -antforms and possessive -ant- derivates as we have shown in Section 4.3, examples (11)–(14). In this section, we set out from the assumption of the participle and -ant- possessive adjectives as containing the same suffix, as we discussed in Section 4.3, and suggest that the individualizing meaning can have originated out of the possessive meaning of the -ant- suffix. In fact, possessors are typically definite, hence more suitable to be individuated, and, as a consequence, animate and agentive. Taylor (1989: 202) views the fact that ‘the possessor is a specific human being’, as the first defining feature of possessors hence also providing a link to the animatizing and agentive function of a possessive morpheme. As is well known, numerous languages, including English, do not allow for the co-occurrence of possessives and determiners (Lyons 1999: 202). Haspelmath (1999: 231) explains this distribution as follows: ‘possessed noun phrases are not necessarily definite, they are only highly likely to be definite. But a high likelihood of (or preference for) definiteness is sufficient for economic motivation.’ Hence, the connection between possession and definiteness is strong enough to have grammatical reflexes cross-linguistically. In Hittite, this characteristic of possessors provides the ground for the agentive function of the free-standing genitive described in Section 4.4.1, as in šāwataraš ‘he of the horn; horn-player’ (KBo 17.1+ i 37, OS), waštulaš ‘he of the ‘sin’; the offender’ (Laws §§147–148) (from Hoffner and Melchert 2008: 256). As shown in Section 4.4.1, free-standing genitives, though often indicating human beings, can also refer to inanimate entities, with the effect that they are conceptualized as individual and definite and in some cases even agentive. Hence, prototypical features of possessors associated with the genitive case have different effects depending on the type of referent. Notably, these effects show a striking similarity with those that are normally ascribed to the -ant- suffix, namely agentivization and individuation. Note that, even though not all -ant- ergative formations can be traced back to possessive, the most frequent one is utneanza. As we discussed in Section 4.3, both Garrett (1990a, 1990b) and Goedegebuure (2018) distinguish two sets of forms of utneant-, and treat some as containing the ergative suffix, while they think that

THE ORIGIN OF ERGATIVE CASE MARKERS

145

other contain the possessive suffix, even though they assume different translations for the possessive lemma. We argued that the distinction is unwarranted: putative differences between the meanings ‘population’ and ‘country’ are better regarded as due to common metonymical processes. Rather than trying to dispose of the forms of this lemma that do not fit expectations based on one’s assumptions, we propose an explanation of their distribution which reconciles all instantiations of utneant- considered a single lemma. The possessive adjective from utne- indicates the population, as commonly assumed. This meaning contains all features of prototypical possession indicated by Taylor (1989): the population is an animate and agentive entity, and it is mentioned in reference to a specific country, hence it is definite and individuated. Given the frequency of this word, the suffix extends to other words that share some of its features, in particular the fact of being neuter and indicating some unbounded entity, and supplies them with a form that can stand in A-function. The feature of individuation becomes salient in the meaning of the suffix in cases such as its use with numerals and perhaps with names of time units, while the resultative meaning that the suffix exhibited with the participle is responsible for its extension to some adjectives. Eventually, -ant- forms split into a group with neuter bases, with which the suffix becomes increasingly grammaticalized and turns into the inflectional ending of a new case, the neuter ergative, and another more heterogeneous group with which the suffix often loses its original meaning and becomes lexicalized.

4.4.3 Other Anatolian evidence: reconstructing alignment in Proto-Anatolian In the preceding sections we have shown that the ergative function of the -antsuffix developed out of a situation in which the nominative of neuter nouns could not encode the A function. At this point, the question of what the original situation in Proto-Anatolian was arises. In fact, no Anatolian language offers evidence for a preceding nominative–accusative alignment system for neuter nouns. Hence, we reconstruct active alignment in the nominal system for Proto-Anatolian, as in Table 4.4.²³ According to Klimov (1974: 13) languages with active alignment are characterized by the presence ‘of a binary division of nouns into active, on the one hand, and inactive, on the other, which is based on the opposition of the referents corresponding to them according to the feature of the presence or absence in them of ²³ According to Garrett (1990a: 279), split ergativity was already established in Proto-Anatolian. Notably, part of the evidence for this state of affairs is argued to be provided by the morphological analysis that views the suffix as deriving from an ablative case marker.

146

SILVIA LURAGHI ET AL. Table 4.4 Alignment in Proto-Anatolian

life activity’. In the case of Hittite, and more in general of Anatolian, active nouns are common gender nouns, while inactive ones are neuter. Notably, all Anatolian languages show cognates of the suffix -ant- at least with its possessive meaning. A possible development towards an ergative system only occurs in Luwian -antis and Lycian -ẽti). According to Goedegebuure (2018) Luwian -ant- displays a distribution comparable to that of OH -ant-. The suffix occurs both with neuter and with common gender nouns, and has individuating force. As in the case of Hittite, neuter nouns in A position consistently take the endings -antis/-antinzi, made up of -ant-, the motion suffix -i- and the nominative common gender endings. However, -ant- neuters in subject position never trigger agreement with adjectival modifiers. Lycian Ṽt- /-Vnt-/ is also predominantly used with numerals when counting individual entities in collectives, as in e.g. nuñtãta am ˜ [ã]ma uwa ‘nine units (of ) cattle as penalty (?)’ (TL 131,3). In addition, -Ṽt- occurs once with the neuter noun tese (n.) ‘oath’ in subject position, as in example (18). Note that in this example tesẽti triggers Ṽt- agreement on the adjective trm͂mili-j-ẽti, showing that the suffix effectively behaved as a fully fledged inflectional ergative marker. (18)

s=ẽne tesẽti: qãñti: CONN=3SG.ACC oath(n).ant.NOM.PL destroy/seize?.PRS.3PL trm͂mili-j-ẽti Lycian-adj-ant.NOM.PL ‘The Lycian oaths will destroy him.’ (TL 149,10)

In Palaic -ant- formations occur sporadically, possibly with possessive meaning (see Valério 2009). However, it is remarkable that the extant texts offer no evidence for neuter nouns in A function. To sum up: evidence from Anatolian languages points towards an early stage with active alignment, with reflexes of the possessive -ant- suffix. In Hittite, Luwian, and Lycian, one finds that the derivational functions of the suffix coexisted with its use with neuter nouns in A position since the earliest documentation. However, only in Hittite and Lycian did the suffix develop into an ergative marker, as evidenced by patterns of adjectival agreement, whereas in Luwian the suffix remains, to the best of our knowledge, derivational.

THE ORIGIN OF ERGATIVE CASE MARKERS

147

While the reconstruction of alignment in PIE falls outside the scope of the present chapter, some remarks on the implications of alignment in ProtoAnatolian are in order (see also Chapter 2 in this volume). Since at least Meillet ([1921] 1948), the inactive nature of PIE neuter nouns has been repeatedly pointed out. In particular, Meillet highlighted the existence of lexical doublets featuring neuter and non-neuter nouns, such as nouns denoting water and fire (211–220), and wrote that non-neuter nouns referred to active entities (hence potential As) while neuter nouns referred to inactive ones. This observation hints to the occurrence of active alignment in PIE as well, a reconstruction fully worked out in Gamkrelidze and Ivanov (1995). In this framework, the Anatolian evidence might be viewed as further supporting the reconstruction of active alignment in PIE as well. A final note is in order concerning language contact. There is a growing body of evidence that the changes undergone by Hittite from OH to NH were partly influenced by extensive contact with Luwian speakers (see discussion in Melchert 2005, Rieken 2006, Yakubovich 2010). Concerning the rise of split ergativity in Hittite, it is interesting to observe that, the Luwian suffix being derivational at all known stages of the language, it is unlikely to have influenced via contact the development of Hittite -ant- into an ergative. On the other hand, as pointed out by Melchert (1994) and Watkins (2001: 54) ergativity is an areal feature of the Indo-European and non-Indo-European languages of Ancient Anatolia, among the latter most notably of Hurrian and possibly of Hattian, see the discussion in Taracha (1995, 1998) and Goedegebuure (2010). Even though the pattern of split ergativity in Hurrian is not gender based (see Campbell 2008), the development of ergativity in some of the Anatolian languages might point towards a process of areal convergence.

4.5 The origin of ergative case markers: a cross-linguistic view Garrett’s proposal of an ablative origin of the Hittite ergative -ants has enjoyed a certain success, and Hittite is often mentioned as instantiating the grammaticalization process ABLATIVE/INSTRUMENTAL > ERGATIVE in linguistic works dedicated to the origin of ergative markers cross-linguistically (see among others Dixon 1994, Heine and Kuteva 2002, Mithun 2005, Legate 2014, Aldridge 2017). In fact, syncretism of instrumental and ergative is not at all uncommon, but, apart from clear cases of reanalysis from passive constructions as attested in Indo-Aryan (Dahl 2016), the diachrony of the possible extension from the former function to the latter is hardly documented. Dyirbal, for example, has an NP split-ergativity syntax with the ergative case ending being the same as the instrumental (Dixon 1976), but nothing is known regarding its development. A development similar to the one reconstructed by Garrett for Hittite has also been assumed by Mithun (2005) for

148

SILVIA LURAGHI ET AL.

the Coosan languages of Oregon. Notably, however, Mithun shows that multiple sources appear to be responsible for the extension of the instrumental ending to the ergative case in Coosan, including reanalysis of an inverse construction. Markers of other cases do by no means exhaust the sources for ergative case markers. In his survey on the origin of ergative case markers McGregor (2017) lists three other sources, including lexical items, indexical elements, and directional elements. While the first and the last type of items are rarely attested, and some alleged examples may be speculative, the role of indexical items in providing the source for ergative case markers, though often underestimated, appears to be prominent in various languages and language families, including several Pama– Nyungan and non- Pama–Nyungan languages of Australian as well as Polynesian, Trans–New Guinea, and North-West Caucasian (McGregor 2017). To the wide variety of sources described thus far in the literature, Hittite provides evidence for yet another possible source, that is, a derivational morpheme, which can be viewed as a special subtype of lexical source. The shift of a morpheme from derivation to inflection is described by Kuryłowicz (1965) as an instance of grammaticalization, whereby an already grammatical morpheme becomes more grammatical: ‘Grammaticalization consists in the increase of the range of a morpheme advancing from a lexical to a grammatical or from a grammatical to a more grammatical status, e.g. from a derivative formant to an inflectional one.’ Hopper and Traugott (1993: 5) highlight the in-between position of derivational morphemes, that ‘serve a role between contentive and grammatical forms’, i.e. between grammatical forms and the lexicon. A key feature of grammaticalization is semantic bleaching, i.e. the loss of a morpheme’s original meaning (Lehmann 2015). This is precisely what happens in the case of -ant-: the suffix originally added specific semantic features to nouns, but once the development into an ergative inflectional marker took place, the meaning of the suffix was progressively lost, as evidenced by the split between its derivational usages, which are largely lexicalized, and its function as inflectional case ending. In the latter function the suffix marked a syntactic role, that is, the role of neuter nouns in A position. By shifting from derivation to inflection, the ergative suffix in Hittite acquired a place in the paradigm of neuter nouns, and became obligatory in a specific syntactic function. As discussed in Section 4.4, however, this shift was gradual, and the change from derivation to inflection was not abrupt. This means that there must have been a period, most likely reflected in the OH and MH texts, in which the derivational and the inflectional function of the suffix were both available for the same morpheme. Garrett (1990a: 268, 271) rejects a possible scenario that he outlines as follows: ‘The “ergative” is morphologically derivational and functionally syntactic. In other words, the endings -anza and -antes consist of the nominative common- gender endings -s and -es preceded by a derivational suffix -ant- … However, its function is not semantic (as expected of derivational morphology), but syntactic’ based on the assumption that ‘a derivational suffix whose function

THE ORIGIN OF ERGATIVE CASE MARKERS

149

is to permit nouns of a particular gender to stand in a particular syntactic function would, as far as I know, have no parallel in Indo-European or elsewhere’. However, as argued in Luraghi (2017: 282–285), borderline phenomena involving derivational affixes that fulfil a syntactic function are found elsewhere in Anatolian, notably in the case of ‘genitival’ adjectives, known from Luwian and, to a lesser extent, from Lycian and Lydian. In Luwian in particular there is no trace of the genitive case, which is systematically replaced by inflected denominal adjectives (see Neumann 1982, Luraghi 1993, 2008, Melchert 2012; for a detailed treatment of nominal modification in Luwian, see Bauer 2014: 117–273), as shown in (19). (19)

iyandu=ku=wa DUMU-annašši[n] annin zaššin go.IMP.3PL=CONN=PTCL this.ADJ.ACC child.ADJ.ACC mother.ACC warallin uwata[ndu] own.ACC bring.3PL.IMP ‘Let them carry this child’s own mother.’ (KUB 35.102+ ii 13’–18’)

In example (19), the noun annin ‘mother’ is modified by the possessive adjective DUMU-annašši[n] ‘belonging to the child’, which is used in the syntactic position in which one usually finds a genitive possessor NP in other IE languages including Hittite. Remarkably, this is the only way in which an adnominal relation can possibly be indicated in Cuneiform Luwian, since neither nouns nor pronouns have a genitive ending. In such a case, one can rightly say that derivation is used in the service of syntax, rather than to enrich the lexicon, in a non-prototypical way. In other words, a suffix which was in origin derivational underwent an evolution by which its function eventually became syntactic. Synchronically, the suffix of genitival adjectives permits nouns to take a specific syntactic function, i.e. that of modifiers. The ergative function of the -ant- suffix may be seen as involving a similar evolution from derivation to the coding of grammatical relations. Remarkably, the extension of possessive adjectives and their replacement of the genitive case have been shown to reflect Hurrian influence (Luraghi 2008, Yakubovich 2008). Hence, both changes in NP structure and the rise of ergativity to different extents can be considered as the outcome of processes of areal convergence. In both cases, derivational morphemes are co-opted to serve syntax, and behave to different extents as inflectional morphemes. A final note is in order concerning the type of change undergone by Hittite -ant-. As remarked in Section 4.2.4, following Garrett’s (1990a) proposal, Hittite ‘ergative’ -ants is regarded as originating out of a previous ablative marker. Within this scenario, -ants was an inflectional morpheme to begin with. At the onset, neuter nouns in the ablative could indicate cause/instrument adjuncts. In specific constructions, in which the -ants marked NP could also be interpreted as an agent, such as the one in (10), the originally cause/instrument expression was reinterpreted as the syntactic subject of the clause and the ablative case ending as an

150

SILVIA LURAGHI ET AL.

ergative case ending. Notably, the process sketched by Garrett instantiates a typical case of syntactic change driven by reanalysis. Reanalysis is commonly defined as a mechanism of language change ‘which changes the underlying structure of a syntactic pattern and which does not involve any modification of its surface manifestation’ (Harris and Campbell 1995: 50; see also Hopper and Traugott 1993, De Smet 2009). What is crucial in cases of reanalysis is that the same surface string of linguistic material is assigned a new syntactic (and semantic) interpretation. Ergative patterns can notoriously arise out of reanalysis of previous constructions. The case of passives developing into ergative constructions is well known, with the oblique agent of the passive being reinterpreted as the ergative NP of the new ergative construction (see, e.g. Dahl 2016 for discussion of this development in Indo-Aryan). However, Garrett’s (1990a) proposal of an ablative origin of -ants is untenable for the reasons discussed in Section 4.2.4. Instead, we suggest that -ant- was originally a derivational morpheme and only secondarily grammaticalized into an inflectional marker. This development does not constitute, strictly speaking, a case of syntactic reanalysis: rather, reanalysis affects the morphological plane, even though it ends up triggering a change in alignment. As a matter of fact, the development does not affect the syntactic representation of the source construction: indeed -ant- nouns in A position were the syntactic subject of the clause to begin with, as evidenced by nominative case marking -s and verbal agreement. The change only affects the combination of the derivational suffix -ant- with the nominative endings -s/-es: -ant- increasingly undergoes semantic bleaching and loses its derivational nature, hence the sequence -ant-s/es is reinterpreted as a single inflectional case ending -ants/-antes. Notably, reanalysis at the morphological level and the shift from derivational to inflectional undergone by the ergative suffix brings about a change in the alignment system of Hittite, which evolves from being active to being split ergative.

4.6 Conclusion In this chapter, we have proposed a reassessment of the development of split ergativity in Hittite. We set out from Laroche’s (1962) famous observation that in Hittite neuter nouns that function as subjects of transitive verbs must occur with a dedicated form -ants/-antes. We have reviewed previous interpretations of -ants, discussing pros and cons of approaches that view the form as featuring a derivational motion suffix -ant- plus the common gender nominative ending -s and those that view -ants as an inflectional ergative case marker. Following Goedegebuure (2018), we showed that a proper understanding of this issue can only be achieved if one takes a diachronic perspective, and considers the chronological distribution of -ant- forms in the corpus constituted by all extant Hittite texts. Goedegebuure’s careful analysis of the distribution of -ant-

THE ORIGIN OF ERGATIVE CASE MARKERS

151

forms in Hittite texts from OH to NH has shown that the suffix displays a derivational behaviour in OH, but slowly turns into a fully fledged inflectional marker by NH, where its inflectional status is indicated by agreement with adjectives in -ants. In the development from derivational -ant- to ergative -ants, we discussed how the complex semantics of the suffix may have played a role. We set out from the assumption that we have to reckon with a possessive suffix, which could attach to different types of bases, in particular nominal and verbal. In the latter case, ant- forms gave rise to verbal nouns that acted as participles. We argued that when occurring with nominal bases, the suffix added features typical of possessors, in the first place definiteness, hence individuation. We then showed that the most frequent denominal -ant- formation, utneanza ‘population’ and metonymically ‘country’ also had other prototypical features of possessors, such as animacy, that favoured its occurrence in A position. This highly frequent item provided a pattern for other neuter nouns to be moved to common gender in order to function as As. As no Anatolian language provides evidence for the occurrence of neuter nouns in A position at any language stage, we have reconstructed an active alignment system for Proto-Anatolian, and have argued that the rise of ergativity in Hittite and possibly in Lycian and Luwian constitutes an areal convergence process with non-Indo-European languages of ancient Anatolia. Finally, we have highlighted the more general significance of the Hittite data for the diachronic typology of ergative case markers, and proposed that Hittite attests to a hitherto neglected path of development of ergative markers, i.e. a shift from a derivational morpheme to an inflectional case ending, thereby enriching our current understanding of how systems of split ergativity come about in the languages of the world. We have further pointed out that developments whereby originally derivational morphemes start being used in a syntactic function are not uncommon in Anatolian languages, as in the case of the Luwian genitival adjectives. Finally, we have also discussed how the development proposed here, contra Garrett’s (1990a) proposal, does not instantiate a typical example of syntactic reanalysis, which is often considered as one of the key mechanisms of change that give rise to new alignment systems in the world’s languages.

Acknowledgements This chapter relies on data carefully extracted from the Hittite texts by Petra Goedegebuure. We wish to thank her for sharing her data with us, and for her willingness to make them available to other scholars through the Appendix to our chapter.

A PPENDI X

Chronologically ordered attestations of -ant- forms Lemma*

Subject of transitive verb

Cases different from nominative

Subject of intransitive verb Thematic forms in -a-

aniur ‘ritualʹ

KUB 41.8 iv 38 (MH/NS), KBo 10.45 iv 40 (MH/NS) KUB 17.10 iv 11 (OH/MS) KU 19.10 i 6 (NH/NS), KUB 19.11 iv 16 (NH/NS), KBo 3.6 i 26 (NH/NS), KUB 1.1 i 30 (NH/NS), KUB 31.66 ii 16ʹ (NH/NS) VBoT 58 i 14ʹ (OH/NS), KBo 25.107:6ʹ (OH/OS)

—

KUB 58.106 iv 8ʹ (MH/NS) —

— —

— —

— —

—

—

—

annaššar ‘pillar (?)ʹ antuḫšatar ‘populationʹ

appuzzi ‘animal fatʹ

aššu ‘goodʹ

KBo 13.68 rev 16ʹ, KBo 41.159+KUB 6.5 obv. 15ʹ, KBo 41.159+KUB 6.5. rev. 36ʹ, KBo 41.174:6, KUB 5.1+ i 52, KUB 5.1+ i 97, KUB 5.1+ i 107, KUB 5.1+ ii 26, KUB 5.1+ iv 68, KUB 5.3+ iv 6, KUB 5.4+KUB 18.53 i 23, KUB 5.4+KUB 18.53 iii 11, KUB 6.7+ iii 11ʹ, KUB 6.7+ iii 13ʹ, KUB 6.7+ iii 16ʹ, KUB 6.7+ iii 22ʹ, KUB 18.21 ii 2, KUB 49.89 r col 5ʹ, KUB 50.58: 11ʹ, KUB 5.11 i 33, KUB 5.11 iii 12ʹ, KUB 5.11 iv 41, KUB 49.21 iii 17, AT 454 i 53, KUB 6.11 + obv. 23ʹ, KUB 16.29+ obv. 17, KUB 16.29+ rev. 11ʹ, KUB 22.61 Continued

Lemma*

aššul ‘favourʹ

erippi ‘cedar(wood)ʹ

erman ‘illnessʹ

Subject of transitive verb

Cases different from nominative

Subject of intransitive verb

Thematic forms in -a-

obv. i 24ʹ, KUB 50.4 rev. iv 9ʹ, KUB 5.17+ ii 41ʹ, KUB 16.53 rev.? 10ʹ, KUB 5.2: 3ʹ, KUB 6.32: 9ʹ, KUB 16.4: 9ʹ, KUB 22.4: 8ʹ, KUB 49.81 obv. 4, KUB 22.64 iii 7ʹ, KUB 50.44 i 12ʹ (all NH/NS) KBo 38.263 + 134 rev. 17ʹ (MH/NS)

—

—

—

—

—

—

—

KUB 49.24 rev. 11 (NH/NS), KBo 6.29 i 10 (NH/NS), KBo 4.12 rev. 9 (NH/NS), KBo 3.6 i 27 (NH/NS), KBo 2.32 iv 1 (NH/NS), KUB 10.91 iii 9 (NH/NS) —

—

—

KUB 24.4 obv. 8 (NH/NS), KUB 24.1 i 12 (NH/NS), KUB 24.2 obv. 10 (NH/NS) Bo 5439 r.col. 8ʹ (OH/NS), KBo 36.53 obv. 4ʹ(OH/NS), KBo 36.53 obv. 6ʹ (OH/NS), KUB 43.4 i 2 (OH/NS), KUB 60.44 obv. 8ʹ (OH/NS), Bo 3428

KBo 12.94:7ʹ (NH/NS), KUB 30.43 iii 24ʹ (NH/NS), KUB 14.15 ii 6 (NH/NS), KUB 19.5 obv. 5 (NH/NS)

ešḫar ‘bloodʹ

ḫaḫḫall- ‘bushʹ ḫanneššar ‘judgmentʹ

rt. col. 2ʹ (OH/NS), KUB 33.121 ii 17ʹ (MH/NS), KUB 6.5 rev. 27ʹ (NH/NS), KUB 8.36 ii 10 (NH/NS), KBo 4.6 obv. 25 (NH/NS), KUB 5.6 ii 38ʹ (NH/NS), KUB 5.6 ii 65ʹ (NH/NS) KUB 30.34 iv 7 (MH/NS), KUB 39.103 rev. 4ʹ (MH/NS), KBo 54.14 iii 10ʹ (MH/NS), KBo 43.319 i 14ʹ (MH/NS), KUB 9.4 i 39 (MH/NS), KUB 9.34 ii 46 (MH/NS) KUB 14.14 rev. 11 (NH/NS), KUB 44.63 ii 3 (NH/NS), KUB 54.1 iv 19 (NH/NS) KBo 14.86 i 4 (MH/NS) KUB 19.67 i 5 (NH/NS)

—

KUB 4.1 ii 22 (MH/NS)

—

— —

— —

— — Continued

Lemma*

Subject of transitive verb

Cases different from nominative

Subject of intransitive verb

Thematic forms in -a-

ḫappeššar ‘limbʹ

KUB 9.4 i 19 (MH/NS), KUB 9.34 ii 36 (MH/NS)

KUB 1.16 iii 40 (OH/NS, acc.), KUB 58.111 rev. 4ʹ (OH/NS, inst.)

—

ḫappir(iy)a- ‘city (c.)ʹ ḫapuš- ‘penisʹ ḫapušašša- ‘shaftʹ

— KUB 9.4 i 31 (MH/NS) KUB 7.1 + KBo 3.8 ii 35 (MH/NS) KBo 53.29:9ʹ (MH/NS) KUB 9.4 i 33 (MH/NS), KUB 9.34 ii 43 (MH/NS) KUB 7.41 i 20 (MH/MS?), KBo 10.45 i 11ʹ (MH/NS) KUB 9.4 i 38 (MH/NS) KBo 13.109 iii 6 (MH/NS) KUB 57.122 r. col. 2ʹ (MH/NS) —

— — —

KUB 41.8 iv 30 (MH/NS) — —

KBo 17.17 rev. 10ʹ (OH/OS), KUB 4.1 iv 31 (MH/NS), KUB 41.1 iii 16 (MH/NS), KBo 3.8 (+ KUB 7.1) ii 31ff. (MH/NS) — — —

—

—

—

—

—

—

— —

— —

— —

—

KBo 18.54 obv. 5 (MH/MS) —

—

ḫaratar ‘offenseʹ ḫargnau- ‘palm, sole (c./n.?)ʹ ḫašša- ‘hearth (c.)ʹ ḫaštai ‘boneʹ ḫattalu ‘boltʹ ḫattulatar ‘healthʹ ḫazgarai ‘h.-womenʹ

KBo 2.7 obv. 26ʹ, KBo 2.13 obv. 10, KBo 2.13 obv. 11, KBo 2.13 obv. 19, KBo 2.13 rev. 3ʹ, KUB 17.35 i 5ʹ, KUB 17.35 i 33ʹ, KUB 17.35 ii 18ʹ, KUB

—

—

ḫilammar ‘gatehouseʹ ḫingan ‘deathʹ

ḫuḫḫa- ‘forefather (c.)ʹ

ḫuidar ‘wildlifeʹ ḫuimpa- ‘? (c.)ʹ

17.35 ii 27ʹ, KUB 17.35 ii 24ʹ, KUB 17.35 iii 8, KUB 17.35 iii 32, KUB 17.35 iii (35), KUB 17.35 iv 13, KUB 17.35 iv 24, KUB 17.35 iv 25, KUB 44.42 obv. 18ʹ, KUB 46.22 i 8ʹ, KUB 41.34 i 6ʹ (joins KUB 46.22), KUB 46.27 obv. 13ʹ (all NH/NS) KUB 17.10 iv 11 (OH/MS) Bo 3428 rt. col. 2ʹ (OH/NS), KUB 24.4 obv. 15ʹ (MH/MS), KUB 24.3+ ii 25ʹ (MH/NS), KBo 7.63:3ʹ (MH/NS), KUB 19.1 + 14.14 + rev. 22ʹ (NH/NS), KUB 19.1 + 14.14 + rev. 24ʹ (NH/NS), KUB 19.1 + 14.14 + rev. 24ʹ (NH/NS), KUB 19.1 + 14.14 + rev. 25ʹ (NH/NS) KUB 21.5 i 14 (NH/NS), KUB 36.87 iv 26ʹ (NH/NS), KUB 36.87 iv 27ʹ (NH/NS) (all in broken contexts) KBo 9.114:7 (OH/MS) KBo 25.184 iii 67 (OH– MH/NS)

— —

— —

— KUB 14.12 obv. 8 (NH/NS), KUB 5.3 + KUB 18.52 i 47 (NH/NS)

KBo 22.129 obv. 5 (MH/MS)

—

—

— —

— —

— —

Continued

Lemma*

Subject of transitive verb

Cases different from nominative

Subject of intransitive verb

Thematic forms in -a-

ḫunḫuneššar ‘waveʹ

KUB 36.12+ ii 16ʹ (NH/NS), KUB 33.89 + KUB 36.21:17ʹ (MH/NS, broken context) — KUB 60.44 obv. 7ʹ (OH/NS), Bo 3428 rt. col. 1ʹ (OH/NS), KUB 5.1 ii 8 (NH/NS), KUB 5.1 ii 23 (NH/NS), KUB 5.1 iii 7 (NH/NS), KUB 5.1 iii 31 (NH/NS), KUB 22.57 obv. 19 (NH/NS), KUB 6.3 obv. 18 (NH/NS), KUB 18.11 + KBo 57.125 obv. 8ʹ (NH/NS), KUB 5.24+ i 50 (NH/NS), KUB 16.29 (+ KUB 16.81) obv. 2 (NH/NS), KUB 5.5 ii 33 (NH/NS), KUB 50.71 obv. 16 (NH/NS), KBo 22.264 (+) KUB 50.23 iii 5ʹ (NH/NS), KUB 16.41 iii 16ʹ (NH/NS), KUB 22.5 obv. 4 (NH/NS), KUB 18.12 + KUB 22.15 i 6 (NH/NS)

—

—

—

— —

KUB 7.58 i 3ʹ (MH/NS) —

— —

ḫuwalliš ‘coal?, pine cone?ʹ idalu ‘evilʹ

inan ‘diseaseʹ išḫaḫru ‘tear(s)ʹ

išḫunau- ‘upper arm (c.)ʹ kaena- ‘in-law (c.)ʹ

KBo 43.319 i 15ʹ (MH/NS) KBo 53.29:9ʹ (MH/NS), KBo 43.319 i 14ʹ (MH/NS) KBo 55.52 obv. 11 (MH/NS) KUB 9.4 i 26 (MH/NS) —

kalulupa- ‘finger, toe (c.)ʹ kamaršu- ‘defecationʹ kašt- ‘hunger (c.)ʹ

KUB 9.4 i 35 (MH/NS) IBoT 1.36 i 43 (MH/MS) —

KÙ.BABBAR ‘silverʹ

KUB 17.4:2ʹ MH NS KUB 17.4:8ʹ (MH/NS), KUB 36.19:9ʹ (MH/NS), Bo 5019:12ʹ (MH/NS) KUB 24.4 obv. 16ʹ (MH/MS), KUB 24.3+ ii 26ʹ (MH/NS), KBo 7.63:4ʹ (MH/NS) KBo 17.22 iii 12 (OH/OS)

išḫieššar

kurur ‘hostilityʹ

laḫḫurnuzzi ‘foliageʹ

lingai- ‘perjury (c.)ʹ

KUB 30.34 iv 7 (MH/NS), KUB 60.44 obv. 9ʹ (OH/NS)

— —

— —

— —

—

—

—

— KUB 13.20 i 33 (MH/NS, dat.) — — KUB 17.10 i 18ʹ (OH/MS), KUB 17.10 i 30ʹ (OH/MS) KUB 36.18 ii 7ʹ (MH/NS, voc.)

— —

— —

— — —

— — —

KUB 17.4:8ʹ (MH/NS), KBo 22.82:6ʹ (MH/NS)

—

—

—

—

—

—

—

—

KBo 14.142 ii 17 (NH/NS), KUB 25.23 i 23, 48, ii 27 (NH/NS), HT 71:6+ IBoT 30.100:11 (NH/NS), KUB 17.36: 11 (NH/NS) —

Continued

Lemma*

Subject of transitive verb

‘having the oathʹ > ‘oath deityʹ

KUB 36.106 rev. 6 (OH/OS), KBo 6.34 iii 16 (MH/NS), KBo 6.34 iv 1(MH/NS) KUB 39.71 ii 44 (NS, but context partly broken) KUB 17.10 iv 10 (OH/MS) KBo 10.15:7ʹ, KBo 10.15:15ʹ (no context) KUB 29.9 i 26 (OH/NS) KUB 15.34 iv 32 (MH/MS), KUB 17.8 iv 9 (MH/NS) KBo 32.14 ii 7 (MH/MS), KBo 32.14 ii 8 (MH/MS), KBo 12.128 rt. Col. 5ʹ (MH?/NS) KUB 29.7 rev. 29 (MH/MS) —

lueššar ‘incense (?)ʹ luttai- ‘windowʹ milid- ‘honeyʹ NÁ- ‘bedʹ nepiš ‘skyʹ paḫḫur ‘fireʹ

papratar ‘impurityʹ peda- ‘placeʹ per ‘houseʹ

KUB 17.10 iv 9 (OH/MS), KUB 41.8 iv 33 (MH/NS), KBo 10.45 iv 35 (MH/NS)

Cases different from nominative

Subject of intransitive verb

Thematic forms in -a-

—

—

—

—

—

—

—

KUB 33.106 iii 40ʹ (MH/NS), KUB 44.4 rev. 2 (NH/NS) —

—

—

—

KUB 32.137 ii 2 (MH?/NS, voc.) —

—

—

KUB 41.8 iv 30 (MH/NS)

—

šagan ‘oilʹ šamalu ‘appleʹ šankunni- ‘priest (c.)ʹ

šankuwai- ‘nailʹ šannapili ‘emptinessʹ

šarawar ‘storm cloud?ʹ šeḫur ‘urineʹ šeḫuwal ‘sharp tool (?)ʹ ŠÈR.ŠÈR ‘chainʹ šuppal ‘beastʹ šušiyazkel (?)

KUB 24.1 i 12 (NH/NS), KUB 24.2 obv. 10 (NH/NS), KBo 10.15:3ʹ KUB 33.9 iii 12ʹ (OH/NS, no context) KUB 1.1 i 19 (NH/NS), KBo 3.6 i 16 (NH/NS), KUB 1.2 i 7 (NH/NS), KBo 16.83 iii 3 (NH/NS) KUB 9.4 i 35 (MH/NS) KUB 5.1 ii 58 (NH/NS), KUB 5.1 iii 27 (NH/NS), KUB 5.1 iii 69 (NH/NS), KUB 5.1 iii 75 (NH/NS), KUB 5.1 iv 10 (NH/NS), KUB 5.1 iv 30 (NH/NS), KUB 49.79 i 17ʹ (NH/NS), KUB 50.14 i 4 (NH/NS), IBoT 1.32 obv 22 (NH/NS) KBo 57.125 + KUB 18.11 obv. 5 (NH/NS) IBoT 1.36 i 34 (MH/MS) KUB 35.145 rev. 18ʹ (MH/NS) FHG 13 iii 7ʹ (NH/NS) KUB 36.32:5 (MH/MS), KUB 36.32:8 (MH/MS) KBo 17.90 ii 12ʹ (NS)

—

—

—

—

—

—

—

—

—

— —

— —

KUB 9.4 i 26 (MH/NS) —

—

—

—

— —

— —

— —

— —

— —

— —

—

—

— Continued

Lemma*

Subject of transitive verb

Cases different from nominative

tapiša- ‘vesselʹ

—

tapuwaš- ‘sideʹ

KUB 9.4 i 28 (MH/NS), KUB 9.34 ii 41 (MH/NS) KBo 25.184 iii 57ʹ (OH– MH/NS) KUB 33.106 i 8 (MH/NS) KBo 6.2 ii 54 (OH/OS) KUB 7.41 i 20 (MH/MS?) HKM 14: 4, HKM 15: 5, HKM 16: 6, HKM 20: 5, HKM 25: 12, HKM 34: 6, HKM 59: 10 (all MH/MS) KUB 23.11 iii 16ʹ (MH/NS), KUB 23.21 ii 30ʹ (MH/NS), KBo 2.5 iii 28ʹ (NH/NS), KBo 14.19 ii 27 (NH/NS) KBo 3.1 i 35ʹ (OH/NS), KUB 29.11 ii 10 (OH/NS), KUB 8.6 obv. 10 (OH/NS), KUB 8.24 ii 3ʹ (OH/NS), KUB 8.24+ 43.2 ii 17ʹ (OH/NS), KBo 13.34 iv?

taru ‘woodʹ tetḫeššar ‘thunderʹ tuek- ‘bodyʹ tunnakkeššar tuppi n.

tuzzi- ‘armyʹ

utne ‘landʹ

Subject of intransitive verb

Thematic forms in -a-

KBo 3.43 rev. 12ʹ (OH/NS, acc.), KUB 34.117:6ʹ (NS?, inst.) —

—

—

—

—

—

— — — —

— KUB 43.62 ii 4 (MH?/NS) — —

— — KBo 10.45 i 11ʹ (MH/NS) —

—

—

KBo 2.5 ii 13 (NH/NS), KBo 2.5 iii 24ʹ (NH/NS)

KUB 12.43: 6ʹ (OH/OS, gen.), KBo 3.22 obv. 25 (OH/OS, gen.), KBo 37.48+KUB 28.8 iii 12ʹ (OH/OS, gen.), KBo 37.48+KUB 28.8 iii 13ʹ

—

—

8ʹ (OH/NS), KUB 14.1 obv. 25 (MH/MS), KUB 24.4+ rev. 7ʹ (MH/MS), KUB 23.11 iii 27ʹ (MH/NS), KUB 23.11 ii 13ʹ (MH/NS), KUB 24.3 ii 49ʹ (MH/NS), DBH 46/2.127:12ʹ

(MH?/NS), KUB 8.35 obv. 16ʹ (MH?/NS), KUB 19.2 rev. 11ʹ (NH/NS), KUB 9.31 iii 54 (NH/NS), KUB 14.3 iii 58 (NH/NS), KBo 15.37 iii 22 (NH/NS)

(OH/OS, gen.), KBo 17.22 iii 15ʹ (OH/OS, gen.), KBo 17.22 iii 16ʹ (OH/OS, gen.), KBo 20.31 obv. 21 (OH/OS, abl.), KBo 17.22 iii 4ʹ (OH/OS, dat.), KBo 20.31 obv. 8 (OH/OS, dat.), KBo 18.151 rev. 15 (OH/OS, gen.), KBo 20.67 iv 22 (OH–MH/MS, gen.), KBo 7.28 + KBo 8.92 obv. 14ʹ (OH/MS, dat.), KUB 31.127 + KUB 36.79 i 16 (OH–MH/NS, gen.), KUB 34.15:10 ‘(OH/NS, dat.), KBo 9.68 rt. col. 10ʹ + KUB 43.2 iii 7ʹ (OH/NS, dat.), KBo 37.48+KUB 28.8 iii 3ʹ (OH/NS, dat.), KUB 12.21 ii 14ʹ (OH/NS, abl.), KBo 47.67 i 7ʹ (MH?/MS, dat.), KBo 47.67 i 16ʹ (MH?/MS, gen.), KUB 23.79 obv. (MH/MS?, dat.), KUB 35.140 i? 3ʹ (MH?/NS, gen.), KUB 35.140 i? 6ʹ (MH?/NS, gen.), KUB 35.140 i? 12ʹ (MH?/NH, gen.), KUB 8.35 obv. 11ʹ (MH?/NS, dat.), Continued

Lemma*

‘having the landʹ > ‘populationʹ

uttar ‘wordʹ

Subject of transitive verb

KUB 36.110:9‘(OH/OS), KUB 36.110:11ʹ (OH/OS), KUB 1.16 ii 62 (OH/NS), KUB 11.1 iii 11 (OH/MS?), KUB 34.19 rev. 9ʹ (MH/NS), KBo 5.6 ii 5 (NH/NS), KBo 3.4 iv 36 (NH/NS), KBo 5.4: 44 (NH/NS) KUB 17.10 iv 6 (OH/MS), KBo 19.92: 3ʹ (OH/NS), KUB 60.156 rev. 14ʹ (OH– MH/NS), KUB 36.91 rev. 8ʹ (+) KUB 43.68 rev. 11ʹ (OH–MH/NS), KBo 59.5 rev. 10ʹ (OH–MH/NS), Bo 27.203 obv. 7 (MH?/MS?), KBo 17.62 + 63 iv 10ʹ (MH/MS?), KUB 33.49 iii 4 (MH/NS), KUB

Cases different from nominative

Subject of intransitive verb

Thematic forms in -a-

—

—

KUB 34.7 iii 13ʹ (MH?/NS, dat.), KUB 24.13 iii 19ʹ (MH/NS, gen.), KUB 36.18 ii 13ʹ (MH/NS, dat.) ‘having the landʹ > ‘populationʹ KBo 17.1+ i 11ʹ (OH/OS, gen.), KUB 24.8+ ii 14ʹ (OH/NS, gen.)

—

walwallawarwalan ‘progenyʹ

watar ‘waterʹ

(MH/NS), KUB 45.20 ii 7ʹ (MH/NS), KUB 17.27 iii 9 (MH/NS), KBo 5.9 ii 25ʹ (NH/NS), KUB 15.28 iii 11ʹ (NH/NS), KUB 15.1 ii 32 (NH/NS) KBo 26.158 i 7ʹ (NH/NS) KUB 21.12 + KUB 23.127 + KBo 6.29 iii 11ʹ, Bo 86/299 ii 98, Bo 86/299 iii 10, Bo 86/299 iv 26, Bo 86/299 iii 10, KBo 4.10 rev. 25 (all NH/NS) KUB 31.127 + ABoT 1.44 iv 22 (OH–MH/NS), KBo 10.45 ii 23 (MH/NS), KUB 60.75:2ʹ (MH/NS), KBo 10.45 ii 24 (MH/NS), KBo 10.45 ii 33 (MH/NS), KUB 41.8 ii 14ʹ (MH/NS), KBo 10.45 ii 50 (MH/NS), KUB 41.8 iv 37 (MH/NS), KBo 10.45 iv 38 (MH/NS), KUB 43.58 i 43 (MH/MS)

— —

— —

— —

—

—

—

* Nouns in the table are neuter gender unless explicitly noted. Neuter nouns that occur as the subject of transitive verbs (column 1) potentially instantiate ergative -ant-. In the case of -ant- forms of linkiya- and utne-, Petra Goedegebuure distinguishes between ergative vs possessive adjectives, while Silvia Luraghi and Guglielmo Inglese do not set up a distinction between pairs of homonymous lexemes. Accordingly, a dashed line is used to indicate potentially different lemmas on which the authors do not agree.

5 Passives and anticausatives in Vedic Sanskrit Synchronic and diachronic perspectives Hans Henrich Hock

5.1 Introduction Since the time of Whitney (1879: §761b) and Delbru¨ck (1888: 267–268) certain Vedic Sanskrit verbs in suffix -ya-, inflected in the middle voice and with accent variation between root and suffix, have received special attention in western linguistics. Most recently, these and other verbs that bear semantic/pragmatic similarities to them have been discussed in a series of publications by Kulikov (1996, 2001 ≈ 2012, 2011a, 2011b, 2013), who argues for an analysis as nonpassive anticausatives which convey spontaneous actions without external agents. The notion anticausative has received various definitions and discussions. Levin and Rappaport Hovav (1995:130–133) discuss related phenomena in the context of the so-called causative alternation, including the fact that for verbs of change of state such as break, the noncausative has inchoative value; but their definitions of causative and noncausative are semantic, rather than morphosyntactic.¹ In approaches that do focus on morphosyntax, Dixon and Aikhenvald (2000b: 7–12) define anticausatives as a special variety of the passive that has no implied agent, and Comrie (1985: 326) claims that they are a category distinct from the passive, not only in terms of lacking an implied agent but also in terms of the fact that the action is conceived of as arising spontaneously. The latter definition is the one adopted by Kulikov. Note that in this context the terms spontaneous(ly) and agent are understood in a specific sense that differs from other uses of the terms. The term agent is used to refer to the agent phrase (a.k.a. underlying subject or logical subject) that may be used in prototypical passives and whose existence may be inferred even if there is no overt agent; and the term spontaneous is used ¹ Sanskrit anticausatives from roots meaning ‘break’ or ‘set free’ are indeed inchoative, but this is not the case for anticausatives from roots like ‘leave over’ or ‘leave behind’, which tend to be resultative. Inchoativity vs resultivity probably are features of the roots, and not of morphology or syntax.

Hans Henrich Hock, Passives and anticausatives in Vedic Sanskrit. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl, Oxford University Press. © Hans Henrich Hock (2022). DOI: 10.1093/oso/9780198857907.003.0005

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

167

to indicate that the action is conceived of as taking place without an implicit agent—in contrast to passives. I accept Comrie’s and Kulikov’s definition of anticausative for Sanskrit, because it finds support in Pāṇini’s native-speaker account; see Section 5.4. (The use of the term spontaneous by Haspelmath (2016) as a characteristic of transitive verbs is orthogonal to this definition.) Anticipating some of the later findings, the Sanskrit anticausatives, which are the major focus of this chapter, can be characterized as follows. In terms of their morphology, they are form-identical with the passives, except for the fact that a subset of anticausatives may have alternating root and suffix accentuation (while passives only have suffix alternation). Syntactically, they differ from passives in the fact that converbs are controlled by the surface subject, while in passives converb control is exerted by the agent as defined above, whether overt or implied, not by the surface subject. Since passives frequently occur without overt agents and anticausatives frequently do not occur with root accent or in structures with converbs, the two categories are in many cases not clearly distinguished and instances of putative passives and anticausatives very often are of ambiguous interpretation (Hock 2019). This chapter critically reviews earlier synchronic accounts of the relationship between Sanskrit anticausatives and passives, both modern western and Pāṇinian, argues for a syntactically based account compatible with that of Pāṇini, and offers an alternative historical perspective to the one proposed by Kulikov. Section 5.2 reviews the earlier accounts by Whitney and Delbru¨ck, as well as Speijer’s somewhat different approach. Section 5.3 is devoted to a critical examination of Kulikov’s evidence and arguments. Section 5.4 shows that Pāṇini’s account provides a better synchronic analysis, an analysis which suggests that the focus of western scholars on morphology and the morphophonology of root/suffix accent alternation may be mistaken and that greater focus on syntactic evidence is required. At the same time, the evidence needed to decide whether given constructions are passive or anticausative commonly is absent, and hence many individual cases must remain ambiguous. Section 5.5 addresses the historical relationship between passives and anticausatives; in this context, the inherent ambiguity of many structures turns out to be an important concern. Section 5.6 presents a summary of the conclusions and their implications for further research.

5.2 Whitney, Delbru¨ck, and Speijer Sanskrit has traditionally been analysed as having two different present formations characterized by the suffix -ya-, the passive in which the suffix bears the accent, and a nonpassive verb inflection in which the root typically is accented, the so-called

168

HANS HENRICH HOCK

Class 4 inflection; see (1a) vs (1b). Some verbs show accent variation between root and suffix and thus seem to straddle the divide between nonpassive Class 4 and Passive; see the root accent on múcyase in (2a) vs the suffix accent on mucyáse in (2b). For ease of reference, the class of verbs with accent variation will be referred to as the muc-class. (1) a. kri-yá-te ‘is (being) made’ (root kṛ) Passive b. mán-ya-te ‘thinks’ (root man) Class 4 (nonpassive) (2) a. ´svātréṇa yát pitrór múc-ya-se pári swelling.INS.SG.N when father.LOC.DU.M release-ya-3SG.MID on (RV 1.31.4c) ‘When you (Agni) get free through swelling on your parents (the kindling sticks)’ b. imā́ uptā́ mṛtyupās´ā́ this.NOM.PL.M cast.NOM.PL.M death.fetter.NOM.PL.M yā́n ākrámya ná mucyáse which.ACC.PL.M enter.CVB NEG release-ya-3SG.MID (AV 8.8.16ab) ‘These fetters of death are cast, having entered which you do not become free.’ ≈ ‘These fetters of death are cast, from which, once you have entered them, you do not become free.’ In western linguistics, Whitney (1879: §761b) apparently was the first to focus on the muc-class.² In his view, verbs of this class are a subset of Class 4 presents in -ya- which ‘have a more or less distinctly passive sense … with change of accent, and sometimes also with assumption of active endings’; see (3) for Whitney’s list of these verbs. Whitney, however, observes that ‘It is not possible to draw precisely the limits of the division’ between passives and Class 4 verbs. Moreover, when discussing the passive, Whitney notes that mriyáte ‘dies’ and dhriyáte ‘maintains itself, is steadfast’, in spite of their passive suffix accent, are not ‘used in a proper passive sense’, adding two other verbs, ā-dri-yá-te ‘heeds’ and ā-pri-yá-te ‘is busy’ as possible further examples (§773); see (3ʹ). Put differently, Whitney admits that in terms of function there is some leakage, both on the side of Class 4 verbs with accent variation and on the side of verbs classified as passives because of their non-varying suffix accent. ² In an even earlier publication, Delbru¨ck considers suffix accent to indicate passive, and root accent Class 4 nonpassive, no matter what the verb. He attributes the distinction to the Indian tradition, and claims that the closeness between passive and middle voice permitted some verbs (our muc-class) to waver between suffix-accent passive and root-accent middle (1874: 167–168).

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT (3)

muc [‘release’]³ kṣi ‘destroy’ jī/jyā ‘injure’ tap ‘heat’ dṛh ‘make firm’ pac ‘cook’ pṛ ‘fill’ mī ‘damage’ ric ‘leave’ lup ‘break’ hā ‘leave’

(3’)

mriyáte ‘dies’ dhriyáte ‘maintains itself, is steadfast’ ā-driyáte ‘heeds’ ā-priyáte ‘is busy’

169

Delbru¨ck (1888: 267–268) accepts Whitney’s interpretation of the muc-class as constituting a special subset of Class 4 verbs, but refines the characterization, especially by providing more detail on the issue of accent variation between root and suffix. These forms, which have a medial [i.e. middle-voice] function, sometimes wavering to a passive one, not rarely show the passive accent [i.e. on the suffix] beside the medial one [on the root], and such that the difference of accentuation seems to be distinguished according to scholarly schools. Thus, T[aittirīya] S[am ˙ hitā] prefers medial accent, M[aitrāyaṇi] S[am ˙ hitā] the passive accent.⁴ (Delbru¨ck 1888: 267)

Delbru¨ck provides a list of verbs belonging to this category, largely overlapping with Whitney’s, but unlike Whitney he provides textual examples (or references to passages in which forms occur); see, e.g. (4). Interestingly, in his glosses, Delbru¨ck sometimes employs a passive rather than a ‘medial’ nonpassive translation for structures with root accent; consider (4a) vs (4b) ³ Whitney did not provide a gloss for this verb. ⁴ ‘Diese Formen, welche medialen, manchmal in’s Passivische hinu¨berschwankenden Sinn haben, zeigen nicht selten neben dem medialen den passivischen Accent, und zwar so, dass die Verschiedenheit der Accentuirung sich nach Gelehrten-Schulen zu scheiden scheint. So hat TS eine Vorliebe fu¨r medialen, MS fu¨r passivischen Accent.’ (English translation by the author.)

170

HANS HENRICH HOCK

(4)

a. ná yásya hanyáte sákhā NEG REL.GEN.SG.M slay.PASS.3SG friend.NOM.SG.M ná jı̄ ́ yate kádā caná (RV 10.152.1c) NEG defeat.ya.3SG ever ‘[D]essen Freund nicht geschlagen, noch je besiegt wird’ [my formatting] (‘whose friend is not slain nor ever is/gets defeated’) b. yádā + éṣām ˙ pramı̄ ́ yeta yádā vā when they.GEN.PL die.ya.PRS.OPT.3SG When or jı̄ ́ yeran(TS 7.2.1.4) defeat.ya.OPT.3PL ‘[S]ollte von ihnen einer sterben oder sie unterliegen’ (‘if one of them should die or they (should) succumb/suffer loss’)

Like Whitney, Delbru¨ck notes that a similar variation between passive and medial, i.e. nonpassive function, holds for some of the verbs that are classified as passives because of their suffix accent. He recognizes a much larger set of such verbs, not only Whitney’s mṛ ‘die’ and dhṛ ‘hold fast’, but also dṛs´ ‘see’ and vid ‘find’ as in (5), as well as dhū ‘shake’, pṛc ‘mix, mingle’, lup ‘break, spoil’, vac ‘speak’, and vañc ‘move to and fro’. (5)

dṛs´-yá-te ‘wird gesehen’ : ‘zeigt sich’ (‘is (being) seen’ : ‘manifests itself, appears’) vid-yá-te ‘wird gefunden’ : ‘findet sich’ (‘is (being) found’ : ‘finds himself/appears’)

Here again, then, we see that the line of division between the muc-class and passives is not clear-cut. Nevertheless, both Whitney and Delbru¨ck maintain that nonpassive muc-class verbs and passives are distinct categories. A very different approach is taken by Speijer (1886: 239–141, 1896: 49–50) who interprets muc-class verbs to be passives: ‘On the other hand, the pure reflexive … is occasionally expressed by a passive; especially mucyate “to release oneself ”.’ Moreover, he notes, it is not ‘possible, even in accentuated texts, to draw everywhere with accuracy the boundary-line between’ Class 4 presents and the passive (1886: 240–241; similarly 1896: 50).⁵ To conclude this section, it is worth noting that Whitney claims that a ‘number [of class 4 verbs of the type muc] are in part evident and in part presumable ⁵ In this context, Speijer refers to Pāṇini’s account, but misinterprets Pāṇini’s notion of karmakartari as passive with ‘intransitive or reflexive meaning’ (1886: 239, with fn. 1). As shown in §4, Pāṇini’s account is different, as well as more insightful.

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

171

transfers from the passive or yá-class’ (§ 761b; formatting adjusted). Anticipated by Delbru¨ck’s 1874 publication, Speijer takes a different historical position, arguing that ‘At the outset, there is likely to have been one conjugation in -yati -yate with intransitive function, whence both the 4th class and the passive have sprung’ (1886: 241). Unfortunately, neither Whitney nor Speijer (or Delbru¨ck) provide any evidence for their claims.

5.3 Kulikov (1996, 2001, 2012, 2011a, 2011b, 2013) In an important series of publications on ya-formations in general (1996, 2001/2012, 2011a), Kulikov starts, like Whitney and Delbru¨ck, with the observation that although muc-class verbs look similar to passives by having the suffix -yaand middle-voice inflection, their accentuation varies between root and affix (as in (2a) vs (2b)), with different branches of the Veda tending to prefer one or the other accentuation.⁶ He also agrees with Whitney and Delbru¨ck in considering this verb type to be nonpassive. He goes beyond Whitney and Delbru¨ck by arguing that the verbs are intransitive anticausatives which convey spontaneous actions, and, importantly, lack any explicit or implicit external agent. Although among the ya-verbs with fixed accent (either on the root or on the suffix), some verbs may occasionally be anticausative, Kulikov claims that among the ya-verbs with variable accent, all verbs are anticausatives, that they are clearly distinct from passives, and that this combination of morphological and morphophonological behaviour sets them apart as a class. Moreover, based on the fact that the earliest Sanskrit text, the Rig Veda, has root accent, while suffix accent is found in the later Atharva Veda (see again (2a) vs (2b)), Kulikov argues that the accent of the Rig Veda must be original and that of the Atharva Veda innovated, through reinterpretation of intransitive ‘becomes free’ etc. as ‘is released’ etc. and due to the ‘increasing productivity of the -yá-passives’ (2011a: 196). In a different paper, published the same year (2011b), however, Kulikov argues for opposite developments of passives to anticausatives through ‘impersonalization’. He finds such developments with verbs of ‘perception and knowledge’, with vac ‘speak’, and with verbs of ‘caused motion’; see e.g. (6a) which is interpreted as passive, beside (6b) which is analysed as nonpassive anticausative.⁷ ⁶ Root accent prevails in RV, TS, TB, and ‘probably’ TA; suffix accent prevails in AV, MS, ŚB, and ‘most likely, in the texts of the Kāṭhaka school’ (Kulikov 2001: 533–538). ⁷ Many of these verbs were already recognized by Delbru¨ck as passives with ‘medial’ function.

172

HANS HENRICH HOCK

(6) a. yát svápne ánnam as´nā́mi if dream.LOC.SG.M food.ACC.SG.N eat.PRS.1SG ná prātár adhigam-yá-te NEG in.the.morning find-PRS.PASS-3SG sárvam ˙ tát astu me ´sivám ˙ all DEM.NOM.SG.N be.IMPV.3SG I.OBL(CLIT) propitious.NOM.SG.N nahı´ tád dṛs´-yá-te dı´vā (AV 7.101.1) for.not DEM.NOM.SG.N see.PRS.PASS.3SG day.INST.SG(ADV) ‘If I eat food in my dream, [and it] is not found in the morning, be all that propitious to me, for that is not seen by day.’ (Kulikov’s translation, my formatting) b. bálād ékam aṇīyaskám child.ABL.SG.M one.NOM.SG.N more.minute.NOM.SG.N utá ékam ˙ ná + iva dṛs´-yá-te (AV 10.8.25ab) CONJ one.NOM.SG.N NEG like see-PRS.PASS-3SG ‘One [thing] is minuter than a child, and another is as if it were invisible.’ (Kulikov’s translation, my formatting) Unfortunately, Kulikov provides no justification for analysing (6a) as passive or (6b) as anticausative. Given the context, nothing would preclude an alternative interpretation of (6a) as anticausative ‘is visible’ and of (6b) as passive ‘is seen’. Kulikov’s admission that anticausative function is not limited to verbs with accent variation thus converges with the observations of Whitney and Delbru¨ck that the line dividing the two categories is porous and hence raises questions about the uniqueness of anticausatives with accent variation. There are, in fact, further issues that raise questions for Kulikov’s analysis. First, and perhaps trivially, there are clear limitations on the evidentiary nature of accent variation. This is true most clearly for those branches of the Veda that prefer suffix accent; but it also holds for main-clause verbs in all branches, since these are unaccented (unless initial). Second, and more important, the absence of agent phrases is not a sufficient criterion to distinguish anticausatives from passives. Agent phrases are exceedingly rare even in structures generally recognized as passives: Out of some 290 (putative) present-system passives found in the Rig Veda,⁸ there are only 20 clear

⁸ My search started out from Avery (1872), and added some items that Avery apparently left out; also included are a few instances of Kulikov’s anticausative verbs or of verbs often classified as nonpassive

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

173

cases with an agent phrase (or about 7 per cent), plus 11 uncertain ones (ambiguous between agent and instrument or with the agent phrase possibly belonging to some other part of the sentence). So, even if everything is included, there are no more than 31 passives with agent phrases, or about 11 per cent of the total. To this must be added the fact, repeatedly acknowledged by Kulikov (e.g. 2001: 218–219, 283), that it is frequently difficult to decide whether a given verb should be interpreted as passive or as nonpassive anticausative. Consider again example (2a). Although the root accent suggests an anticausative reading, captured in the first translation (by me), Jamison and Brereton 2014, Geldner 1951, and even Kulikov (2001: 316) provide passive readings—‘are set free’, ‘entbunden wardst’, ‘were released’. (2) a. ´svātréṇa yát pitrór múc-ya-se pári swelling.INS.SG.N when father.LOC.DU.M release-ya-3SG.MID on (RV 1.31.4c) ‘When you (Agni) get free through swelling on your parents (the kindling sticks)’ Jamison and Brereton: ‘when through your swelling in your two parents [=the kindling sticks] you are set free …’ Geldner: ‘Sobald du mit Kraft in den Eltern entbunden wardst …’ (‘as soon as you were released in your parents with strength’) Kulikov: ‘When you (sc. Agni) were released from both parents by force …’ Similar variations in judgement can be found elsewhere. For instance, the passage in (4a), repeated for convenience, is interpreted as passive by Delbru¨ck, Geldner, and Jamison and Brereton, but as anticausative by Kulikov (2001: 283). Here, again, the root accent on jı̄ ́ yate evidently did not prevent a passive interpretation by most scholars. Perhaps the parallel passive form hanyáte lit. ‘is slain’, with suffix accent and hence classified as passive, influenced judgements in this case. However, an alternative, which pays closer attention to the accent of jı̄ ́ yate, would be possible, namely that hanyáte is anticausative too, meaning something like ‘comes to ruin’, and that the class of anticausatives with fixed suffix accent may be larger than Kulikov (2011b) realized. In that case, the alternative translation of the last gloss would be possible.⁹ (As we will see in Section 5.4, such an account would be perfectly possible under Pāṇini’s account.) class IV verbs: jyā, tap, dhṛ, mī, muc, śṛ-, sṛj, hā. The verb jan ‘be born’ is not included, since its syntactic behaviour is not passive. ⁹ Kulikov (2001: 227–229) considers the possibility that, in combination with prefixes, hanyáte may have had anticausative functions, but does not allow for anticausative function of unprefixed hanyáte.

174

HANS HENRICH HOCK

(4) a. ná yásya sákhā hanyáte hanyáte NEG REL.GEN.SG.M slay.PASS(?).3SG friend.NOM.SG.M ná jī́yate kádā caná (RV 10.152.1c) NEG defeat.ya/ANTICAUS.3SG ever Delbru¨ck: ‘[D]essen Freund nicht geschlagen, noch je besiegt wird’ [my formatting] (‘whose friend is not slain nor ever is defeated’) Geldner: ‘[D]essen Freund nicht geto¨tet noch jemals vergewaltigt wird’ (‘whose friend is not killed, nor ever scathed’) Jamison and Brereton: ‘[W]hose comrade is not slaid nor is he ever conquered’ Kulikov: ‘[W]hose (sc. Indra’s) friend is not killed, nor ever suffers loss’ to ruin ruin nor ever suffers loss’ Alternative: ‘whose friend does not come ruin, come to Possible evidence for such an interpretation can be found in later Vedic passages such as (7), where Eggeling provides a nonpassive interpretation of hanya̱te, entirely parallel to mīya̱te, which belongs to the class of anticausative verbs with accent variation. An alternative, more explicit rendering might be the second translation. (7)

ya̱d u ca + REL.NOM.SG.N PTCL CONJ ātmano̱ ’caraṇena

eva̱ + asya + EMPH 3SG.GEN hanyate vā

a̱tra here

self.GEN.SG.M non.conduct.INS.SG.N slay.PASS.3SG or mīya̱te vā (S´B 2.6.1.3) lessen.ya.3SG or injury or loss he suffers through his own Eggeling: ‘And whatever injury unrighteous conduct (…)’ to ruin ruin or suffers for him here because of his ‘And whatever goes goes to own misconduct’ Further, comparison of Geldner’s and Jamison and Brereton’s translations of the Rig Veda shows that variations in judgement are quite common. Consider the passage in (8). (8)

samudrám āsām áva tasthe agrimā́ sea.ACC.SG.N they.GEN.PL(CLIT.) descend.PRF.MID.3SG. first.NOM.SG.F ná riṣyati sávanaṃ yásmin ā́yatā NEG fail.PRS.ACT.3SG pressing.NOM.SG.N RP.INS.SG.N arrived.NOM.SG.F átrā ná hā́rdi kravaṇásya rejate here NEG heart.NOM.SG.N kravaṇa.GEN.SG.M tremble.PRS.MID.3SG

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

175

yátrā matı´r vidyáte pūtabándhanī (RV 5.44.9) where thought.NOM.SG.F find.PRS.PASS(?).3SG purified.connected.NOM.SG.F Geldner: ‘Ihre Anfu¨hrerin kam hinab zum Meere; nicht mißlingt die Somapressung, zu der sie sich hingezogen fu¨hlt. Da zittert nicht das Herz des Kravaṇa(?), wo die Dichtung sich befindet, die mit dem gela¨uterte (Soma) verbunden ist.’ (‘Their leader descended to the sea; the soma pressing that she is attracted to does not fail. The heart of Kravaṇa (?) does not tremble there where the poetry finds itself/occurs that is connected with the purified (soma).’ ) Jamison and Brereton: ‘The foremost of these (females) has stepped down into the ocean (of soma). The pressing is not harmed into which she is guided. Here the heart of the working poet [?] does not tremble, where the thought is found that is his bond to the purified (soma/fire).’ As in the case of hanyáte, the evidence of the Rig Veda does not permit a clear decision between the two different interpretations. However, in this case the later history of the language offers strong evidence that vidyáte did acquire anticausative function. Only under this assumption is it possible to explain that the padapāṭha of Pāṇini’s late Vedic grammar lists vid ‘sattāyām = in the sense of being’ (4.62) as a Class 4 verb different from Class 6 vid ‘lābhe = in the sense of obtaining’ (6.138) (and Class 2 vid ‘jñāne = in the sense of knowing’ (2.255)). Evidently, by Pāṇini’s time an anticausative vidyáte ‘finds itself, occurs’ had been lexicalized as a verb of existence. See Figure 5.1.

Early Vedic

vidyáte (passive and potentially anticausative)

Later Vedic

vidyáte ‘is seen’ (passive)

vidyáte ‘finds itself, occurs’ (anticausative)

Late Vedic

vidyáte ‘is seen’ (passive)

vidyáte ‘is, exists’ (lexicalized)

Fig. 5.1 The development of vidyáte

Perhaps most important, Kulikov fails to draw on a more reliable, SYNTACTIC criterion to define anticausatives,¹⁰ namely converb control. As noted in Hock 1986: 87, passives and anticausatives¹¹ differ syntactically: in passive structures, converbs are controlled by the underlying agent, whether overt or implicit, as in (9a); in anticausative structures, it is the surface subject that controls converbs, (9b).

¹⁰ Except for a brief reference in 2001: 316 to Hock (1986). ¹¹ In 1986, I used the term intransitive.

176

HANS HENRICH HOCK

(9) a. tau vā etau Pragāthāv … DEM.NOM.DU.M PTCL DEM.NOM.DU.M pragātha.NOM.DU.M punarādāyam ´sasyete (KS 28.2) punarādāyam ˙˙ again.take.up.CVB sing.PRS.PASS.3DU ‘Those two pragāthas are sung (by one who is) taking them up again’ Not: ‘The two pragāthas, having taken them [= themselves?!] up again, are sung’ b. ubhau saha + uttamam ˙ sam ˙ vatsaram eva both.NOM.DU.M together last year.ACC.SG.M EMPH saha vimucyete (AB 3.17.3) kalpayitvā kalpayitvā + ubhau establish.CVB both.NOM.DU.M together free.ya/ANTIPASS.PRS.DU.3 ‘Both together having established the last year, both get free together’ Not: ‘Both are freed together [by some other two] having (i.e. who have) established the last year together’ The test of converb control makes it possible to firmly establish anticausative status for many constructions, including ones that involve verbs not considered by Kulikov, or his predecessors, such as (10d) and (10e). (10) a. imā́ DEM.NOM.PL.M yā́n REL.ACC.PL.M

uptā́ cast.NOM.PL.M ākrámya ná ākrámya enter.CVB NEG

mṛtyupās´ā́ death-fetter.NOM.PL.M mucyáse release.ya.3SG.MID (AV 8.8.16ab = (2a) above) entered which you do not ‘These fetters of death are cast, having having entered become free.’ (≈ ‘These fetters of death are cast, from which, once you have entered them, you do not become free.’ ) b. ´ındro vṛtrám ˙ hatvā́ Indra.NOM.SG.M Vṛtra.ACC.SG.M slay.CVB devátābhis´ ca + indriyéṇa ca vy ā̀rdhyata (TS 2.5.3.1) deity.INS.PL.F & power.INS.SG.N & deprive.ya.IMPF.3SG having slain slain Vṛtra, was bereft of/lost the deities and power.’ ‘Indra, having c. tās sīmānam eva+ ūrdhvā udīrya udīrya + DEM.NOM.PL.F parting.ACC.SG EMPH above.NOM.PL.F rise.CVB asṛjyanta (JB 3.104) release.ya.IMPF.3PL ‘They, rising rising up up above (his) hair parting, came into existence.’ (Sim. JB 1.72, S´B 11.1.6.7, 8, 4.1, etc.) d. sarva̱m ˙ sūy(ai) (S´B 5.2.3.1; sim. S´B 5.2.3.6) parigṛhya parigṛhya all.ACC.SG.N encompass.CVB impel.ya.SBJV.1SG Having encompassed encompassed everything, may I undergo the (rājasūya) ‘Having ‘consecration.’

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

177

e. sa bhṛgum ˙ ’ntaradhīyata (GB 1.1.4) sṛṣṭvā sṛṣṭvā he.NOM.SG.M Bhṛgu.ACC.SG.M create.CVB hide.ya.IPF.3SG having created created Bhṛgu, became hidden/hid himself.’ ‘He, having A potential additional problem is the claim that anticausatives refer to spontaneous actions, without an implied agent. In examples like (2) = (10a) it would be possible to postulate an agent who makes the release possible, just as in (9a) it is possible to infer the existence of an implicit agent (see Koontz-Garboden 2005, fn. 13 for such an argument in a different context). However, this approach would fail to account for the systematic difference in converb control between anticausatives and passives. Moreover, in many instances the approach also fails on semantic grounds. Consider (10b) and (10e). In (10b) it was Indra’s action of having slain Vṛtra that deprived him of the deities and of power; there is no implied outside agent. Similarly, in (10e) Brahman went into hiding, not by somebody else’s action but on his own. In many such cases, a reflexive reading is possible; but in others such a reading can be safely ruled out. Thus, in a lengthy passage (S´B 11.4.1.5), the life of a human being is characterized as being born toothless, growing teeth, losing the teeth (prabhid-ya̱-nte split.forth-ANTICAUS-3PL.PRS), and in the final stage of life again losing the teeth (prabhid-ya̱-nte split.forth-ANTICAUS-3PL.PRS). Clearly, the loss of one’s first teeth simply happens, without outside or some kind of reflexive agency. Finally, we must take seriously the characterization of our anticausatives by the Indian grammatical tradition as indicating an action conceived of as spontaneous, without an external agent (see the next section). To sum up: Kulikov has made an important contribution by introducing the term anticausative, rather than earlier, more nebulous terms such as medial or intransitive, and specifically by focusing on the fact that the muc-class verbs convey spontaneous actions and lack any explicit or implicit external agent. His morphological/morphophonological focus on verbs with alternating accent, however, covers only part of the territory and ignores important aspects of the SYNTACTIC behaviour of anticausatives. At the same time, the fact that only some anticausatives have alternating accent may raise questions as to what semantic/pragmatic or historical factors might be responsible for this restriction. This issue is addressed in Section 5.5.

5.4 Pāṇini’s account Pāṇini’s grammar operates with a system of kārakas (roughly, underlying cases), which by other rules are realized as overt cases. Kārakas are assigned to participants in an action, such as karaṇa ‘means, instrument’, karman ‘underlying object, patient’, and kartṛ ‘independent agent, underlying subject’. See, e.g. (11).¹² ¹² This is a condensed version of the more detailed discussion in Hock (2016); see Cardona (1974, 1976, and 1978) for more detailed discussion.

178

HANS HENRICH HOCK

(11)

devadatto ’sinā vṛkṣam ˙ bhinátti D.NOM.SG.M axe.INS.SG.M tree.ACC.SG.M split.PRS.3SG KARTṚ KARAṆA KARMAN (marked for KARTṚ) ‘Devadatta splits the tree with an axe.’

Marking of the karman or kartṛ on the verb affects case assignment; compare (11) where the verb is marked for the kartṛ, with the passive (12) where it is marked for karman. (See Pāṇini 3.1.62 and 3.1.67.) (12)

devadattenā ’sinā vṛkṣo bhidyáte D.INS.SG.M axe.INS.SG.M tree.NOM.SG.M split.PRS.3SG KARTṚ KARAṆA KARMAN (marked for KARMAN) ‘The tree is split by Devadatta with an axe.’

The grammar also has to allow for the fact that speakers may choose to treat what is classified as karaṇa or karman in the kāraka section as kartṛ ‘independent agent’. In the majority of cases, this has no consequence on verb structure, as in (13), where the participant classified as karaṇa ‘instrument’ is treated as independent agent, kartṛ. (13)

asir vṛkṣam ˙ axe.NOM.SG.M tree.ACC.SG.M KARTṚ KARMAN ‘The axe splits the tree.’

bhinátti split.PRS.3SG (marked for KARTṚ)

However, when the karman is treated as kartṛ, in what is called karma-kartari ‘under the condition of a karman-kartṛ’, the verb is marked the same way as in the passive, including accent on the suffix -yá-; see (14) and Pāṇini 3.1.62, 3.1.67, and 3.1.87. (14)

vṛkṣaḥ bhidyáte (svayam eva) tree.NOM.SG.M split.PRS.SG.3 by itself EMPH KARMAKARTṚ (marked for KARMAKARTṚ) ‘The tree splits (on its own).’

Patañjali’s commentary makes it clear that in such cases, the action takes place on its own, without any outside agency: But the grain pot that breaks without wind, without rain (falling) on (it), suddenly, of that (pot) there is no other kartṛ, other than this grain pot.¹³ (Patañjali’s Mahābhāṣya on P. 3.1.87)

¹³ yas tu khalu nivāte nirabhivarṣe acirakālakṛtaḥ kuśūlo bhidyate tasya nānyaḥ kartā bhavati anyad ataḥ kuśūlāt.

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

179

This, of course, is a major criterion in Kulikov’s definition of anticausatives. Put differently, on this count there is no conflict between Pāṇini and Kulikov, contrary to Kulikov’s (2013) claim.¹⁴ Further, by limiting the use of converbs to the condition that their kartṛ and the kartṛ of the matrix verb must be identical (3.4.21), Pāṇini accounts as follows for the difference in converb control in structures like (9), repeated with some modification. In both cases, the converb is controlled by the kartṛ, but in (9a) that kartṛ is an (overt or implicit) agentive agent, meeting the prototypical definition of kartṛ, in (9b) it is the karmakartṛ, that is a prototypical karman that is treated by the speaker as kartṛ. pragāthāv (9) a. tau vā etau … DEM.NOM.DU.M PTCL DEM.NOM.DU.M pragātha.NOM.DU.M punarādāyam ˙ Ø ´sasyete (KS 28.2) again.take.up.CVB (somebody) sing.PRS.PASS.3DU KARTṚ (marked for KARTṚ) ‘Those two pragāthas are sung (by someone who is) taking them up again’ b. ubhau saha + uttamam ˙ sam ˙ vatsaram eva both.NOM.DU.M together last year.ACC.SG.M EMPH kalpayitvā ubhau saha vimucyete (AB 3.17.3) establish.CVB both.NOM.DU.M together free.ya/ANTIPASS.PRS.DU.3 KARMAKARTṚ (marked for KARMAKARTṚ) ‘Both together having established the last year, both get free together’ Pāṇini’s approach, thus, provides an account for all the syntactic facts. However, as in other approaches, this is an account in principle. Without the presence of converbs,¹⁵ of overt agent phrases, or of root accentuation, structures such as (15) remain ambiguous between passive and karmakartari/anticausative interpretation. (15)

vṛkṣaḥ bhidyáte tree.NOM.SG.M split.PRS.PASS.SG3 KARMAN (marked for KARMAN) ‘the tree is (getting) split (by somebody)’ or tree.NOM.SG.M split.PRS.ANTICAUS.SG3 KARMAKARTṚ (marked for KARMAKARTṚ) ‘the tree splits’

Finally, the morphophonological issue of accent variation is covered much later in Pāṇini’s grammar, as an optional rule (6.1.195), which significantly affects only ¹⁴ Kulikov simply misunderstood Pāṇini; see Hock (2016) for further discussion. ¹⁵ Or reflexives, which also are controlled by the kartṛ.

180

HANS HENRICH HOCK

some classes of verbs. Accent variation, thus, is not a primary factor in Pāṇini’s account for anticausatives. This is an important point, which suggests that Whitney’s, Delbru¨ck’s, and Kulikov’s focus on accent variation may have been a mistake. Rather, it is the syntactic and associated pragmatic behaviour of anticausatives that seems to be of primary significance.

5.5 The historical relationship between anticausatives and passives The major points of this chapter so far can be summarized as follows. Kulikov provides a more satisfactory account of muc-class verbs and others with similar behaviour than his western predecessors by introducing the term anticausative, in the sense of nonpassive structures that convey spontaneous actions without external agents. However, his and his predecessors’ focus on accent variation leads them to focus on only part of the territory; in fact, especially in his 2011b paper, Kulikov acknowledges that there are anticausatives that do not exhibit accent variation. In all of these respects, Pāṇini provides a more satisfactory account by separating syntactic behaviour from the optional morphophonological one. Especially important in this regard is the issue of converb control, adumbrated in Hock 1986, but more explicitly accounted for much earlier, in the work of Pāṇini. Converb control establishes a robust syntactic test for anticausativehood, while semantic/pragmatic criteria run into the difficulty, acknowledged by Kulikov and his predecessors, and confirmed through varying interpretations by the translators, that the boundary between passives and anticausatives is porous. Importantly, where muc-class verbs are not root-accented, and in the absence of converb evidence or of agent phrases, many structures remain of ambiguous interpretation. This very ambiguity, however, combined with the fact that most structures traditionally classified as passives typically do not have overt agents, offers ample opportunities for reinterpretations of passives as anticausatives—or vice versa.¹⁶ Some cases of this sort are covered in Kulikov 2011b; others include the development of vidyáte ‘is found’ to ‘appears, occurs’, and further (by lexicalization) to ‘is, exists’, and possibly examples such as sūyai ‘may I undergo the consecration’ and antar (a)dhīyata ‘hid himself, went into hiding’ in (10d) and (10e). Before pursuing this line of thought further, let us take a brief look at earlier historical perspectives. ¹⁶ The latter possibility is explored in more detail in Hock (2019).

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

181

5.5.1 Earlier historical perspectives The idea that the difference between anticausatives and passives may not be original has been around at least since the time of Delbru¨ck 1874, but opinions differ on the nature of the ancestral structures and the developments that led to the distinction. Delbru¨ck (1874) suggests that the passive developed out of intransitive Class 4 verbs; Whitney proposes exactly the opposite; and Speijer suggests an original class of intransitives in -ya- that split into Class 4 presents and passives; see the following citations. Thus one sees that an original difference between passive and medium [i.e. middle voice] does not exist. But the reason why the slowly developing passive function is especially connected with the medium of the ya-class [i.e. Class 4 presents] has for a long time been found in the fact that especially this class for the most part consists of intransitives, i.e. of verbs that least clearly express an action by the subject.¹⁷ (Delbru¨ck 1874: 168) It is not possible to draw precisely the limits of the division between passives and Class 4 verbs. A … number [of Class 4 verbs] … are in part evident and in part presumable transfers from the passive or yá-class, with change of accent … (Whitney 1879: § 761; Whitney’s formatting) At the outset, there is likely to have been one conjugation in -yati -yate with intransitive function, whence both the 4th class and the passive have sprung. (Speijer 1886: 240; similarly 1896: 49–50)

A major obstacle to Whitney’s proposal is the fact that the passive inflection in -ya- is an innovation in Indo-Iranian (and Armenian) and cannot be traced back to Proto-Indo-European (PIE). His proposal, thus, can be safely ruled out. Delbru¨ck’s and Speijer’s hypotheses can be reconciled by noting that both operate with the assumption that the ancestral construction was intransitive, and in fact, we can reconstruct PIE verbs in *-ye/yosome of which are intransitive. However, as we will see, many details remain unexplained. ¹⁷ ‘Somit sieht man, dass ein urspru¨nglicher Unterschied zwischen Passiv und Medium nicht vorhanden ist. Der Grund aber, warum der allma¨hlich entstehende Passivbegriff sich gerade an das Medium der ya-Classe anlehnt, ist schon la¨ngst in dem Umstande gefunden, dass gerade diese Classe zum gro¨ssten Theil aus Intransitivis besteht, also aus solchen Verben, welche am wenigsten deutlich eine Handlung des Subjects ausdru¨cken.’ (English translation by the author.)

182

HANS HENRICH HOCK

5.5.2 Kulikov’s accounts In his various publications, Kulikov goes farther in the direction of an explanation. Two accounts must be distinguished. One is the Vedic-internal history of anticausatives; the other concerns the developments from PIE. Let us first look at Kulikov’s Sanskrit-internal proposals. First, as regards mucclass anticausatives, Kulikov argues that root accent is original, that the verbs are Class 4 presents in origin, and that suffix accent is secondary. Kulikov finds support for this hypothesis in the fact that root accent is attested in the Rig Veda, while suffix accent first appears in the Atharva Veda, a later Vedic text (e.g. 2011a: 196). As Kulikov sees it, suffix accentuation results from reinterpretation of intransitive ‘becomes free’ etc. as ‘is released’ etc. and is supported by the ‘increasing productivity of the -yá-passives’. For other anticausatives, with invariable suffix accent, Kulikov (2011b) argues for reinterpretations in the opposite direction, from passive to anticausative. Regarding the prehistoric origins of passives and Class 4 presents, Kulikov 2011a offers an extensive discussion of earlier views, most of which argue for derivation from PIE nonpassive -ye/yo-verbs with suffix accent. A strong argument in favour of this view is that the root of Sanskrit ya-verbs, whether passive or nonpassive, is overwhelmingly in the zero-grade (or in the synchronically weakest form acceptable in the environment before y). In Indo-European terms, accent on a zero-grade root would be anomalous, so from this perspective, suffix accent would be preferred. Kulikov finally comes to the following conclusion: The difference in accentuation between (middle) class IV presents and -yápassives is clearly secondary. We can only speculate why the passive subclass has generalized the suffix accentuation (which probably was original), while nonpassives have retracted accent to the root. This accent shift may have started in a few old non-passive -ya-presents, in which zero and full grade could not be distinguished (e.g. mányate), or where the full grade was introduced instead of the phonetically impossible zero grade (as in padyate ← **pdyáte) or in order to avoid morphological opacity (nahyati/te ← **ahyáti/te < *n̥hi áti/-tai (?) or ̑ asyati/te ← *si áti/-tai). ̑ (Kulikov 2011a: 197)

His final, speculative scenario is summarized in Figure 5.2, where I have highlighted the relevant parts by means of an added oval and circle. Although Kulikov’s complex scenario is admittedly speculative, it makes an important contribution by attempting to provide a more comprehensive account than his predecessors.

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

183

Proto-Indo-Aryan I. Accent retraction in non-passive -ya-presents. Accentual differentiation of -yá-passives and (middle) non-passive -ya-presents

II. Accent shift in the type mriyáte: *Cŕ̥ -i̯ a- — → Criyá-

√ø-i̯ a-te

√ø-i̯ á-te (e.g. hanyáte ‘is killed’)

√ø-i̯ á-te (hanyáte)

√ǿ-i̯ a-te (e.g. mányate ‘thinks’)

Criyá-te (mriyáte)

√ǿ-i̯ a-te (mányate)

Old Indo-Aryan (Vedic) dialects III. Accent shift in some (middle) non-passive -ya-presents (in the dialects √ø-i̯ á-te of the AV, MS-KS and ŚB) (hanyáte)

Criyá-te (mriyáte)

√ø-i̯ á-te (mucyáte AV, MS)

√ǿ-i̯ a-te (mányate)

Fig. 5.2 Kulikov’s speculative scenario (2011a: 198) (with oval and circle added by author)

5.5.3 Problems with Kulikov’s accounts Nevertheless, problems remain. One of these concerns Kulikov’s account for mucclass anticausatives, the last development in Figure 5.2. First, I find it difficult to understand how an ‘increasing productivity of the -yá-passives’ would have influenced the accentuation of nonpassive anticausatives. Second, and more important, Kulikov himself notes that root vs suffix accent is a matter of differences between Vedic schools. The difference between the root accent of the Rig Veda and the suffix accent of the Atharva Veda, therefore, need not be a matter of chronology but may reflect differences in school affiliation. Finally, there is independent evidence for reinterpretation of passives as anticausatives, including the evidence adduced in Kulikov 2011b; on the other hand, there is no historically verifiable evidence for developments in the opposite direction. If we take these findings seriously, we would have to conclude that the muc-class anticausatives are more likely to have originated by the same path—reinterpretation of original passives. In that case, the fact that unlike other, apparently later reinterpretations, the muc-class anticausatives adopted the root accent of Class 4 verbs in some Vedic schools can be accounted for by assuming that in these schools the muc-class anticausatives originated early enough to participate in the developments that led to the fixation of root accent in nonpassive Class 4 verbs vs suffix accent in the passives.

184

HANS HENRICH HOCK

What complicates matters is that as observed repeatedly in the earlier sections of this chapter, the distinction between muc-verbs and passives is not clear-cut. In fact, as argued in Hock 2019, there is a (near-)systematic ambiguity between these two types of structures, permitting reanalyses in either direction, without there being clear evidence as to the direction.

5.5.4 Towards an alternative account Whatever the nature of the Vedic/Sanskrit developments may have been, one thing is clear. The passive in *-ye/yo- > -ya- and its contrast with an anticausative in -ya- are innovations; PIE did not have a separate passive morphological or anticausative category. Vedic/Sanskrit traces of an earlier stage can be found in a few nonpassive yapresents with suffix accent. These include Whitney’s examples in (3ʹ), repeated for convenience, as well as vacyáte ‘moves in a wavering fashion; gallops’. Of these, both mriyáte and vacyáte are found in the Rig Veda. These forms may be considered archaic relics of the original suffix accent of all ya-verbs with zero-grade root, whether passive or anticausative, or nonpassive/nonanticausative for that matter. (3’)

mriyáte ‘dies’ dhriyáte ‘maintains itself, is steadfast’ ā-driyáte ‘heeds’ ā-priyáte is busy’

Kulikov (e.g. 2011a: 187–188) tries to account for the suffix accent in mriyáte phonologically from an original root accentuation: ‘The only way to preserve the transparency of the form was to introduce the accent on the suffix: *Cŕ̥-i a- → ̑ Criyá- . . ’. But he does not explain why the change of r̥ to ri should have entailed accent shift. As regards RV vacyáte,¹⁸ Kulikov (2011b: 245) proposes reinterpretation of an original passive; but there is no corresponding active in the Rig Veda and the Class 1 verb vañcati (etc.), attested in the later Atharva Veda, has the same semantic range. Further, the contexts in which vacyáte and related forms occur do not require passive interpretations; Geldner, and Jamison and Brereton consistently offer nonpassive translations; and Grassmann 1873, while listing the verb as passive, adds the comment ‘depon[ent]’. The common passive interpretation (e.g. Avery 1872, Lubotsky 1997) seems to be based on no evidence other than the suffix accent. In fact, LIV2 (s.v. ?*uenk) states ‘In spite of suffix accent [this is] ̑ ¹⁸ Attestations of vacyáte and related forms are found at 1.46.3a, 1.142.4d, 1.184.3c, 3.6.1a, 3.6.2d, 3.39.1a, 9.2.2a, 9.97.2c, 9.108.10a, and 10.47.7c.

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

185

NO passive; as in the case of mriyáte the old place of accent of the ie-presents is ̑ exceptionally retained.’¹⁹ What, then, might account for the accent difference between nonpassive Class 4 presents and the passives? One possibility, which like Kulikov starts with the reconstruction of all PIE -ye/yo-presents with suffix accent, would be to assume that the root accent of the nonpassive Class 4 presents is modelled on the root accent of the likewise nonpassive Class 1 presents (as suggested by Kuryłowiz 1952, as well as Gonda 1951, referred to in Kulikov 2011a). Especially important in this regard is that Class 1 verbs not only have root accent but also constitute the most productive category of nonpassive verb inflection. There is, however, a possible alternative, adumbrated by Kulikov but not further pursued. This is the possibility that beside -ye/yo-presents with zero-grade root and suffix accent, PIE also had -ye/yo-presents with full-grade root and accent on the root. This possibility is widely embraced in recent Indo-Europeanist publications, including Sihler 1995: 503, Meier-Bru¨gger 2000: 160, Fortson 2004: 89, Ringe 2006: 28, and the important LIV2 : 19. Opinions differ as to which verbs belonged to the root-accented full-grade class, and Sihler raises some important questions regarding the antiquity of this formation. If accepted, however, this reconstruction would make it possible to derive at last some Class 4 verbs from this formation. These include the verbs in (16). (16)

pádyate ‘falls; walks’ < ?*péd-ye- (or *pe d-yé-?) pás´yati ‘sees’ < *(s)péḱ-yerā́yasi ‘you bark’ < *léh2 -ye háryati ‘rejoices’ < ?*ǵhér-ye-²⁰

If PIE did indeed have two types of -ye/yo-verbs, one with full-grade root and root accent, the other with zero-grade root and suffix accent, that would provide a more tangible motivation for the split between Class 4 and the passives. The variation in accent would have been amenable to polarization, such that root accent was specialized for nonpassive Class 4 verbs, modelled on the root accent of the nonpassive and highly productive Class 1 verbs, and suffix accent for the passives. Putting together the various alternatives to Kulikov’s account considered here results in the scenario in Figure 5.3. The most important steps are the pre-Vedic accent polarization, with support from the nonpassive Class 1 verbs; the fact that muc-class verbs (Anticausatives I) participated in the specialization of root accent in some of the Vedic schools; and the assumption that later reinterpretations of

¹⁹ ‘Trotz Suffixakzent KEIN Passiv, wie bei mriyáte ist ausnahmsweise die alte Akzentstelle der iȇ Pra¨sentien erhalten.’ ²⁰ LIV2 adds the note ‘Ved. hárya la¨ßt sich kaum als Neubildung erkla¨ren und deutet daher auf hohes Alter der vollstufigen Bildung …’ (‘Ved. hárya is hardly explainable as an innovation and therefore points to high antiquity of the full-grade formation …’).

186

HANS HENRICH HOCK PIE

Pre-Vedic polarization with influence from nonpass. Class 1

ë-ye/o-

√-yé/ó-

nonpass.

ë-ya-

pass. √-yá-

muc-class anticaus. with accent shift to root in some ‘schools’, vs. nonshift in others

nonpass.

ë-ya-

√´-ya-/√-yá(Anticaus. I)

Later anticausatives (e.g. vidyáte)

nonpass. ë-ya

pass. √-yá-

√´-ya-/√-yá √-yápass. √-yá(Anticaus. I) (Anticaus. II)

Fig. 5.3 Alternative speculative scenario

passives as anticausatives (Anticausatives II), came too late to participate in the root-accent specialization.

5.6 Conclusions and outlook Kulikov’s arguments for a distinction between Sanskrit passives and anticausatives constitute a definite improvement over earlier western approaches. Further support for such a distinction comes from the syntactic evidence of converb control, as well as the account of the Sanskrit grammatical tradition going back to Pāṇini. However, the distinction is one in principle only; where accentual or converb evidence is lacking, the formal identity of passives and anticausatives leads to near-systematic ambiguity, especially because even passives generally occur without overt agent phrases. This ambiguity, in turn, has consequences for any attempt to investigate the relationship between anticausatives and passives, for in principle it would allow for reinterpretations in either direction. The consequences are especially relevant for attempts to unravel the development of both passives and anticausatives out of PIE nonpassive, nonanticausative -ye/yo-presents. Kulikov argues for an early development of anticausatives as a subcategory of root-accented nonpassives. This chapter offers an alternative account according to which anticausatives developed from earlier passives through reinterpretation of ambiguous structures.²¹ Note, however, that both accounts are highly speculative. The ultimate relation between passives and anticausatives may not be resolvable based on the evidence of Sanskrit; the formal aspects of the Indo-European antecedents are not fully settled; and, as noted, the development of a distinct passive is an innovation (shared ²¹ According to Haspelmath (1990), anticausatives can change to passives, but not vice versa. The present account, if correct, suggests that this generalization may need to be reconsidered.

PASSIVES AND ANTICAUSATIVES IN VEDIC SANSKRIT

187

with Iranian and apparently also with Armenian) whose earlier history cannot be documented. This conclusion may come as a disappointment, but given the fact that even in Vedic (and later Sanskrit) the boundary between passives and anticausatives is porous, it should perhaps not be surprising. At any rate, it is best to acknowledge these problems, so that they remain as a challenge for future research.

Acknowledgements Earlier versions of this chapter were presented at the 2016 East Coast IndoEuropean Conference in Athens, Georgia, and at the 2017 International Conference on Historical Linguistics, San Antonio, Texas. See also Hock 2016 and 2019, where some of the issues addressed in this chapter are dealt with in a different context. I am grateful to Leonid Kulikov for sending me a copy of his 2001 dissertation. Comments by an anonymous reviewer have helped me improve both presentation and argumentation.

6 Non-nominative arguments, active impersonals, and control in Latin Michela Cennamo and Claudia Fabrizio

6.1 Introduction This chapter discusses some active impersonal constructions with nonnominative (accusative and/or oblique) arguments in Latin, in relation to (i) the presence of the accusative with some verbs (e.g. me pudet I.ACC shame.PRS.IND.3SG ‘it shames me’), (ii) the alternation accusative/dative with other types of verbs (e.g. me/mihi decet I.ACC/DAT become.PRS.IND.3SG ‘it becomes me’), (iii) the accusative/oblique realization of the verb’s argument(s), according to the verb/predicate and the type of construction (e.g. me eius miseret I.ACC he.GEN take.pity.PRS.IND.3SG ‘I take pity on him’, me-I.ACC latet.escape.PRS.IND.3SG ‘it escapes me’, mihi-I.DAT liquet.be.clear.PRS.IND.3SG ‘it is clear to me’, nivit sagittis snow.PRS.IND.3SG lightning.ABL.PL ‘it snows (with) lightning[s]’, (iv) the relationship among the impersonal active, personal active, and passive patterns, which are sometimes available for one and the same verb (e.g. fallo ‘to deceive’). The discussion is organized as follows. Section 6.2 illustrates the role played by the notion of control in some voice alternations in Latin. Section 6.3 describes impersonal constructions and the strategies realizing them, focusing on some ‘impersonal’ patterns with active verb morphology and accusative/oblique arguments. Evidence is provided in Section 6.4 for their witnessing the existence of a dependent-marked active/agentive-coding subsystem, already attested in Early Latin in some grammatical domains, whereby inactive arguments are coded in the accusative,¹ the case of canonical objects/patients of transitive clauses and of ‘inert’

¹ In Early Latin inactive/nonagentive arguments in the accusative rather than the expected nominative case obtain in nominal clauses involving ellipsis of the verb sum ‘to be’ (i), exclamative–presentative patterns introduced by eccum, eccillam, eccillum (through univerbation of the adverb ecce ‘here’ and the accusative of the pronouns is, ille ‘he’, iste ‘this’ (ii) (Cennamo 2001, 2009: 311–313), and with second declension thematic/weak ‘neuters’ of the o-stem (e.g. uterum ‘belly’, corium ‘skin’, caelum ‘sky’) (with corresponding ‘secondary’ forms of animate masculine/feminine gender, depending on the noun), occurring in syntactic contexts with an inactive S (i.e. unaccusative structures), illustrated in (iii) for a passive construction, with the neuter variant witnessing the early occurrence of the accusative to mark inactive arguments (Lazzeroni 2002a, 2002b, Rovai 2007a; 2007b, Cennamo 2009: 314)): Michela Cennamo and Claudia Fabrizio, Non-nominative arguments, active impersonals, and control in Latin. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl, Oxford University Press. © Michela Cennamo and Claudia Fabrizio (2022). DOI: 10.1093/oso/9780198857907.003.0006

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

189

(Collinge 1978), less affected arguments (the dative for experiencers), the verb reverting to the default third-person singular ‘impersonal’ active form. Finally, Section 6.5 provides the conclusions.

6.2 Control and voice alternations in Latin: active–medio-passive, active–impersonal Control, the semantic spectrum reflecting the degree of ‘primary responsibility of a participant over the verbal process’ (Lakoff 1977), plays an important role in the encoding of transitivity and of the argument structure of the clause in Latin, both synchronically and diachronically (Cennamo 1998: 83–88, 2001: 54–58, 2009, 2011, 2016: 967–971, 2020, also for the transition from Latin to Romance).² This notion involves various transitivity features such as agentivity, volitionality, individuation of the clause nuclear participant(s) (e.g. animacy, definiteness, referentiality), and the aspectual nature of the predicate (i.e. the types of eventualities) (Timberlake 1977: 162, Lehmann 1988: 57–61, Comrie 1989: 61–62, Klaiman 1991). For instance, with animate subjects it determines fluctuations between the active voice and the medio-passive -r form,³ that acts as a syntactico-semantic detransitivizer, turning a transitive causative verb into an intransitive one, marking the affectedness/lack of control of the subject over the verbal process, as shown in (1)–(2): (i)

(ii)

(iii)

fortunatum Nicobolum lucky.ACC Nicobolus.ACC ‘How lucky is Nicobulus.’ sed eccum Amphitrionem, but here.he.ACC Amphitruo.ACC ‘But here comes Amphitruo.’

(Plaut. Bacch. 455) advenit come.PRS.IND.3SG

detegetur corium uncover.MPASS.PRS.IND.3SG skin.N/ACC ‘The skin is uncovered’ Glossing of the examples follows the Leipzig Glossing Rules (http://www.eva.mpg.de/lingua/files/morpheme.html)

(Plaut. Amph. 1005)

(Plaut. Epid. 65)

² The data analysed consist of literary and non-literary texts (including the inscriptions) from the earliest attestations to Late Latin, following the conventional periodization of Latin: Early/PreClassical Latin (250–81 BCE), Classical Latin (81 BCE–14 CE), Post-Classical/Imperial Latin (14–180 CE), Late Latin (180–600 CE), Medieval Latin (end of 500 CE–700 CE) (Feltenius 1977, Cuzzolin and Haverling 2009, Gianollo 2014: 949, note 3, Pinkster 2015: 5–6, Vincent 2016 for a more recent discussion in relation to the boundaries between Latin and Romance). ³ The original function of the -r suffix, either an impersonal (Ernout 1908–1909, Lindsay 1895: § 21) or a medio-passive suffix (Bassols de Climent 1948: § 5, Leumann, Hofmann and Szantyr 1965: § 162, note a), is controversial (Kurzová 1993: 157–171). In Latin it may be regarded as the marker of the non-active voice (Cennamo 1998: 78), employed for different types of intransitive structures, including middles, anticausatives, passives, and impersonals. This form only occurs in the tenses of the so-called infectum (i.e. in imperfective tenses, present, imperfect, future). In the tenses of the perfectum (i.e. in forms expressing perfective aspect, perfect, pluperfect, future perfect), a syntactic construction is employed, consisting of a form of the verb sum ‘to be’ + the past participle of the lexical verb (see Pinkster 1988: 220ff., 2015: 230–242, Cennamo 2005: 178–179, Gianollo 2014: 949–951).

190

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

(1)

a. quaeso ne me e somno beg.PRS.IND.1SG NEG 1SG.ACC from sleep.ABL excitetis wake up.PRS.SUBJ.2PL ‘I beg you not to wake me up.’ (Plaut., Merc.1, 2, 48) b. excitor illo (sc. dolore) wake.up.PRS.IND.MPASS.1SG that.ABL pain.ABL ‘I wake up owing to that (sc. pain).’ (Ovid., Ep. (vel Her.) 10, 33)

(2)

a. vipereas rumpo verbis et carmine viperous.ACC.PL.F break.PRS.IND.1SG word.ABL.PL CONJ verse.ABL.SG fauces gullet.ACC.PL ‘I break the viperous gullets with verses and poems.’ (Ovid., Met. 7, 203) b. rumpor et ora mihi break.PRS.IND.MPASS.1SG CONJ face.N.PL 1SG.DAT cum mente tumescunt with mind.ABL swell.up.PRS.IND.3PL ‘I burst and my face and mind swell up with anger.’ (Ovid., Epist.(vel Her.) 8, 57)

The notion of control also appears to be involved in the alternation between the personal vs ‘impersonal’ (active) encoding of (in)transitive situations in Latin. For instance, with some activity (e.g. iuvo ‘to help’, fallo ‘to deceive’, delecto ‘to amuse’), stative (namely, albeit not exclusively, experiencer verbs) (e.g. pudeo ‘to shame’, doleo ‘to grieve’) and, marginally, change-of-state verbs (e.g. contingo ‘to reach, attain, befall’), the use of the default third-person singular active impersonal form appears to denote the taking place of an eventuality, its spontaneous manifestation. The pattern optionally involves a (non-agentive) participant, encoded as an accusative and/or an oblique argument (e.g. dative, ablative, according to the verb and its valency), as shown in (3a) for the verb delecto ‘to amuse’ and further discussed in Sections 6.3.1.3–6.3.1.4. Thus, with the activity verb delecto ‘to amuse’ the ‘impersonal’ active with an accusative experiencer argument, me ‘I’ in (3a), can be contrasted with a corresponding active transitive structure with an A argument,⁴ the stimulus, in the nominative (the subject ista … fama in (3b)), as well as with the reflexive (3c) (nos delectabimus) and medio-passive forms (3d) (delectamur), where the different voice patterns (reflexive vs the R-ending) reflect a difference ⁴ S, A, O/P are syntactico-semantic categories, referring to the clause nuclear arguments, following a well-established terminology (Dixon 1979, 1994, Comrie 1989 (who has P for O), Bickel and Nichols 2009, Bickel 2011). S is the sole argument of an intransitive verb/predicate. A and O/P are mnemonic for the Agent and Object/Patient arguments of a transitive verb/predicate. They may correlate highly with the semantic roles of Agent and Patient and coincide with the grammatical categories of Subject and Object in the languages/constructions where these relations obtain. They correspond to the semantic macroroles of Actor and Undergoer (Van Valin and LaPolla 1997, Van Valin 2005), which are rooted, however, in the lexico-aspectual characteristics of verbs/predicates (see Mithun and Chafe 1999 and Haspelmath 2011 for a critical discussion of these notions and their applicability).

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

191

in control, marking a volitional (3c) and non-volitional action (3d), respectively (Cennamo 1998: 83–88, Cennamo, Barðdal, and Eythórsson 2015: 699–700): (3)

a. me magis de Dionysio delectat 1SG.ACC more about Dionysus.ABL delight.PRS.IND.3SG ‘I prefer Dionysus (lit. (it) delights me more about Dionysus).’ (Cic., Q. Fr. II, 13) b. non tam ista sapientiae … fama me NEG so DEM.NOM 1SG.ACC wisdom.GEN fame.NOM delectate… delight.PRS.IND.3SG ‘I am not so much delighted with that reputation for wisdom …’ (Cic., Amic., 15, 11) c. interea … nos delectabimus in.the.meantime 1PL.ACC delight.FUT.IND.1PL ‘In the meantime we shall organize our own pleasure’. (Cic. Att. II, 4.2) (Cennamo 1998: 84) d. et enim si delectamur cum scribimus and indeed if delight.PRS.IND.MP.1PL when write.PRS.IND.1PL ‘Indeed, if we enjoy writing.’ (Cic., Fin., I, 3)

Lack of control of the A/S argument over the verbal process can be marked not only by the R-ending, as illustrated in (1b) (excitor ‘I wake up’), (2b) (rumpor ‘I burst’), and (3d) (delectamur ‘we enjoy’), but also through the default thirdperson singular active voice in so-called impersonal function, as in (3a), where me delectat literally means ‘it delights me, I happen to be delighted’, the pattern underlining the involitionality, ‘inertness’ of the verbal argument and the ‘happenstance’ nature of the verb eventuality (Barðdal 2004, 2008, 2014, Cennamo, Barðdal and Eythórsson 2015: 700) (Sections 6.3.1.2, 6.3.1.3). In the following discussion, we investigate (i) the verbs entering this construction and their lexico-aspectual characteristics,⁵ (ii) the different coding of the optional ⁵ In our discussion we follow the Vendler (1967)/Dowty (1979) four-way classification of the inherent temporal properties of verbs, subsequently refined within different frameworks (see Levin and Rappaport Hovav 2005, Van Valin 2005, Rappaport Hovav 2008, 2014, Ramchand 2008, 2019, Beavers and Koontz-Garboden 2020, and related references): states (non-dynamic, durative eventualities) (know, stay), activities (dynamic, durative eventualities lacking a final point/result state (i.e. atelic) (sing, walk, work), semelfactives (instantaneous eventualities lacking a final point/result state (i.e. punctual activities) (cough, jump), achievements (dynamic, instantaneous eventualities inherently encoding a final point/result state) (i.e. punctual) (break, explode), accomplishments (dynamic, durative eventualities lexicalizing a final point/result state) (i.e. telic) (change, sink, appear), degree achievements (Hay et al. 1999 / gradual completion verbs (cool, grow) (Bertinetto and Squartini 1995) (dynamic, durative eventualities denoting the gradual approach to a final point along a scale, which may or may not be attained and which can be ‘the final goal or a further stage’ (Bertinetto and Squartini 1995: 13) (i.e. verbs of variable telicity). The last class of verbs instantiates so-called non-quantiszed change, since the final point they lexically entail is non-specific, unlike achievements and accomplishments, which realise quantiszed change, as they lexicalize a specific final state (Beavers and Koontz-Garboden 2017: 855). A verb’s meaning consists of two components, a structural aspect (i.e. its event structure template) and a root (i.e. its idiosyncratic aspect), which differentiates it from other verbs of the same aspectual class, i.e. sharing the same event structure template (Levin and Rappaport Hovav 2005).

192

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

argument(s) occurring in this pattern, and (iii) its ‘impersonal’ status vis-à-vis the other impersonal strategies in Latin.

6.3 Impersonals, detransitivization, and control in Latin The third-person singular of the active and passive voice are the two impersonal strategies appearing in various types of subjectless constructions in Latin, optionally figuring with a non-nominative argument, depending on the construction (Cuzzolin and Napoli 2010, Pinkster 2015). Passive and Impersonal are marked systems of correlations among (morpho)syntactic, semantic, and pragmatic features, realizing different points along a Detransitivization continuum, illustrated in Table 6.1 (Givòn 1984, Shibatani 1985, also Cennamo 1997, 2005, 2010, 2016 for Latin and Romance). Passives, O-oriented patterns with defocusing of the A argument, optionally surfacing as an adjunct (oblique/prepositional phrase) (Shibatani 1985, 1994, Givón 1984: 565–572, Siewierska 2008), share with impersonals the pragmatic notion of agent defocusing, and differ in the extent to which the agent (either S or A, according to the syntactic valency of the verb) is either lacking or implied and syntactically expressed (see also Cennamo 1997, 2003: 57–58, 2011 and discussion in Malchukov and Ogawa 2011). Therefore, Impersonality can be conceived of as a cline, whereby one goes from a logically implied (but unexpressed) argument (A/S/O) to a situation where the eventuality described by the verb is seen as taking place by itself, with no underlying argument (Cennamo 1993, 1997, 2003, and 2016 for Romance). The extent to which the underlying argument (when there is one) (either A, S, or O) is either understood or syntactically expressed, varies within languages (see Malchukov Table 6.1 The passive–impersonal continuum Agent-defocusing (e.g. Agent suppression) > Stativization (Perfective–resultative perspective on the event/ Marked verbal morphology) > Subjectization of a non-Agent (Patient/Benefactive/ Recipient, an original DO/IO) > Topicalization of a non-Agent > Affectedness of surface subject Source: Cennamo (1997: 145, 2006: 313, 2016: 967). A recent revision of the distinction is put forward by Beavers and Koontz-Garboden (2020: 227–234), who propose the presence of templatic entailments (e.g. notions such as causation, change, possession, co-location) also in the root component of some verbs. As we shall see (Section 6.3.1.4), this proposal leads to interesting results for the analysis and interpretation of verbs acquiring a different meaning in the involitionality/impersonal alternation in Latin, an issue that however we only tangentially address in the present chapter and that we leave for further investigation.

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

193

Table 6.2 Synthetic and analytic passives SYNTHETIC (infectum)

ANALYTIC (perfectum)

laudatur (present) ‘he is (being) praised’

laudatus est (perfect) ‘he was praised’/ ‘he has been praised’/ ‘he is (a) praised (man)’ (i.e. extolled, praiseworthy) (adj.) laudatus erat (pluperfect) ‘he had been praised/he was praised’ (adj.)

laudabatur (imperfect) ‘he was (being) praised’

and Ogawa 2011 for an overview of the different functional varieties of impersonal constructions across languages). In Latin the voice strategies employed for the passive–impersonal continuum reflect the morphological aspectual cleavage in the verbal system between forms expressing imperfective aspect (roughly an ongoing, continuous, repetitive action) and forms conveying perfective aspect (roughly a completed action) (Cennamo 2005: 178–179, 2020: 111, Pinkster 2015: 230–257). The former are synthetic (realized by an inflexional ending added to the verb stem, the -r suffix), the latter are analytic (instantiated by a syntactic construction, a form of sum ‘to be’ + past participle), as illustrated in Table 6.2 for the first conjugation verb laudo ‘to praise’. Thus, passives are instantiated by the -r suffix in imperfective tenses (4a) (impediretur), and by a form of the verb sum ‘to be’ + the past participle of the lexical verb in perfective tenses (4b) (dies datus … est). The agent is optionally expressed, surfacing in the ablative if [–animate] (4a) (his rebus) and by means of a prepositional phrase introduced by the prepositions a, ab + ablative (more rarely per + accusative) if [+ animate], as shown in (4b) (ab dis) and (4c) (ab hostibus) (Cennamo 1998: 80, Pinkster 2015: 245–250). (4)

a. his rebus cum iter … impediretur DEM.PL.ABL thing.PL.ABL if path block.MP.IPF.SBJ.3SG ‘If the way is blocked by these things.’ (Caes. Gall. 2, 17, 5) b. optatus hic mi ⁄ dies desired.M.SG.NOM DEM.M.SG.NOM 1SG.DAT day datus hodie est ab dis give.PRF.PTCP.M.SG.NOM today be.PRS.IND.3SG by gods.PL.ABL ‘This day of my desire has been given to me today by the gods …’ (Plaut. Per. 773b–4) (Pinkster 2015: 245) c. ab hostibus conspiciebantur by enemy.PL.ABL recognize.MP.IPF.IND.3PL ‘They were recognized by the enemies.’ (Caes. Gall. 2, 26, 3)

As for the impersonal pole of the passive–impersonal continuum, in the imperfective aspect, i.e. in the tenses of the infectum (present, imperfect, future), there occurs a synthetic form, the unmarked third singular of the -r form (amatur, itur)

194

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

(5a–b). Some experiencer verbs exhibit the third singular of the active inflection (e.g. pudet ‘it fills with shame/one feels ashamed’) (5c), in alternation with the -r form with other experiencer verbs (e.g. miseret/miseretur ‘it feels pity for’) (see Flobert 1975: 499, Fedriani 2014: 154–158, Pinkster 2015: 132–135 for the different chronology of attestation of the active impersonal form with these verbs): (5)

a. amatur love.PRS.IND.MP.3SG ‘One loves, we/you/I love.’ b. itur go.PRS.IND.MP.3SG ‘One goes; we/you/I go.’ c. pudet feel.shame.PRS.IND.3SG ‘There is shame/One feels ashamed (lit. (it) shames).’

The same variation obtains with these verbs in the perfective aspect, i.e. in the tenses of the perfectum (perfect, pluperfect, future perfect), where there occur either the default third-person singular active, as in (6a) for the verb pudeo ‘to be ashamed, to feel shame’, or a form of the verb sum ‘to be’ in the third-person singular + the (neuter singular) past participle of the lexical verb, as in (6b–d), or of the gerundive (formed from the stem of the infectum with the -nd-infix and the inflectional ending of first and second declension adjectives) (with a deontic value) (e.g. amandus ‘to be loved’) (Pinkster 2015: 62–63), in the neuter singular form, as in (7) (Woodcock 1959: 167, Flobert 1975: 499–500, Pinkster 1992, 2015: 290–300, Cennamo 2005, 2020): (6)

a. (me) puduit 1SG.ACC feel.shame.PRF.IND.3SG ‘I was ashamed; there was shame (on me).’ b. puditum est feel.shame.PRF.PTCP.N.SG be.PRS.IND.3SG ‘One was ashamed.’ c. amatum est love.PRF.PTCP.N.SG be.PRS.IND.3SG ‘One has loved; we/you/I have loved (indef.).’ d. itum est go.PRF.PTCP.N.SG be.PRS.IND.3SG ‘One ran; running took place.’

(7)

amandum est love.GER.N.SG be.PRS.IND.3SG ‘One has to love; loving is to take place’

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

195

On the basis of their characteristics and function (e.g. whether (i) lacking a participant, (ii) implying an S/A argument, or (iii) instantiating an existential– presentative pattern), the strategies employed (the active and/or passive verbal morphology) and the marking of arguments with monovalent/divalent verbs (e.g. accusative and/or dative, ablative), a number of impersonal constructions can be identified in Latin (some of which instantiate different diachronic stages), summarized in Section 6.3.1 and illustrated in Sections 6.3.1–6.3.3.

6.3.1 Active impersonal verbs and constructions The third-person singular active inflection is a common agent-defocusing strategy, already attested in Early Latin, used to denote the existence and taking place of an eventuality, discussed under the heading ‘impersonal’ verbs in traditional reference grammars and analyses (Lo¨fstedt 1936, Woodcock 1959: 166–171, Leumann, Hofmann and Szantyr 1965: §221, Ronconi 1968: 14, Bauer 2000: 103–120 and the recent discussion in Pinkster 2015: 94–97, 192–195, 750–552). Since its earliest occurrences this pattern is found with different verb classes, of different argument structures, and in different constructions, including weather verbs (Section 6.3.1.1), impersonal and existential uses of BE and HAVE (Section 6.3.1.2), impersonal forms of divalent verbs (e.g. fixed forms such as inquit ‘it is said’), modal verbs (e.g. oportet ‘it is proper, necessary’, potest ‘it is possible’, licet ‘it is possible’, libet ‘it pleases’), other impersonal uses of verbs attested in Late Latin (e.g. horret ‘it is dreadful’, valet ‘it is possible’, dicit ‘it is said’) as well as a number of experiential verbs (e.g. pudet ‘it fills with shame’, miseret ‘it moves to pity for’, paenitet ‘it causes regret’, placet ‘it pleases’, delectat ‘it delights’) (Section 6.3.1.3). The third-person singular active form is also characteristic of existential–presentative constructions in Late Latin, most typically with an accusative S/O argument (Section 6.3.3).

6.3.1.1 Weather and natural conditions verbs Verbs denoting atmospheric and natural events and conditions most typically occur in the default third-person singular active (8), sometimes with an overt dummy subject, the demonstrative neuter pronoun hoc ‘this’, depending on the verb, as in (8b) (that could exemplify, however, the adverb hoc ‘here’) (Leumann, Hofmann and Szantyr 1965: §§ 45c, 220, Neue-Wagener 1985: 647–650, Pinkster 2015: 193–195): (8) a. ut multum pluerat as a.lot rain.PLPF.IND.3SG ‘A s it rained heavily.’ b. hoc… lucebit DEM.N dawn.FUT.3SG ‘It will become light.’

(Plaut., Men. Prol., 63)

(Plaut., Curc., 1, 3, 26)

196

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

With attestations already from Early Latin, the default third-person singular may also be found in monovalent patterns with the ablative (9a) or, more rarely, the accusative of the substance participant (9b). The two case forms may alternate within one and the same author (e.g. Livy, Classical age) (9c–d), also in the same type of construction (e.g. the accusative + infinitive) as shown for the verb pluo ‘to rain’, as for lapidibus pluisse in (9e) vs lapides pluere in (9f ) (Leumann, Hofmann and Szantyr 1965: §§ 45c, 220, Pinkster 2015: 193–194, Dahl 2020: 132–133): (9) a. nivit sagittis snow.PRS.IND.3SG shaft.of.lightning.ABL.PL ‘It is lightning (lit. (It) snows with lightning).’ (Pacuv., Praetext. 4) b. … ninxerit caelestium molem mihi snow.FUT.PERF.3SG heavenly.GEN.PL weight.ACC 1SG.DAT ‘The sky will fall upon me like snow (lit. (It) will snow the heavenly weight to me).’ (Acc., Trag., 101) c. quod sanguine per biduum pluisset in area that blood.ABL for two.days rain.SBJV.PLPF.3SG in area.ABL Vulcani Vulcan.GEN ‘That for two days it rained blood in the area of Vulcan.’ (Liv., AbUrbe Cond. 39. 46) d. In area Volcani et in area.ABL Volcan.GEN and Concordiae sanguinem pluit Concordia.GEN blood.ACC rain.PRS.IND.3SG ‘In the area of Volcan and Concordia there rained blood.’ (Liv. Ab Urbe Cond. 40, 19, 2) (sanguine in Briscoe 2007: 459, Pinkster 2015: 194) e. nuntiatum regi patribusque est in report.N.PRF.PTCP king.DAT senator.PL.DAT-and be.PRS.IND.3SG in monte Albano lapidibus pluisse mountain.ABL Alban stone.ABL.PL rain.PRF.INF ‘It was reported to the king and the senators that it had rained (with) stones in the Alban hill.’ (Liv. Ab Urbe Cond. 1.31.1) f. lapides pluere … vos portenta esse stone. ACC.PL. rain.PRS.INF you.NOM omens.N.PL be.PRS.INF putatis think.PRS.IND.3PL ‘You take as omens that it rains (with) stones.’ (Liv. Ab Urbe Cond.28.27.16) (Dahl 2020: 132)

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

197

The different case marking for the substance participant is viewed as reflecting its argument (if in the accusative) vs adjunct (if in the ablative) status (‘satellite’ in Pinkster 2015: 194, Dahl 2020). For some verbs (e.g. ningo ‘to snow’) the ablative and the accusative are found already in Early Latin (9a– b). For other verbs (e.g. pluo ‘to rain’), the two case forms are well attested in Classical Latin, as shown in (9d–f ), with the accusative for the substance participant being less common than the ablative (Pinkster 2015: 193, Dahl 2020: 133). In Classical authors weather and natural condition verbs are also found in monovalent patterns with the source of the eventuality as subject (instantiated by a [± animate] Nominal) (10) (see Neue-Wagener 1985: 647–650, Pinkster 2015: 193–195, Dahl 2020, and further examples therein): (10) a. dies illuxisset day.NOM dawn.SBJV.PLPF.3SG ‘It had dawned.’ (Cic., Diu., I, 50) b. caelum tonat sky.NOM thunder.PRS.IND.3SG ‘The sky is thundering.’ (Verg., Aen., 9, 541) c. tonans Juppiter thunder.PRS.PTCP.NOM Jupiter:NOM ‘Jupiter thundering.’ (Horat. Carm. 3, 5, 1) (Neue-Wagener 1985: 648) Examples of the substance participant as subject are attested at a later stage, as shown in (11a–b) (from the first century CE), often in a non-literal meaning of the verb (e.g. FALL for pluo ‘to rain’ in (11) (Pinkster 2015: 193): (11) a. pluit … nimbus … teretis mali … rain.PRS.IND.3SG cloud.M.NOM round.N.GEN.SG apple.N.GEN ‘A shower of shapely apples rains down’ (Col. 10.364–365) (Pinkster 2015: 193) b. stridentia funda saxa pluunt whistle.PRS.PTCP.F.SG.ABL sling.F.SG.ABL stone.N.PL rain.PRS.IND.3PL ‘From a whistling sling stones are raining.’ (Stat. Pap. Thebais 8.416–417) (1st cent. CE) A later development is also the divalent, causative use of some of these verbs (e.g. pluo ‘to rain’), with the NOM–ACC case frame, first attested in Christian Latin and in Bible translations (Pinkster 2015: 194–195) as shown in (12a), a word-by-word translation of the Hebrew original (Dahl 2020: 133) and in (12b), with an unexpressed object and the modal verb possum ‘can’ (example from Pinkster 2015: 195):

198

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

(12) a. Igitur Dominus pluit super Sodomam therefore Lord.NOM rain.PRS.IND.3SG over Sodoma et Gomorram sulphur et ignem a Domino and Gomorra sulphur.N and fire.ACC by Lord de caelo from heaven.ABL ‘The Lord let it rain sulphur and fire over Sodoma and Gomorra, from the Lord out of heaven.’ (Vulg., Gen., 19, 24) b. Agricola … pluere vinae suae non farmer rain.PRS.INF vineyard.DAT own.DAT not potest can.PRS.IND.3SG ‘A farmer … He has no power to send rain on his vineyard (lit. he cannot make rain on his vineyard.’ (August. Psal. 66.1) (Pinkster 2015: 195) As for the lexico-aspectual properties of the atmospheric and natural condition verbs found in the active impersonal construction in Latin, they include most typically verbs of emission, that can be viewed as intermediate between states and activities (Levin and Rappaport Hovav 1995: 91, 169, 138–142, 237–238 and discussion of weather verbs in Levin and Krejci 2019, Eriksen, Kittila¨, and Kolehmainen 2015 for a typological overview). They range from the more stativelike light emission (e.g. luceo ‘to be light’) (13a), to the more processual-like substance (e.g. pluo ‘to rain’, roro ‘to drop, distil dew, bedew, moisten, wet’) (13b), and sound (e.g. tono ‘to thunder’) (13c), bucino ‘to sound on a trumpet’) (13d) emission. Also other activity and change-of-state verbs may be used to denote atmospheric/natural conditions in this pattern: adflo ‘to blow’ (> adflat ‘it is blowing’) (13b), lapido ‘to throw stones’ (> lapidat ‘it is raining stones’) (13f ), vesperasco ‘to get dark’ (> vesperascit ‘it is getting dark’) (12e), sicco ‘to dry’ (> siccat ‘it is dry’) (12f ), dissereno ‘to become clear’ (> disserenat ‘it becomes clear’) (13h) (examples from Pinkster 2015: 193, see Neue-Wagner 1985: 647–650, see Pinkster 2015: 192–195 for a detailed analysis and further examples and lists of verbs occurring in this pattern as well as Section 6.3.1.4). (13) a. priusquam lucet, adsunt before be.light.PRS.IND.3SG be.present.PRS.IND.3PL ‘They are present before it is light.’ (Pl. Mil. 3, 1, 115) b. si roravit quantulumcumque imbris if bedew.PRF.3SG how.small how little heavy.rain.GEN.M aut si adflavit or if blow.PRF.3SG ‘If the smallest sprinkle of rain has fallen or if it has been blowing’ (Plin. Nat. 17, 74)

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

199

c. ut valide tonuit! how heavily thunder.PRF.3SG ‘How heavily it thundered!’ (Pl. Amph. 5, 1, 10) d. Saepe declamante illo ter often speak.PRS.PTCP.ABL he.ABL three-times bucinavit trumpet.PRF.3SG ‘Often while he was speaking the trumpet would blow three times.’ (Sen. Con. 7pr.1) (Pinkster 2015: 195) e. Et vesperascit et non and get.dark.PRS.IND.3SG and not noverunt viam know.PRF.IND.3PL way.ACC ‘And it is getting dark and they do not know the way.’ (Ter. Hau. 248) f. Reate imbri lapidat Reate.ABL rainstorm.ABL throw.stones.PRS.IND.3SG ‘At Reate there was a rainstorm of stones.’ (Liv. Ab Urbe Cond. 43, 13, 4) g. ubi pluerit et siccaverit where rain.FUT.PRF.3SG CONJ dry.FUT.PRF.3SG ‘Where it will rain and it will dry up.’ (Cato, Agr. 112, 2) h. … hiemabit … disserenabit be.winter.weather.FUT.PRS.3SG clear.up.FUT.PRS.3SG ‘It will be winterish and it will clear up.’ (Plin. Nat. 18, 356) The S argument, when present at argument structure. is either thematically underspecified or non-agentive. Alongside avalent patterns, also monovalent and even divalent ones can be found (see Late Latin transitive pluo ‘to rain’, lapido ‘to throw stones’ (attested from the Augustan age) (Lewis and Short 1942, s.v. lapido) (e.g. transitive sicco ‘to dry up’, dissereno ‘to clear up’), their degree of syntactic elasticity varying according to the verbs, their argument structure and uses in different authors and genres (e.g. poetry vs technical works or legal texts) as well as the semantic, lexico-aspectual, and syntactic changes these verbs undergo in the course of time.

6.3.1.2 Active impersonal constructions with BE and HAVE The third-person singular active is also found with the verb sum ‘to be’, in existential patterns with infinitives and subordinate clauses (the accusative and infinitive or finite clauses introduced by the conjunction ut ‘that, so that’ + the subjunctive), occurring also with a modal meaning, ‘it is possible’, ‘it is allowed’ (14a–b). This construction is rare in Early Latin and becomes more common in (post-) Augustan poetry and Christian writers, probably owing to the influence of an analogous Greek pattern (example (14a) and discussion from Pinkster 2015: 95).

200

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

(14) a. Scire est liberum ingenium understand.PRS.INF be.PRS.IND.3SG children.GEN nature.N.SG atque animum CONJ mind.M.ACC ‘It is obvious that their natures and inclinations are fundamentally honourable …’ (Ter. Ad. 828–829) b. non erat ut fieri posset NEG can.IMPF.IND.3SG that happen.PRS.INF can.IMPF.SBJV.3SG ‘It could not happen.’ (Lucr. De Rerum Natura 5, 979) There also occur impersonal patterns consisting of an adjective in the unmarked neuter singular form or an adverb + the third person singular of the verb sum ‘to be’: manifestum est ‘it is clear’, recte est ‘it is right’, bene est ‘it is good’, sero est ‘it is late’, etc. (Lindsay 1907: 52–53, Bassols de Climent 1948: 94, Ronconi 1968: 13, Pinkster 2015: 96). Analogous adverbial existential patterns are found in Classical Latin as well as in Late Latin, with the third-person singular active of the verb habeo ‘to have’ (15a–b) (Leumann, Hofmann, and Szantyr 1965: §221, c, Pinkster 2015: 97): (15) a. Bene habet. Iacta sunt good have.PRS.IND.3SG lay.PRF.PTCP.F.PL be.PRS.IND.3PL fundamenta defensionis foundation.PL.N defense.GEN ‘That’s good. The foundations of his defence have been laid.’ (Cic. Mur. 14) b. Ostria vero necesse habet … permittere oysters certainly necessary have.PRS.IND.3SG allow.PRS.INF interdum occasionally ‘Occasionally it is necessary … to allow oysters.’ (Anthim. 49) In Late Latin the third-person singular active of the verb HAVE is frequently found in existential constructions, with a non-agreeing nominal (16), a pattern also occurring with other verbs (Svennung 1935: 475–477, 572–573, Leumann, Hofmann, and Szantyr 1965: §221, c, Cennamo 2011: 177–179, Pinkster 2015: 97 and Section 6.3.3): (16) …habet in biblioteca Ulpia have.PRS.IND.3SG in library.ABL of.Ulpian.ABL in armario sexto librum elephantinum … in chest.ABL sixth.ABL book.ACC.SG consisting.of.ivory.tablets.ACC.SG ‘There is a book consisting of ivory tablets in the sixth chest in Ulpian’s library.’ (Hist. Aug. Tac. 8, 1)

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

201

6.3.1.3 Active impersonals of (in)transitive verbs and control The third-person singular of the active voice in impersonal function, to denote the taking place of an event, is also found with non-meteorological, (in)transitive (i.e. divalent/monovalent) verbs, already in ancient legal texts (e.g. the XII Tables) (17a), and other Early Latin texts (17b–c) (Lindsay 1907: 52–53, Bassols de Climent 1948: 94, Ronconi 1968: 13). Indeed, the third-person singular active in impersonal function, to mark a generic, indefinite human participant, is also attested with an accusative argument in some early authors (e.g. Caecilius, Cato, Varro), as shown in (17b–c) (Woodcock 1959, Ronconi 1968: 13, Rosén 1970, Cennamo 2009): (17) a. si in ius vocat if in law:ACC call.PRS.IND.3SG ‘If one (the person so entitled or authorized) calls (a person) to court.’ (Leg. XII Tab.I,1) (Rosén 1992: 388) b. multa quae non volt, videt many: ACC.PL REL.ACC.PL NEG want.PRS.IND.3SG see.PRS.IND.3SG ‘One sees a lot of things/several things that one would not like to see.’ (Caec. 175 Ribb.) (Ronconi 1968: 14) c. selibram tritici… indat, half.a.pound.ACC wheat.GEN take.PRS.SBJV.3SG bene lavet well wash.PRS.SBJV.3SG ‘One (the farmer) should take half a pound of wheat, one (he) should wash it well.’ (Cato, Agr., 86) However, patterns such as in ius vocat in (17a) and selibram indat in (17c) may also exemplify, instead, the omission of a third-person subject/agent (the plaintiff in (16a), the farmer in (17c), that is implied and contextually recoverable (Bassols de Climent 1948: 94, Bauer 2000: 107, Pinkster 2015: 750–751). In point of fact, in Latin the third-person singular active (alongside the thirdperson plural) is commonly employed for a participant that has not been mentioned in discourse, whose identity is recoverable either from the preceding context or from ‘general knowledge’ (Meillet and Vendryes 1924: 306, Herman 1991: 416–425, Bauer 2000: 107, Pinkster 2015: 750–754, also Section 6.3.1.2). This characteristic, related to the pro-drop nature of the language, decreases in Late Latin (Herman 1991: 416–418), when ‘zero subjects’ are replaced by ‘pronominal (chiefly demonstrative) elements’ (Herman 1991: 417, Pinkster 2015: 750–752 for discussion). In technical texts (e.g. legal documents, medical, veterinary, and agricultural treatises), the subject is generally unexpressed, already in Early Latin (e.g. the XII Tables, and Cato’s De Agricultura, Varro’s De Re Rustica), in patterns which may be ambiguous between a generic, referential indefinite ‘impersonal’ interpretation and a referential definite one. The unexpressed agent/subject may refer to the participant featuring in particular genres (e.g. the farmer, the patient, an animal), as also illustrated in (16a, c) (Pinkster 2015: 750–751), or its identity

202

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

is recoverable from the wider context. At times it is indeed difficult to detect the function of the third-person singular active form, and generally to differentiate the generic indefinite use of the pattern from its reference to an unexpressed, contextually recoverable subject/agent (Pinkster 2015: 750–751 for a recent and nuanced discussion). For instance, in (18) the verb forms est, sentit, sapit, appear to have a clear generic indefinite reference (‘someone’), denoting a participant whose identity is unknown, unnecessary to define, and who does not comprise either speaker or hearer (i.e. the Speech Act Participants (SAPs)): (18) Senex quom extemplo est, iam nec old.man.NOM when immediately be.PRS.IND.3SG now neither sentit nec sapit, / aiunt hear.PRS.IND.3SG neither know.PRS.IND.3SG say.PRS.IND.3PL solere eum rursum repuascere be-accustomed 3SG.ACC back renew.childhood.PRS.INF ‘A s soon as someone is an old man and no longer has his senses or wits about him, they say that he enters his second childhood.’ (Pl. Mer., 295–296) (Pinkster 2015: 751) As pointed out by Pinkster (2015: 750), in Classical Latin the indefinite pronoun aliquis ‘someone’ would be used as the subject of the clause introduced by the conjunction quom ‘when’ in (18), unlike the subjectless third-personal active forms of the verbs. By contrast, in Late Latin texts the generic indefinite reference of the thirdperson singular active is more clearly identifiable and accompanied also by a different syntax of the pattern, as for impersonal debet (= decet, oportet) ‘it is necessary’, attested in early Latin (e.g. Varro) without a complement (19a), occurring in Late Latin texts such as the Mulomedicina Chironis (end of the fourth century CE) with an active infinitive and an O argument, as in (19b) (Lo¨fstedt 1936: 136–137, Bauer 2000: 122 for a discussion of the issue and further examples): (19) a. ut debuit… as must.PRF.IND.3SG ‘A s it ought to be.’

(Varro, L.L. 10, 1, 1)

b. sanguinem emittere … de capite debet blood.ACC let.INF from head.ABL must.PRS.IND.3SG ‘One ought to let blood from its head.’ (Mul. Chir. 33) Further evidence for the early occurrence of the third-person singular active as an agent-defocusing strategy comes from fixed impersonal forms of divalent verbs such as inquit (< inquam ‘to say’) (and more rarely ait (< aio ‘to assent, affirm’) ‘it is said, someone says/objects’, used when quoting a saying, the opinion of an ‘imagery opponent’ (Ronconi 1968: 14, Pinkster 2015: 753–754). The vitality of this strategy already at an early stage, is witnessed by the occurrence of several divalent (NOM–ACC/DAT case frame) and monovalent verbs, also in the

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

203

third-person singular active impersonal form, optionally accompanied by either an ACC or a DAT argument (patient–theme/experiencer), according to the syntactic valency of the verb, and an infinitive or accusative and infinitive clause as subject, as in (20a–b) (e.g. decet (me-ACC/mihi-DAT) ‘it becomes, it befits (me)’, delectat (me-ACC) ‘it delights (me), placet (mihi-DAT) ‘it pleases (me)’ (Woodcock 1959: 168, Leumann-Hofmann and Szantyr 1965: § 60, 221, and Table 6.1): (20) a. oratorem irasci minime decet speaker.ACC lose.his.temper.INF not.at.all befit.PRS.IND.3SG ‘It is not at all fitting for a speaker to lose his temper.’ (Cic., Tusc., 4, 25) b. quam delectabat eum defections solis how delight.IPF.IND.3SG 3SG.ACC eclipse.ACC.PL. sun.GEN praedicere foretell.INF.PRS ‘How it delighted him to foretell eclipses of the sun.’ (Cic., Sen., 49) With some verbs, e.g. deceo ‘to suit, to become, to befit’, the verbal argument may alternate between the accusative and dative cases, in a seemingly free alternation, with no detectable difference, as shown in (21) (Bennett 1914: 106, 212): (21) a. facis ut te decet make.PRS.IND.2SG as 2SG.ACC become.PRS.IND.3SG ‘Do what becomes you.’ (Ter., Andr., 2, 5, 10) b. ita nobis decet thus 1PL.DAT become.PRS.IND.3SG ‘It becomes us thus.’ (Ter., Ad., 5, 8, 5) Other (in)transitive verbs that are used impersonally in the third-person singular active voice, optionally taking a dative argument and followed by either an infinitive or an accusative and infinitive, as shown in (22) for the modal verb licet ‘it is permitted’, are consto ‘to be in agreement with’ (e.g. constat ‘it is agreed’), praesto ‘to stand out’ (e.g. praestat ‘it is preferable’), appareo ‘to appear’ (e.g. apparet ‘it is apparent’), liqueo ‘to be clear, apparent’ (e.g. liquet ‘it is clear’), modal verbs (e.g. licet ‘it is permitted’, libet ‘it is agreeable, it pleases’, oportet ‘it is proper, necessary’, potest ‘it is possible’) (Woodcock 1959: 170–171, Neue-Wagener 1985: 659–662, Bauer 2000: 95–97, 121–129, and Table 6.2): (22) a. licuit esse otioso Themistocli (infinitive) be.allowed.PRF.3SG be.INF idle.DAT Themistocles.DAT ‘It was allowable for Themistocles to be at leisure.’ (Cic., Tusc., 1, 33) b. te liquet esse meum (accusative and infinitive) 2SG.ACC clear.PRS.IND.3SG be.INF POSS.1SG.ACC ‘It becomes clear that you are mine.’ (Ov., Tr., 1, 1, 62) In Late Latin this pattern is also found with verbs with which it is unattested at earlier stages (e.g. horret ‘it is dreadful’ (< horreo ‘to shudder’), valet ‘it is possible’ (< valeo ‘to be worth’), dicit ‘it is said’ (< dico ‘to say’) (Lo¨fstedt 1911: 44–47, 1936:

204

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

131–142, Svennung 1935: 472–475, 612, Bauer 2000: 108–129, and discussion in Pinkster 2015: 134 for impersonal horret). Also some divalent experiencer/psychological verbs occur in the default thirdperson singular (both in the infectum and in the perfectum) (e.g. piget ‘annoyance is at work’, pudet ‘shame is at work’, paenitet ‘remorse is at work’, taedet ‘weariness comes on’, miseret ‘pity is at work’) with O/S (the Experiencer) in the accusative and the A argument (the Stimulus) optionally expressed in the genitive (Woodcock 1959: 167, Fedriani 2013, 2014, Pinkster 2015: 132–135): (23)

(tui) me miseret/pudet 2SG.GEN 1SG.ACC pity/shame.PRS.IND.3SG ‘I pity you/I am ashamed of you (lit. It pities/shames/ me of you).’

These verbs show alternation between the impersonal and personal pattern, rarely also the -r form for some verbs (e.g. misereo ‘to take pity’) (24d), already in Early Latin (Bennett 1914: 91, Woodcock 1959: 167, Ronconi 1968: 17), although mainly/only with the Stimulus as subject, depending on the verb and realized as a neuter pronoun (see haec ‘these’ in (24a) (Fedriani 2014: 139–140):⁶ (24) a. non te haec pudent (personal) (Stimulus subject) NEG 2SG.ACC DEM.N.PL shame.PRS.IND.3PL ‘You are not ashamed of these things.’ (Ter., Ad., 754, 9) b. tui me pudet (impersonal) 2SG.GEN 1SG.ACC shame.PRS.IND.3SG ‘I am ashamed of you.’ (Plaut., As., 933, 6) c. patris me miseretur (impersonal) father.GEN 1SG.ACC pity.PRS.IND.MP.3SG ‘I take pity on my father.’ (Turp., 55) d. aliquando miseremini sociorum (personal: misereor) some.time pity.IMP.2PL ally.PL.GEN ‘Take pity for sometime on the allies.’ (Cic., Verr., 1, 72) e. ipse sui miseret 3SG.NOM REFL.3SG.GEN pity.PRS.IND.3SG ‘He pities himself.’ (Lucr., De Rerum Natura, 3, 881) A non-agreeing argument is also frequently attested (already in Early Latin) with the impersonal gerundive, as in (25) (Ernout 1908–1909: 297, Ronconi 1968: 200, Pinkster 2015: 290–291): (25) a. (ut) vasa vinearia et olearia faciendum in.order.to container of.wine and of.oil make in.order.to container.N.PL of.wine.N.PL CONJ of.oil.N.PL make.GER.N.SG ‘In order to make containers for wine and oil.’ (Varr., r.r. 1, 13) ⁶ We do not address in our discussion the syntactic status of the non-nominative arguments occurring with the verbs illustrated in (20)–(24) and their pivot/subject-like behaviour, investigated by Fedriani 2009, 2014: 123–24, Dahl 2012, Fabrizio forthcoming.

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

205

b. poenas timendum=st punishment.ACC.PL fear.GER.N.SG=be.PRS.IND.3SG ‘One should fear punishments.’ (Lucr., De Rerum Natura, 1, 111)

6.3.1.4 Active impersonals, non-nominative arguments, and control Several verbs are attested in the pattern illustrated in Sections 6.3.1.2–6.3.1.3, both bivalent (in two case frames, NOM–ACC and NOM–OBL) and monovalent, belonging to all conjugations and to different aspectual classes: states (e.g. attineo ‘to pertain’, deceo ‘to become’, poeniteo ‘to repent’, doleo ‘to grieve’, consto ‘to agree with’, placeo ‘to please’, liceo ‘to be permissible’, resto ‘to remain’, vaco ‘to be free (from labour), leisure’), activities (e.g. iuvo ‘to delight’, lateo ‘to conceal’, fugio ‘to escape’, fallo ‘to deceive’, lapido ‘to throw stones’) and changes of state (e.g. illucesco ‘to throw light upon’, gelo ‘to freeze’, contingo ‘to reach’, accido ‘to happen’, appareo ‘to appear’, venio in mentem ‘to come to one’s mind’), with different attestations in the history of the language. States (namely experiencer verbs) appear to instantiate the core of this type of ‘impersonal’ structures (Tables 6.3–6.5) (Barðdal 2004, Barðdal and Eythórsson 2009). Table 6.3 Personal~impersonal alternation with divalent verbs (NOM–ACC) Verb classes

Divalent verbs. Case frame: NOM–ACC Personal

Activity

iuvo ‘to delight’ lateo ‘to conceal’ delecto ‘to amuse’ fugio ‘to flee, run away, escape’ praetereo ‘to pass by, omit, forget’ fallo ‘to deceive’ lapido ‘to throw stones (at someone)’

State

attineo ‘to hold, concern, pertain’ deceo ‘become’ paeniteo ‘repent, displease’ (caus.) misereo ‘feel pity’ (caus.) pudeo ‘ashame’ (caus.) pigeo ‘to trouble’

Change of state

illucesco ‘to throw light upon’ gelo ‘to freeze’ (caus.)

~

Impersonal me iuvat ‘it is useful, it pleases me’ me latet ‘it escapes me’ me delectat ‘it delights me’ me fugit ‘it escapes me’ me praeterit ‘it escapes me’ me fallit ‘I am wrong/I happen to be wrong’ lapidat (Ø) ‘stones fall (lit. it falls stones)’ me attinet ‘it concerns, pertains to me’ me decet ‘it becomes me’ me paenitet ‘it displeases me’ me miseret ‘it pities me’ me pudet ‘it ashames me’ me piget ‘it irks, disgusts me’ illucescit (Ø) ‘it daybreaks’ gelat (Ø) ‘it freezes’ (Imperial age, Plinius)

206

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

Table 6.4 Personal~impersonal alternation with divalent verbs (NOM-OBL) Verb classes

Divalent verbs. Case-frame: NOM-OBL (DAT/GEN) Personal

~

Impersonal

Activity

expedio ‘to help out, promote’

expedit mihi ‘it is useful, it helps me’

State

doleo ‘grieve’ (caus.) consto ‘to agree with’ placeo ‘to please, like’ liceo ‘to be permissible’

mihi dolet ‘it grieves/pains me’ mihi constat ‘it is certain to me’ mihi placet ‘it pleases me’ mihi licet ‘it is permissible to me’

Change of state

contingo to ‘touch, reach’

mihi contingit ‘it happens to me’

Table 6.5 Personal~impersonal alternation with monovalent verbs (NOM (–DAT)) Verb classes

Monovalent verbs. Case frame: NOM (–DAT) Personal

Activity State

Change of state

~

Impersonal

roro ‘to fall, drop, distil dew’ liqueo ‘to be liquid, clear’

rorat ‘it drizzles, dew falls’

resto ‘to remain’

mihi restat ‘it remains to me’

oporteo ‘to be necessary’

oportet ‘it is necessary’

vaco ‘to be free, have time, leisure’

mihi vacat ‘it lacks me, there is time, leisure’

hiemo ‘to winter, be cold, wintry’ sto ‘to remain’

(Ø) hiemat ‘it is wintry, cold’

accido ‘to fall upon, happen’

mihi accidit ‘it happens to me’

dissereno ‘to clear away’

(Ø) disserenat ‘it clears away, it is clear’

venio in mentem ‘to come to one’s mind’

mihi venit in mentem ‘it comes to my mind’

mihi liquet ‘it is clear to me’

(Ø) stat ‘it is agreed/decided’

The distribution and different interpretations of the various voice forms in which ‘impersonal’ verbs may occur, reveal that patterns such as me pudet ‘it shames me’, me miseret ‘it pities me’, me libet ‘it pleases me’, me fallit ‘it deceives me’, me fugit ‘it escapes me’ instantiate structures reflecting the degree of control

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

207

of the A/S argument and the spontaneous manifestation of the verb’s eventuality, its happenstance nature. Indeed, lack of control/involitionality/spontaneous manifestation of the verb’s eventuality appears to be the unifying parameter for the occurrence of the accusative with so-called impersonal verbs of traditional grammar, accounting for the occurrence of the accusative and/or the dative with impersonal constructions such as mihi libet ‘it pleases me’, me fallit ‘it deceives me’, me fugit ‘it escapes me’, me pudet ‘it shames me’, me miseret ‘it pities me’, as also illustrated in (26)– (27). For example, the impersonal form me fallit in (26d) can be contrasted with the personal passive form fallor in (26c). In both patterns the S/O argument is affected by the verbal process, but in the impersonal encoding the focus is on the taking place of the event itself, as involving an inactive argument, marked in the accusative case, the canonical case for ‘inert’ participants (Collinge 1978). The difference between the two patterns can be provisionally described as being one of control/spontaneous manifestation of the event, whereby me fallit in (25d) would mean ‘I happen to be wrong’ while fallor in (25c) would just denote the affectedness of the S/O argument: ‘I am wrong, I am deceived’ (Ronconi 1968: 16–17, Cennamo 2010, Cennamo et al. 2015: 700): (26) a. fallo ‘to deceive’: active transitive use: ‘deceive somebody’ Tibi videor esse quem tam aperte 2SG.DAT seem.PRS.IND.MP.1SG be.INF REL.ACC.SG so openly fallere incipias dolis deceive.INF begin.PRS.SBJ.2SG fraud.ABL.PL ‘Do I seem to you to be one whom you can begin to deceive so openly with fraud?’ (Ter. Andr. 493) nisi memoria me fallit if.not memory.NOM 1SG.ACC deceive.PRS.IND.3SG ‘If memory does not deceive me.’ (Au.Gel., NA, 20, 1, 14, 3) b. reflexive: me fallo: nisi me forte fallo if.not 1SG.ACC accidentally be.in.error.PRS.IND.1SG ‘If I am not wrong (lit. If I am not accidentally deceiving myself ).’ (Cic., Phil., 12, 21, 8) c. medio-passive -r form: fallor (‘I am deceived (passive), I am mistaken (middle)’) nisi fallor if.not be.in.error.PRS.IND.MPASS.1SG ‘If I am not mistaken.’ (Cic., Att., 4, 19, 1–4)

208

MICHEL A CENNAMO AND CL AUDIA FABRIZIO d. impersonal: me fallit ‘I am wrong (I happen to be mistaken) (lit. me deceives).’ quod me non fefellit as.far.as.this 1SG.ACC NEG be.in.error.PRF.3SG ‘I was not (I did not happen to be) mistaken as far as this is concerned.’ (Cic., Ver., 19, 2, 1, 19, 3–4)

With some activity verbs in the corresponding active impersonal form the verb acquires a different meaning, denoting an atmospheric event, as with roro ‘to drop, dew’, already attested in Early Latin (27a) and lapido ‘to throw stones’ (unattested before the Classical age), that in the corresponding active impersonal form means ‘it rains/falls stones’, occurring with an argument in the ablative in (27b) (see Tables 6.3–6.5): (27) a. ante rorat before drizzle.PRS.IND.3SG ‘It drizzles before raining.’ b. Reate imbri Reate.ABL rain.ABL ‘At Reate it rained heavily.’

quam pluit than rain.PRS.IND.3SG (Varr., De lingua latina, 7, 58) lapidavit throw.stones.PRF.3SG (Liv., Ab Urbe Condita, 43, 13)

This pattern seems to be only marginally attested with change-of-state verbs (e.g. illucesco ‘to throw light upon’, gelo ‘to cause to freeze’, accido ‘to fall upon, happen’, contingo ‘to reach, attain, befall’), at times with an accusative and/or a dative argument (Tables 6.3–6.5), as shown for the verb venire ‘to come’ in conjunction with a prepositional complement introduced by the preposition in ‘in(to)’ + an abstract nominal in the accusative, mentem.ACC ‘mind’ in (28), in the expression venire in mentem ‘to come/fall to one’s mind’, with O in the dative and A in the accusative, as in (28b), the predicate denoting a mental process. The event structure template of the verb venio (accomplishment) remains the same in the new meaning acquired by the predicate, whereas its root loses its idiosyncratic meaning of ‘motion towards a goal’, while retaining its templatic meaning of ‘transition into a state’. (28) venire in mentem ‘come to one’s mind’ a. active intransitive use: istuc mihi venit in mentem DEM.N.SG 1SG.DAT come.PRS.IND.3SG in mind.ACC (Ter., Hau., 888–889) ‘This comes to my mind.’ b. ei venit in mentem hominum 3SG.DAT come.PRF.3SG in mind.ACC man.GEN.PL fortunas (DAT-ACC) fate.ACC.PL ‘Men’s fate came to his mind (lit. to him-DAT came to mind men’s destinies-ACC).’ (Cn. Naev., Pun., 20, 1)

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

209

Not only mental process and experiencer verbs, but also stative verbs denoting a physical state (e.g. resto ‘to stand firm, be left, remain’, sto ‘to stand, remain standing’, vaco ‘to be empty, be void’, lateo ‘to lurk, be concealed’) can occur in the impersonal active pattern, at times acquiring a different meaning, occurring in different registers and with different times of attestations. Stat ‘it is agreed/decided’ occurs in Classical Latin (e.g. Cicero) (29a), vacat ‘there is time’ (29b), attested in Early Latin, is subsequently found in poetry and the post-Augustan prose (Lewis and Short 1942, s.v. vaco). Hiemat ‘it is wintry, cold’ (< hiemo ‘to winter’ with a [+animate] subject), and ‘to be wintry, cold’, with a [− animate] subject), on the other hand, appears to be a late development, attested during the Imperial age (e.g. in technical works) (29c): (29) a. neque adhuc stabat quo potissimum, neither until.now stay.IMPF.IND.3SG where especially sed scies but know.FUT.2SG ‘And it has not been settled yet where exactly, but you will know.’ (Cic., Att., 3, 14, 2, 8) b. quo magis te cui vacat even more 2SG.ACC REL.DAT.SG be.time.PRS.IND.3SG hortor urge.PRS.IND.MP.1SG ‘In as much as/even more so I urge you, since you have time (lit. to whom there is time).’ (Plaut., Ep., 1, 10, 11, 3) c. vehementer hiemat extremely be.cold.PRS.IND.3SG ‘It is extremely cold.’

(Col., Re Rust. 11, 2, 4)

As the data clearly show, the pattern under investigation was not confined to a few ‘impersonal’ mental process/emotion verbs, so-called affective verbs, with the experiencer in the accusative and/or the dative case, as usually assumed in the literature and reference grammars (Woodcock 1959, Leumann, Hofmann, and Szantyr 1965: §165, Bauer 2000, Fedriani 2014). It was instead a pervasive construction, attested throughout the history of the language, also in use during the Classical age (although its productivity at the various stages needs, however, further investigation). Indeed, the examination of the verbs attested in the me/mihi decet, me pudet type, however, suggests that they represent a different clause type if not a distinct voice strategy, rather than a type of impersonal construction (similarly to analogous constructions in some Paman languages (Australia) (Verstraete 2011: 609).

210

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

As pointed out above, they could be better described as patterns denoting the lack of control of the A/S argument/s and the involitionality/spontaneous manifestation of the verb eventuality. The latter is realized in its taking place and as affecting a core argument, expressed either in the accusative, the canonical case for objects, i.e. inactive arguments or in the dative, the case of experiencers/beneficiaries and generally for arguments with a lower degree of affectedness (Næss 2009: 573–574), depending on the verb and its original case frame. It is control, therefore, that seems to be involved in the personal vs impersonal encoding of some (in)transitive predicates in Latin. Some of them involve mental process verbs, while others belong to different subclasses, e.g. activity verbs (e.g. fallo ‘to deceive’, delecto ‘to delight’, fugio ‘to pass by’), states (e.g. attineo ‘to pertain’), and changes of state (e.g. venio in mentem ‘to come to one’s mind’, illucesco ‘to throw light upon’). A better characterization of the predicates figuring in this construction could be cast in lexico-aspectual terms: only activity, state verbs, and accomplishments appear to allow this type of (in)transitive alternation (albeit the issue needs further investigation). Thus, other verb classes alongside emotion and mental process verbs allowed this alternation in early Latin. Such forms as me delectat, me fallit, me poenitet, mihi libet/dolet, therefore, i.e. the various subclasses of third-person singular impersonal verbs usually listed in traditional grammars, may be regarded as the crystallization of a usage that must have been productive at earlier stages of the language, an issue that we leave for further investigation. This interpretation accounts for the coexistence, in Early Latin, of the personal and impersonal forms.

6.3.2 Impersonals with passive morphology/syntax Already in Early Latin, the impersonal active could alternate with the impersonal passive pattern, the -r form, in the unmarked third-person singular, in the tenses of the infectum, and with a synthetic construction, the neuter form of the past participle of the lexical verb + the third singular of sum ‘to be’ in the tenses of the perfectum, in a number of impersonal constructions, covering different types and degrees of agent defocusing. The impersonal passive is employed for atmospheric and natural conditions verbs, where no participant is involved, as in (30a–b) and exceptionally with one argument verbs such as bucino ‘to sound, give signal with a trumpet’, mostly used in the impersonal form (30c) (Perseus, s.v. bucino): (30) a. Ubi nubilabitur where be/become.cloudy.FUT.PRS.MP.3SG ‘Where it will get cloudy.’

(Cato, Agr. 88, 2)

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL b. caletur be.hot.PRS.IND.MP.3SG ‘It is hot.’ c. cum bucinatum when trumpet.PRF.PTCP.N.SG ‘When it trumpeted.’

211

(Plaut., Capt., 80) est be.PRS.IND.3SG (Varr. R. R. 2, 4, 20)

This pattern also occurs with monovalent (31a–b) (e.g. itur, egetur) and, more rarely, divalent verbs (31b) (e.g. agitur, amatur) most typically without an overt object (as well as with trivalent verbs) (Ernout 1909: 18, Napoli 2009, 2010, 2013, Pinkster 1992, 2015: 267–272 for further discussion, examples, and references). The implied agent can be referential, indefinite (‘one, people’) as in (31a–b), or referential, definite, as in (31c–f ) (examples from Pinkster 2015: 267): (31) a. sic itur ad Astra in.this.way go.PRS.IND.MP.3SG to star.N.ACC.PL ‘In this way one goes to Heaven.’ (Verg., Aen., IX, 641) b. quid agitur, Calidore? amatur what do.PRS.IND.MP.3SG Calidore? love.PRS.IND.MP.3SG atque egetur acriter CONJ be.poor.PRS.IND.MP.3SG highly ‘How goes it, Calidore? One loves and is extremely insolvent.’ (Plaut., Pseud., 273) c. propter ipsam viam qua Assoro itur near same.ACC road.ACC that.ABL Assorus.ABl go.PRS.IND.MP.3SG Hennam Henna.ACC ‘Close to the road that people take from Assorus to Henna (lit. near that road from which one goes from Assorus to Henna).’ (Cic. Ver. 4.96) (Pinkster 1992: 163) d. Itur ad te, Pseudole go.PRS.IND.MP.3SG to 2SG.ACC Pseudolus.VOC ‘You are being approached, Pseudolus (lit. One goes/there is going towards you).’ (Pl. Ps. 453–454) e. Eatur. Sequere hac go.PRS.SBJ.MP.3SG follow.IMP.MP.2SG. this.way ‘Let’s go, then. Follow me this way.’ (Ter. Hau. 743) f. bene ambulatum est? well walk.PRF.PTCP.N.SG be.PRS.IND.3SG ‘Did you have a good walk?’ (Plaut. Tru. 369–370) (Pinkster 1992: 170) The overt expression of the Agent (through a prepositional phrase introduced by the preposition a(b) + the ablative case) (32a–b), can include the speaker/hearer, as in (32b) and (32c), from Late Latin (eighth century CE) (Lo¨fstedt 1942: 205), but is rare in Early Latin, and not very frequent also at later times (Pinkster 1992,

212

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

Napoli 2010, 2013 for a perceptive and nuanced analysis of the lexico-aspectual constraints on its occurrence): (32) a. cum a Cotta resisteretur when by Cotta:ABL resist.SBJV.IMPF.MPASS.3SG ‘If there was resistance on the part of Cotta.’

(Caes., BG, 531, 1)

b. peccatum a me maxime-st sin.PRF.PTCP.N.SG by I.ABL highly-be.PRS.IND.3SG ‘I have been grievously in fault.’ (Ter., Hau. 632) (Pinkster 1992: 167) c. interrogatum est ei a nobis interrogate.PRF.PTCP.N.SG be.PRS.IND.3SG 3SG.DAT by 1PL.ABL ‘He was interrogated by us (lit. was interrogated him by us).’ (Form. Marc. Suppl 2. 108, 1) The impersonal passive is also found with some experiencer verbs most typically occurring in the third-person singular active impersonal form (miseret, pudet, taedet, etc.), as shown in (33a–b) for miseret ‘it feels pity’ (see Flobert 1975: 499– 500, Fedriani 2014 for the different times of attestation of the impersonal active and passive forms with these verbs): (33) a. quam matris nunc patris me miseretur more mother.GEN now father.GEN 1SG.ACC feel.pity.PRS.IND.3SG magis more ‘Now I pity my father more than my mother.’ (Turp. 55 Ribb. Ap. Non. 477.15) (Ronconi 1968: 17) b. me eius 1SG.ACC 3SG.GEN ‘I pitied him.’

miseritum est pity.PRF.PTCP.N.SG be.PRS.IND.3SG (Pl., Tr., 430)

This pattern also occurs with transitive verbs taking a non-accusative object (e.g. noceo ‘to harm’, invideo ‘to envy’, resisto ‘to resist’), verbs which cannot occur in a corresponding personal passive, but only in the impersonal passive pattern with the agent optionally expressed as a prepositional phrase, as shown in (34) (Michaelis 1993): (34) a. a nobis non parcetur labori by 1PL.ABL NEG spare.FUT.MPASS.3SG toil.DAT ‘Toil will not be spared by us.’ (Cic., Att, 2, 14, 2) b. omnibus his resistitur all.DAT DEM.DAT.PL resist.PRS.IND.MP.3SG ‘A ll these are resisted.’ (Caes., BC, 4, 1)

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

213

The impersonal passive, therefore, may mark either maximal agent defocusing (i.e. lack of a participant), as with weather verbs (30) or different degrees of (semantic, pragmatic, syntactic) defocusing of the A/S participants, as in (31)–(34). In the latter constructions A/S may be referential indefinite, denoting a participant in the universe of discourse whose identity is unknown to both speaker and hearer, as in (30a) or as participant with a specific reference, referring to any participant in the universe of discourse, including also SAPs, either contextually recoverable or optionally expressed by means of a prepositional phrase, as in (32) (see further discussion in Pinkster 1992, 2015, Pieroni 2000, Napoli 2009, 2010, 2013, and further references therein):

6.3.3 Existential–impersonal constructions In Late Latin the impersonal passive and active forms often occur with a nonagreeing argument (in case, number, and/or gender), as shown in (35). Example (35a) exhibits lack of number agreement of the verb (in the default third-person singular) (imponatur) with the preverbal plural argument in the accusative (ipsos ficos), conveying given information, while (35b) displays lack of gender agreement of the past participle (factum) with the verbal argument in the accusative (missam), conveying new information. The non-agreeing argument in the accusative case can occur in pre/postverbal position, conveying non-topical information, as illustrated in (35b–c) (Cennamo 2000, 2009, 2011): (35) a. Ipsos ficos imponatur (+given) DEM.ACC.PL fig:ACC.PL gather.PRS.SBJ.MP.3SG ‘One should gather these figs.’ (Ruf. De Pod. 35) b. cum factum fuerit missam (+new) when make.PRF.PTCP.N.SG be.PRF.SBJ.3SG Mass.ACC ‘When the Mass is over.’ (Per. Aeth., 32, 2) c. bonum aerem facit (+new) good.ACC weather.F.ACC make.PRS.IND.3SG ‘It is good weather (lit. it makes good weather).’ (Sal. Vit. patr. 5, 11, 51) This pattern appears to develop rare analogous (albeit mostly philologically uncertain) Early Latin constructions (Ernout 1908–1909, Bauer 2000: 109–110, Cennamo 2011: 178 and, more recently, Pinkster 2015: 268–269, who denies the existence of impersonals with accusative arguments in Early Latin, in line with Calboli 1962: 7–56, who views the construction as a Late Latin development; see also Adams 2013: 240–242 for a discussion of Late Latin examples): (36) a. me … despicatur 1SG.ACC despise.PRS.IND.MP.3SG ‘I am despised/one despises me.’

(Plaut. Cas. 185)

214

MICHEL A CENNAMO AND CL AUDIA FABRIZIO b. agitandum est do.GER.N.SG be.PRS.IND.3SG ‘One must be on guard.’ c. vitam vivitur life.ACC live.PRS.IND.MP.3SG ‘One lives life.’

vigilias sentinel.ACC.PL (Plaut., Trin., 869)

(Enn. Trag. 202)

As pointed out (Section 6.3.1.2), existential habeo ‘to have/hold’ also occurs in this construction in Late Latin, as in (36), clearly anticipating later Romance developments (Cennamo 2011: 177–178): (37) in Hebraeo… non habet hunc numerum in Hebrew.ABL NEG have.PRS.IND.3SG this.ACC number.ACC ‘In Hebrew this number does not exist (lit. not has this number).’ (Hier., Ezech., 11. 297B)

6.4 The impersonal active, non-nominative arguments, and alignment As illustrated in Section 6.3, the active impersonal in Latin is not always employed as an agent-defocusing strategy, unlike the passive impersonal form. Whereas the use of the latter in such patterns as curritur ‘running takes place’, pugnatum est ‘fighting took place’, seems to be a means of foregrounding the event and of defocusing the agent (only rarely overtly expressed) both in the infectum and in the perfectum, the use of the third singular active with an accusative argument in O/S function, rather seems to be a strategy for signalling lack of control/involitionality of the participant over the verb eventuality, which is portrayed as affecting it. The verbal argument occurs either in the accusative, the inactive case, or in the dative, the case of experiencers/beneficiaries, and generally for arguments with a lower degree of affectedness (Næss 2009: 573–574). Active impersonal patterns with an accusative/oblique argument like me pudet, me fugit, me delectat, me/mihi decet, therefore, show a striking similarity with analogous constructions in languages with semantic alignment (Donohue 2008, Malchukov 2008, Mithun 2008, and contributions in Malchukov and Siewierska 2011). For instance, in several Australian languages (e.g. Murrinh-Patha and Waray (Walsh 1989: 428–429, 432, Evans 2004: 178), Iwaidjan (Evans 2004), Umpithamu and the Lamalamic languages (Verstraete 2011)), involuntary physical processes occur in the impersonal form, characterized by the lack of cross-referencing bound pronouns prefixed to the verb root, marking the subject and object status of verbal arguments (Walsh 1987: 426, Verstraete 2011: 607), the ‘experiencer object construction’, as in (38a) vs (38b), the plain intransitive form (Walsh 1987: 429, Evans 2004: 178):

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

215

(38) a. dam -ngi -kule 3SG.SUBJ -3.subj 1SG.OBJ.cough (experiencer–object construction) (Murrinh-Patha) ‘I feel like coughing; I am going to cough.’ b. ngi -kulurrk -nu (Murrinh-Patha) 1SUBJ -cough -FUT ‘I’ll cough.’ c. pulnu pan -laki -nj (Waray) sickness 1.OBJ-push-REALIS ‘I am sick (lit. sickness it pushes/tosses me).’ This pattern is at times undistinguishable from an objectively inflected intransitive. For instance, in (38c), exemplifying a noun–verb construction in Waray (Walsh 1987: 432, Evans 2004: 178), a third-person singular subject is not marked (i.e. there is zero marking on the verb). The pattern is therefore ambiguous between a quasi-transitive construction (with pulnu ‘sickness’ as subject and pan ‘me’ as object) or an objectively inflected monovalent verb, illustrated in (38a) for the verb COUGH. Comparable involitionality constructions from Western Indo-European languages such as Icelandic and Lithuanian are illustrated in (39): (39) a. mig dreymdi o¨mmu (Icelandic) 1SG.ACC dream.PST.3SG grandma.ACC ‘I dreamt about grandma.’ (Barðdal 2004: 108) b. Joną purto (nuo šalcˇio) (Lithuanian) Jonas.ACC shaking.PRS.3SG from frost.GEN) ‘Jonas is shaking (from the cold).’ (Wiemer and Bjarnadóttir 2014: 306) The patterns in (38)–(39) are reminiscent of the Latin ‘impersonal’ verbs optionally taking a non-nominative argument, as in (40) (Cennamo 2011): (40) a. me (ACC) pudet (3SG.IMPERS) (fratris) (GEN) (transitive impersonal) b. me (ACC) pudet (3SG.IMPERS) (intransitive) ‘I am ashamed (of my brother) (lit. (it) shames me of my brother).’ The Latin construction exemplified in (40), discussed in Section 6.3.1.4 for the types of eventualities involved, indeed are similar to structures found in languages with semantic alignment, where impersonal verb forms correlate with involuntary/unintentional eventualities (physiological and/or psychological processes, according to the language(s)) (see Walsh 1989, Evans 2004, Verstraete 2011 for Australian languages; Roberts 2001 for Amele, Papua New Guinea, Klamer 2008 for Kambera, Austronesian; Mithun 2008 for American Indian languages; Malchukov 2008, Malchukov and Ogawa 2011 for a general discussion). Analogous patterns are also found in the coding of experiencers in Tibeto-Burman and Indo-Aryan languages of the Himalayas (Bickel 2004), South-Asian languages (Verma and Mohanan 1990) as well as a number of Western Indo-European

216

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

languages (e.g. Icelandic, Lithuanian) (39) (Barðdal 2004, Barðdal et al. 2012, Wiemer and Bjarnadóttir 2014, Holvoet 2016, Lavine 2016) and in the Balkans (i.e. Eastern Indo-European) (see Friedman and Joseph 2018 for a recent discussion). Also in Latin, therefore, there appears to be the same correlation between impersonal verb forms and involuntary/unintentional eventualities that one finds in several semantically aligned systems, a phenomenon that has been widely investigated for its semantics, syntax (especially in relation to the subject status of non-nominative S/A arguments) and areal distribution across languages, both synchronically (see contributions in Bhaskararao and Subbarao 2004) and, more recently, diachronically (see Barðdal and Eythórsson 2009, Barðdal et al. 2012, Montaut 2013 and other contributions in Serzˇant and Kulikov 2013). Unlike in other early Indo-European languages (e.g. Sanskrit (Hock 1990) and Hittite (Luraghi 2010, Inglese 2020: 33), displaying the same type of patterns, the construction in Latin is not confined to few experiencer/psychological verbs/predicates (i.e. states), but it also occurs with other types of eventualities, including activities and accomplishments, as illustrated in Section 6.3.1.4.⁷ The lack of control/affectedness of S, however, may also be conveyed by the passive voice (the -r form in the infectum, sum ‘to be’ + past participle in the perfectum), that may act as a detransitivizer, turning a transitive verb into an intransitive one, marking an inactive subject, as pointed out in Section 6.2, and further exemplified below (Cennamo 1998: 81, 2020): (40) a. b. c. d. e.

excito ‘I awaken’ > excitor ‘I wake up’ gravo ‘I oppress’ > gravor ‘I have difficulties’ rumpo ‘I break’ > rumpor ‘I burst with envy’ me praecipito ‘I throw myself ’ > praecipitor ‘I fall down’ augeo ‘I increase’ > augeor ‘I grow’

The difference between the two strategies seems to lie in the defocusing of the S/A argument and the foregrounding of the event in the impersonal active pattern, unlike in the personal passive construction, that only signals affectedness of the subject.

6.5 Conclusions The notion of control plays a major role in the encoding of transitivity in Latin, determining voice fluctuations with animate subjects and the ‘impersonal’ encoding of eventualities. ⁷ The issue, however, needs further quantitative and qualitative investigation, with a more finegrained analysis of the lexico-aspectual and thematic constraints on this type of intransitive alternation/voice strategy, as resulting from the interplay of the templatic and idiosyncratic aspects of verb meaning and their internal structure, in line with recent approaches (Beavers and Koontz-Garboden 2020), as well as a study of the diachronic development of this construction in Latin.

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

217

In particular, we have shown that some impersonal constructions represent a different clause type, and could be defined as involitionality/lack of control patterns, similar, in their semantics and formal marking, to analogous involuntary, ‘impersonal’ constructions in languages with semantic alignment (sub)systems (e.g. (Northern and North-Western) Australian languages such as (non-)Pama–Nyungan (Murrinh-Patha, Waray) and Iwaidjan), Kambera— Austronesian, Amele—Papua New Guinea, Tunica for American Indian languages, as well as Indo-Aryan, Tibeto-Burman languages of the Himalayas, South-Asian, and Western Indo-European languages). We have also demonstrated that active and passive impersonals, although overlapping in some of their functions, are not equivalent in Latin. Whereas impersonal passives foreground the event and defocus the agent, that may be lacking at argument structure, unexpressed, or realized as an oblique (dative/prepositional phrase), in some of its uses the impersonal active pattern (with an optional nonnominative argument, according to the syntactic valency and degree of syntactic elasticity of the verb) points to the existence of a dependent-marked subsystem of active–inactive alignment in early Latin, probably in use throughout the history of the language, sensitive to the notion of control and to the lexico-aspectual characteristics of verbs. The data investigated, therefore, give further evidence for regarding Latin as a language with syntactically based (nominative–accusative) and semantically based alignment patterns, while also pointing to the usefulness of recent nuanced approaches to verb meaning for a better understanding of involitionality patterns in Latin and their diachronic development.

Acknowledgements Preliminary versions of the present work were presented at the workshops ‘Indo-European Case and Argument Structure in a Typological Perspective’, Bergen, 20–21 August 2008, ‘Subject and Transitivity in Indo-European and Beyond: A Diachronic Typological Perspective’, 43rd Annual Meeting of the Societas Linguistica Europaea, Vilnius, 2–3 September 2010, ‘Non-Canonically Case-Marked Subjects, within and across Languages and Language Families’, Reykjavik-Eyjafjallajo¨kull, 4–8 June 2012, ‘Argument Structure and Syntactic Reconstruction’, Bergen, 11–12 June 2015. We wish to thank the audiences and especially Jóhanna Barδdal, John Beavers, Bridget Drinka, Thórhallur Eythórsson, Brian Joseph, Leonja Kulikov, Andrej Malchukov for interesting and enlightening remarks. We are also very grateful to two anonymous reviewers and to the editor, Eystein Dahl, for most helpful comments and suggestions on an earlier draft of this chapter, which helped to improve and clarify some of the claims put forward. All shortcomings are, of course, our own. The research has been funded by the Italian Ministry of Education and Research, grant n. 20159M7X5P_002 PRIN 2015

218

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

‘Transitivity and argument structure in flux’, PI Michela Cennamo (University of Naples Federico II).

Latin sources Acc., Trag.: L. Accius, Tragoediae, in Scaenicae Romanorum Poesis Fragmenta, ed. O. Ribbeck, Leipzig, Teubner, 1897, vol. I. Anthim.: Anthimi, De Observatione Ciborum ad Theodoricum Regem Francorum Epistula, ed. E. Lichtenhan, Leipzig, Teubner, 1928 (Corpus Medicorum Latinorum VIII, I). Au. Gel., NA: A. Gellius, Noctes Atticae, ed. K. Marshall, Oxford, Oxford Clarendon Press, 1968. August. Psal.: Augustinus Hipponensis. In Psalmum 66 enarratio. Patrologiae Cursus Completus Series Latina, ed. J-P. Migne 1841, vol. 36. Caec.: Caecilius Calactinus Fragmenta, ed. E. Ofenloch, Leipzig, Teubner, 1967. Caes, BG: C. Iulius Caesar, De Bello Gallico, in C. Iulii Caesaris Commentarii Rerum Gestarum, ed. O. Seel, Teubner, Lepzig, 1961, vol. I. Caes, BC: Bellum Civile, ed. A. Klotz, Leipzig, Teubner, 1964. Cat., Agr.: M. Porcius Cato, De agricultura fragmenta, in Scriptorum Romanorum De Re Rustica Reliquiae, ed. F. Speranza, Messina, Biblioteca di Helikon, vol. 1, 1974. Cic., Amic.: M. Tullius Cicero, Laelius de Amicitia, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, Part 4, vol. 3, ed. C. F. W. Mueller, 1890. Cic., Att.: M. Tullius Cicero, Epistulae ad Atticum, in Cicero’s Letters to Atticus, ed. D. R. Shackleton Bailey, Leipzig, Teubner, 1965–1968. Cic., Q. Fr.: M. Tullius Cicero, Epistulae ad Quintum Fratrem et M. Brutum, ed. D.R. Shackleton Bailey, Cambridge University Press, 1890. Cic., De Fin.: M. Tullius Cicero, De Finibus Bonorum et Malorum, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, ed. T. Schiche, Leipzig, Teubner, 1915. Cic. Sen.: M. Tullius Cicero, De Senectute, De Amicitia, De Divinatione, ed. W. A. Falconer, Cambridge, MA, Harvard University Press, 1923. Cic., Div. Caec.: M. Tullius Cicero, Divinatio in Q. Caecilium, in M. Tulli Ciceronis Orationes, ed. W. Peterson, Oxford, Clarendon Press, 1917, vol. III. Cic., Mur.: M. Tullius Cicero, Pro Murena, in M. Tulli Ciceronis: Orationes, ed. A. C. Clark, vol. I, Oxford University Press, 1905. Cic., Phil.: M. Tullius Cicero, Philippicae, in M. Tulli Ciceronis Orationes, ed. A. C. Clark, Oxford, Clarendon Press, 1918, vol. II. Cic., Tusc.: M. Tullius Cicero, Tusculanae Disputationes, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, ed. M. Pohlenz, Oxford, Oxford Classical Texts, 1918.

OBLIQUE ARGUMENTS, ACTIVE IMPERSONAL, CONTROL

219

Cic., Verr.: M. Tullius Cicero, In Verrem, in M. Tulli Ciceronis Orationes, ed. W. Peterson, Oxford, Oxford Classical Texts, 1917, vol. III. Col., RR.: L. Iunius Modeatus Columella, De Re Rustica, libri 1–2: in L. Iuni Moderati Columellae Opera Quae Exstant, ed. V. Lundstro¨m, vol. II, 1917; libri 3–5: ibid., ed. V. Lundstro¨m, A. Josephson, and S. Hedberg, 1968, vol. III; libri 6–7: ibid., ed. V. Lundstro¨m, 1940, vol. IV; libri 8–9, ibid., ed. V. Lundstro¨m and A. Josephson, 1955, vol. V; libri 10–11: ibid., ed. V. Lundstro¨m, 1902–1906, vol. VI–VII; libro 12, ibid., ed. S. Hedberg, 1958, vol. VIII, Uppsala, Libraria Lundequistiana; Leipzig, Harrassowitz. Enn., Ann.: The Annals of Q. Ennius, ed. O. Skutsch, Oxford, Clarendon Press, 1985. Hier. Ezech.: Hieronymus Commentariorum in Hiezechielem libri XIV, ed. F. Glorie, 1964. Corpus Cristianorum Series Latina 75 (CCSL) Hist. Aug.: Historia Augusta. Scriptores Historiae Augustae, ed. E. Hohl, vol. 1, Leipzig, Teubner, 1971. Horat., Carm: Q. Horatius Flaccus, Carmina, in Opera, ed. F. Klingner, De Gruyter, 1959. Liv., Ab Urbe Cond.: T. Livius, Ab Urbe Condita, libri 1–5: ed. R. S. Conway and C. F. Walters, 1955; libri 6–10, 21–25: ed. C. F. Walters and R. S. Conway, 1919– 1950; libri 26–30: ed. R. S. Conway and S. K. Johnson, 1953; libri 31–35: ed. A. H. McDonald, 1969; libri 36–40: ed. W. Weissenborn and M. Mu¨ller; libri 41–45: ed. W. Weissenborn and W. Heaeus, 1908, Oxford, Clarendon Press. Lucr., De Rerum Natura: T. Lucretius Carus, De Rerum Natura Libri Sex, ed. J. Martin, Leipzig, Teubner, 1969. Mul. Chir.: Claudii Hermeri Mulomedicina Chironis, ed. E. Oder, Leipzig, Teubner, 1901. Naev, Pun.: Cn. Naevius, Bellum Punicum, in Fragmenta Poetarum Latinorum Epicorum et Lyricorum praeter Ennium et Lucilium, ed. W. Morel, Leipzig, Teubner, 1927. Ov., Met.: P. Ovidius Naso, Metamorphoses, in Metamorphoses in Two Volumes, ed. F. J. Miller and G. P. Goold, London, Loeb Classical Library, 1977–1984. Ov., Ep. (vel Her.): P. Ovidius Naso, Epistulae (vel Heroides), in Ovid in Six Volumes, ed. G. Showerman and G. P. Goold, Harvard, Loeb Classical Library, 1977, vol. I. Ov., Tr.: P. Ovidius Naso, Tristia, ed. Gluck, Heidelberg, 1977. Pac., Trag.: M. Pacuvius, Tragoediae, in Scaenicae Romanorum Poesis Fragmenta, ed. O. Ribbeck, Leipzig, Teubner, 1897, vol. I. Per. Aeth.: Silviae vel potius Aetheriae peregrinatio ad loca sancta, ed. W. Heraeus, Heidelberg, Carl Winter, 1929. Plaut., Amph.: T. Maccius Plautus, Amphitruo, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol I.

220

MICHEL A CENNAMO AND CL AUDIA FABRIZIO

Plaut., Cas.: T. Maccius Plautus, Casina, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol I. Plaut, Curc.: T. Maccius Plautus, Curculio, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol I. Plaut., Men.: Plautus, Menaechmi, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol I. Plaut., Mer.: T. Maccius Plautus, Mercator, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol I. Plaut., Miles: T. Maccius Plautus, Miles Gloriosus, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol II. Plaut. Per.: T. Maccius Plautus, Persae, in T. Macci Plauti Comoediae, ed. G. Goetz and Fr.Schoell, Leipzig, Teubner, 1896, vol 5. Plaut., Ps.: T. Maccius Plautus, Pseudolus, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol II. Plaut., Trin.: T. Maccius Plautus, Trinummus, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol II. Plaut., Truc.: T. Maccius Plautus, Truculentus, in Plauti Comoediae, ed. F. Leo, Berlin, Weidmann, 1895, vol II. Plin., Nat.: C. Plinius Secundus, Naturalis Historia, ed. C. Mayhoff, Leipzig, Teubner, 1892–1909, voll. I-V. Sal. Vit. Patr.: Salonius, A. H. 1920. Vitae Patrum: Kritische Untersuchungen u¨ber Text, Syntax und Wortschatz der spa¨tlateinischen Vitae patrum (B. II, V, VI, VII). Lund: Gleerup. Stat., Thebais: P. Papini Stati Thebaidos Libri XII, ed. D.E. Hill, Brill, 1983. Ter., And.: P. Terentius Afer, Andria, in P. Terenti Afri Comoediae, ed. R. Kauer, W. M. Lindsay, and O. Skutsch, Oxford, Oxford Clarendon Press, 1958. Ter., Ad.: P. Terentius Afer, Adelphoe, in P. Terenti Afri Comoediae, ed. R. Kauer, W. M. Lindsay, and O. Skutsch, Oxford, Oxford Clarendon Press, 1958. Ter., Hau: P. Terentius Afer, Heauton Timorumenos, in P. Terenti Afri Comoediae, ed. R. Kauer, W. M. Lindsay, and O. Skutsch, Oxford, Oxford Clarendon Press, 1958. Turp.: Turpilii Comici Fragmenta, ed. L. Ryclewska, Leipzig, Teubner, 1971. Varro, L.L.: M. Terentius Varro, De lingua latina quae supersunt, ed. G. Goetz and F. Schoell, Teubner, 1910. Verg., Aen.: P. Vergilius Maro, Aeneis, in P. Vergili Maronis Opera, ed. R. A. B. Mynors, Oxford, Clarendon Press, 1972. Vulg. Gen: Vulgata, Genesis.

Digital resourses Perseus = Perseus Digital Library and Dictionaries, www.perseus.tufts.edu

7 Infinitives and subjecthood between Latin and Old Italian Claudia Fabrizio

7.1 Introduction Infinitives can appear in subject function both in Latin and in Italian (Pinkster 2015: 756, 796, Skytte and Salvi 1991, Egerland 2010a, 2010b), as the following examples show. In (1a), from Cicero, the infinitive peccare ‘to err’ is the subject of est utile ‘it is useful’; in (1b) the Italian infinitive errare, with the same meaning, plays an identical role with respect to the finite predicate è umano ‘it is human’: (1) a. Numquam est utile peccare never be. IND.PRS.3SG useful.NOM err.INF ‘To err is never useful.’ (Cic., De Off., 3,64) b. Errare è umano, perseverare diabolico err.INF be.IND.PRS.3SG human persist.INF diabolical ‘To err is human, but to persist in error is diabolical.’ In spite of its superficial likeness, however, the syntax of subject infinitives in these two languages obeys radically different constraints, and this fact suggests that the distribution of subject infinitives might have undergone a deep reshaping in the transition from Latin to Italian. After discussing the relevant data, I account for the diachronic change leading from the restricted use of Latin infinitives to the unconstrained Italian distribution. In my account, the diachronic path at issue turns out to be concerned with the domain of alignment and the notion of subjecthood. In Section (7.2) the results of a corpus-based investigation from Early and Classical Latin are presented and discussed. After a complete survey of the data, it is shown that the distribution of subject infinitives in this language is governed by semantic constraints, in that infinitives, unlike other action nouns, can only appear in subject function when the subject itself has the semantic macrorole of undergoer (see also Chapter 3 this volume). The topic of subject infinitives in Old Italian is addressed in Section 7.3. The main hypothesis is that their unrestricted distribution is due to the reanalysis of some experiencer predicates allowing an accusative argument, which might have paved the way for the spread of the subject infinitive construction to all kinds of biargumental clauses, regardless of the Aktionsart Claudia Fabrizio, Infinitives and subjecthood between Latin and Old Italian. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl. Oxford University Press. © Claudia Fabrizio (2022). DOI: 10.1093/oso/9780198857907.003.0007

222

CL AUDIA FABRIZIO

of the finite predicate and the consequent macrorole of the subject. Further, an interesting, and so far unknown, phenomenon of differential marking of subject infinitives in Old Italian is pointed out.

7.2 Latin In Latin, an infinitive may realize the subject of a finite predicate, as in the following example:¹ (2) senem oppugnare certumst consilium old.ACC.SG attack.INF firm.NOM.SG.be.IND.PRS.3SG decision.NOM.SG mihi 1SG.DAT ‘It is my firm decision to attack the old man.’ (Pl., Epid., 163) However, this use of the infinitives is not, as it were, constraint-free. In this section I aim to show that Latin infinitives are denied subjecthood when specific conditions concerning the finite verb of the clause are not met, and that they are, conversely, likely to function as subjects only when those conditions are fulfilled. I also argue that the constraints governing the distribution of infinitives suggest a semantically oriented pattern of alignment in the grammatical domain of nominalizations. In previous works (Fabrizio 2015, 2018), I have observed that infinitives in subject function can only surface in Latin when the finite verb of the clause belongs to a set of intransitive predicates, namely states, achievements, and passive predicates; on the other hand, infinitives do not occur as subjects of finite transitive and unergative predicates, such as accomplishments and activities.² The examples in (3)—all taken from the same author and from the same work—show that errare ¹ Unless specified by further glossing, the abbreviation INF is intended as ‘present tense infinitive, active voice’. ² I adopt here a four-type Aktionsart classification, mostly drawing upon Vendler (1967), Dowty (1979), as well as Foley and Van Valin’s (1984). The Aktionsart of a predicate is independently established on the basis of an array of relevant tests. Briefly, states and activities are the basic types; they are [-telic] and [-punctual]. States represent non-dynamic situations (such as ‘be tall’, ‘have’, ‘like’), while activities denote non-telic dynamic processes (such as ‘sing’, ‘drive’, ‘pray’). Achievements and accomplishments denote the attainment of a resultant state, but while achievements (such as ‘die’, ‘fall’) encode instantaneous changes, accomplishments express telic processes happening over some temporal span followed by a resultant state, mostly showing an agentive and volitional feature. Van Valin and LaPolla (1997: 90–129) and Van Valin (2005: 32) also add the non-Vendlerian class of semelfactives. Moreover, they distinguish between accomplishments and active accomplishments. For an in-depth classification of Latin predicates based on a more fine-grained Aktionsart distinction, I refer to Fabrizio (forthcoming). In the Role and Reference Grammar framework, the semantic macroroles of actor and undergoer are assigned to the subject argument depending on the Aktionsart of the predicate. With a monoargumental predicate, (i) if the verb has an activity predicate in its logical structure, the macrorole is an actor, and (ii) if the verb has no activity predicate in its logical structure (as with states, accomplishments, and achievements), the macrorole is an undergoer (Van Valin 2005: 63).

INFINITIVES AND SUBJECTHOOD

223

‘to err’ is the only argument of the stative predicate tolerabile sit ‘let it be tolerable’ (3a), but cannot occur as the subject of, say, a transitive accomplishment as deduxit ‘it led’ (3c)—unlike the cognate noun error ‘error’ (3b): (3) a. semel errare sane tolerabile sit once err.INF let tolerable.NOM.SG be.SBJ.PRS.3SG ‘Let to err once be tolerable.’ (Quint., Decl. Min., 310, 4) b. eo me deduxit error ut there 1SG.ACC lead.IND.PRF.3SG mistake.NOM.SG so that filios meos occiderem son.ACC.PL 1SG.POSS.ACC.PL kill.SBJ.IMPF.1SG ‘The mistake led me to the point of killing my sons’ (Quint., Decl. Min., 337, 11) c. *eo me deduxit errare ut filios there 1SG.ACC lead.IND.PRF.3SG err.INF so that son.ACC.PL meos occiderem 1SG.POSS.ACC.PL kill.SBJ.IMPF.1SG More detailed evidence on the distribution of infinitives in subject function in Latin is given in Section 7.2.1. It is already clear, however, that the state of affairs shown by (3) is slightly unexpected in a language otherwise exhibiting a consistent nominative–accusative alignment, whereby grammatical relations are independently established regardless of the Aktionsart of the predicate. Then, it should be verified whether the syntax of infinitives in subject function obeys a (local) manifestation of a different coding strategy.

7.2.1 The data In this section, the results of a corpus-based investigation on Latin are taken into account.³ The occurrences showing an infinitive in subject function have been grouped according to the type of finite verb of the main clause. In point of fact, Latin infinitives are attested in subject function only with some finite verbs, namely stative predicates (among which, nominal predicates), achievements (denoting telic change of state), and activity predicates in the marked voice, i.e. passive. For the sake of convenience, in the following examples subject infinitives are underlined, while finite predicates are in bold: ³ The Latin corpus has been built up as follows: (a) review of the occurrences of subject infinitives quoted in reference grammars (Draeger 1878, Bennett 1914, Ronconi 1946, Woodcock 1959, Ernout and Thomas 1964, Leumann, Hofmann, and Szantyr 1965, Pinkster 1991, 2015); (b) corpus-driven investigation: examination of all the contexts displaying an Infinitive argument in a corpus consisting of 12 works of seven authors through the PHI-5.3 corpus, including all Latin literary texts from the second century BCE to the second century CE.

224

CL AUDIA FABRIZIO

(4) Stative predicates a. Salutem scriptam dignum est salutation.ACC.SG written. ACC.SG worthy. NOM.SG be.IND.PRS.3SG mittere dignis mittere worthy.DAT.PL send.INF ‘It is worth sending a written salutation to worthy people.’ (Pl., Ps., 1013) b. In re tam usitata satis est in thing.ABL.SG such ordinary.ABL.SG enough be.IND.PRS.3SG ostendere show.INF dixisse omnis antea ius ita dixisse everybody.ACC.PL before law.ACC.SG so say.INF.PRF ‘In such an ordinary matter, it is enough to show this is the way that everybody judged in the past.’ (Cic., Ver., 2, 1, 114, 13–14) c. Nonne emori virtutem praestat emori per NEG die.INF through virtue.ACC.SG be.preferable.IND.PRS.3SG quam vitam miseram atque inhonestam … rather than life.ACC.SG miserable.ACC.SG CONJ dishonest.ACC.SG per dedecus amittere? through shame.ACC.SG lose.INF ‘Isn’t it preferable to die honourably, rather than to lose your miserable and dishonest life shamefully?’ (Sall., Cat., 20, 9) (5) Achievements (change of state/place) a. Qui in mentem venit tibi istaec how in mind.ACC.SG come.IND.PRS.3SG 2SG.DAT DEM.ACC.PL dicta dicere? dicere dicere word.ACC.PL say.INF ‘How does it occur to you to say such things?’ (Pl., Trin., 77) b. Si contigerit fundos Apuliae iungere… iungere iungere if happen.SBJ.PRF.3SG farm.ACC.PL Apulia.DAT join.INF ‘If it were to happen that I join my farms to Apulia… (lit.: If it happened to join my farms to Apulia).’ (Petr., Sat., 77, 3, 1) (6) Passive predicates a. Latine est in magna laude loqui loqui Latin:ADV speak.INF be. IND.PRS.3SG in great.ABL.SG praise.ABL.SG ponendum put.GER.NOM.SG ‘Speaking Latin must be greatly praised (lit. Speaking Latin must be put in great praise).’ (Cic., Brut., 140)

INFINITIVES AND SUBJECTHOOD b. Posse eripitur loqui Posse loqui be able.INF speak.INF take away. IND.PRS.3SG.MP ‘The faculty of speech is taken away.’

225

(Ov., Met., 2, 483)

c. Nec ripas datur horrendas et NEG shore.ACC.PL allow.IND.PRS.3SG.MP terrifying.ACC.PL CONJ rauca fluenta transportare transportare roaring.ACC.PL stream.ACC.PL go beyond.INF ‘It is not allowed to go beyond the terrifying shores and the roaring streams.’ (Verg., Aen., 6, 327–328)

7.2.2 Constraints on subject infinitives The data presented above allow a preliminary generalization: in Latin, infinitives are prevented from being the subject of transitive clauses. Their «bias» against transitivity is unambiguously exemplified by the following example:⁴ in a context like (7a), only the noun potio ‘drink’ can be the subject of the verb efficere ‘make’. Significantly, I have found no trace of infinitives functioning as actors of biargumental constructions like (7a)—witness the unattested (7b), which is an exemplum fictum (like (3c)).⁵ The same holds for the contrast between (8a) and (8b): (7) a. umidum corpus efficit multa humid.ACC.SG body.ACC.SG make.IND.PRS.3SG huge.NOM.SG potio drink.NOM.SG ‘Drinking a lot makes the body moist (lit.: A huge drink makes the body moist).’ (Cels., Med., 1, 3, 30–31) b. *umidum corpus efficit multum bibere moist.ACC.SG body.ACC.SG make.IND.PRS.3SG much drink.INF (8) a. aquae frigidae potio adsumpta ulcera water.GEN.SG cold.GEN.SG drink.NOM.SG drunk.NOM.SG sore.ACC.PL adstringit reduce. IND.PRS.3SG ‘To drink a cold beverage reduces sores (lit.: A drink of cold water reduces the sores).’ (Cels., Med., 4, 22, 4–8) b. *aquam frigidam bibere ulcera adstringit water.ACC.SG cold.ACC.SG drink.INF sore.ACC.PL reduce.IND.PRS.3SG ⁴ In Fabrizio (forthcoming) I list and discuss the only four (out of more than 300) occurrences of my whole corpus in which an infinitive happens to realize the first argument of a biargumental predicate. For the sake of space, I do not treat them here. At any rate, their scarceness does not run counter to the generalization I put forward. ⁵ Notice that in a sentence like (7a) the action noun potio acts, from a semantic point of view, as it were an infinitive, in that it does not denote a punctual action (a single act of drinking), but the process itself (the fact of drinking) (for a classification of action nouns illustrating this semantic contrast, see Simone 2003).

226

CL AUDIA FABRIZIO

Table 7.1 Aktionsart, macrorole, and infinitive subjects in Latin

So far, the evidence suggests that a constraint operates on the distribution of infinitives in subject function. A closer look reveals that it depends on the Aktionsart of the finite verbs and its argument structure, and allows us to refine our first generalization. First of all, predicates taking an infinitive as their subject argument only belong to three main Aktionsarten: states, activities (albeit only in the passive voice), and achievements, i.e. telic changes of state. Second, Latin infinitives can only appear as subjects of intransitive clauses, displaying a single argument in their argument structure (Jackendoff 1976, Dowty 1979, Van Valin 2005, Van Valin and LaPolla 1997: 82–138). The logical structures in (9a) and (9b) exemplify stative and achievement predicates; (9c) represents an accomplishment: (9) a. deficere dignum esse b. contingere c. eripere

lack ́ (x) be ́ (x, [worthy ́]) INGR happen ́ (x) do ́ (x, [tear ́ (x, y)]) & [BECOME tornʹ (y)]

Just like the second argument of a transitive pattern, the sole argument of states and achievements reflects the properties of an undergoer, i.e. of a participant ‘which does not perform, initiate, or control any situation but rather is affected by it in some way’ (Foley and Van Valin 1984: 29, Van Valin and LaPolla 1997: 146, Van Valin 2005: 60–67). Henceforth, I shall refer to the sole argument of a state or an achievement predicate as UI , and to the second argument of a biargumental predicate as UT . As shown in Table 7.1 in the language under examination infinitives only realize one macrorole,⁶ the undergoer, therefore behaving as non-canonical subjects (Bayer 2004, Mahajan 2004, Tsunoda 2004; Bentley 2006: 10, Bentley 2010): ⁶ Semantic roles tend to group into two natural macroroles, called actor and undergoer (see also Chapter 3 in this volume). A macrorole is therefore a generalization across the argument types, mediating between semantics and syntax (Van Valin and LaPolla 1997: 146).

INFINITIVES AND SUBJECTHOOD

227

Further evidence is provided by the fact that unergatives do not allow infinitives to code their sole argument, insofar as they take an actor subject (henceforth, AI ), i.e. a participant ‘which performs, effects, instigates, or controls the situation denoted by the predicate’ (Foley and Van Valin 1984: 29; Van Valin and LaPolla 1997: 146). This fact is extremely important vis-à-vis the claim that infinitives only realize undergoers, as well as with regard to the assumption that split intransitivity is revealed by a number of phenomena in the ancient stages of Indo-European languages (Benedetti 2002; Gianollo 2005, Cennamo 2001, 2009, 2011, Rovai 2005, 2007a, 2007b, 2012, 2014, Lazzeroni 2002a, 2002b, 2009, 2013) (Table 7.2). Table 7.2 Intransitive finite predicates with infinitives as subjects in Latin Infinitives as subjects

It is very significant, in my view, that activity predicates can license an infinitive as their subject argument only in the passive voice, passive being a strategy whereby the UT is assigned the grammatical relation subject, and is marked accordingly (Table 7.3). Table 7.3 Constraints on activity predicates with infinitives as subjects in Latin Infinitives as subjects

Moreover, Latin infinitives behave both as UI and UT , without morphological declension being needed (Ronconi 1946: 161, Leumann, Hofmann, and Szantyr

228

CL AUDIA FABRIZIO

1965: 343, Pinkster 2015: 756, 796).⁷ Examples in (10) show that the infinitives can be used as UT of biargumental clauses, even in coordination with an accusative argument (10a): (10) a. duas res … persequitur, two.ACC.F.PL thing.ACC.F.PL look for.IND.PRS.3SG.MPASS rem militarem et argute loqui thing.ACC.F.SG military.ACC.F.SG CONJ wittily speak.INF ‘It (sc. the Gaulish people) looks for two things, military life and wit language (lit.: wittily speaking).’ (Cat., Orig., 34) b. At ego amo hanc. At ego AND 1SG.NOM love. IND.PRS.1SG DEM.ACC.F.SG AND 1SG.NOM esse et bibere eat.INF CONJ drink.INF ‘A s for me, I love this one.’ ‘A s for me, I love eating and drinking.’ (Pl., Poen., 313) Note that this is a restricted domain of Latin grammar where UI and UT pattern together from a morphological point of view. This behaviour is radically inconsistent with a canonical nominative–accusative system, which strives to neatly contrast subjects and objects. In the light of what we have observed so far, I conclude that Latin infinitives only realize one semantic macrorole, the undergoer, thus resulting in a direct mapping of semantic roles (and not in their neutralization) in the morphosyntax of the language. In other words, their distribution obeys a semantically based pattern of alignment.

7.2.3 Comparing infinitives and deverbal action nouns Given that animacy is often implied in semantically based alignments, one could be tempted to hypothesize that it is involved in determining the distribution of the infinitives, which cannot but refer to inanimate notions. This section tests this assumption by contrasting infinitives with a different nominalizing strategy, namely deverbal action nouns, showing that animacy is irrelevant in accounting for their distribution. Unlike infinitives, suffixed action nouns do not show any restriction in their distribution.⁸ In the following examples, the deverbal nouns oppugnatio ‘attack’ ⁷ Indo-European languages show a clear-cut distinction between nominal and verbal morphology. Infinitives can therefore be said to be atypical items from a lexical and morphosyntactic point of view, displaying both nominal and verbal features, and being used in predicative as well as in argumental slots with no overt morphological derivation (Wackernagel 1928: 259). ⁸ To be sure, an action noun may also refer to a concrete manifestation or result of a process (functioning therefore as nomen rei actae). Notice, however, that in the examples provided here the action

INFINITIVES AND SUBJECTHOOD

229

and cogitatio ‘thought’ encode the subject arguments of the accomplishments vicit ‘it won’ and excitavit ‘it moved’, i.e. their AT : (11) a. oppugnatio … vicit pertinaciam attack.NOM.F.SG win.IND.PRF.3SG obduracy.ACC.F.SG Macedonum Macedonian.GEN.M.PL ‘The attack defeated the Macedonians’ obduracy.’ (Liv., Ab Urbe Cond., 32, 15, 2) b. te … alia cogitatio … excitavit 2SG.ACC another.NOM.F.SG thought.NOM.F.SG move.IND.PRF.3SG ‘A different thought moved you.’ (Cic., Planc., 55, 10–11) The data in (11) show that the constraint at work with the infinitives does not hold for other kinds of nominalizations, as is the case with deverbal action nouns, which can express both the actor and the undergoer macroroles. They are assigned the grammatical relation subject, as conceived in a standard nominative–accusative system. Crucially, a comparison between the distribution of abstract deverbal nouns and that of infinitives proves that while the former behave as every other lexical item provided with the nominative case in a nominative–accusative language, the latter manifest a (local) semantically oriented alignment, the contrast thus resulting in a split in the verbal classes: unaccusative predicates admit infinitives as subjects, while the AI argument of unergatives and the AT argument of biargumental predicates can only be realized by deverbal suffixed action nouns. Now, it is worth noting that the infinitives do not differ from action nouns as for the feature of animacy, since they both refer to inanimate, abstract notions; rather, infinitives differ from deverbal nouns in that they cannot be actors. The irrelevance of animacy to the grammatical relation subject is probably inherited from the Proto-language, where, according to Luraghi (1995), both animate and inanimate entities could function as agentive subjects (see, however, Chapter 2 in this volume for a different perspective). It is of interest that in the languages under investigation, on the contrary, biargumental predicates normally allow inanimate, indefinite, abstract, and non-referential subjects. In a word, Kuryłowicz’s (1964: 158) statement (‘The overall distinction between the two categories [infinitives and verbal abstracts] ought to be established on the basis of their syntactical behaviour’) is confirmed by my analysis. Finally, taking a broader view, according to Bauer (2000: 85, 337–349), in many ancient Indo-European languages the nominal forms of the verb seem to be a distinctive aspect of grammar with their own syntax, not always fitting the canonical nominative–accusative alignment. nouns retain a full verbal value, in that they denote abstract processes (and not concrete manifestations, nor results).

230

CL AUDIA FABRIZIO

7.2.4 Interim conclusions From the perspective taken in this chapter, the distribution of the infinitives fits with a coding pattern whereby AT and AI are coded alike (i.e. as deverbal action nouns), differently from undergoers (i.e. as infinitives). To put it differently, infinitives do not neutralize the actor and the undergoer macroroles, but only realize the latter. A coding pattern which contrasts actors and undergoers is inconsistent with a nominative–accusative system. In a nominative–accusative system, the only argument of an intransitive predicate is coded with the nominative case, regardless of its semantic macrorole, in order to provide the sentence with the grammatical relation subject. This implies that both the actor and the undergoer are equally eligible for subjecthood with monovalent predicates. The morphosyntax of the language consistently contrasts subjects and objects, while the semantic opposition between actors and undergoers is neutralized. In a semantically oriented system, on the other hand, arguments are marked according to their position in the logical structure of the predicate. More exactly, active–inactive alignment is a kind of coding system whereby arguments are flagged according to whether they are actors or undergoers.⁹ The distribution of Latin infinitives, instantiating a scheme of split intransitivity, suggests a semantically based pattern of alignment in the domain of nominalizations. The main types of argument alignment are represented in Table 7.4. Table 7.4 Infinitive subjects instantiating active-inactive alignment in Latin

⁹ See Donohue (2008) for a general outline of semantic-alignment-related phenomena in a large number of languages. I also refer to Rovai (2012: 51–52) for an exhaustive survey of the typological literature on semantic alignment and a rich exemplification from different languages. It is debated whether semantic alignment is a consistent linguistic type, or, rather, an epiphenomenal feature that emerges only in some language, resulting from the direct coding of semantic macroroles in the morphosyntax of the language (see Chapter 1 in this volume). The discussion is not relevant, however, for the sake of my argument, since I simply maintain that Latin simply shows some patterns of semantic alignment, being as for the rest a nominative–accusative language.

INFINITIVES AND SUBJECTHOOD

231

It is widely recognized that languages generally display multiple splits in their argument realization, in that some patterns might follow a peripheral coding scheme and, to some extent, deviate from the prevalent alignment (Nichols 2008). To put it differently, alignment is construction-specific, rather than languagespecific.

7.3 Old Italian Modern Italian commonly allows infinitives to be the subjects of finite biargumental predicates denoting accomplishments (Vanvolsem 1981: 14–15, Skytte and Salvi 1991: 483–485, 538 ff., Kerleroux 1990, Simone 2003, Pérez Vázquez 2007, Fabrizio 2017): (12)

a. Fumare gli ha danneggiato i polmoni smoke.INF 3SG.DAT have:PRS.3SG damage.PP DEF.ART lungs ‘Smoking has damaged his lungs.’ b. danneggiare damagedʹ (y)]

‘damage’ do ́ (x, [damage ́ (x, y)]) & [BECOME

The following sections tackle the diachronic path leading to the syntactic change which took place between Latin and Italian, concerning the distribution of subject infinitives.

7.3.1 The data To begin with, let us consider whether the syntax of subject infinitives in Old Italian (more precisely, Old Tuscan, 1200–1370 CE)¹⁰ still reflects the Latin distribution. The following examples show the contexts allowing a subject infinitive in Old Italian; once again, they are grouped according to the Aktionsart of the finite predicate of the clause, the finite predicate is in bold and the subject infinitive is underlined: (13)

Stative predicates intrar a. M’ è uopo intrar ne l' 1SG.ACC be.IND.PRS.3SG need enter.INF into DEF.ART aringo rimaso competition left ‘It is necessary for me to face the pending trial.’ (Dante, Pd., I, 18) versare versare b. Molte parole è mattezza many.PL word.PL waste.INF be.IND.PRS.3SG folly ‘It is foolish to waste many words.’ (Alb. da Brescia, Tratt. Dilez., 26, 170)

¹⁰ The Italian corpus has been built up through the corpus M.I.DIA (=Morfologia dell’Italiano in Diacronia), www.corpusmidia.unito.it.

232

CL AUDIA FABRIZIO c. Più grave è male dato dato, avere avere dato more reproachable be.IND.PRS.3SG have.INF badly given che niente avere ricevuto than nothing have.INF received ‘It is more reproachable to give inappropriately, than to receive nothing.’ (Fiore di fil., 57)

(14)

Achievements (change of state/place) a. Né m’ è cura perché il mio NEG 1SG.ACC be.IND.PRS.3SG care so that DEF.ART my agli uomini non pervenga parlare parlare speak.INF to man.PL NEG arrive.SBJ.PRS.3SG ‘I do not care that my words will not reach men.’ (Boccaccio, Eleg. Madonn., 2) b. E quando 'l mi ven maginar maginar CONJ when DEF.ART imagine.INF 1SG.DAT come.IND.PRS.3SG ben fiso, … well fixed ‘And when the imagining becomes fixed in me,…’ (Dante, Vita Nuova, XXXI, 6) (15) Passive predicates a. E non serà smarruto lo mio acquistar acquistar CONJ NEG be.IND.FUT.3SG lost DEF.ART my acquire.INF ‘And my acquire will not be lost.’ (B. Orbicciani, Poesie, II, 25–26) b. Molto fu da ciascuna delle donne e greatly be.IND.PRF.3SG by every of.the woman.PL CONJ degli uomini il di madonna parlar parlar of.the man.PL DEF.ART speak.INF of lady Oretta lodato Oretta praised ‘Lady Oretta’s words were greatly praised by every woman and every man.’ (Boccaccio, Decam., VI, 2) (16)

Activities a. V’ ha sì punto come dimostra 2PL have.IND.PRS.3SG to that point as show.IND.PRS.3SG 'l vostro buon trovare trovare DEF.ART your good create rhymes.INF ‘It wounded you so deeply, as your rhymes show.’ (Dante, Rime, XVI, 11) pugnare onora b. Ben pugnare well fight.INF give-honour.IND.PRS.3SG ‘To fight well gives honour.’ (Guittone, Rime, VIII, 29)

INFINITIVES AND SUBJECTHOOD (17)

233

Accomplishments a. ché troppo fa gli òmini amare amare since too much love.INF make.IND.PRS.3SG DEF.ART man.PL stolti stupid.PL ‘Since loving too much makes men become stupid.’ (C. Angiolieri, Rime, 67, 12) b. il cui m’ inonda parlar parlar DEF.ART REL.GEN speak.INF 1SG.ACC inundate.IND.PRS.3SG e scalda CONJ heat.IND.PRS.3SG ‘Whose words inundate and heat me’ (Dante, Pd., IV, 119–120) partir c. M’ ancide il doloroso partir 1SG.ACC kill.IND.PRS.3SG DEF.ART leave.INF painful ‘This painful departure kills me.’ (O. da Bologna, Rime, IV, 4–5) Pianger di sospirar d’ angoscia mi d. Pianger doglia e sospirar cry.INF of sorrow CONJ sigh.INF of anguish 1SG.ACC strugge 'l core consume.IND.PRS.3SG the heart ‘To cry out in sorrow and to sigh of anguish are consuming my heart.’ (Dante, Vita Nuova, XXXI, 7)

Unsurprisingly, in Old Italian subject infinitives may well appear with stative finite predicates, as those in (13): m’è uopo ‘it is necessary’, è mattezza ‘it is foolish’, or grave è ‘it is reproachable’. Like in Latin, they can also appear with achievements (examples in (14)), as pervenga ‘it arrives’, and passive predicates, as serà smarruto ‘it will be lost’ in (15a) and fu lodato ‘it was praised’ in (15b). Unlike Latin, nevertheless, Old Italian also licenses subject infinitives with finite predicates belonging to activity and accomplishment Aktionsarten. Evidence is provided by the examples in (16) and (17). Let us concentrate on accomplishments. Look how some highly transitive, causative predicates like fa stolti ‘make stupid’ (17a) or inonda ‘it inundates’ and scaldare ‘it heats’ (17c), or ancide ‘it kills’ (17e) can all take subject infinitives. We can conclude that already in Old Italian infinitives are eligible to subjecthood regardless of the Aktionsart of the finite verb of the clause they belong to.¹¹ Interestingly, determiners are not needed for infinitives to be the subjects of biargumental finite predicates: ¹¹ The impression of an increased frequency of nominal infinitives in the Romance languages, with respect to Latin (Vanvolsem 1981: 13), is then simply due to the enlargement of the Aktionsart classes allowing infinitives as subjects.

234

CL AUDIA FABRIZIO

(18)

Lo dolce parlare isvellia e nutrica parlare the sweet speak.INF awake.IND.PRS.3SG CONJ feed.IND.PRS.3SG amore love ‘A sweet discourse awakes and feeds love.’ (A. da Brescia, Tratt. Dilez., II) b. Dolce parlare acquista e conserva parlare sweet speak.INF procure.IND.PRS.3SG CONJ preserve. IND.PRS.3SG amore love ‘A sweet discourse procures and preserves love.’ (A. da Grosseto, Volgar., II) a.

Table 7.5 shows that in a neutral alignment arguments are not differentiated at a morphological level, neither with respect to their syntactic nor semantic macrorole. The behaviour of subject infinitives in Old Italian conforms to this argument scheme.¹² Table 7.5 The behaviour of subject infinitives in Old Italian

7.3.2 The ‘bridge’ of experiencer predicates My hypothesis is that the change involving the syntactic status of nominal infinitives from Latin to Old Italian (i.e. from marked arguments to unconstrained, unmarked subjects) might have started from clauses with a finite experiencer predicate. According to Croft (1993), an experiencer is precisely characterized as being in a mental state regarding the stimulus. Experiencers have been also depicted as ‘the locus of an internal event’ (Van Valin 1993:42), and as ‘sentient beings that experience internal states, such as perceivers, cognisers and emoters’ ¹² The same actually happens with the whole lexicon in the transition from Latin to its daughter languages. When the accusative comes to mark all sort of subjects, a neutral alignment is finally achieved through the generalization of one single case. This eventually results in a neutral alignment, wherein case marking no longer expresses argument distinctions (Va¨a¨na¨nen 1966, Pensado 1986, Zamboni 1997, 2002, Cennamo 2009).

INFINITIVES AND SUBJECTHOOD

235

(Van Valin and La Polla 1997: 85; see also Van Valin and Wilkins 1996).¹³ In Bentley’s (2006: 95ff.) account, the defining property of this class of predicates is that they have an experiencer animate argument, although they can vary as to whether they show an experiencer subject, a causer subject, or a theme subject. Some Latin experiencer predicates allow a subject infinitive, playing the thematic role of stimulus, and take an experiencer (animate) argument, encoded with the dative or the accusative case, depending on the verb: (19) a. qui convenit in minore negotio how be.suitable.IND.PRS.3SG in smaller.ABL.SG affair.ABL.SG timere, legem quom eam in maiore timere, law.ACC.SG respect.INFwhile 3SG.F.ACC in greater.ABL.SG neglegeris? disregard.SBJ.PRF.2SG ‘How could it be suitable to respect the law in minor details, and to disregard it in major ones?’ (Sall., Cat., 51, 24) b. Quid tibi nunc misero prodest how 2SG.DAT now miserable.DAT be.helpful.IND.PRS.3SG grave carmen? dicere dicere serious. ACC.SG recite.INF poem.ACC.SG ‘How could you, miserable man, benefit from reciting a serious poem now? (lit.: How could reciting a serious poem be helpful for you now?).’ (Prop., Eleg., 1, 9, 9–10) prodire (20) a. Non te pudet in prodire NEG 2SG.ACC feel.ashamed.IND.PRS.3SG appear.INF in conspectum meum? view.ACC.SG POSS.1SG.ACC.SG ‘Don’t you feel ashamed to come into my view?’ (Pl., Men., 708) b. Oratorem minime irasci decet irasci speaker.ACC.SG not.at.all get.angry.INF be.becoming.IND.PRS.3SG ‘It is unbecoming for a speaker to get angry.’ (Cic., Tusc., 4, 55, 1) Some oblique argument-taking experiencer predicates in Old Italian are listed in (21). They all can take a subject infinitive, and they all encode the animate experiencer in the dative case. In (21a), lo tuo parlar is the subject of non mi piace (literally, ‘is not pleasant to me’).

¹³ While causative experiencer predicates (as ‘to bore’, ‘to scare’, ‘to irritate’, ‘to surprise’) have a complex event structure, non-causative ones (as ‘to love’, ‘to fear’, ‘to envy’, ‘to like’) are in nature very close to states. As for experiencer predicates in Latin, I refer to Fedriani’s (2014) comprehensive study, and to Dahl and Fedriani (2012) for a comparative (Latin, Greek, and Vedic) Indo-European approach. For a survey of patterns of experiencer predicates in modern Indo-European languages, see Bossong (1998).

236

CL AUDIA FABRIZIO

(21) Dative-taking experiencer predicates a. Lo tuo parlar parlar, rispuosi, non mi parlar DEF.ART your speak.INF reply.IND.PRF.1SG NEG 1SG.DAT piace be.pleasant.IND.PRS.3SG “‘I don’t like”, I replied, “your discourse” (lit.: Your speaking is not pleasant to me).’ (F. degli Uberti, Ditt., 4, 80) Mangiar conviene b. Mangiar all’ uomo acciò eat.INF be.essential.IND.PRS.3SG to.the man in.order che viva that live.SBJ.PRS.3SG ‘It is essential for a man to eat, in order to live.’ (B. Giamboni, Fiore, 27) rimembrar c. Questo è 'l che più rimembrar this be.IND.PRS.3SG the remember.INF REL the most mi dole 1SG.DAT hurt.IND.PRS.3SG ‘This is the memory which hurts me the most.’ (Petrarca, Rime Extr., 19, 14) pingere d. E pure li dispiace lo pingere che and yet he.DAT be.unpleasant. IND.PRS.3SG the paint.INF REL face make.IND.PRS.3SG ‘And yet the picture that he paints saddens him.’ (G. da Lentini, Poesie, 1, 43–44) mordere altrui motteggiando e. Non le si disdisse il mordere NEG she.DAT be.unbecoming.IND.PRF.3SG ‘To castigate people by teasing them was not unbecoming to her.’ (Boccaccio, Decam., VI, 3) Object-taking experiencer predicates are also well attested with subject infinitives. In (22a), Vostro plangere m’allegra (‘your crying makes me happy’), vostro plangere ‘your crying’ is the subject infinitive of m’allegra ‘makes me happy’, and me is a direct object. The same holds for (22b) and (22c). (22) Object-taking experiencer predicates plangere (…) m’ a. Vostro plangere allegra your cry.INF 1SG.ACC make.happy.IND.PRS.3SG ‘Your crying makes me happy.’ (D. da Maiano, Rime, 41, 22) b. Mica mi spaventa l’ amoroso NEG 1SG.ACC frighten.IND.PRS.3SG DEF.ART loving volere volere desire.INF ‘The loving desire does not frighten me.’ (G. da Lentini, Poesie, 3, 55)

INFINITIVES AND SUBJECTHOOD

237

c. Mi diletta alla singulare venire venire 1SG.ACC delight.IND.PRS.3SG come.INF to.the unique grandezza della tua felicitade hugeness of.the your happiness ‘To deal with your happiness, so uniquely huge, delights me.’ (Alb. della Piag., Boez. volg., II, 3) Now, Latin experiencer predicates such as those of the pudet type,¹⁴ selecting an accusative experiencer, might have undergone a transitive reanalysis, whereby the infinitive acts as AT and the animate argument as UT . Crucially, already in the Early and Classical language, some of the verbs entering the pudet-type construction can also surface with personal patterns, with the stimulus encoded by the nominative case (Fedriani 2014: 142–149); the transitive pattern finally prevails in Late Latin (Fedriani 2014: 177–183); I quote the following example from her list: (23) a. Non te haec pudent? NEG 2SG.ACC DEM.NOM.PL. shame.IND.PRS.3PL ‘Don’t these things shame you?’ (Ter., Ad., 754) b. Quod me minime paeniteat REL.NOM.SG 1SG.ACC very.little repent.SBJ.PRS.3SG ‘This may cause very little repentance to me.’ (Cic., Ad Att., 13, 28, 2) Personal patterns whereby the experiencer is encoded by the nominative case are also attested, although more rarely (Fedriani 2014: 178): (24)

Ipse sui miseret he.NOM REFL.GEN take pity.IND.PRS.3SG ‘He takes pity on himself.’ (Lucr., De rerum Nat., 3, 881)

The pattern in (25) does not override the stimulus-as-subject one. It is likely that dative-taking experiencer predicates provide and reinforce the constructional pattern whereby the stimulus, and not the experiencer, is encoded as subject. See what happens with piacere ‘to be pleasant’, which both in Latin and in Italian only admits stimuli (occasionally represented by infinitives) as subjects: (25) a. Placuit caeleste numen et precari precari seem.advisable.PRF.3SG divine.ACC.SG invoke.INF will.ACC.SG CONJ quaerere quaerere auxilium per sacras sortes help.ACC.SG through holy.ACC look-for.INF oracle.ACC.PL ‘It seemed advisable to invoke the god’s will and to look for help by means of holy oracles.’ (Ov., Met., 1, 367–368) ¹⁴ Since it may select an accusative experiencer, decet ‘it is becoming’ is part of this group of predicates as well (see Chapter 6 this volume).

238

CL AUDIA FABRIZIO b. Il gelato piace a tutti the ice.cream please.IND.PRS.3SG to everybody ‘Everybody likes ice cream.’ c. *Tutti piacciono il gelato Everybody please.IND.PRS.3PL the ice ream *‘Ice cream likes to everybody’ (lit.: *‘Everybody pleases ice cream’) d. Sciare piace ski.INF please.IND.PRS.3SG ‘Everybody likes to ski.’

a tutti to everybody

e. Sciare diverte ski.INF amuse.IND.PRS.3SG ‘Skiing amuses everybody.’

tutti everybody

It is not by chance, perhaps, that in Old Italian, in about 60 per cent of the contexts with infinitives employed as subjects of biargumental finite predicates, the second argument happens to be an animate experiencer, as those in (26). This amounts to saying that the majority of biargumental contexts with a subject infinitive select experiencer predicates: poter (26) a. Lo non mi nonpoter the NEG-be.able.INF 1SG.ACC ‘To be unable upsets me.’

turba upset.IND.PRS.3SG (G. da Lentini, Poesie, I, 41)

b. Lo meo gran sospirare porrìa sospirare vi the my big sigh.INF 2PL.ACC can.COND.PRS.3SG certa fare de l’ amorosa flamma certain make.INF of the amorous flame ‘My sighs could make you certain of my love.’ (G. delle Colonne, Rime, 6, 30–31) pensare c. Tanto m' angoscia 'l prefondo pensare so much 1SG.ACC anguish.IND.PRS.3SG the deep think.INF ‘My deep thoughts anguish me so much.’ (G. Guinizzelli, Rime, IX, 7) To sum up, experiencer clauses might have been the starting point of the gradual expansion of subject infinitives to the domain of canonical transitivity. The grey areas in Table 7.6 iconically represent the experiencer constructions with infinitive-as-stimulus as the bridge towards the domain of biargumental accomplishment clauses with subject infinitives, i.e. the novelty of Romance infinitival syntax.

INFINITIVES AND SUBJECTHOOD

239

Table 7.6 Experiencer constructions with infinitive as stimulus in Latin and Old Italian

7.3.3 Differential marking of subject infinitives An interesting coding property¹⁵ of infinitives in Old Italian lies in an optional differential marking of postverbal subjects according to their semantic macrorole. It has already been noted that when the subject infinitive is postverbal, three options are available: it can be a bare infinitive, or an infinitive marked by the prepositions di or a (Ageno 1964, Ageno 1978: 276–278, Egerland 2010a: 822–826): (27) a. Ma è buono seminare seminare but be.IND.PRS.3SG good seed.INF ‘But it is good to seed.’ (Pucci, Libro, 2, 18) b. …a la quale era malagevole d’ d’ to the REL be.IND.IMPF.3SG hard PREP li versi latini intendere intendere understand.INF the verse.PL Latin.PL ‘For whom it was hard to understand Latin verses.’ (Dante, Vita Nuova, XXV, 1) ¹⁵ I assume here that coding properties identify grammatical relations by dint of case marking, verbal agreement, and word order (Keenan 1976, Haspelmath 2001, 2010).

240

CL AUDIA FABRIZIO c. Più beata cosa è che a aa dare dare more happy thing be.IND.PRS.3SG PREP give.INF than PREP tòrre take.INF ‘It is even better to give than to take.’ (Fiore di virtù, XI)

In (27a), the postverbal subject seminare ‘to seed’ is a simple infinitive; in (27b), intendere ‘to understand’ is a postverbal infinitive subject marked by di; in (27c), under the same circumstances, dare ‘to give’ is marked by a. The rationale underlying the alternation of di and a seems rather unclear. The same author, in the same work, can adopt all of the three strategies alternatively: (28)

a. Male è con coloro che aa parlare parlare bad be.IND.PRS.3SG PREP speak.INF with those REL parlano più a volontà che a ragione speak.IND.PRS.3PL more To caprice than to reason ‘It is a bad thing to speak to those who speak according to their caprice, rather than to their reason.’ (Fiore di fil., 43) b. Propia cosa è del savio esaminare esaminare i typical thing be.IND.PRS.3SG of.the wise examine.INF the consigli advice.PL ‘It is typical of a wise man to examine advice’ (Fiore di fil., 48) c. Feminile cosa è il litigare e contendere litigare contendere feminine thing be.IND.PRS.3SG the quarrel.INF CONJ fight.INF ‘It is a feminine thing to quarrel and to fight.’ (Fiore di fil., 41)

In a larger comparison, however, let us observe that the preposition di encodes both the undergoer of an intransitive clause (UI ) and the undergoer of a transitive pattern (UT ), while a can only mark the former (UI ). In (29a), acquistare ‘to acquire’ is the intransitive subject of abisogna ‘it is needed’. In (29b), audire ‘to hear’ is the direct object of sofferino ‘they bear’. Both can be marked by di. The intransitive subject of meravigliosa cosa è ‘it is a wonderful thing’ can also occasionally be marked by a: (29) a. Abisogna dd’’ acquistare acquistare benivolenzia be.needed.IND.PRS.3SG PRP acquire.INF sympathy ‘To acquire sympathy is needed.’ (UI ) (B. Latini, Rett., 162) b. …che mi sofferino dd’’ audire audire that 1SG.ACC bear.SBJ.PRS.3PL PRP hear.INF ‘That they bear to hear me.’ (UT ) (Dante, Vita Nuova, VII, 7)

INFINITIVES AND SUBJECTHOOD

241

c. Maravigliosa cosa è aa pensare pensare quanto wonderful thing be.IND.PRS.3SG PREP think.INF how sieno difficili a investigare le forze d' be.SBJ.PRS.3PL difficult PREP investigate.INF the force.PL of amore love ‘It is wonderful to think how the forces of love can be difficult to be investigated.’ (UI ) (Boccaccio, Decam., IV, 8) What counts more, the actor subject (AT , AI ) is left unmarked, as examples in (17)–(18) clearly show. This mechanism of differential marking has escaped scholars’ attention so far. Although irregularly, it serves to contrast actors and undergoers, hence conforming to a semantically oriented pattern (Table 7.7). Interestingly, the differential marking of subject infinitives interacts with word order, in that it persists with postverbal subjects. It then gradually gets lost. However, the postverbal preposition di can still be found with some specific predicates in Modern Italian to encode undergoers.¹⁶ In (30a), the postverbal infinitive essere felici ‘to be happy’ is the subject of succede ‘it happens’, and is marked by di. In (30b), the infinitive far visita ‘to visit’ is the second argument of sopporterò (‘I shall bear’), and, again, it is marked by di. di essere felici felici di essere (30) a. Succede all’improvviso happen.IND.PRS.3SG PRP be.happy.INF suddenly ‘It suddenly happens to be happy.’ (UI ) visita ai b. Sopporterò tuoi di far di far visita bear.IND.FUT.1SG PRP visit.INF to.the your ‘I shall bear to visit your parents.’ (UT )

genitori parents

It must be stressed that in Modern Italian, the encoding with di of infinitives playing the macrorole of UT and UI is lexically constrained, and much less frequent

Table 7.7 Differential marking of infinitival subjects in Old Italian active–inactive

AT AI

UT UI optionally marked by di, a

¹⁶ See Fabrizio (2017). I refer to some predicates taking object infinitives marked by di, such as finire di (‘to cease’) or smettere di (‘to stop’). Moreover, some nominal phrases are preceded by di when surfacing in a postadjectival position, functioning as the only argument of an (implicit) stative predication, as in quello sciocco di Luigi (lit.: ‘that stupid of Luigi’) (Kayne 1994, Moro 2010: 217–218).

242

CL AUDIA FABRIZIO

than in Old Italian. As far as I know, the above-mentioned residual traces of dimarking of subject infinitives have never been traced back to an earlier, tentative differential marking coding pattern, contrasting actors and undergoers.

7.4 Conclusions and further research The transition from Latin to Italian, concerning the syntax of subject infinitives, involves the loss of the constraints governing the distribution of such non-canonical subjects in Latin, and the concomitant expansion towards all kind of constructions. While the syntax of subject infinitives in Latin obeys an active–inactive coding scheme, the above-mentioned diachronic change develops alongside the emergence of a neutral alignment, whereby arguments are not differentiated by morphological devices according to their semantic macroroles. Experiencer predicates have been shown to be a plausible ‘bridge’ leading to a gradual expansion of subject infinitives. In Old Italian, they are the most representative ones, from a quantitative point of view, allowing a subject infinitive. Finally, there is some evidence of a differential marking of postverbal subject infinitives in Old Italian. Further research should in my view address the syntax of subject infinitives in Late and Vulgar Latin, and in non-Tuscan varieties, since non-Classical Latin and non-standard Old Italoromance vernaculars, as well as other early Romance languages, might provide evidence for different starting points, and different developmental paths.

Acknowledgements I would like to thank the reviewers who commented upon an earlier draft of this chapter. Their suggestions greatly improved its overall quality. Actually, some of the issues they raised were wide-ranging enough to justify (at least) another paper, and thus will have to be dealt with in future work.

Sources A. da Brescia, Tratt. Dilez. = Il trattato della dilezione d’A lbertano da Brescia, Arrigo Castellani (ed.), Firenze, Accademia della Crusca, 2012. A. da Grosseto, Volgar. = Andrea da Grosseto, Volgarizzamento del Liber de doctrina loquendi et tacendi di Albertano da Brescia in L’edizione del volgarizzamento di Andrea da Grosseto del Liber de doctrina loquendi et tacendi di Albertano da Brescia e il codice G, L. Guia (ed.), Roma, Biblioteca Italiana, 2005.

INFINITIVES AND SUBJECTHOOD

243

Alb. della Piag., Boez. volg. =“Il Boezio” nella traduzione trecentesca di Alberto della Piagentina, S. Battaglia (ed.), Torino, Utet, 1929. B. Giamboni, Fiore = Bono Giamboni, Fiore di rettorica, G. Speroni (ed.), Pavia, Universita degli studi, Dipartimento di scienza della letteratura e dell’arte medioevale e moderna, 1994. B. Latini, Rett. = La Rettorica di Brunetto Latini, Francesco Maggini (ed.), Stab. Galletti e Cocci, Firenze 1915 B. Orbicciani, Poesie = Bonagiunta Orbicciani, Poesie, in Rimatori siculo-toscani del Dugento, G. Zaccagnini and A. Parducci (ed.), Bari, Laterza, 1915. Boccaccio, Decam. = Giovanni Boccaccio, Decameron, in Tutte le opere di G. Boccaccio, V. Branca (ed.), IV, Milano, Mondadori, 1976. Boccaccio, Eleg. = Giovanni Boccaccio, Elegia di Madonna Fiammetta, in Tutte le opere di G. Boccaccio, G. Padoan (ed.), V, II, Milano, Mondadori, 1994. C. Angiolieri, Rime = Cecco Angiolieri, Le rime, A. Lanza (ed.), Roma, Archivio Guido Izzi, 1990. Cat., Orig. = Marcus Porcio Cato, Origines. In: Historicorum Romanorum Reliquiae. Vol. 1, H. Peter (ed.), Leipzig, Teubner, 1914. Cels., Cels., Med. = A. Cornelii Celsi De medicina libri octo, C. Daremberg (ed.), Leipzig, Teubner, 1891. Cic., Ad Att. = M. Tullius Cicero, Epistulae ad Atticum, in Cicero’s Letters to Atticus, D. R. Shackleton Bailey (ed.), Leipzig, Teubner, 1965–1968. Cic., Brut. = M. Tullius Cicero, Brutus, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, E. Malcovati (ed.), Leipzig, Teubner, 1970. Cic., De Off. = M. Tullius Cicero, De Officiis, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, C. Atzert (ed.), Leipzig, Teubner, 1932. Cic., Planc.= M. Tullius Cicero, Pro Plancio, in M. Tulli Ciceronis Orationes, A. C. Clark (ed.), Oxford, Clarendon Press, 1911, vol. VI. Cic., Tusc. = M. Tullius Cicero, Tusculanae Disputationes, in M. Tulli Ciceronis Scripta Quae Manserunt Omnia, M. Pohlenz (ed.), Oxford, Oxford Classical Texts, 1918. Cic., Ver. = M. Tullius Cicero, In Verrem, in M. Tulli Ciceronis Orationes, W. Peterson (ed.), Oxford, Oxford Classical Texts, 1917, vol. III. D. da Maiano, Rime = Dante da Maiano, Rime, R. Bettarini (ed.), Firenze, Le Monnier, 1969. Dante, Pd. = Dante Alighieri, La Commedia secondo l’antica vulgata. Paradiso, G. Petrocchi (ed.), Firenze, Casa Editrice Le Lettere, 1994. Dante, Rime = Dante Alighieri, Rime, G. Contini (ed.), Torino, Einaudi, 1965. Dante, Vita Nuova = Dante Alighieri, Vita nuova, M. Barbi (ed.), Firenze, Le Monnier, 1932. F. degli Uberti, Ditt. = Fazio degli Uberti, Il Dittamondo e le Rime, G. Corsi (ed.), Bari, Laterza, 1952.

244

CL AUDIA FABRIZIO

Fiore di fil. = Fiore di filosofi e di molti savi, Testo in parte inedito, citato dalla Crusca e ridotto a miglior lezione da Antonio Cappelli, presso Gaetano Romagnoli, Bologna, Regia Tipografia, 1865. Fiore di virtù = Fiore di virtù, A. Gelli (ed.), Firenze, Le Monnier, 1856. G. da Lentini, Poesie = Giacomo da Lentini, Poesie, R. Antonelli (ed.), Roma, Bulzoni, 1979. G. delle Colonne, Rime = Guido delle Colonne, Rime, in Poeti italiani della Corte di Federico II, B. Panvini (ed.), Napoli, Liguori, 1994. G. Guinizzelli, Rime = Guido Guinizzelli, Rime, L. Rossi (ed.), Torino, Einaudi, 2002. Guittone, Rime = Guittone d’Arezzo, Le Rime, F. Egidi (ed.), Bari, Laterza, 1940. Liv., Ab Urbe Cond. = T. Livius, Ab Urbe Condita, books 1–5: R. S. Conway and C. F. Walters (ed.), 1955; books 6–10, 21–25: C. F. Walters and R. S. Conway (ed.), 1919–1950; books 26–30: R. S. Conway and S. K. Johnson (ed.), 1953; books 31–35: A. H. McDonald (ed.), 1969; books 36–40: W. Weissenborn and M. Mu¨ller (ed.); books 41–45: W. Weissenborn and W. Heaeus (ed.), 1908, Oxford, Clarendon Press. Lucr., De rerum Nat.: T. Lucretius Carus, De Rerum Natura Libri Sex, J. Martin (ed.), Leipzig, Teubner, 1969. Mart., Ep. = M. Valerii Martialis Epigrammaton Libri, W. Heraeus and I. Borovskij (ed.), Leipzig, Teubner, 1982. O. da Bologna, Rime = Le rime di Onesto da Bologna, S. Orlando (ed.), Firenze, Sansoni, 1974. Ov., Met.: P. Ovidius Naso, Metamorphoses, in Metmorphoses in Two Volumes, F. J. Miller and G. P. Goold (ed.), London, Loeb Classical Library, 1977–1984. Petr., Sat. = Petronius, Satyrica, K. Mu¨ller and W. Ehlers (eds), Berlin, de Gruyter, 1983. Petrarca, Rime Extr. = Francesco Petrarca, Rime Extravaganti, in Id., Opera omnia, P. Stoppelli (ed.), Roma, Lexis Progetti Editoriali, 1997. Pl., Epid.= T. Maccius Plautus, Epidicus, in Plauti Comoediae, F. Leo (ed.), Berlin, Weidmann, 1895, vol I. Pl., Men. = T. Maccius Plautus, Menaechmi, in Plauti Comoediae, F. Leo (ed.), Berlin, Weidmann, 1895, vol I. Pl., Poen.= T. Maccius Plautus, Poenulus, in Plauti Comoediae, F. Leo (ed.), Berlin, Weidmann, 1895, vol II. Pl., Ps. = T. Maccius Plautus, Pseudolus, in Plauti Comoediae, F. Leo (ed.), Berlin, Weidmann, 1895, vol II. Pl., Trin. = T. Maccius Plautus, Trinummus, in Plauti Comoediae, F. Leo (ed.), Berlin, Weidmann, 1895, vol II. Prop., Eleg.=: Sex. Propertius, Elegiae, in Elegies, G. P. Goold (ed.), London, Loeb Classical Library, 1990.

INFINITIVES AND SUBJECTHOOD

245

Pucci, Libro = Antonio Pucci, Libro di varie storie, A. Varvaro (ed.), Palermo, Accademia di Scienze, Lettere ed arti, 1957. Quint., Decl. Min. = M. Fabius Quintilianus, The Minor Declamations Ascribed to Quintilian, M. Winterbottom (ed.), Berlin, de Gruyter, 1984. Sall., Cat. = C. Sallustius Crispus, Catilinae Coniuratio, in C. Sallusti Crispi Catilina, Iugurtha, Fragmenta Ampliora, A. Kurfess (ed.), Leipzig, Teubner, 1957. Ter., Ad.: P. Terentius Afer, Adelphoe, in P. Terenti Afri Comoediae, R. Kauer and W. M. Lindsay and O. Skutsch (ed.), Oxford, Oxford Clarendon Press, 1958. Verg., Aen.: P. Vergilius Maro, Aeneis, in P. Vergili Maronis Opera, R. A. B. Mynors (ed.), Oxford, Clarendon Press, 1972.

8 Alignment changes with Spanish experiential verbs Chantal Melis

8.1 Introduction Experiential verbs across languages are known to be particularly versatile from a syntactic point of view. As discussed in many studies, the reason for the unusual grammatical behaviour of these verbs has to do with the ambivalent properties of the central participant on which they predicate. The entity experiencing what the verb denotes, indeed, is typically a human being, similar in this respect to the agentive subject of other verbal classes, but associated with a notion of affectedness, which is characteristic of participants cast in the role of patientive undergoers. Although constructions involving experiencers follow the general rules of the grammatical organization of a language in many instances, it happens to be the case that experiential types of events exhibit a certain tendency to correlate with minor alignment patterns (Nichols 2008: 129). All languages have syntactic templates, specific to individual verbs or verbal sets, which deviate from the prevailing model of transitive or intransitive argument coding (Bickel and Nichols 2009), and semantic experiencers are likely attractors of such ‘deviations’. It is plausible to assume that communicative needs stimulate the search for experiencer-coding alternatives. Speakers are interested in expressing nuances of meaning which they feel cannot be rendered with the conventional structures. The dual figure of the experiencer presents a challenge and the avenues of resolution call for something beyond the categorical opposition of subject and object. In this chapter I focus on a particular class of experiencer verbs in Spanish. The lexical items under study, such as alegrar ‘make happy’, asustar ‘frighten’, entristecer ‘sadden’, enfurecer ‘infuriate’, irritar ‘irritate’, preocupar ‘worry’, or regocijar ‘delight’ are psychological predicates, denoting emotions or feelings, which in their basic valence frame assign the subject function to the stimulus of the mental state and encode the experiencer as direct object:

Chantal Melis, Alignment changes with Spanish experiential verbs. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl. Oxford University Press. © Chantal Melis (2022). DOI: 10.1093/oso/9780198857907.003.0008

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

247

(1) a. La boda alegró a muchos. the wedding.NOM gladden.PFV.3SG ACC many ‘The wedding made many happy.’ (21st c., CORPES) b. El aspecto d-el hombre lo asustó. the look.NOM of-the man he.ACC.SG frighten.PFV.3SG ‘The man’s looks frightened him.’ (21st c., CORPES) With these predicates (e.g. English X frightens Y ), the accent is put on the emotion engendered in the experiencer, as a result of the action of the stimulus (Talmy 1985), and in the Hispanic literature the members pertaining to this class of psychological verbs are referred to as ‘emotional causatives’ (Cano Aguilar 1981). Object experiencer verbs of this type may be viewed as giving maximal expression to the phenomenon of semantic ‘inversion’ which according to Bossong (1998) underlies all experience events, in the sense that, contrary to the prototypical transitive scenario in which the action is initiated by a human participant and is transferred to a second participant which may or may not be animate, in mental experiences the prominent, obligatory, and typically human participant, instead of functioning as instigator, is the target or endpoint of the process designated by the verb. Nevertheless, as observed by Malchukov (2008), it is precisely because object experiencer verbs assign the position of targeted patient to the human protagonist, against the expected alignment of human beings with the initiating subject role, that these verbs tend to be diachronically unstable. The changes they undergo may come in different forms, but the direction in which they push is constant: upgrading the experiencer constitutes the central and common goal. The Spanish verbs under study will give us the opportunity to verify this tendency. They presently display three alternative encoding patterns, documented in other languages, which help to promote the experiencer syntactically, while signalling that the foregrounded human participant, unlike the canonical agent, is affected by the event. As illustrated in (2), one of these devices is a middle voice type of construction which assigns the subject role to the experiencer (2a). Another option corresponds to a dative alignment pattern with the experiencer occupying the initial topic position (2b). Finally, reaching the pole of highest markedness, we have a phenomenon of split intransitivity, peculiar to stative–active languages and not expected to surface in a nominative–accusative one like Spanish (2c) vs (2d). (2) a. Cosme se alegró con mi visita. Cosme.NOM REFL.3SG gladden.PFV.3SG with my visit ‘Cosme was happy about my visit.’ (21st c., CORPES) b. A Rosa le alegra la noticia. DAT Rosa CL.DAT.3SG gladden.PRS.3SG the news.NOM.SG ‘Rosa is happy about the news.’ (21st c., CORPES)

248

CHANTAL MELIS c. Me alegro. REFL.1SG gladden.PRS.1SG ‘I’m glad.’ d. Me alegra. ACC/DAT.1SG gladden.PRS.3SG ‘I’m glad.’

(21st c., CORPES)

(21st c., CORPES)

The main objective of this chapter is to trace the history of the three valencyaffecting strategies developed as an alternative to the object experiencer case frame of the Spanish emotional causatives. But prior to the historical account, a global overview of the relevant constructions will be presented, with the intention of providing an adequate background for the exposition of the Spanish data. The chapter is organized as follows. In Section 8.2, I characterize the experiencer role (Section 8.2.1), and then proceed to describe the properties of the three experiencer constructions pertinent to Spanish—semantically aligned subject (Section 8.2.2), topical dative or oblique (Section 8.2.3), and middle subject (Section 8.2.4)—as they are treated in the typological literature. I review some of their universal distributional tendencies (Section 8.2.5) and comment on the historical changes in which they have been involved (Section 8.2.6). In Section 8.3 I turn to Spanish. I begin by situating the subclass of emotional causatives within the overall panorama of the language’s psychological predicates (Section 8.3.1), move to the description of their basic transitive uses (Section 8.3.2) and contrast their accusative object with the dative experiencer required by other verbs (Section 8.3.3). The historical account of the experiencer-promoting strategies comes next, focusing on the inherited reflexive middle construction (Section 8.3.4), the development of a dative alignment pattern at a later stage (Section 8.3.5), and the rise of a phenomenon of split intransitivity derived from the previous change (Section 8.3.6). Finally, a short conclusion is presented in Section 8.4.

8.2 Experiencer-coding strategies This section has a threefold aim. First, it delimits the experiencer role (8.2.1). Second, it provides a typological outline of the three main experiencer constructions found in Spanish (8.2.2–8.2.4). Third, it reviews the distribution of these three constructions in typological perspective (8.2.5) and explores their historical development (8.2.6).

8.2.1 The experiencer role Experiential verbs, as I mentioned in the introduction, have a tendency to appear in constructions which deviate from the basic coding paradigms of a given lan-

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

249

guage (Nichols 2008: 129). The complexity of the experiencer role explains this behaviour. In terms of the ‘inherent’ properties (Fedriani 2014: 18) associated with this role, the feature of obligatory animacy approximates the experiencer to canonical agents and is expected to propitiate his or her appearance in subject position (Lehmann 1991: 219). Far less clear, on the other hand, are the event-based ‘relational’ properties (Fedriani 2014: 18) used to define thematic roles, such as ‘agency’ or ‘affectedness’, which in the case of the experiencer are neither homogeneous nor transparent (for a good overview of this topic and a rich bibliography, see Fedriani 2014: 16–29). On occasion, in relation to specific constructions, scholars feel motivated to identify some ‘volitional’ or ‘controlling’ dimension. More generally perhaps, there is a sense in which experiencers (always) participate in the coming about of the described situation. According to Brekke (1988: 171), indeed, the experiencer is an individual ‘whose mental/emotional faculties are uniquely involved in the psychological state denoted by the predicate’. The labels ‘conscious’ (Givón 1984: 84) or ‘sentient’ (Dowty 1991: 573) are used to capture this particular type of involvement. More specifically, the internal process arises from the mental contact the human participant establishes with the stimulus of the experience (Langacker 1991, Croft 1993, Kemmer 1993, Ruwet 1994, Barðdal 2001), and through the means of this contact becomes ‘actively’ engaged in what it experiences (Langacker 1991: 239). At the same time, however, experiencers are considered to be ‘affected’ to some degree by the process they undergo, and thus share one feature in common with semantic patients (Givón 1984, Lehmann 1991, Croft 1993, Bossong 1998, Næss 2007). Around this relational property of the experiencer role there exists little disagreement. Yet, unfortunately, as Beavers (2011) argues, what it means for a participant to be conceived of as more or less affected by the event has suffered from the lack of a rigorous definition: the notion is intuitively connected to ideas about undergoing some change and is applied to a whole range of event types without taking into account the differences mediating between events in terms of the semantic parameter in question. All I can say, by way of responding to one of the anonymous reviewers, is that my appeals to the notion of affectedness will perpetuate the shortcomings indicated by Beavers, because of the impossibility of dealing with this interesting but delicate issue within the scope of the present chapter.¹ I will assume, alongside many other authors, that the expression of a mental condition qualifies as an instance of (intuitively defined) affectedness, and I will rely

¹ Beavers (2011) himself develops a very suggestive model of analysis which interacts with the aspectual properties of predicates and relates modalities of affectedness to the crucial concept of progression along some type of scale. As a way of testing the theory, the author chooses to focus on durative dynamic predicates, acknowledging that non-physical as well as non-scalar punctual changes (both characteristics apply to the events examined in the present chapter—see Melis 2019) are matters for future investigation.

250

CHANTAL MELIS

on common claims concerning the overt manifestation of this feature with certain voice patterns (middle vs active) and certain case markings (accusative or dative vs nominative) when I discuss the alignment phenomena, in Spanish and in other languages, which constitute my object of study. To sum up, the dual configuration of the experiencer role—active involvement and affectedness—may be seen as lying at the root of the variable and often ‘deviant’ encoding strategies which the human participant cast in this role tends to promote both across and within languages. As mentioned in the introduction, three such strategies linked to Spanish emotional experiencer verbs are the focus of the present chapter. In all three, the experiencer is given the treatment of a syntactically prominent argument, while being simultaneously endowed with a feature of affectedness. The device recruited for the purpose of conveying this feature— (i) accusative agreement, (ii) dative case, (iii) middle voice morphology—is where the essential difference between the three constructions lies. The first part of the chapter will be devoted to characterizing the alignment patterns under study, both typologically and diachronically, with a view to facilitating the subsequent analysis of the Spanish data.

8.2.2 Patient subject experiencer It is appropriate to begin with the device more closely linked to semantically aligned systems, considering that the three strategies in focus are ways of coming to terms with the complex semantics of the experiencer role. In these systems, called ‘active’ or ‘agentive’, the intransitive verbs exhibit a split in the behaviour of their subject (S): with some verbs S is cross-referenced by means of the set of forms also used for the transitive subject (A), while others mark S in the same way as the object in transitive clauses (P). Languages vary in the organization of these splits. Verbs are distributed in non-homogeneous ways between the two marking strategies and there are also many differences with respect to the size of the established verbal classes. The semantic parameters on which the splits depend are not identical either, varying as to whether the aspectual properties of the verb or the relational properties of the subject play the determining role. Yet, on the whole, it is possible to state that subjects which do not exercise control over the event and/or are construed as affected in some way tend to attract the patientive encoding (Merlan 1985, Van Valin 1990, Mithun 1991, Arkadiev 2008). Since emotional experiencers fit relatively well into these categories, they often appear with patient forms. For instance, in Lakhota, I am mad, I am happy, and I blew up in anger have the first person designated by an object pronoun (ma-, instead of agentive wa-) (Mithun 1991: 515). The following example illustrates a similar use in another language:

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS (3) Gopnyan galak-geuh that he (P) happy-3SG very ‘He is very happy.’

251

[Acehnese] (Durie 1985: 46)

Of relevance to our analysis of Spanish (Section 8.3.6) is the fact that speakers sometimes have the option to choose between forms depending on their point of view. Thus, in Central Pomo, where affectedness plays an important role, first persons motivate the use of the patient form on a more regular basis than third ones: (4) šyácˇ’aṱo I (Patient Case) am afraid šyácˇ’amul He (Agent Case) is afraid

(Mithun 1991: 522)

As Mithun (1991: 522) notes, having no direct access to what other individuals feel, speakers refrain from casting third parties as grammatical patients.

8.2.3 Dative experiencer The second strategy corresponds to a non-canonical type of alignment pattern much discussed in the literature. The defining feature of this pattern lies in the presence of a typically human entity which occurs in initial position and appears to represent the entity the proposition is ‘about’, as if it were the subject, but does not bear nominative case: it is often marked as dative, although genitive or accusative encodings are also possible. Given that in Spanish experiencer arguments entering this construction are always dative, I will refer to the second strategy as the ‘dative experiencer’ variant. This construction exists in languages of different areas and families. In (5) some examples with emotional verbs are presented. In addition to the dative-marked experiencer they contain a stimulus argument, which may display the coding properties of canonical subjects (it is nominative and controls verb agreement)—(5a) and (5b)—or may be realized as an oblique within the contours of a syntactically ‘impersonal’ template (5c): (5) a. Henni leiddust strákarnir her.DAT bored.3PL boys.the.NOM ‘She found the boys boring.’ [Icelandic] (Sigurðsson 2002: 692) b. hamko yah film bahut pasand aaii we.DAT DEM film.F.SG much taste come.SG.past.F.SG ‘We liked this film very much.’ [Hindi] (Montaut 2004: 42) c. Uzu-z Xu-jir-i-q-an gucˇ’ura I-DAT dog-PL-OBL-POST-EL be afraid ‘I am afraid of dogs.’ [Tabasaran] (Ganenkov 2006: 187)

252

CHANTAL MELIS

The special case marking on the human entity is interpreted by most scholars as motivated by the role of the referred participant in the event denoted by the verb. Cross-linguistically, indeed, this type of non-canonical marking tends to occur with similar event classes—mental experiences, physiological states, desire, need, possession, and existence (Onishi 2001)—which share the property of involving an individual devoid of control over the verbal situation and in some sense ‘affected’ by it. Far more controversial is the grammatical status of the non-canonically marked argument. A substantial amount of literature has been devoted to this issue. Studies explore to what degree the argument may be said to behave like a subject, beyond the single fact of occurring in topic position, with no consensus reached so far.² It appears that languages vary considerably in the syntactic treatment of the thematically prominent entity (Bickel 2004, Chapter 3 this volume). With regard to the so-called Standard Average European languages—excepting Icelandic (Sigurðsson 2002, 2004)—the claim is that the diagnostics for subjecthood yield unconclusive evidence (Haspelmath 2001), but not everyone endorses this view (see, for example, Barðdal 2006). This issue will remain unaddressed here since it is not crucial to the diachronic analysis of the Spanish emotional verbs presented here. The dative case used for the non-canonical marking in languages like Spanish, on the other hand, deserves a brief characterization. Datives cover a range of functions which differ from one system to another (Næss 2009). Typically, though, datives embody semantic ‘recipients’ (Næss 2009), that is, human participants functioning as ‘goals’ in acts of transfer.³ Recipients do not instigate the transfer—the agent does—but they are actively involved in the event and contribute to its successful outcome by accepting the transferred object (Newman 1996). At the same time, recipients are associated with a notion of affectedness, as new objects are introduced into their sphere of possession for their enjoyment or benefit (Newman 1996, Næss 2007). Viewed from this perspective, the dative marking typically found on the recipient seems particularly suited to encode the experiencer role: recipients and experiencers are active, yet non-instigating, as opposed to the agent; they are affected, but to a lesser degree, in comparison with the canonical patient. In the experiential domain, the dative case can therefore be used to create oppositions to both canonical subjects and accusative objects. In languages where verbs alternate ² The more widely used tests for subjecthood involve anaphoric binding of reflexive pronouns, raising, and controlling or undergoing coreferential deletion in different types of clauses; see the papers in Verma and Mohanan 1990, Aikhenvald, Dixon, and Onishi 2001, Bhaskararao and Subbarao 2004, among others. ³ The locative component of the dative case has recurrently been acknowledged. There is evidence of a historical connection between dative and locative morphology in Indo-European and in other languages (Aristar 1996). Not surprisingly, therefore, experiencers, whose correlation with the dative case is cross-linguistically very strong, are interpreted as mental ‘locations’ in some studies (Bickel 2004, Butt, Grimm, and Ahmed 2006, Landau 2010).

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

253

between a nominative experiencer and a dative one, the contrast is seen as involving a notion of volition or control, with the nominative implying some degree of responsibility for one’s feelings and the dative highlighting the involuntary aspect of the mental experience (Verma and Mohanan 1990, Onishi 2001). A shift from accusative to dative, on the other hand, signals a less wholly affected and more actively engaged experiencer (Lehmann 1991, Palmer 1994, Garcı´a-Miguel 1995, Chapter 6 this volume). In short, the dative construction promotes the experiencer to the rank of primary and salient clausal topic, while indicating that the topical participant is also affected to some degree, unlike the prototypical agentive subject.

8.2.4 Middle subject experiencer There is still another way in which the experiencer may be promoted to a grammatically prominent position, with a concomitant feature of affectedness. The instrument is a middle voice construction, provided the verbal category exists in the language. In her typological study on the middle voice, Kemmer (1993: 131) includes Quechua among the modern languages which possess an ‘emotion middle’. Consider this example from one of the Quechuan dialects, in which -ku operates as a middle suffix: (6) y tsayshi pipis tsayman chaptin piña-ku-rku-ya-n anger-MID-PVF-PL-3 ‘And if someone arrives, they suddenly get angry.’ [Quechua] (Villari and Menacho López 2015: 311) As pointed out by Dixon and Aikhenvald (2000b: 11), the term ‘middle’ ‘is used with a frightening variety of meanings in present-day linguistics’. For a solid point of reference we have to turn to the original active–middle opposition of IndoEuropean, where the active voice was used ‘to portray the subject as simply the doer of the action’ (Siewierska 1984: 163), while the middle voice, covering a range of semantic domains (Siewierska 1984: 163), served to indicate, according to Lyons’s classic definition (1969: 373), that ‘the “action” or “state” affects the subject of the verb or his interests’. The definition, as we see, remains silent about the agentive properties which the subject may or may not exhibit, contrary to the common assumption that middle subjects are low in agentivity (Viti 2016: 46) or otherwise inactive (Fedriani 2014: 61). In fact, the representative examples cited in grammars of Ancient Greek (Smith 1956), such as wash myself (§356a), crown oneself (§1717), he sent for hoplites (§1721), or they routed the cavalry (§1720), show middle subjects performing actions like agents. Kemmer (1993) supports this view. She identifies the semantic domains which are typical attractors of the middle category (grooming

254

CHANTAL MELIS

actions, changes in body posture actions, nontranslational motion actions, etc.), and on the basis of her findings states that ‘one of the main functions of a middle marker is to code the affectedness of an initiating entity’ (Kemmer 1993: 130). Note the reference to the ‘initiating’ character of the middle subject. In comparison, she adds, emotion events, relatively low in volitionality, are ‘less like the other types of middle situation types we have examined’ (Kemmer 1993: 130). However, the extension to this less prototypical domain comes as no surprise since the essential feature the middle ascribes to its nominative subject is that of being affected in some way by the situation. In the ancient Indo-European languages, subject experiencer verbs had a tendency to occur in the middle voice (Dahl and Fedriani 2012). Emotional processes or states, such as envy, be angry, worry, or become furious in Hittite (Luraghi 2010), be happy, find pleasure, love, be disgusted, be ashamed, or repent in Tocharian (Viti 2016), be glad, enjoy, love, fear, be ashamed, or be angry in Sanskrit (Hock 1990), have pity or fear in Classic Armenian (Viti 2016), enjoy, stand in awe of, be ashamed, or mourn in Homeric Greek (Dahl and Fedriani 2012), were expressed with middle verbs. Latin, too, inherited an emotion middle from the mother language (Section 8.2.6), and this will be of importance for our analysis of Spanish.

8.2.5 Typological distribution of experiencer-coding strategies The three strategies we are concerned with have been shown to respond to similar motivations. They are devices put at the disposal of speakers to indicate the affectedness of the experiencer (patient agreement, dative case, middle inflection), without having to degrade it to object status. All three bestow primary salience on the human participant. In principle, there is nothing that should impede the coexistence of semantically related strategies in a single language. We have known for a long time that partial synonymy permeates grammars, and we now also know that all languages are mixed types, featuring construction-specific alignment patterns that do not conform to the general rules of their syntax, but are shared in common with other systems (Bickel and Nichols 2009). Nevertheless, the truth is that the three strategies under scrutiny project some interesting distributional tendencies. One of these tendencies opposes patient subject experiencers to dative (oblique or non-nominative) experiencers. As discussed in Nichols (2008), the former are much more common in the Americas and the Pacific, whereas the latter are typical of Eurasia, and the reason for this has to do with basic structural properties of the languages involved in the distributional phenomenon. Indeed, given that patient subjects (SP ) are coded through verbal agreement, it is natural that they should occur with much higher frequency in the head-marking languages that prevail in

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

255

the Americas and the Pacific. By contrast, the typically dependent-marking languages of Eurasia are more likely to encode experiential semantics via case forms on the arguments. Additionally, Nichols takes note of the fact that the split-S languages in question tend to be primary object languages, so that, at a deeper level, there is a sense in which the two strategies converge: experiencers are assimilated to goal-like recipients through the dative in languages with ‘indirective’ alignment, or through object agreement in languages where recipients align with patients (Nichols 2008). Other distributional data relevant to our purposes involve the ancient IndoEuropean languages. Here it is observed that non-canonically marked experiencers are more widely attested in the Northern and Western branches, as opposed to the Southern and Eastern languages, in which nominative experiencers, occurring in the active or middle voice, appear to predominate (Viti 2016). It is likely that non-canonical alignment existed in the proto-language (see Viti 2016 for references). In fact, it is sometimes claimed that at its earliest stage IndoEuropean functioned as an ‘active’ type of system, with case roles being assigned on a semantic base, and that the development of its transitive syntax was posterior and gradual (Seefranz-Montag 1984, Bauer 2000, Hewson and Bubenik 2006, Chapter 2 this volume). On this view, oblique experiencers, where they appear, are to be interpreted as residues of the ancestral period when semantic alignment governed argument relations. Scholars who disagree with the ‘active’ hypothesis, on the other hand, emphasize the relative scarcity of non-canonically marked experiencers in the earliest attested and most conservative languages such as Hittite, Vedic, and Ancient Greek (Viti 2016), and argue that the frequency with which experiencers trigger non-canonical alignment across languages makes it more reasonable to assume that Indo-European included minor encoding patterns restricted to specific classes of semantic roles (Luraghi 2010). However one approaches the data, the greater productivity of oblique experiencers noticeable in some languages (Germanic, Baltic, Slavic), as compared to other ancient members of the Indo-European family, calls for an explanation, which Viti (2016) proposes to relate to the fact that the languages in question are known to have lost the inflectional middle category early in their history. In other words, the suggestion is that oblique experiencers proliferated as a means of compensating for the decay of middle subject experiencers.

8.2.6 Changing strategies In this section, historical changes involving the three experiencer-coding strategies under analysis are briefly discussed, with a view to the evolution of the emotion verbs of Spanish to be examined below.

256

CHANTAL MELIS

8.2.6.1 Loss of dative (oblique) experiencers The more familiar of these changes is the shift whereby oblique experiencers gave way to canonical nominative subjects in most of the modern European languages (Haspelmath 2001). The Germanic family, and English, in particular, have served as the focal points for insights into the pathway of development, along with the lexical, morphological, and syntactic parameters implied in bringing about the change (Jespersen 1927, Lightfoot 1979, Fischer and van der Leek 1983, Allen 1995). The accounts differ in details and perspectives, but the change is unanimously defined as a process moving towards uniformity, by which ‘deviant’ structures were made to conform to the model of the basic transitive paradigm. As a result of the change, non-canonical patterns of alignment vanished, or were significantly reduced, in the majority of the modern European languages.⁴ 8.2.6.2 Expansion of dative (oblique) experiencers Changes in the opposite direction—introducing irregularities and strengthening areas of ‘deviant’ syntax—have not been as thoroughly investigated, but they do occur. For instance, it is known that the non-canonical experiencers of English, before dying out, were extended to new verbs, both borrowed and native, which previously required nominative subjects (Allen 1995). Similarly, in Latin, regular nominative verbs developed oblique structures over time (Bauer 2000). Moreover, Latin handed down non-canonical alignment to the Romances. The mother language possessed a small group of impersonal verbs of feeling built with an accusative experiencer and a stimulus in the genitive case.⁵ These particular verbs fell into disuse or turned personal. But the dative experiencer of the Latin ‘like’ verbs (e.g. mihi placet ‘it pleases me’) was seized upon and used as a (partial) model for the creation of a non-canonically aligned dative structure, which was extended to a much wider range of semantically related verbs in, for example, Italian (Fedriani 2014), French (Mathieu 2006), and Spanish (Folgar 1993). French eventually lost the construction (Seefranz-Montag 1984), in accordance with the general European tendency mentioned above, whereas Spanish, unexpectedly, allowed for dative alignment to grow and spread to distinct ⁴ The movement towards uniformity has been approached from two different vantage points. When it is argued that the older oblique arguments functioned as topicalized objects, the analysis focuses on how these gradually acquired behavioural subject properties, before turning into full subjects coded in the nominative and triggering verb agreement (Cole, Harbert, Hermon, and Sridhar 1980). Scholars who, on the contrary, maintain that the obliques had subject-like properties from the earliest attested Germanic period onwards equate the change with a process of decrease in frequency, stimulated by the analogical extension of nominative coding, and leading in some cases to total evanescence (Barðdal and Eythórsson 2003, Eythórsson and Barðdal 2005). ⁵ For example, eos (ACC) infamiae suae (GEN) non pudet (3 SG) ‘they are not ashamed of their ill repute’ (Cic. 1 Verr. 35). See Fedriani (2014) for an extensive discussion of these constructions and for some evidence regarding the subject-like behaviour of the (usually) initial accusative experiencer (Fedriani 2014: 123–124). Behavioural properties of canonical subjects have also been ascribed to the dative experiencer of Latin placet referred to in what follows (Fedriani 2014: 98–99).

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

257

event types in the course of time (Elvira 2011, Melis and Flores 2013, and see Section 8.3.5). A similar evolution seems to have taken place within the Indo-Aryan family. Some of the modern languages have thematically prominent dative experiencers— viewed as one of their typical features (Verma and Mohanan 1990)—which do not go back to Sanskrit.⁶ The non-canonical marking patterns arose through a series of changes, involving lexical shifts, morphological innovations, and syntactic reanalyses which were instrumental in guiding their development (Butt, Grimm, and Ahmed 2006, Montaut 2013).⁷

8.2.6.3 Increase of middle subject experiencers The Indo-European inflectional category of the middle voice, as mentioned, was lost early in some of the branches of the family. In Latin, traces of the middle survived in the verbal -r forms, which in some of their uses conveyed middle semantics (Kemmer 1993, Cennamo 1998, Fedriani 2014). Various of the so-called deponent verbs—appearing only with the -r forms—denoted emotions, such as laetor ‘I rejoice’, patior ‘I suffer’, miror ‘I am astonished, amazed’, and vereor ‘I feel awe for’; other emotional verbs were inflected as middles in their subject experiencer variant, such as delector ‘I enjoy, delight myself ’, angor ‘I feel distressed’, and perturbor ‘I am disturbed’ (Fedriani 2014; see also Chapter 6 in this volume).⁸ In the course of time, however, the use of a reflexive marker with active voice began to encroach on the semantic middle territory of the -r form (e.g. delector vs me delecto). Classical Latin offers examples of the competition between the two strategies (Cennamo 1998, Fedriani 2014, Chapter 6 this volume), a phenomenon that paved the way for the eventual triumph of the reflexive pattern, at the cost of the older inflectional option unknown to the Romance languages. ⁶ Sanskrit had non-canonically marked possessors (Hock 1990, Montaut 2013) of very remote origin (for comparative Iranian evidence, see Haig 2008), but shows little evidence of non-canonically marked experiencers (Hock 1990, among others, against Hook 1990, who relies on the dative of ruc ‘please, like’ to uphold a notion of inheritance). ⁷ Contact with neighbouring Dravidian languages, where non-canonical datives encode a range of semantic roles, has been invoked to explain the change (Hock 1990, Viti 2016). However, recent findings have brought to light that topical dative-marked experiencers were as marginal in ancient Dravidian as in Sanskrit (see Montaut 2013 for references). These data constitute the base for Montaut’s (2013) hypothesis about a phenomenon of converging innovations which may have occurred under the influence of a semantically aligned Austro-Asiatic substratum. Nevertheless, she admits that the dative object experiencer of Sanskrit ruc ‘please, like’ (see note 6 above) ‘may be considered as a possible source’ for the future development of non-canonically aligned experiencers in the daughter languages (Montaut 2013: 100). This echoes what we saw above in connection to the role played by Latin placet ‘please’ in the emergence of the Romance dative structures. ‘Like’ verbs requiring a dative are found in a variety of ancient Indo-European languages and are likely to descend from Proto-Indo-European (Fedriani 2014). ⁸ In her study of the Latin deponent verbs, Gianollo (2005) observes that the human subject is generally affected by the verbal process but may show a certain degree of agentivity in some cases. This is consistent with the semantics of the Indo-European middle voice, as pointed out earlier in this chapter (Section 8.2.4).

258

CHANTAL MELIS

The reflexive middle inherited by the Romance languages became very productive and was extended to a gradually widening set of middle situation types. This evolution has been described in some detail for French (Hatcher 1942, Kemmer 1993). It took place in Italian as well (Fedriani 2014), and is reflected by presentday Spanish, where the reflexive form covers a broad range of middle functions (Maldonado 1999). Note that the recruitment of reflexive morphology for expressing middle meanings is a robust typological tendency (Kemmer 1993). Some of the Germanic languages (Gothic, High German, Dutch) too developed a reflexive middle at some point in their history (Kemmer 1993). In French, according to Seefranz-Montag (1984), the productivity of the reflexive middle increased during the period in which topical dative experiencers were on the wane. Under her hypothesis, the middle offered the advantage of a structure whose semantic load was akin to that of the deviant patterns, but whose form fitted the requirements of transitive syntax more adequately in having a nominative subject. This hypothesis once again underscores the functional proximity of non-canonical dative arguments and middle subjects in the experiential domain.

8.2.6.4 Emergence of patient subject experiencers The last change I will discuss, foreshadowing the discussion of Spanish data, concerns the development of active–inactive alignment patterns in accusative systems. The possibility of a change of this nature receives support from the results of typological surveys of recent date which suggest that ‘active’ languages do not constitute a unitary class. What is instead observed is a distribution of split-intransitivity phenomena along a continuum of alignment types extending from ergative to accusative languages (Nichols 2008, Bickel and Nichols 2009). The idea that active alignment patterns were introduced in Late Latin has been defended in a number of studies (La Fauci 1994, Cennamo 1998, 2000, 2009) on the basis of the following data: (1) extension of accusative instead of nominative marking, which affects inactive intransitive subjects at first and then progressively generalizes across the subject category until the final dissolution of the Latin case system;⁹ (2) rise of a pleonastic use of reflexives with intransitive predicates, linked to the demise of the medio-passive -r forms, with sibi favouring patient-like subjects and se more biased towards agentive ones; (3) participle agreement in periphrastic verbal forms, cross-referencing the direct object of the newly formed active perfects (auxiliary habēre) and the subject of perfective passives and deponents (auxiliary esse). All these phenomena underwent processes of regression with the birth of the Romance languages, whose coding systems signalled the revival of a basic ⁹ But see Adams (2013: ch. 12, s. 6), for a more balanced assessment of the overlaps between nominative and accusative uses in Late Latin, with some reservations about viewing the accusative as having become the ‘default’ case in the post-Classical era.

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

259

nominative–accusative type of alignment, and most clearly so, it is argued, in Spanish, which eliminated the contrast between auxiliaries, as opposed to Italian or French, for example, and re-established a nominative–accusative opposition via the prepositional marking (a) of some of its direct objects (La Fauci 1994). Viewed from this perspective, the fact that modern Spanish should evidence an innovating pattern of split-S alignment, as I will try to show, is unexpected, but not anomalous given the history of the mother language. Typological shifts of accusative systems to active alignment have also been the topic of studies centred on non-Indo-European languages. Some accusative languages from California, examined by Mithun (2008), for instance, manifest the existence of entrenched or incipient patterns of semantic alignment, which the author attributes to the influence of geographically proximate active systems. A similar change has also been reconstructed for North Halmaheran languages, which now show robust patterns of semantic alignment but derive from a nominative–accusative proto-stage (Holton 2008). Of particular relevance to my analysis of the Spanish data (Section 8.3.6) is the proposal, outlined in both of these studies, that transitive sentences containing a human pronominal argument in the accusative case and an unspecified thirdperson subject could have facilitated a reanalysis of the transitive object as the patientive subject of an intransitive verb. Schematically, the hypothetical scenario is as follows (Mithun 2008: 309): (7) (subject) object transitive verb ‘It scared me (object)’ > (it) scared me me patient intransitive verb ‘I (patient) was scared’ The purported source of the reanalysis corresponds to a type of construction labelled ‘transimpersonal’ in some works (see Malchukov 2008) and exemplified in (8): (8) a. i-hi-maata 3A-1SG.U-cold ‘I am cold.’ [Tobelo] (Holton 2008: 261) b. Ija maja-ni t-ei-a 1SG shame-1SG.POS 1SG.DO-3SG.SU-TODP ‘I was ashamed.’ [Amele] (Roberts 2001: 226) c. I-nga-mardalkanyi-ny arrkbi 3MO-3FA-sneeze-P man ‘He is sneezing, he has the flu. [lit.: it has made him sneeze].’ [Ilgar] (Evans 2004: 182) The transimpersonal construction is defined by three essential features: (1) the verbal morphology is transitive; (2) an experiencer argument is indexed or crossreferenced on the verb by an object agreement affix; and (3) the agreement slot

260

CHANTAL MELIS

designed for the agent is occupied by a third-person singular form which does not refer to any specifiable entity, having the character of a ‘dummy’ index (Evans 2004, Holton 2008, Reid 2000, Roberts 2001).¹⁰ Researchers who discuss this construction (without calling it ‘transimpersonal’) bring out a couple of issues that will be relevant to the analysis of Spanish. First, regarding its origin, they suggest a potential derivation from transitive sentences with an inanimate subject bearing the semantic role of a ‘causer’, which was later incorporated into the verb or omitted altogether.¹¹ Second, they draw attention to the fact that on occasion it is difficult to decide between assuming that the thirdperson actor index is a dummy or considering that it might refer anaphorically to a causer participant recoverable from the context (Evans 2004: 185). As said earlier, the transimpersonal construction is supposed to act as a mediating step in the development of split-intransitivity patterns from accusative sentences. One necessary condition for the change to occur is ‘the availability of object agreement on a transitive verb’ (Malchukov 2008: 98). The other requirement is a process of intransitivization: the actor slot of the transitive verb has to be eliminated. Here, languages may vary as to the mechanisms that come into play. With respect to the North Halmaheran languages, Holton (2008) argues that the pleonastic subject prefix of the transimpersonal construction suffered a process of morphological erosion. Regarding the Californian languages, on the other hand, Mithun (2008) conjectures that the contact-induced reanalysis of O as SP was facilitated by the fact that third-person agents are not usually indexed on the ¹⁰ For a distinct analysis of this type of construction, see, for example, Amberber 2005. In this study on Amharic, the subject marker of the transimpersonal is taken to index a silent argument of the verb associated with the semantic role of ‘ambient cause’ (Pesetsky 1995). ¹¹ As shown in Reid (2000), Ngan’gityemerri has some transimpersonal constructions which include an incorporated body-part term, not susceptible of being analysed as an argument of the verb. The constructions are monoreferential (O experiencer); the body part denotes the ‘locus of sensation’ in the experiencer’s body. For example (Reid 2000: 349): danging-ngi-ge-da 3SGA.poke-1SGO-belly-hit ‘I’m feeling sad.’ The transimpersonal constructions of Ilgar examined in Evans (2004) help us visualize a possible path of evolution: some appear with a nominal which could in principle be functioning as the subject, but the nominals admitted in the construction are highly restricted (i); in (ii) the nominal is fixed; and (iii) no longer includes a nominal that can be taken as the referent of the transitive subject prefix (Evans 2004: 170): (i)

Ngan-ni-mi-ny wurrwiny 1O-3MA-do-PAST shame ‘I got ashamed (lit.: shame did me).’ (ii) Nga-ni-mi-ny ngok 1O-3MA-get-PAST [meaning unclear] ‘I´m full.’ (iii) I-ni-marruku-n 3MO-3MA-make.wet-NP ‘He is sweating.’

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

261

verb in the borrowing languages. It will be of interest to explore how Spanish met both requirements—O agreement and A loss—in the development of its patient subject experiencers.

8.3. Alignment changes with the emotional causatives of Spanish This section explores the behaviour of emotional causatives in Spanish. First, it gives an overall panorama of the psychological predicates in the language (8.3.1), describes their basic transitive uses (8.3.2), and contrast their accusative object with the dative experiencer selected by other verbs (8.3.3). Second, it explores the historical development of experiencer-promoting strategies, with particular focus on the inherited reflexive middle construction (8.3.4), the development of a dative alignment pattern (8.3.5), and the rise of a split intransitivity subsystem within the psychological predicates in Spanish (8.3.6).

8.3.1 Preliminaries Experiential verbs referring to emotions or feelings in Modern Spanish divide into three syntactic subclasses (Di Tullio 1996). One of these integrates transitive experiencer-subject verbs such as amar ‘to love’, of which nothing more will be said in this chapter. The largest group in terms of membership—around 200 items (Di Tullio 1998)—comprises verbs like alegrar ‘gladden, make happy’, asustar ‘frighten’, entristecer ‘sadden’, enfurecer ‘infuriate’, irritar ‘irritate’, preocupar ‘worry’, and regocijar ‘delight’, which are the focus of the present diachronic study. As mentioned in the introduction, when retaining their basic transitive case frame, these verbs occur with the stimulus argument in subject position and the experiencer functioning as direct object. The third class is represented by gustar ‘to please, to like’, a dative experiencer verb addressed in a considerable number of publications, whose argument structure patterns with that of the non-canonically behaving predicates discussed in Section 8.2.3. Gustar is of concern to us, because in the course of time, as we shall see, the alegrar-type verbs borrowed its template as an alternative strategy for the encoding of their arguments.

8.3.2 The emotional causatives The transitive alegrar-type verbs of Spanish are similar to items like please in English (e.g. it pleases me), referred to as stimulus-subject verbs in Talmy (1985: 99–100). From the perspective of these verbs, the internal process is construed as an emotional reaction induced by the stimulus (subject) in the experiencer

262

CHANTAL MELIS

(patient-like direct object). Many languages have experiential verbs behaving this way (Bossong 1998, Haspelmath 2001), a phenomenon which could be seen as semantically driven by the fact that experiencers are not initiators but targets of the denoted mental process, and in this sense favour various types of ‘inverse’ construction (Bossong 1998; Fedriani 2014: 37). Nonetheless, within the framework of current theories on the interface between semantics and syntax, the assignment of the direct object role to the experiencer is considered to go against the predicted mapping of the human (sentient) participant of higher thematic prominence onto the primary grammatical function of subject (Lehmann 1991: 219; Butt, Grimm, and Ahmed 2006: 14). Different accounts have been proposed to justify the irregular behaviour of direct object experiencer verbs. For instance, in the well-known and much-cited paper by Belletti and Rizzi (1988), the verbs belonging to their second group and represented by Italian preoccupare are denied transitive status and are analysed as double object unaccusatives with a derived external argument. In Grimshaw (1990), on the other hand, the exceptional linking encounters an explanation in the higher position which the ‘causer’ (stimulus) occupies, relative to the experiencer, on the Aspectual Hierarchy, and this approach has been adopted as the right solution by a number of scholars. Thus, focusing on Italian, Bentley (2006) demonstrates that the members of Belletti and Rizzi’s second group of verbs have a causative component in their semantics. And with respect to the Spanish verbs under study, proof as to their causative nature is presented in Bogard (1993) and Di Tullio (1996, 2004), among others. Hence, the label ‘emotional causatives’ by which they are referred to in the Hispanic literature (Cano Aguilar 1981). While the causative meaning of object experiencer verbs sheds light on their argument structure from a theoretical standpoint, what has been observed in their regard is a certain degree of diachronic instability (Malchukov 2008: 93), attributable precisely to the fact that these lexical items counter ‘the well-known functional tendency to align the most prominent argument within a clause with the subject’ (Malchukov 2008: 90). They are, consequently, suitable targets of valencyaffecting strategies that specialize in upgrading the experiencer, as the history of the Spanish emotional causatives will illustrate very nicely. The point to be stressed in relation to the verbs we are concerned with is that their basic transitive template tends to favour human stimuli. As is known, in contrast to the necessarily animate experiencer participant, stimuli in emotional events may designate individuals, inanimate things, or situational triggers. The tendency for the emotional causatives of Spanish to select human subject stimuli can be perceived in the oldest available texts (9a) and continues to manifest itself in data of Modern Spanish (9b):¹² ¹² When the stimulus refers to a human being, the basic transitive construction is used in 80 per cent or more of the recorded instances through time (Melis, Flores, and Bogard 2003: 8, 15; see Melis

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

263

(9) a. El sabio fijo alegra a-l padre the wise son.NOM.SG gladden.PRS.3SG ACC-the father ‘The wise son makes his father happy.’ (13th c., CORDE) b. La Beba entra envuelta en una sábana y lo asusta. and he.ACC.SG frighten.PRS.3SG ‘His Baby enters wrapped in a sheet and scares him.’ (21st c., CORPES) The examples in (9) illustrate typical instances of the basic transitive use under discussion. The human stimulus subject may be anaphorically encoded in the verbal ending (9b) or take the form of a lexical noun phrase, usually located in initial position (9a). The experiencer, on the other hand, occurs as an accusative-marked noun phrase (9a), regularly postverbal, or as an accusative clitic pronoun (lo ‘him’, la ‘her’), which in Spanish obligatorily precedes the verb (9b). The prevalence of human stimuli in the transitive uses of the emotional causatives has to be interpreted in the context of the grammatical system as a whole. Spanish syntax is very sensitive to semantic distinctions and shows a host of phenomena hinging on a human vs non-human opposition (Narbona Jiménez 1989), considered by some scholars to represent an essential and defining property of the language (Lapesa 1968). What appears to underly this opposition is the influence of the familiar Animacy Hierarchy, according to which the higherranked human beings motivate various instances of differential treatment with respect to the lower-ranked inanimate entities. Of particular interest here is the phenomenon whereby two-place verbs involving a human participant and a nonhuman argument, with the non-human one predisposed to function as subject by virtue of its semantic role, tend to stimulate the use of alternative constructions allowing for the human participant to gain in prominence (Melis and Flores 2013). The emotional causatives will corroborate this tendency. By contrast, when a human subject acts on another human being, the effects of the Animacy Hierarchy are neutralized, and this explains why the emotional causatives throughout the history of the language hold on to their basic valency frame preferably in environments where the stimulus embodies a human participant.

8.3.3 Transitivity in Spanish Before turning to the valency-changing constructions developed by the emotional causatives, some remarks on transitivity in Spanish are in order. We are indeed 1999a: 54, for data from the late twentieth century). Non-human stimuli do appear in the transitive constructional pattern, especially if they are topical, but in general promote alternative, i.e. middle or dative, encoding options (Melis 1999a, Melis, Flores, and Bogard 2003, and see below).

264

CHANTAL MELIS

dealing with a language in which the distinction between accusative and dative objects is not sharp (for a full discussion, see Garcı´a-Miguel 2015). First, Spanish is known for its system of differential object marking (DOM), which roughly opposes marked human direct objects to unmarked non-human direct objects by means of the prepositional form a, homophone of the case marker that introduces all indirect object noun phrases. In Modern Spanish, the distinction between direct and indirect objects marked with a has partially been re-established through a phenomenon of clitic doubling—the use of a coreferential dative clitic pronoun with the dative noun phrase—still optional for ditransitive predicates but obligatory in the case of bivalent verbs such as gustar ‘to please, to like’. Second, the clitic pronouns of first and second persons display a single form covering both accusative and dative functions (me ‘me, to me’, te ‘you, to you’, etc.). Lastly, although the clitic pronouns of third person have preserved the accusative (lo ‘him’, la ‘her’) vs dative (le ‘to him, to her’) contrast, in some dialectal varieties a dative le is employed where an accusative lo would have been expected (a phenomenon known as leı´smo). Nevertheless, attending to factors of case marking and word order, it is possible to distinguish the base transitive use of the emotional causatives from the dative experiencer construction exhibited by verbs such as gustar ‘to please, to like’. Thus, on the one hand, the transitive causatives, as already illustrated in (9), take an accusative experiencer. The human referent cast in this role appears as a noun phrase DOM-marked with a, usually following the verb, or as a clitic pronoun, which in the third person will show the accusative case form lo or la in a majority of dialects. Moreover, when the stimulus subject is lexical, it tends to occur in sentence-initial position. On the other hand, verbs like gustar instantiate a construction-specific type of encoding, identified as such in grammars (Alcina and Blecua 1975: 895) and other studies of Spanish (Vázquez Rozas 2006, Elvira 2011), and associated with predicates from different semantic domains (Melis and Flores 2013). The construction has a number of defining traits observable in these examples featuring gustar: (10) a. A mi hijo le gustó mucho tu novela. DAT my son CL.DAT.3SG like.PFV.3SG much your novel.NOM.SG ‘My son liked your novel a lot.’ (21st c., CORPES) b. A los dos les gustaban las carreras de caballo. DAT the two CL.DAT.3PL like.IPFV.3PL the races of horse.NOM.PL ‘Both liked horse races.’ (21st c., CORPES) c. Si les gusta el jazz, lo promueven. if they.DAT.PL like.PRS.3SG the jazz.NOM.SG it promote.PRS.3PL ‘If they like jazz, they promote it.’ (21st c., CORPES) The dative construction usually contains a human participant (here the experiencer of gustar) and a non-human second argument (here the stimulus of gustar).

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

265

The first is sentence-initial (the unmarked position for this participant) and always bears dative case. In (10a) and (10b) the lexical noun phrase is preceded by the dative case form a and is additionally cross-referenced by means of the dative clitic le(s), considered by many scholars to be a marker of grammatical agreement between the dative object and the verb (Suñer 1988, Garcı´a-Miguel 1991, Franco 2000). In (10c), the dative form les ‘to them’ functions as a cataphoric pronoun. The non-human argument, on the other hand, controls verbal agreement, like a canonical subject, but appears after the verb in the slot normally reserved for the direct object.¹³ The distinction between the transitive template of the emotional causatives (9) and the dative construction of gustar-type verbs (10) will be important for a correct appraisal of the valency-changing strategy examined in Section 8.3.5. This said, I will now proceed to discuss the alternative construals developed by the emotional causatives in the course of time.

8.3.4 The inherited middle construction The oldest alternative, registered in early corpus data, is a middle voice type of construction—a heritage from the Indo-European ancestor, mediated through the Latin deponent verbs, and passed on to the Romance languages in the form of a reflexive marker (Section 8.2.6). The examples given in (11) are representative of different epochs. They show that we are dealing with an intransitive structure, whose single required argument is the human experiencer promoted to subject (11a), although a reference to the stimulus can be introduced into the clause, as an oblique (prepositional) noun phrase (11b), an infinitive (11c), or a finite clause (11d): (11) a. E digo-vos verdat, que yo fallé que me alegré and (I) tell you truth that I found that REFL.1SG gladden.PFV.1SG ‘Verily I say unto you that I felt that I rejoiced.’ (13th c., CORDE) b. Metello se contristaua de la muert jinjusta de su amigo M. REFL.3SG sadden.IPFV.3SG of the death unjust of his friend ‘Metellus was grieving over the unfair death of his friend.’ (14th c., CORDE) ¹³ The dative construction in Spanish has generated the same kind of debate as similar constructions from other languages have (Section 8.2.3). Scholars have sought to demonstrate that the prominent human participant has subject properties (Fernández Soriano 1999, Rivero 2004, Cuervo 2010), and others have refuted the subject analysis (Gutiérrez-Bravo 2005, Vázquez Rozas 2006). The consensus, though, is that the collocation of the dative at the beginning of the sentence represents the unmarked word order. In Spanish, ordinary dative participants do not occupy the initial slot unless they are topicalized.

266

CHANTAL MELIS c. Más, señor, me marauillo de ver como os reýs more sir REFL.1SG amaze.PRS.1SG of see.INF how you laugh ‘Even more, sir, I am amazed at seeing how you laugh.’ (16th c., CORDE) d. Estos hijos de carpintero, que se avergu¨enzan de que these children of carpenter who REFL.3PL shame.PRS.3PL of COMP su padre sea carpintero! their father be carpenter ‘Children of carpenters who are ashamed that their father is a carpenter!’ (19th c., CORDE)

In contrast to the basic transitive pattern of the causatives, the middle makes it possible to depict the emotion as something emanating from within the experiencer, abstracting away from the external forces that may have played a role in triggering the mental state in the experiencer (11a). When reference is made to the stimulus (11b–d), the middle removes this element from focal attention and leaves the way it was involved in the emotional process quite undetermined. The obvious effect of the middle construal is a loss of the causative features borne by the stimulus in the base transitive construction.¹⁴ As suggested by the examples in (11), the middle tends to be preferred to the transitive in contexts implicating non-human stimuli (for quantitative data supporting this diachronic tendency, see Melis, Flores, and Bogard 2003). This is consistent with the influence of the Animacy Hierarchy on Spanish syntax as discussed in Section 8.3.2. Speakers continue to use the reflexive emotion middle nowadays, but to a lesser extent, due to the development of another construction competing with the middle for the expression of events involving non-human stimuli. The new construction, modelled on the syntactic behaviour of gustar ‘to please, to like’, is the topic of the next section.

8.3.5 Origin and development of the dative alignment pattern The dative experiencer construction of present concern differs from the middle in being the product of a change that took place at a comparably late stage in the history of the Spanish language (see below). Examples of the innovating use display the characteristic features outlined above in relation to gustar, which very clearly acted as the analogical model for the extension of the dative alignment pattern to the emotional causatives (Melis, Flores, and Bogard 2003, Vázquez ¹⁴ It has been proposed that the middle variant of the emotional causatives could be viewed as a type of anticausative construction (Moreno Cabrera 1984). Scholars who oppose the idea of a derived structure emphasize the lack of a one-to-one correspondence between transitive and reflexive: some emotional verbs are deponents (e.g. arrepentirse ‘regret’) and for some causatives the reflexive is not available (e.g. fascinar ‘fascinate’). For further discussion, see Maldonado 2009.

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

267

Rozas and Rivas 2007, Elvira 2011). As shown in (12), when framed like gustar, the causatives appear with a dative-marked experiencer in preverbal topic position, cross-referenced by the dative clitic le if expressed lexically, and a typically inanimate stimulus, relegated to the postverbal object slot, yet controlling verb agreement: (12) a. No le asustó la memoria de su pérdida NEG he.DAT.SG scare.PFV.3SG the memory.NOM.SG of his loss ‘He was not scared by the memory of his loss.’ (17th c., CORDE) b. A don Ramiro le contentaron mucho estas noticias DAT Mr. R. CL.DAT.3SG content.PFV.3PL much these news.NOM.PL ‘Mr. Ramiro was very happy with this news.’ (19th c., CORDE) Gustar itself has an interesting history, which can only briefly be sketched here. From denoting a physical experience of ‘tasting’ (< Latin gustāre ‘taste’) in medieval times, gustar shifted to the expression of a notion of ‘liking’ (Vázquez Rozas and Rivas 2007, Melis and Flores 2018) via a well-documented path of metaphorical mapping (Sweetser 1990: 36–37). At the beginning of its mental trajectory, gustar continued to function as an experiencer-subject verb (yo gusto ‘I take pleasure, I like’) and then became attracted to the non-canonical dative experiencer construction, which was adopted as the preferred argument structure of gustar at the turn of the eighteenth century (Melis, Flores, and Bogard 2003, Melis and Flores 2018).¹⁵ In this process of attraction, the emotional verb placer ‘please, like’ played an important part (Vázquez Rozas and Rivas 2007, Elvira 2011, Melis and Flores 2018). The descendant of Latin placēre ‘please’, registered with high frequency in medieval texts (Melis 1999b, Elvira 2011), takes a dative experiencer, usually clause-initial, and has the postverbal stimulus realized as a prepositional phrase, commonly introduced by de ‘of ’ and in this way reminiscent of the genitive of the Latin impersonal verbs of feeling (Section 8.2.6). Among the medieval verbs which exhibit the dative alignment pattern, few share this impersonal template with placer (Folgar 1993: 122–123, Elvira 2011). Generally, the second argument in postverbal position is nominative-like (it agrees with the verb), and such is the model which gustar will adopt, departing from placer. Unquestionable, on the other hand, is the fact that dative gustar came to relieve placer in usage, considering that the gradual expansion of dative gustar occurred in parallel to the gradual decline in frequency of placer (Melis and Flores 2018).¹⁶ ¹⁵ Note that the evolution of Spanish gustar from a regular subject experiencer to a dative alignment pattern contrasts with the opposite change undergone by English like, representative of the general trend that took place in most of the modern European languages (Section 8.2.6). ¹⁶ The verb pesar ‘grieve, not like’, used as an antonym of placer in medieval texts, behaved similarly (Melis 1999b), and like placer suffered a process of decline during the post-medieval era, being replaced by no gustar ‘not like’.

268

CHANTAL MELIS

The extension of the dative alignment pattern to the emotional causatives is a phenomenon that comes into view once dative gustar has become established as a highly prominent emotional verb in language use. The driving force behind the process of diffusion was analogy, a fundamental cognitive principle, whose pervasive influence in language change has stirred renewed interest among historical linguists (Fischer 2013, Traugott 2008, De Smet 2012). Since analogy works on the basis of perceived similarities in meaning, form or function, it comes as no surprise that the semantically related causatives were subjected to the power of analogical attraction exercised by gustar, similarly to how gustar had been attracted by placer at an earlier historical stage. Examples of emotional causatives entering the dative experiencer construction sporadically show up in textual sources prior to the turn of the eighteenth century—interpretable as occasional exploitations of an old constructional pattern—but it is only after dative gustar has become firmly entrenched that a productive use of the construction with the causatives emerges, in contexts where the middle variant could have been chosen instead. In order to sustain this claim, in Table 8.1 I present quantitative data which show the distribution between topical dative experiencers and middle subjects with some causatives across the centuries of the history of Spanish.¹⁷ The fractions indicate the number of dative tokens in relation to the total of examples containing one or the other variant, and the figures on the right give the percentages corresponding to the dative uses. As the data show, dative experiencers occur very sporadically during the medieval period, they gain presence in the following centuries (Renaissance Spanish), but it is not until the nineteenth century that a real encroachment of dative alignment on the older middle territory can be witnessed. The temporal coincidence between the entrenchment of dative gustar and the explosion of dative subjects with the causatives is what gives support to the hypothesis about the process of analogical attraction.¹⁸

¹⁷ The examples corresponding to the sampled periods between 1200 and 1950 were extracted from the electronic database CORDE and the contemporary ones from CORPES, via a manual search limited to written prose. For each of the verbs listed in Table 8.1, I retrieved dative uses and middle constructions that included a stimulus argument (middles without stimulus were excluded since they could not be viewed as competing with the bivalent dative variant). All conjugated verbal forms in first-person singular and third-person singular and plural were excerpted, independently of tense or mood. The medieval tokens represent the entirety of the relevant data that were available. Starting with Renaissance Spanish, due to a notable increase in the use of some of the emotional verbs, specific decades were selected within the outlined periods for the purpose of my sample. ¹⁸ There are several verb-specific phenomena, reflected in Table 8.1, which I can only touch on very briefly here. Some verbs enter the language at later stages, and the ones attested from the beginnings yield scanty data due to the nature of the medieval texts that have come down to us (mostly historical and legal). The evolutionary paths are not homogeneous. On the whole, the negative emotions, such as fear, sadness, worry, or anger, were quicker in attracting the dative pattern, as compared to the positive emotions of happiness or content. The two verbs expressing anger, however, appear to have checked the progress of the dative variant, while contentar is unique in its resistance to the newer construction. In any case, broadly speaking, the growing visibility of dative alignment in this area of Spanish grammar is unquestionable

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

269

Table 8.1 Competition between MIDDLE and DATIVE

asustar ‘frighten’ avergonzar ‘shame’ desesperar ‘make despair’ entristecer ‘sadden’ espantar ‘frighten’ preocupar ‘worry’ enfadar ‘anger’ enojar ‘anger’ maravillar ‘amaze’ alegrar ‘make happy’ contentar ‘content’

Medieval 1200–1350

Renaissance 1500–1650

Modern 1800–1950

Present day 2000–2015

#

#

#

%

#

%

72/199 24/268 15/43 52/73 77/174 109/246 10/30 18/46 20/137 37/439 6/426

36 9 35 71 44 44 33 39 15 8 1.5

265/402 106/284 28/80 62/66 82/129 280/467 5/34 16/47 27/93 169/514 3/169

66 37 33 94 64 60 15 34 29 33 2

0/11 0/59 0/22 8/81

%

%

0 0 0 9

2/109 2 1/27 4 13/69 19 44/364 12

4/35 11 0/38 0 5/162 3

21/104 20 11/160 7 2/407 0.5 15/402 4 92/528 17

# Occurrences of Dative/Total of Middle + Dative % Frequency of Dative

The ensuing question is why a situation of competition in experiencer coding was generated between two similar strategies, as defined in the first part of the chapter, namely, between two devices of syntactic promotion (nominative case or initial topic position), coupled with a signal of affectedness (middle morphology or dative marking). Now, it is possible that the dative construction—with a postverbal but nominative-marked stimulus—was welcomed because it conserved a veiled allusion to the role of the stimulus in inducing the emotion, as opposed to the middle variant which defocalized the participation of the stimulus in a more radical way (oblique or absent). More significant, I believe, is the fact that the feature of affectedness carried by the reflexive middle had become somewhat opaque. The extension of the middle to a wide range of semantic domains, which took place in the Romance languages (Section 8.2.6), brought as a consequence that the nominative subjects of the middle verbs were sometimes difficult to distinguish from (unaffected) nominative se fue ayer ‘he agents (e.g. fue a la universidad ‘he went to the university’ vs se left yesterday’). The dative marker, by comparison, conveyed affectedness more overtly. Evidence in support of this view comes from looking at the progression of the dative subject construction once a real competition has been established with specific verbs. As shown in Table 8.2, which outlines the distribution between variants corresponding to the later stages of evolution (modern and present-day Spanish), the choice of the dative is always higher when the speaker communicates something about his or her emotional state. The preference for the dative in these

270

CHANTAL MELIS

Table 8.2 Emotion of EGO favours DATIVE

asustar ‘frighten’ avergonzar ‘shame’ desesperar ‘make despair’ entristecer ‘sadden’ espantar ‘frighten’ preocupar ‘worry’ enfadar ‘anger’ enojar ‘anger’ maravillar ‘amaze’ alegrar ‘make happy’

Modern 1800–1950

Present day 2000–2015

1p

1p

3p

3p

#

%

#

%

#

%

#

%

41/65 14/93 11/18 31/36 45/65 33/48 3/7 10/12 14/29 24/259

63 15 61 86 69 69 43 83 48 9

31/134 10/175 4/25 21/37 32/109 76/198 7/23 8/34 6/108 13/180

23 6 19 57 29 38 30 24 6 7

158/178 65/123 11/20 45/47 37/43 136/183 2/7 9/12 13/29 141/318

89 53 55 96 86 74 29 75 45 44

107/224 41/161 15/60 17/19 45/86 144/284 3/27 7/35 3/64 28/196

48 25 25 90 52 51 8 20 22 14

# Occurrences of Dative/Total of Middle + Dative % Frequency of Dative

contexts makes sense if we consider that expressing affectedness presupposes the ‘internal’ view of someone who has direct access to the experiencer’s self and is able to evaluate how he or she feels. Naturally, speakers are in a better position to do so when talking about personal emotions than when referring to what other people feel (Mithun 1991). As the quantitative data presented in Table 8.2 thus suggest, it may well be that an important motivation for bringing the dative construction into the domain of the emotional causatives was the availability of a device superior to the middle in highlighting the experiencer’s affected condition. The diluted semantics of the older construction would explain the change.

8.3.6 Rise of a split-intransitive pattern What I want to show next is that the extension of the dative experiencer construction to the emotional causatives gave rise to a phenomenon of split intransitivity in Spanish. Since active alignment patterns are susceptible of being attested, to variable degrees of frequency, in all languages (Nichols 2008, Bickel and Nichols 2009), and considering that Late Latin developed a variety of grammatical structures related to this phenomenon (Section 8.2.6), the change to be discussed does not come as a total surprise. For the sake of my analysis, the central point to remember is that, in some studies devoted to the topic at hand (Section 8.2.6), a transimpersonal construction

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

271

is assumed to play a fundamental role in promoting the reanalysis of transitive objects (O) as intransitive subjects (SP ), although concrete data supporting this hypothesis are not provided. The interest in documenting what happened in Spanish is precisely tied to the fact that it will be possible to observe how the A argument of a bivalent verb can turn into something indefinite, leaving a trace of its original presence in a verbal suffix that no longer agrees with a specific participant, and how the transimpersonal construction thus formed can lead to the expected reanalysis. What I have to mention before I begin, however, is that the evidence pointing to this process is restricted to a couple of emotional verbs, as far as I have been able to detect, and is found in association with singular first-person experiencers only (see Mithun 2008 for similar first-person constraints on incipient patterns of split intransitivity). The process of change I will sketch has its point of departure in dative constructions with emotional causatives featuring a first-person experiencer, as in (13): (13) a Por mi parte … a mı´ no me asustan for my part DAT me NEG CL.DAT.1SG scare.PRS.3PL esas cosas these things.NOM.PL ‘For my part, I don’t fear such things.’ (19th c., CORDE) b. Me extraña tu actitud, CL.DAT.1SG surprise.PRS.3SG your attitude.NOM.SG me alegra tu interés CL.DAT.1SG gladden.PRS.3SG your interest.NOM.SG ‘I’m surprised at your attitude, I’m pleased with your interest.’ (20th c., CORDE) In (13a), the first-person experiencer is realized as an (emphatic) full pronominal phrase (a mı´), obligatorily cross-referenced by means of an unstressed clitic pronoun (me); in (13b) the clitic pronoun stands alone. As I have indicated (Section 8.3.3), Spanish clitics doubling dative nominals are generally held to behave as agreement indexes, and arguments in favour of the agreement analysis can be extended to the clitic pronouns of first and second persons in all their uses (Ormazabal and Romero 2013). Note furthermore that in (13) me is interpreted as dative, but the same form would serve to encode the accusative relation (Section 8.3.3). Both data—the agreement function of the clitic and the formal identity of dative and accusative me—will be crucial to the necessary reanalysis leading to split intransitivity. The examples in (13) show that the emotional causatives alternate between third-person singular (13b) and third-person plural (13a) depending on the number features of the nominative postverbal stimulus. Other persons are rare, since the dative construction, as noted, favours inanimate stimuli. An additional point

272

CHANTAL MELIS

to be stressed is that the dative uses of the causatives are characterized by the predominance of the third-person singular, due in part to the rather uncommon occurrences of plural nominatives agreeing with the verb, and also following from the fact that the stimulus often refers to a situation, encoded in the form of an infinitive (14a) or finite clause (14b), and even, on occasion, as a type of adverbial adjunct establishing no agreement relation whatsoever with the verb (14c):¹⁹ (14) a. Me asusta tener que edificar de nuevo CL.DAT.1SG scare.PRS.3SG have-to.INF build again mi existencia my existence ‘I’m scared of having to rebuild my life.’ (20th c., CORDE) b. por eso me alegra que usted hable de mı´ for this CL.DAT.1SG gladden.PRS.3SG COMP you speak of me ‘that is why I am happy that you talk about me’ (20th c., CORDE) c. me espanta cuando pienso que, quizá, CL.DAT.1SG frighten.PRS.3SG when (I) think that perhaps estas muñecas […] ‘I feel terrified when I think that, perhaps, these dolls […].’ (20th c., CREA) The predominance of the third-person singular, in conjunction with the recurrent non-nominal character of the stimulus, incline the dative construction towards the behaviour of a type of impersonal clause. The sense in which the emotional verb ‘agrees’ with the stimulus is weakened; the argumental function of the stimulus recedes into the background. As it turns out, the third-person singular in Spanish is the default form used, for instance, with the impersonal weather verbs (e.g. nieva ¹⁹ The quantitative data to be presented confirm this observation. Three verbs were chosen—alegrar ‘make happy’, asustar ‘frighten’, preocupar ‘worry’—and all their dative uses, documented in a sample of materials from CREA (books in prose; years 1980–1999) and appearing in three tenses of the indicative (present, imperfective, and perfective past) were retrieved. The figures below summarize the distribution between recorded forms of the stimulus argument, i.e. between nominative noun phrases (NP), infinitives (INF), finite clauses introduced by que ‘that’ (COMP), adverbial clauses (ADV ), and zeromarked stimuli (ø) referring anaphorically to a situation mentioned in the preceding discourse (as in example (16) examined in what follows). alegrar 3SG

(179 tokens) (176 = 98%) NP (24 = 13%), INF (81 = 45%), COMP (55 = 31%), ADV (2 = 1%), ø (14 = 8%) 3PL (3 = 2%) asustar (223 tokens) 3SG (190 = 85%) NP (122 = 55%), INF (42 = 19%), COMP (18 = 8%), ADV (1), ø (7 = 3%) 3PL (33 = 15%) preocupar (433 tokens) 3SG (388 = 90%) NP (198 = 46%), INF (60 = 14%), COMP (82 = 19%), ADV (15 = 3%), ø (33 = 8%) 3PL (45 = 10%)

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

273

‘[it] snows’). And from this perspective, the third-person singular suffix on the verb of the dative construction could easily be reanalysed as an empty morpheme which no longer indexes a participant (see also Chapter 6 this volume). In other contexts, the stimulus is omitted. Since Spanish is a pro-drop language, there are cases in which the person suffix on the verb refers anaphorically to a previously mentioned entity cast in the stimulus role: (15) El El mensaje mensaje era pesimista, pero Ø no me asustóó3SG ‘The message was pessimistic, but it did not scare me.’

(20th c., CREA)

But sometimes the anaphoric link involves a more diffuse kind of antecedent embracing a whole chunk of previous discourse, as in (16): (16) — ¿Puedo saber quién es o lo considerarı´a una indiscreción? ‘May I know who it is or would you consider I’m being indiscreet?’ — Un señor de Barcelona que conocı´ ayer en el baile. Tiene una mercerı´a y no imagina lo bien que nos encontramos juntos, como si nos conociéramos de toda la vida. ‘A gentleman from Barcelona I met yesterday at a dance. He owns a notions store and you cannot imagine how well we get along, as if we had known each other our whole life.’ — ¡Cuánto meDAT.1SG alegra3SG , Julia! ‘I’m so glad, Julia!’ (21st c., CORPES) In (16), alegra denotes a feeling of joy stimulated by the story of an encounter, which abounds in descriptive details and subjective appreciations, such that the possibility of analysing this chunk of discourse as the constituent with which the verb ‘agrees’ seems out of reach. Note that if we were to separate this example from its context, the expression cuánto me alegra would start to look very much like one of the transimpersonal constructions discussed in the literature. It could be claimed that the clitic me indexes the experiencer argument on the transitive verb and that the A slot is occupied by a non-referring third-person subject (I am so glad = ‘it’ makes me so glad). Here it is useful to recall that in the literature on transimpersonal constructions it is suggested that these may have arisen from sentences with an inanimate causer (Section 8.2.6). Examples of the type illustrated in (16) are a recent phenomenon. My inquiry into earlier periods confirmed that for similar cases of unbounded, discursive stimuli the middle construction was used: (17) Perico. — Que están ahı´ mis amos, y quieren ver por la puerta de la alcoba este retablo. ‘My masters are here, and they want to look at this altarpiece through the door of the bedroom.’ Santos. — ¿De veras? ¡Cuánto meREFL alegro1SG ! ‘Really? I’m so glad!’ (18th c., CORDE)

274

CHANTAL MELIS

Today, with a couple of emotional causatives, the transimpersonal construction (as I will call it) and the middle variant alternate in this type of context, which suggests that speakers perceive them as near equivalents. On the basis of these alternations, and taking into account that the middle is an intransitive (monovalent) construction, we could then propose that the transimpersonal is being reanalysed as an intransitive sentence, giving way to a phenomenon of fluid-S alignment, as in (18) and (19): (18) a. —Muy bien, me alegr-o. ‘Very well, I’m glad [A].’ b. —Me alegr-a, me alegr-a. ‘I’m glad [P], I’m glad [P].’

(21st c., CORPES) (21st c., CORPES)

(19) a. —Me preocup-o, señora, me preocup-o y mucho. ‘I worry [A], lady, I worry [A], and very much so.’ (21st c., CORPES) b. —Y me preocup-a. ‘And I worry [P].’ (21st c., CORPES) In (18a) and (19a), the experiencer is indexed on the intransitive (reflexive middle) verb by means of the first-person suffix -o, which is used for the marking of A in transitive clauses. In (18b) and (19b), the third-person verbal suffix -a can be reinterpreted as carrying information relative to Tense/Aspect/Modality but no person feature, since the same default suffix also appears with impersonal weather verbs.²⁰ On this reading, the single argument selected by the verb is the experiencer, indexed via the agreement clitic me. Spanish me is dative in some contexts, but accusative in others, as stated earlier. It can therefore be regarded as formalizing a SP . What must be postponed for future research are the discourse-related factors orienting speakers’ choices in the direction of an SA or SP pattern. Mithun (1991, 2008) reports similar cases of alternation with first persons in the ‘active’ systems she examines and suggests that internally perceived differences about degrees of control and affectedness may account for the variation. Here, however, we have a more subtle type of contrast, pinning a patient-like experiencer (SP ) against one (SA ) that shares its nominative coding with canonical agents but carries the implicit affectedness feature associated with middle voice subjects, so that the motivating forces behind the alternation are less transparent and would require fine contextual analyses. In sum, it appears that Spanish exhibits split intransitivity in at least one small area of its accusative grammar. By tracing the emergence of the phenomenon,

²⁰ It is interesting to note that there are other languages where the dummy A suffix of the transimpersonal construction is identical to the form used with weather verbs (Evans 2004).

ALIGNMENT CHANGES WITH SPANISH E XPERIENTIAL VERBS

275

insight was gained as to the mediating role of transimpersonal constructions in this process. The data helped to strengthen the hypothesis regarding the diachronic path along which semantic alignment might develop from accusative languages (Malchukov 2008). The data also confirmed that certain conditions have to be met for the reanalysis of the transimpersonal as an intransitive sentence. The literature includes object agreement and erosion of A among the indispensable requirements. It just so happens that Spanish satisfies both conditions for the reanalysis: its clitic pronouns index participants on the verb and the thirdperson singular used for A coincides with the default morpheme of impersonal sentences.

8.4 Conclusions Many languages allow for a definition in terms of a prevailing basic alignment type, without excluding the display of a variety of minor patterns which deviate from the central paradigm. Experiential verbs have a tendency to show up in these minor patterns, because of the known ambiguity attached to the role of the prominent entity on which they predicate. What these minor patterns usually reflect is a predilection for alternative expressions, tuned to subtle nuances of meaning that might be difficult to convey otherwise. The chapter focused on three semantically motivated strategies of experiencer coding, revolving around a single class of basic transitive verbs (stimulus/S + experiencer/DO) and developed in one and the same language. The strategies were compared in order to detect similarities and differences in alignment. The comparison revealed that in all three the experiencer was given primary syntactic status, in association with a feature of affectedness, and the differences emerged in relation to the expression of the semantic feature: middle subject, topical dative, and a kind of SP . It was observed that the three strategies have parallels in many other languages, with a tendency to be in complementary distribution geographically or to replace each other over the course of time. This made the Spanish situation look somewhat surprising and by the same token more interesting. The chapter explored how the three strategies came to converge in Spanish. The changes were examined and various mechanisms of detransitivization were brought to the fore. Additionally, and perhaps most significantly, the Spanish data provided evidence to the effect that a potential way for semantic alignment to emerge in accusative systems crucially depends on the mediating role of a ‘transimpersonal’ construction, as recently argued in a number of studies (see the papers in Donohue and Wichmann 2008).

276

CHANTAL MELIS

Corpus CORDE: real academia española. Corpus Diacrónico del Español, www.rae.es CORPES: real academia española. Corpus del Español del Siglo XXI, www.rae.es CREA: real academia española. Corpus de Referencia del Español Actual, www.rae.es

9 Armenian morphosyntactic alignment in diachrony Robin Meyer

9.1 Introduction Classical Armenian was a language at the crossroads, both linguistically and culturally. Beyond its Indo-European heritage and, arguably, phylogenetically close relationship to Greek, Armenian was heavily influenced by West Middle Iranian languages (most significantly Parthian), Classical and early Byzantine Greek, and (to a lesser extent) Syriac.¹ The influence of the Iranian languages went so far as to impact not only the lexicon and phraseology of Armenian, but its syntax as well. This syntactic influence is most patent in a set of diachronic alignment changes in the morphosyntax of the Classical Armenian periphrastic perfect, a discussion of which forms the core of this chapter. It is argued that, on the basis of or at least in convergence with the West Middle Iranian tense-sensitive split-ergative verbal system, Armenian ‘borrowed’ the construction of the Iranian ergative–absolutive past tense onto its periphrastic perfect by means of pattern replication. Owing to morphosyntactic pressures, this replicated pattern was adapted and resulted in the largely tripartite alignment of the perfect attested in the earliest Classical Armenian texts. Over the course of time, this split-tripartite alignment was ousted in favour of nominative–accusative alignment, which dominated the non-periphrastic tenses. Section 9.2 outlines the basic alignment structures of Classical Armenian, focusing on two contrasts: that between periphrastic tenses based on the -eal participle and the synthetic tenses; and that between definite and indefinite objects, which influences surface interpretations of the alignment pattern. In Section 9.3, traditional explanations of the development and loss of this split-alignment pattern are discussed with a view to their explanatory power and potential problems. An alternative approach is presented in Section 9.4: it is argued that the primary cause of this alignment pattern is to be found not language-internally, but ¹ There are further a number of loanwords from Hurro-Urartian and the Kartvelian languages; their numbers are, however, so small that the influence of those languages, compared to the others named above, is negligible. See Greppin and Diakonoff (1991), Greppin (1996) on Hurro-Urartian, Deeters (1927: 111–114), Vogt (1938), Djahukian (2003), and Gippert (2005: 153–155) on Kartvelian. Robin Meyer, Armenian morphosyntactic alignment in diachrony. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl. Oxford University Press. © Robin Meyer (2022). DOI: 10.1093/oso/9780198857907.003.0009

278

ROBIN MEYER

in language contact with West Middle Iranian. This proposal is backed up both by other syntactic similarities between the two languages as well as typological observations. The discussion ends in an outline of the loss of split alignment in late Classical Armenian. Section 9.5 addresses the subsequent developments in Medieval and Modern Armenian, all of which show nominative–accusative alignment without significant exception.

9.2 Morphosyntactic alignment in Classical Armenian Before going medias in res of morphosyntactic alignment in Classical Armenian, it is worth briefly outlining its morphosyntactic categories. Both the verbal and nominal systems are derivable from Proto-Indo-European without too much effort and compare readily to those of other Indo-European languages; they have, however, undergone simplification and much syncretism, broadly speaking. Armenian nouns and verbs are differentiated for two numbers (singular and plural); a dual does not exist. The case system consists of seven cases (NOM, ACC, GEN, DAT, LOC, ABL, INS), some of which have, to a greater or lesser degree, syncretized with one another.² There is no gender category. The verbal system differentiates three persons, three moods (IND, IMP, SBJV) and two voices (ACT, MP). There are three synthetic tenses (PRS, PST, AOR) next to two analytical tenses (PRF, PLPF) formed with a participle and a copulative verb;³ the PST only occurs in IND. A consistent voice distinction exists only in the AOR; the future is expressed by means of the SBJV. Table 9.1 1.SG.IND forms of sirem ‘to love’ and hełum ‘to pour; to flow’

² The details of case syncretism are dependent on number and declension; for a general overview, see Jensen (1959: 49–67), for more historical detail, see Godel (1975: 92–107), Matzinger (2005), and Schmitt (2007: 89–114). ³ Armenian possesses other participial forms which are of no interest here, however, since they are not systematically used to form a specific tense; for an overview of these forms, see Stempel (1983).

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

279

Table 9.1 illustrates that the voice distinction in Classical Armenian is inconsistent; only -em verbs show separate synthetic MP forms outside the AOR.⁴ This, in turn, results in ambiguities: hełu (3.SG.PRS.IND) can be read as ACT ‘he pours (transitive); he flows (intransitive)’, but equally as MP ‘he is poured’. This inconsistency has been afforded great importance by some scholars in the question of the diachronic development of morphosyntactic alignment in Armenian, as detailed in Section 9.3.

Table 9.2 Declension of the noun am ‘year’, and the personal pronouns of the 1.SG and 3.SG

Table 9.2, in turn, exemplifies the types of syncretism prevalent across the nominal and pronominal system. In general, NOM and ACC have syncretized across both systems in the singular, with the exception of the 1.SG and 2.SG personal pronouns; they remain distinct in the plural, however. Other types of syncretism are common too (GEN=DAT and often also =LOC in the singular; GEN=DAT=ABL and ACC=LOC in the plural), but depend on declension class. Again, the formal identity of some NOM and ACC forms has an impact on questions of alignment as discussed below. In the case of Classical (and indeed modern) Armenian, the three categories of syntactic interest are subject (S), agent (A), and direct object (O). For the present purpose, further differentiation, e.g. of more or less goal-like objects in three-place arguments, is not necessary, nor is the consideration whether S is more agentive or more patientive.⁵ As might be expected of an Indo-European language, for the

⁴ These MP forms in PRS are supplied by i-stem forms, which also exist in isolation, consisting of intransitive verbs without transitive counterparts; for historical notes on these verbs, see Meillet (1936: 107–108), Godel (1975: 120), Klingenschmitt (1982: 9–11). ⁵ More detailed analyses of alignment structures, such as presented in Dowty (1991) and Bickel and Nichols (2009), are of course available and could be applied to Armenian, too; they would, however, not shed any further light on the alignment pattern of Armenian at least given the present state of research.

280

ROBIN MEYER

most part Classical Armenian expresses S and A as NOM, and O as ACC. The finite verb agrees with S or A in number and person. This general rule must be qualified in two ways, however. First, since NOM and ACC are not morphologically differentiated in all instances, and given that there is no fixed constituent order,⁶ Armenian effectively exhibits both direct (or neutral) alignment (S=A=O) as well as NOM–ACC alignment (S=A≠O). The frequency of direct alignment patterns is diminished, however, by the second qualification, namely differential object marking. Definite, or at least individuated objects are commonly marked by the proclitic z=.⁷ The following examples illustrate the relevant sentence types for the synthetic tenses: active, intransitive (1); passive, no agent (2); passive, agent expressed (3); active, transitive, indefinite object (4); active, transitive, definite object (5).⁸ (1)

S = NOM (ACT) hasanēr aṙ is hraman aṙn arrive.3.SG.PST to 1.SG.ACC order.NOM/ACC.SG man.GEN.SG mioy patuakani Yovsēp‘ kocˇ‘ec‘eloy INDF.GEN.SG venerable.GEN.SG PN called.GEN.SG ‘There arrived for me the order of a venerable gentleman called Yovsēp.’ (Kor. 1.1) (2) S = NOM (PASS) ayl ibrew satakec‘aw kaysr=n CONJ when kill.3.SG.AOR.PASS emperor.NOM/ACC.SG=DET darjan amenayn hoviwk‘ episkoposk‘n ork‘ return.3.PL.AOR all shepherd.NOM.PL bishop.NOM.PL REL.NOM.PL ak‘sorealk‘ ēin exile.PTCP.NOM.PL be.3.PL.PST ‘But when the emperor was killed, all shepherding bishops returned who had been exiled.’ (PB IV.13) (3)

S = NOM (PASS); PP (agent/instrument) Zi gitac‘ t‘ē y=Astucoy patrastec‘aw CONJ understand.3.SG.AOR COMP by=god.ABL.SG prepare.3.SG.AOR.PASS

⁶ Little research has been done on this subject. Grammatical surveys largely reiterate the same point, that constituent order is largely pragmatically motivated; see Meillet (1936: 138), Schmitt (2007: 158), Klein (2017:1109). ⁷ It should be noted that even inherently more individuated or definite entities (e.g. personal pronouns, personal names) can be and frequently are marked by this proclitic, but that this is not always the case. For the present purpose, [±DEF] refers to the presence or absence of this proclitic. ⁸ In all examples, constituents fulfilling S/A function are marked in bold, while those with O underlined function are underlined underlined.

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

281

kocˇ‘el z=na ǰatagov ˇcšmartut‘ean=n aṙ i call.INF OBJ=3.NOM/ACC.SG protector.NOM/ACC.SG truth.GEN.SG=DET to tal patasxani give.INF answer.NOM/ACC.SG ‘For he understood that it was ordained by God to summon him as protector of the truth to give a rebuttal [to the heretics]’ (PB IV.8) (4)

A = NOM; O = ACC-DEF ew ard kamik‘ p‘oxanak K‘ristosi ararcˇ‘i=n CONJ now wish.2.PL.PRS instead Christ.GEN.SG maker.GEN.SG=DET jeroy caṙayel anastuac=n magut‘ean jez jez 2.PL.POSS.GEN.SG enslave.INF 2.ACC.PL 2.ACC.PL godless=DET magism.DAT.SG ‘And now, instead of Christ [your] maker, you wish to enslave yourself to the godless religion of the Magians …’ (PB IV.51)

(5)

A = NOM; O = ACC+DEF ayl tēr Astuac z=zōrutiwn=n z=zōrutiwn=n CONJ lord.NOM/ACC.SG God.NOM/ACC.SG OBJ OBJ=might. =might.NOM/ACC.SG=DET NOM/ACC.SG=DET iwr yayteac‘ own show.3.SG.AOR ‘But the Lord God showed his might …’ (PB III.3)

It is worth noting that Armenian is a pro-drop language; both S and O can be left unexpressed where the context allows for it. In PASS, the agent is most commonly expressed in a PP (i + ABL); INS is used for instruments. In the analytical tenses, viz. PRF and PLPF, the picture is more complex. S and O remain marked by NOM and ACC, respectively, with the same qualifications as above regarding definiteness and case syncretism. What changes are the case of A and verbal agreement. A is marked as GEN, wherefore in the analytical tenses, Armenian exhibits both ergative alignment (S=O≠A) and tripartite alignment (S≠A≠O), depending on the type and definiteness of O.⁹ Further, in the analytical tenses, the finite verb agrees only with S; with transitive verbs, the copula appears in the 3.SG form independent of the number or person of A or O. These patterns are exemplified by the following sentences: active, intransitive (6); passive, no agent (7); passive, agent expressed (8); active, transitive, indefinite object (9); active, transitive, definite object (10). (6)

S = NOM (ACT) ew mincˇ‘cˇ‘ew na ekeal CONJ before 3.NOM.SG arrive.PTCP ‘And before he had arrived, …’

ēr be.3.SG.PST (PB III.20)

⁹ As a result of the morphological details described above, the ergative pattern is effectively restricted to O-DEF in SG in settings not involving speech acts and thus pronominal references.

282

ROBIN MEYER

(7)

S = NOM (PASS) zi yaṙaǰ nax and šineal ēr CONJ first first there build.PTCP be.3.SG.PST ekełec‘i=n church.NOM/ACC.SG=DET ‘For there was built for the very first time the holy church.’

surb holy

(PB III.14)

(8)

S = NOM (PASS); PP (agent/instrument) z=surb uxt ekełec‘woy=n or OBJ=holy covenant.NOM/ACC.SG church.GEN.SG=DET REL.NOM/ACC.SG ocˇ‘ ēr heṙac‘eal i zawrakenē=n NEG be.3.SG.PST abandon.PTCP by soldiery.ABL.SG=DET ‘The holy covenant of the Church, which had not been abandoned by the soldiers.’ (Eł. p. 106)

(9)

A = GEN; O = ACC-DEF oroy yankarc uremn ēr gteal REL.GEN.SG unexpectedly somewhere be.3.SG.PST find.PTCP ałp‘ap‘etac‘ hayerēn lezui nšanagirs nšanagirs language.GEN.SG sign. sign.ACC.PL ACC.PL alphabet.GEN.PL Armenian ‘Who somewhere had unexpectedly discovered alphabetic signs for the Armenian language.’ (Kor. VI.3)

(10) A = GEN; O = ACC+DEF ew ēr sora ənkaleal z=k‘orepiskoposut‘ean z= z= OBJ==rural-bishop.GEN.SG CONJ be.3.SG.PST 3.SG.GEN receive.PTCP OBJ asticˇani=n i jerac‘ jeṙnadrut‘iwn jeṙnadrut‘iwn consecration. consecration.NOM/ACC.SG NOM/ACC.SG rank.GEN.SG=DET from hand.ABL.PL meci=n Grigori great.GEN.SG=DET PN.GEN.SG ‘And he had received the consecration to the rank of bishop from the hands of the great Grigor.’ (PB III.14) Four complexities need to be added. First, in PRF, the finite copula is optional, meaning that the participle may stand on its own as a full verb (11). Second, the tripartite alignment pattern dominant in the analytical tenses has exceptions: occasionally, S is expressed as GEN (12) or A as NOM (13); these seem to be free, unconditioned variants. Third, the participle may be used converbially with other tenses, and in such instances A may be expressed as NOM or GEN (14).¹⁰ Finally, since NOM=ACC in most instances (as mentioned above) and the two cannot be ¹⁰ A corpus analysis of fifth-century texts indicates that statistically the converbial use is, by far, the single most common application of the participle, accounting for 40 to 68 per cent of all participles in the texts surveyed (Meyer 2017: 196).

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

283

distinguished except if O is definite, the alignment of some sentences may appear as ERG–ABS (15; see also Table 9.3). (11)

PTCP as full verb without copula ew ert‘eal i kołmans Arami i k‘ałak‘s erkus CONJ arrive.PTCP to side.ACC.PL PN.GEN.SG in city.ACC.PL two Asorwoc‘ Assyrian.GEN.PL ‘And he arrived in the region of Aram, in two cities of the Assyrians …’ (Kor. VII.1)

(12)

S = GEN apa ekeal ēr ark‘ayi=n i tełi then come.PTCP be.3.SG.PST king.GEN.SG=DET to place.NOM/ACC.SG ˇcakatu=n ew ənd nma mec front.GEN.SG=DET CONJ with 3.DAT.SG great episkoposapet=n Hayoc‘ chief-bishop.NOM/ACC.SG=DET Armenian.GEN.PL ‘Then the king came to the frontline and with him the great chief-bishop of the Armenians …’ (PB III.7)

(13)

A = NOM ew cˇ‘aragorc=n meławor Merzˇuan=n z=iwr z=iwr OBJ=REFL.POSS OBJ=REFL.POSS CONJ evil-doer.NOM/ACC.SG=DET sinful PN=DET ew ew z=zēn=n zard z=zēn=n zard OBJ= OBJ=arms. arms.NOM/ACC.SG=DET NOM/ACC.SG=DET CONJ ornament. ornament.NOM/ACC.SG NOM/ACC.SG CONJ z=nšan=n saławarti=n bazmac‘ z=nšan=n OBJ OBJ=insignia. =insignia.NOM/ACC.SG=DET NOM/ACC.SG=DET helmet.GEN.SG=DET many.DAT.PL edeal ēr z=noyn ōrinak give.PTCP be.3.SG.PST OBJ=same fashion.NOM/ACC.SG ‘And the sinful evil-doer Merzˇuan had given as identical copies his arms, ornaments, and the insignia on his helmets to many [people].’ (PB V.43)

(14)

PTCP used as CVB; S unexpressed ew haseal gayr handēp Gardmanakan joroyn CONJ arrive.CVB come.3.SG.PST opposite Gardman valley.GEN.SG ‘And he arrived opposite the valley of Gardman …’ (Kor. XVIII.1)

(15)

O-DET = NOM/ACC hraman hraman ew nora tueal ark‘ayagund 3.GEN.SG give.PTCP order. CONJ order.NOM/ACC.SG NOM/ACC.SG royal-guard banakac‘=n army.DAT.SG=DET ‘And he gave an order to the royal army …’ (Ag. §829)

284

ROBIN MEYER

Table 9.3 ventures to summarize the alignment system of Classical Armenian at its most conservative, viz. as represented in the earliest attested texts dating to the fifth century CE. It does not take into account the diachronic trends leading to the loss of tripartite alignment in the analytical tenses over the course of the following three centuries, which result in the predominantly NOM–ACC alignment of late Classical Armenian and subsequent forms of the language. These developments are discussed in Sections 9.4–9.5. On the surface, viz. from a solely morphological perspective, the synchronic data suggests that Classical Armenian shows a twodimensional alignment split: tense-sensitive alignment (TSA) between synthetic and analytical tenses; and a split along lines of definiteness. From a broader, morphosyntactic perspective, and considering accounts of contemporary Armenian grammarians, it seems more appropriate, however, to treat the formal identity of NOM and ACC in some categories as coincidental. For the purposes of morphosyntactic alignment, it is after all the morphosyntax that plays the pivotal role. The result is that, generally speaking, Armenian is best understood as showing only TSA along the lines mentioned above; for this purpose, the solidly shaded column in Table 9.3 need not be seen as a separate dimension. For a satisfactory diachronic explanation of this pattern, any explanation needs to answer or at least address the following questions:

(a) (b) (c) (d)

How and why did TSA arise? Why does GEN mark A? Why is the finite copulative verb in PRF optional? Why does the finite copulative verb in PRF.TR show Ø-agreement?

These questions will be addressed in the following two sections.

Table 9.3 Summary of constituent marking and morphosyntactic alignment

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

285

9.3 Traditional explanations of this alignment pattern Over the course of the twentieth century, there have been different attempts at explaining this split-alignment pattern. The following selection is representative if not exhaustive: (1) (2) (3) (4) (5)

a nominal construction (nomen actionis + genitivus auctoris); a ‘have’-perfect with a genitivus possesivus; the genitive is a prototypical agent case; language contact with Caucasian languages; the result of analogical shifts.

While there is an ongoing debate whether (Pre-)Proto-Indo-European was a language with an alignment pattern other than NOM–ACC, this is of no consequence for the present question, as the type of alternative alignment envisaged by some revolves around questions of animacy or gender rather than tense or definiteness.¹¹ In what follows, the historical explanations listed above are discussed briefly with a view to demonstrating why they fall short of explaining alignment in Classical Armenian to a sufficient degree.¹²

9.3.1 Meillet and the nominal construction The first attempt at explaining GEN-marking in PRF goes back to Antoine Meillet, who in the 1936 edition of Esquisse d’une grammaire comparée de l’arménien classique proposes that the construction be construed as a genitivus auctoris with a nomen actionis: [L]’emploi au premier abord étrange, du génitif dans les tours [participaux] provient sans doute de ce que les participes en -eal représentent d’anciens substantifs: nora bereal ē ‘il a porté’ a dû signifier originairement ‘il y a porter de lui’, c’est-à-dire que l’infinitif et le participe seraient des formations également nominales, mais de structure distincte.¹³ (Meillet 1936:129; italics added) ¹¹ For a recent and thorough discussion, see Willi (2018: 504–546). ¹² For a more detailed discussion of these and other approaches, see Meyer (2017: 113–131). In particular, a discussion of the historical morphology and its implications for the valency of the participle in -eal has been avoided for reasons of space; on this matter, see Meyer (2014; 2017: 39–82) with bibliography. ¹³ ‘While strange at first glance, the usage of the genitive in all [participles] doubtlessly originates in the fact that the participles in -eal represent old nouns: nora bereal ē ‘he has brought’ must originally have meant ‘there is a bringing of his’, that is, that the infinitive and the participle were equally nominal forms, but with different structure.’

286

ROBIN MEYER

While it addresses all the questions asked above,¹⁴ this explanation—also presented in the first edition of Meillet’s Esquisse from 1903—had already been refuted by Deeters on different grounds: Meillet does not explain the difference between the intransitive PRF (S = NOM) and the transitive PRF (A = GEN). In polemic terms: why is it ‘there is my carrying him’ but not ‘there is my coming’ (Deeters 1927: 80)? Benveniste (1952: 58) further objects that Meillet’s analysis requires a different morphological history for the intransitive and transitive participle, respectively, since the -eal participle would have to be a participle sensu stricto in the intransitive construction, but a nomen actionis in the transitive sense.

9.3.2 Benveniste and the ‘have’-perfect Trying to improve on Meillet’s explanation, Benveniste (1952) suggested that the Armenian perfect should be construed as a ‘have’-perfect with its agent in the genitivus possessivus; the construction starts out with the possessed entity in NOM (‘I have a written letter’ ~ ‘There is unto me a written letter’), which is over time reanalysed as the object of a transitive sentence and expressed as ACC. This explanation is still accepted even in more recent scholarship, so Schmitt (2007: 152). For Benveniste, the Armenian construction is similar to its Old Persian counterpart, the so-called taya manā kr̥tam construction.¹⁵ Pointing out the possessive function the Armenian GEN/DAT fulfils,¹⁶ Benveniste sees the perfect as ‘une expression possessive baˆtie en arménien même sur un modèle idiomatique pour rendre ce qui était apparemment le sens propre du parfait transitif ’ (1952: 60).¹⁷ The fact that Armenian, as opposed to Old Persian, marks O as ACC is explained as a cogent development of its transitive nature (Benveniste 1959). The 3.SG copula, in turn, Benveniste sees as part of the possessive construction: ‘I have’ is the same as ‘there is unto me’. His approach does, however, not explain why possessed entities in the plural do not receive a 3.PL copula (‘there are unto me’); equally, it offers no cogent explanation as to why, in a diachronically increasing trend, A can be expressed by NOM in some instances, or why, at an early stage, S can be expressed by GEN on occasion.¹⁸ ¹⁴ In Meillet’s eyes, this alignment pattern is the result of a retained nominal construction; the copula does, historically, agree with the nomen actionis and is optional because the copula generally is in Armenian and many other Indo-European languages. ¹⁵ Exhaustive discussions of this construction can be found in Haig (2008:23–88) and Ju¨gel (2015). ¹⁶ GEN and DAT are morphologically indistinguishable in Armenian except in pronouns. The perfect must construe with GEN, however, since it never occurs with a DAT pronoun. ¹⁷ ‘a possessive expression built in Armenian itself on an idiomatic model in order to express that which seemingly was the sense of the transitive perfect.’ ¹⁸ See (12)–( 13). The proposal that GEN is ‘préférée parce qu’elle faisait mieux ressortir le rapport d’antériorité’ [‘preferred because it better brought to light the anteriority relation’] (Benveniste 1959: 63) is ad hoc and lacks any reasoning.

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

287

Finally, Benveniste does not take into account that the copula is an innovation and unlikely to have been a part of the original construction, as evidenced by its overwhelming absence in the earliest texts.¹⁹

9.3.3 Schmidt and the prototypical genitive agent Since the role of GEN as the case marking A is pivotal in the discussion of the origin of the split alignment in Classical Armenian, the question whether GEN may simply be a prototypical agent case is worth asking. Such a point was made by Schmidt (1963), pointing to a number of other Indo-European languages in which GEN fulfils such or similar functions when combined with verbal adjectives or participles, like for instance Latvian, Lithuanian, Vedic, and Tokharian.²⁰ He admits, however, that in each case this agentive function is a secondary development based on the original role of GEN with the nominal, non-finite forms of the verb (1963: 11). Hettrich (1990: 94, 97) adds that the use of GEN as an agent was originally restricted to qualifications of verbal adjectives with PASS force as evidenced by Vedic, Greek, and Old Persian data; it cannot be considered a separate function of that case, but is a manifestation of its basic appurtenative use, wherefore any agentive uses must have developed einzelsprachlich and thus independently.²¹ Since even the closest comparandum to the Armenian construction, namely Tokharian verbal adjectives, has a different aetiology,²² the idea of a prototypically agentive GEN can be rejected.

¹⁹ Schmidt (1962: 231–232) elaborates on Benveniste’s approach and does take into account the copula-less, appositive, or converbial use of the participle; he argues that the participle is ‘prima¨r unempfindlich gegen eine Diathesenunterscheidung [und] stimmt [darin] mit anderen armenischen … Verbalnomina u¨berein’ [‘fundamentally insensitive towards a diathesis opposition [and thereby] concurs with other Armenian … verbal nouns’] and that, despite its passive origins, the construction derives its transitive use from the appositive, copula-less variant. This assumption is, however, unmotivated; if at all, a transitive reinterpretation could only arise from the ‘have’-perfect use, and the argument runs the risk of becoming circular. For a more detailed discussion, see Meyer (2017: 116–117). ²⁰ For Latvian, see Endzelīns (1923: §774); for Tokharian, Thomas (1952); for Vedic, Jamison (1979: 133–137). ²¹ An earlier analysis by Jamison (1979: 133–137) corroborates Hettrich’s position. Jamison shows convincingly that so-called agentive uses of GEN in Greek and Vedic are so marginal as to be virtually non-existent. ²² In Tokharian, the deontic verbal adjective in TA -l, TB -lle/-lye < PIE *-lo-, like its Armenian counterpart, is used primarily as a passive, agreeing with its patient, while the agent is found in GEN (Thomas 1952: 19). The fact that the construction is largely passive, expresses a deontic modality, and that the Tokharian GEN has through syncretism taken on functions of the lost DAT (Zimmer: 1985: 568– 569; Pinault 2008: 463, 2011: 383) suggests that there is no relation to the Armenian construction; in fact, closer comparanda can be found in the deontic constructions of other Indo-European languages, e.g. Hittite, Vedic, Avestan, Greek, and Latin (Luraghi 1995: 262; Hettrich 1990: 64–66).

288

ROBIN MEYER

9.3.4 Lohmann and contact with languages of the Caucasus The proposal that the construction of the Armenian perfect should be due to contact with other languages of the Caucasus, many of which have or had ergative alignment, goes back to Meillet who suggests that the Armenian construction ‘rapelle […] le “caractère passif du transitif dans les langues du Caucasus”’ (1899– 1900: 385).²³ Despite clear and convincing refutations by Deeters (1927), who makes the point that constructions with neither agent nor patient in NOM are equally unusual in Caucasian languages, this line of argumentation was further pursued by Lohmann (1937). The latter adopts Meillet’s historical analysis of the construction (see Section 9.3.1) and seeks to explain the ACC object as the equivalent of NOM in the Kartvelian transitive perfect constructions (effectively ABS in ergative alignment; 1937: 53). The agent expressed by GEN in Armenian is analogous to the Kartvelian DAT–ACC; since Armenian does not distinguish GEN and DAT consistently, Lohmann believes the DAT functions to be sufficient for, in modern terms, polysemy copying.²⁴ This comparison with the Kartvelian ergative construction and the suggestion of a historical link with the Armenian construction have also been advocated by Tumanyan (1974). The suggestion is, however, untenable. The contact situation between Armenian and other languages of the Caucasus except Greek and the West Middle Iranian languages was not sufficiently well developed to result in such significant changes: morphological and syntactic borrowings are not found; lexical borrowings are not numerous, and even they have been called into question recently.²⁵

9.3.5 Stempel and analogical shift An entirely different approach is presented by Stempel, who rejects the attempts outlined above for similar reasons. Together with Benveniste and Schmidt, Stempel assumes that the perfect construction arose from an earlier passive construction of the type *nora gorceal ē gorc ‘the deed was done by him’, where gorc ‘deed’ is the clausal subject; according to Stempel, this initial stage fits best with the intransitive-passive and adjectival nature of the -eal participle (1983: 83). Instead of the possessive explanation of Benveniste, Stempel provides an ‘innerarmenisches Motiv’ for the diathetic shift from passive to active. ²³ ‘is reminiscent of […] the “passive character of the transitive in the languages of the Caucasus”’. ²⁴ For the term polysemy copying, see note 28 below and Matras and Sakel (2007: 852), Heine and Kuteva (2005: 100), Heine (2012). ²⁵ Deeters (1927: 111–114) and Vogt (1938) present some findings, suggesting that the pre-literary contact between the Kartvelian languages and Armenian was not comparable in degree to the latter’s contact with Iranian languages or Greek. Even the elements listed there are, however, problematic (Gippert 2005: 153–155).

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

289

Under the assumption that the participle is originally intransitive-passive, he proposes that, at least initially, the perfect only existed in a passive sense (1983: 84). He explains the GEN agent as inherited from the proto-language, comparable to similar usages in Tokharian, Lithuanian, and in remnants of Greek. Since the agent in passive constructions was otherwise expressed by i + ABL, the advent of the perfect passive led to the coexistence of two types of agent marking in that tense. It is at this point that Stempel suggests an analogical shift: the coexistence in the perfect passive of two possibilities of marking the agent taken together with the system pressure exerted by the synthetic tenses, in which an active diathesis existed next to the passive, left open the possibility of one of the two agent-marking patterns being reanalysed. He suggests that the original perfect passive construction *nora gorceal ē gorc, whose function was now performed by the more common i + ABL agentive expression, was reinterpreted as an active according to the pattern illustrated in Table 9.4. While the perfect had thus acquired an active diathesis as well, the participle itself had not yet followed suit. Further steps were required to arrive at the attested grammatical status quo: in the new perfect active, the former grammatical subject in NOM had to be reanalysed as the logical object in ACC, a process simplified by the identity of NOM and ACC in the singular of nouns and most pronouns. Subsequently, any potential congruence in number marking between former subject and participle was likely eliminated; since adjectives preceding their head nouns do not normally agree with the latter in Armenian anyway, and only optionally do so when following them, this pattern may have been adopted for postposed predicative adjectives, too (Stempel 1983:85). In a final analogical step, and owing to the reinterpretation of NOM subject as ACC object, the latter could be marked by the definite object proclitic z=, as is frequently the case in all other tenses. Table 9.4 Analogical shift explanation of morphosyntactic alignment in Classical Armenian (Stempel 1983)

AOR PRF

ACT

PASS

na gorceac‘(z=)gorc Ø

gorc gorcec‘aw i nmanē gorc ē gorceal i nmanē *nora gorceal ē gorc

↓ analogical shift ↓

AOR PRF

ACT

PASS

na gorceac‘(z=)gorc nora gorceal ē (z=)gorc

gorc gorcec‘aw i nmanē gorc ē gorceal i nmanē

290

ROBIN MEYER

In the course of time, so Stempel, these processes allowed for an active interpretation of even attributively used participles, which in turn required the maintenance of a morphologically marked difference between active and passive participle, thus PASS na teseal ē ‘he has been seen’ vs ACT nora teseal ē ‘he has seen [something]’ (1983: 86). Stempel’s explanation cannot work, however. To begin with, it is based on the erroneous assumption that genitive agents are an inherited Indo-European feature; this has been rejected in Section 9.3.3. Second, the extent of the analogical remodelling proposed is implausible. While the reinterpretation of NOM subject as ACC object may have taken place as such, two questions arise: given that an alternative agent marking for PASS already existed, and that nonNOM subjects are not otherwise found in Armenian, why was the GEN-marked agent not simply lost? Similarly, why is the copula a fossilized 3.SG form, and only arises after the establishment of the perfect, as borne out by the earliest texts? As in the other explanations discussed, these questions remain unresolved.

9.4 A language contact approach While none of the approaches outlined above wholly explain the development of the Armenian perfect to a satisfying degree, many partly plausible suggestions have been made. In what follows, a different approach is laid out which aims to answer the four questions posed at the end of Section 9.2. The background of this approach is the firmly established contact relationship between Armenian and the West Middle Iranian languages, particularly Parthian, which is the result of extended Iranian rule over the Armenian kingdom and finds its most well-known and patent expression in the great number of Iranian lexical loans in Armenian.²⁶ Furthermore, extralinguistic as well as linguistic evidence strongly speaks in favour of this contact situation having been strong enough as to have gone beyond lexical influence and having an impact on Armenian syntax as well.²⁷ With this in mind, an Iranian origin of the Armenian perfect construction must be considered. The West Middle Iranian languages also show TSA, with the tenses ²⁶ For an overview, see Schmitt (1983), Meyer (forthcoming (a)); a more thorough, though dated discussion can be found in Bolognesi (1960). ²⁷ In particular, this refers to the strong social ties between the Arsacid Parthian rulers of Armenia and the nobility otherwise in charge of the region. These close relationships find an expression in Iranian–Armenian intermarriage and tutelage of young nobles in other families; the Christianization of the Iranian rulers together with their Armenian subjects; and frequent political and military struggles with the Sasanian neighbours. A more detailed treatment of this question and of other Armenian syntagmata replicated on the basis of Iranian models can be found in Meyer (2017: 295–339), Meyer (2022).

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

291

based on the present stem aligning NOM–ACC, whereas the participle-based tenses show ERG–ABS alignment.²⁸ Examples (16)–(19) illustrate this briefly. (16)

PRS, explicit A and O ’w ’m’ẖ hrw’yn bwxtqyft bwxtqyft wynd’m CONJ 1.PL all salvation salvation seek.1.PL.SBJV ‘And we all shall seek salvation.’ (BBB 302–303; Parthian)

(17)

PRS, explicit A, O enclitic ky=m wyš’h’ẖ ’c hrwyn gryhcg w m INTERROG=1.SG.OBL abyss and 1.SG.OBL liberate.3.SG.SBJV from all zynd’n prison ‘Who will free me from all [these] abysses and prisons …?’ (H/IVa/1a = Durkin-Meisterernst 2014: 292)

(18) PST, copula with S-agreement; A as enclitic, copula with O-agreement, O marked DIR qd tw ’br sd ’yy ’w=t ’z’z 1.SG.DIR when 2.SG up ascend.PTCP be.2SG.PRS CONJ=2.SG.OBL 1.SG.DIR hyšt hym syywg leave.PTCP be.1.SG.PRS orphan ‘When you ascended and left me [as] an orphan …’ (M42/R/i/14–16 = Durkin-Meisterernst 2014: 394) (19)

PST, A as enclitic, O in 3.SG, no copula cy kw ‘ym, cy cy=m ’c Tw wx’št, ’w mn COMP DEM REL REL=1.SG.OBL from 2.SG request.PTCP to 1.SG.OBL REL w’c’ẖ say.2.SG.SBJV ‘That which I requested from you, may you tell me [it]!’(MKG 1610–1611)

When comparing the West Middle Iranian PST to the Armenian PRF, however, there are a number of notable differences: the copula agrees in person and number with O, except in 3.SG where it is absent;²⁹ West Middle Iranian has lost most of its case morphology, and with the exception of 1.SG pronouns and clitic pronouns for all persons, constituent function is derived from word order; the alignment of the

²⁸ The nominal morphology of West Middle Iranian is very limited: a case distinction is only made in the 1.SG pronoun and, in early sources, in some kinship terms (Skjærvø 1983: 49, 176); enclitic pronouns, when used, only code OBL functions, i.e. O and A as well as possessive marking. The standard constituent order is SOV. ²⁹ The absence of the 3.SG copula applies only to PST, where the copula would be a PRS form of ’h ‘to be’; in the PLPF, a PST copula or auxiliary in 3.SG can occur (Durkin-Meisterernst 2014: 246–247, 376–377).

292

ROBIN MEYER

West Middle Iranian PST is ERG–ABS, with S=O. How, then, can this situation be related to the tripartite alignment pattern of Classical Armenian presented above? As the data suggest, the tripartite pattern is a secondary development of a previous, less clearly attested ERG–ABS alignment in Armenian, too; see (15). At first, it is this alignment pattern which Armenian imitates, likely by extending the usage of the participle beyond its basic attributive use, copying the usage of the West Middle Iranian participle. Examples (20)–(21) illustrate this usage in both languages.³⁰ (20)

Attributive PTCP in Armenian ənt‘erc‘uack‘ [ēin] surb groc‘ katareal lesson.NOM.PL be.3.PL.PST holy scripture.GEN.PL complete.PTCP uraxut‘iwnk‘ happiness.NOM.PL ‘The lessons in holy scripture were supreme happiness.’ (Eł. 125)

(21)

Attributive PTCP in Parthian hrw ’st’r ’ndryn ’wd b‘yn ’ndyš’d w’xt ’wd each sin internal and external think.PTCP speak.PTCP and myhg’r qyrd cy myhg’r do.PTCP INTERR damage damage ‘Each sin, external and internal, thought, spoken, and done [entails] what damage?’ (HLS 284–287)

The ensuing process of pivot matching, in which the functions of the key components of the model construction (the West Middle Iranian PST) are mapped onto Armenian counterparts,³¹ accordingly needs to find appropriate expressions for the involved constituents and the copula. The initial matching of S and O, where the Pth. DIR is mapped onto the Arm. NOM, is trivial; as regards the marking of A, however, the mapping of Pth. OBL onto Arm. GEN is not immediately obvious. It stands to reason, however, that in another case of polysemy copying, it was the Pth. OBL enclitic pronouns which made the Arm. GEN the obvious choice; the former frequently occur as A, but are also used for marking possessive and appurtenative relationships, like the Arm. GEN.³² This possessive function is shown in (22). ³⁰ Ju¨gel (2015: 273–274). ³¹ Particularly in bilingual speakers, syntagmata can be copied in a language-processing mechanism which identifies ‘a structure that plays a pivotal role in the model construction, and matching it with a structure in the replica language, to which a similar, pivotal role is assigned in a new, replica construction’ (Matras and Sakel 2007: 830). This process is called pivot-matching; the resulting replica constructions do not usually involve borrowing of lexical items at the same time (Matras 2009: 26–27). It relies on another process termed polysemy (or polyfunctionality) copying, in which the bilingual identifies an element in the replica language that shares some semantic or functional features with a pivot element in the model language and extends (or ‘copies’) some or all of the model language features onto the replica element. ³² Given the absence of morphological distinction in the WMIr. nominal system, the pronominal system is the logical locus for polysemy copying. The choice of GEN, beyond the reason mentioned here, is surely also structurally motivated: large-scale isomorphism in the Arm. oblique cases results in the

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY (22)

Parthian enclitic pronoun expresses possession ’w whyšt ’nwšg gy’n=wm gy’n=wm j’m soul=1.SG 1.SG lead.IMP to paradise immortal soul= ‘Lead my soul to immortal Paradise!’

293

(MMiii 887)

The copula is a different matter. Given that in the earliest sources, the copula is still a relatively uncommon occurrence, it must be assumed that its increasing presence is an Armenian innovation. The West Middle Iranian model, then, is likely the 3.SG in which the copula is not found anyway; this also goes to explain why, when the copula is introduced later, it exhibits a different agreement pattern than its model. The rise of the copula and its Ø-agreement in transitive expressions is likely the result of system pressure from the synthetic tenses, in which all verbs are finite and marked for person and number. This does not pose a problem for the intransitive PRF, where S-agreement is readily marked since S is expressed by NOM. In the transitive PRF, however, where A is marked by GEN, agreement does not appear to be licensed and the copula defaults to the 3.SG;³³ this kind of behaviour can be observed in other languages undergoing similar alignment changes, too, which lends further credence to this suggestion.³⁴ These are the mechanisms most likely at play leading to the replication of the West Middle Iranian PST alignment pattern in Armenian and to the development of a fossilized 3.SG copula. This accounts for the right-most column in Table 9.3. To account for the tripartite alignment pattern, however, one further step of reanalysis is required. As pointed out in Table 9.1, the voice distinction in Classical Armenian is inconsistent; only -em verbs show separate synthetic MP forms outside the AOR. This, in turn, results in ambiguities: hełu (3.SG.PRS.IND) can be read as ACT ‘he/she/it pours (transitive); it flows (intransitive)’, but equally as MP ‘it is poured’. This inconsistency has been afforded great importance by some scholars in the question of the diachronic development of morphosyntactic alignment in Armenian, as detailed in Section 9.3. Table 9.2 illustrates that there is considerable isomorphy between NOM and ACC in Armenian, particularly in the SG. Since O is marked as ACC in the synthetic undifferentiability of, e.g. GEN, DAT, and often ABL and their associated functions (e.g. appurtenance, recipient, origin marking). The exceptions are the majority of instances of GEN in the pronominal system (all personal pronouns; SG of demonstratives, etc.), which exhibit different forms. Therefore, the choice of GEN is motivated not only by functional equivalence with its Pth. model, but by unambiguous and economic coding of grammatical functions as well. ³³ It is worth noting that even in the non-standard patterns mentioned above (where unexpectedly S=GEN or A=NOM), copula agreement with either A or O did not arise. The fossilized 3.SG copula appears to be an Armenian-internal development, with incidence rising sharply in the course of the fifth century (Meyer 2017: 185–189). Were a model like that of Benveniste to be correct (see Section 9.3.2), traces of O-agreement with plural objects might be expected and a greater initial incidence, if not obligatoriness, of the copula. ³⁴ See Pirejko (1966) and Payne (1979: 442) on Talyši, Comrie (1978: 342) on Dānesfāni, Anand, and Nevins (2006: 7) on Hindi.

294

ROBIN MEYER

tenses, a reanalysis in the PRF of O as being marked ACC rather than NOM is not difficult to imagine, particularly if taking into account the role of the proclitic z= in marking definite O. This situation prevails for the majority of the history of Classical Armenian; as pointed out above (see (12)–(13)), the grammaticalization process behind the establishment and later ousting of the tripartite alignment pattern is evidenced already in certain early uses not conforming to the perceived standard, thus e.g. NOM-marked A or GEN-marked S; these alternative constructions indicate that the split-alignment system strove to normalize the NOM–ACC alignment of the synthetic tenses, no matter which case took S=A function; system pressure resulted in NOM winning over GEN. The statistical data gleaned from a corpus analysis shows that the NOM-marked A rises significantly already over the course of the fifth century (Meyer 2017: 182–184); by the end of the eighth century, NOM–ACC alignment has been established, though in more elevated literature, some remnants of the old construction still crop up owing to literary imitation.

9.5 Morphosyntactic alignment in Medieval and Modern Armenian After the loss of the split-alignment system as outlined above, the NOM–ACC alignment of post-classical Armenian has remained stable throughout the Middle Ages in all variants of the language. The Armenian verbal system, however, and to a lesser extent the nominal system have undergone significant morphological and, subsequently syntactic, changes.³⁵ Both major variants of modern Armenian—Modern Eastern Armenian as spoken in the Republic of Armenia and Modern Western Armenian as spoken in the diaspora³⁶—have developed a highly analytical verbal system with few synthetic forms remaining.³⁷ As an example, consider the development of the PRS as outlined in Table 9.5. ³⁵ There is next to no literature on the linguistic changes in Middle Armenian; the standard reference remains Karst (1901). ³⁶ Owing to the nature of the diaspora, Modern Western Armenian is pluricentric and for a large number of speakers a heritage language, a linguistically unified description of which is difficult to achieve; cf. Chahinian and Balakian (2016). For a map and general description of dialect distinctions prior to the emergence of the modern diaspora in the late nineteenth and early twentieth centuries, see Adjarian (1909); for a more general and modern discussion of Armenian dialects, see Greppin and Khachaturian (1986). ³⁷ Most verbal forms are composed of a non-finite form of the lexical verb and a finite form of an auxiliary; the remaining synthetic forms are found in AOR, (dynamic) SBJV, (dynamic) COND, and IMP (in the terms of Dum-Tragut 2009: 214–277).

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

295

Table 9.5 Diachronic development of the Armenian PRS as seen in the 1.SG.ACT of grem ‘I write’ Classical Arm.

Early MArm.

Late MArm.

MEA

grem IND

grem IND/SBJV

→

grem SBJV

→

grem SBJV

~

grem SBJV

↳ ↱

ku grem IND

→

kgrem COND/FUT

≠

kə grem IND

→

(modal particle)

↱ (PTCP periphrasis)

MWA

grum em IND

Next to these changes, modern Armenian has developed a standard, unmarked constituent order, SOV; this order can, however, be altered for pragmatic reasons, e.g. to emphasize one specific constituent;³⁸ particularly with O+DEF , the order SVO is similarly very common (Dum-Tragut 2009: 562). S and O continue to be expressed as NOM and ACC, respectively; as in Classical Armenian, these two cases are morphologically identical in the nominal system but have different expressions in the pronouns.³⁹ Owing to its greater uniformity and better linguistic description, only Modern Eastern Armenian will be discussed in what follows. It is worth noting that, like Classical Armenian, neither S nor A need not be expressed explicitly if they can be inferred from context (pro-drop). The following examples illustrate this alignment pattern: (23) intransitive; (24) transitive; (25–27) PASS without agent, with agent, and with instrumental expression, respectively. (23)

intransitive ACT aysōr, uriš hogsi aṙǰew enk‘: today another concern.GEN.SG before be.1.PL.PRS.IND en socˇiner=n ˇc‘oranum. pine.NOM.PL=DET be.3.PL.PRS.IND wither.IPFV.PTCP ‘Today, we are faced with another concern: the pines are drying up.’ (Aṙawōt, 08/06/2002)⁴⁰

³⁸ Compare, for instance, es namakə grec‘i ‘I wrote the letter’ (SOV, unmarked) vs es grec‘i namakə ‘It is I who write the letter’ (SVO, marked); Dum-Tragut (2009: 555–644). ³⁹ The standard grammar of Modern Eastern Armenian (Dum-Tragut 2009) does not operate with a category ACC for the nominal system. Instead, it differentiates O into NOM objects [-human, -definite] and DAT objects [+human, +definite]; cf. the situation in Spanish He visto el libro [-human] vs He visto a Juan [+human]. At the same time, the pronominal system retains a category ACC which is formally identical to DAT. For the purpose of this chapter, any O with the feature [-human] will be glossed as ACC in continuation of the practice found in Classical Armenian. ⁴⁰ Where not otherwise specified, the Modern Eastern Armenian examples are taken from the Eastern Armenian National Corpus (EANC), www.eanc.net.

296

ROBIN MEYER

(24)

transitive ACT ir hert‘in K‘ank‘anyan=ə vrdovvac LOG.GEN.SG turn.DAT.SG PN.NOM.SG=DEF outrage.RES.PTCP ē grum. namak namak letter. letter.ACC.SG ACC.SG be.3.SG.PRS.IND write.IPFV.PTCP ‘In turn, K‘ank‘anya writes an outraged letter.’ (Harut‘yunyan and Melik‘yan, Cicałum en vanic‘nerə, 2006)

(25)

PASS, no agentive expression ayo, petk‘ē uraxanal, or karmir lenter yes it-is-necessary be-happy.PRS.INF COMP red lens.NOM.PL en ktrvum. be.3.PL.PRS.IND cut.IPFV.PASS.PTCP ‘Yes, one can be glad that red lenses are cut.’ (Aṙawōt, 05/12/2002)

(26)

PASS, with agentive expression verǰini=s ordi=n […] spanvel latter.GEN.SG=DET son.NOM.SG=DET kill.PF.PASS.PTCP ē erku ambastanyalneri kołmic‘ be.3.SG.PRS.IND two defendant.GEN.PL side.ABL.SG ‘The son of the latter was killed by the two defendants.’ (Aṙawōt, 16/01/2003)

(27)

PASS, with instrumental expression spanut‘yun=n irakanac‘vel ēr murder.NOM.SG=DET carry-out.PF.PASS.PTCP be.3.SG.PST.IND danakov knife.INS.SG ‘The murder was carried out with a knife’ (Aṙawōt, 29/07/2003)

Like Modern English, then, the alignment of Modern Armenian is largely direct/neutral from the perspective of nominal morphology and its usage, but word order, verb agreement, and pronominal morphology evidence that NOM– ACC is a more accurate description from a morphosyntactic perspective.⁴¹ Similarly, like many other languages, Armenian makes an animacy-based differentiation in the optional argument found with PASS predicates: animate agents are marked by a postpositional phrase (GEN + kołmic‘), while inanimate instruments are expressed as NPs in INS.⁴² As mentioned above, however, pragmatic considerations and definiteness can affect word order. The resulting variants, largely SVO or OVS, may be caused by ⁴¹ Compare the similar situation in Classical Armenian, briefly discussed at the end of Section 9.2. ⁴² Particularly with affective verbs, this postpositional phrase can often be replaced by an NP in ABL of the agent (Dum-Tragut 2009: 94).

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

297

dislocating O+DEF , ‘heavy’ O-DEF , or indeed pragmatically marked S—to name but a few—to the right. These scenarios are exemplified in (28)–(30). (28)

SVO, O+DEF na tesel ē miayn zˇołovrdi 3.NOM.SG see.PF.PTCP be.3.SG.PRS.IND only people.GEN.SG storac‘um=ə storac‘um=ə humiliation.ACC.SG=DET ACC.SG=DET humiliation. ‘He saw only the humiliation of [his] people.’ (Sovetakan dproc‘, 17/02/1956)

(29)

SVO, ‘heavy’ O-DEF gorc es kardac‘el em erku gorc 1.NOM.SG read.PF.PTCP be.1.SG.PRS.IND two work. work.ACC.SG ACC.SG oronk‘ šat hetak‘rk‘rakan ēin REL.NOM.PL very interesting be.3.PL.PST.IND ‘I have read two pieces which were very interesting.’ (Azg, 11/12/2004)

(30)

VS, emphasizing S Moskvayum ē tesel loys loys Moscow.LOC.SG light. light.ACC.SG ACC.SG be.3.SG.PRS.IND see.PF.PTCP Karen Brutenc‘i nor girk‘=ə PN.GEN.SG new book.NOM.SG=DET ‘In Moscow, Karen Brutenc’s new book has appeared.’ (Azg, 19/05/2006)

In (30), VS word order is presumably the result of the topicalization of the subject, which continues to be of relevance for the ensuing paragraph. Beyond matters of word order changes, the only further complication in constituent marking lies in the so-called partitive subject, which is marked by ABL. These partitive subjects only occur with passive verbs, only refer to non-human actants, and denote that ‘the action is only carried out on one part, or partially [on] this grammatical subject’ (Dum-Tragut 2009: 313).⁴³ An example is provided in (31): (31)

partitive S Ṙadijov heṙarjakvum ēin radio.INS.SG broadcast.IPFV.PASS.PTCP be.3.PL.PST.IND Hovhannes T‘umanyani patmvack‘neric‘ PN.GEN.SG story.ABL.PL ‘On the radio, [some] of Hovhannes T‘umanyan’s stories were being broadcast.’ (see Dum-Tragut 2009: 313)

⁴³ See further Abrahamyan (2004: 40), Papoyan and Badikyan (2003: 142–143).

298

ROBIN MEYER

Since this type of expression is highly constrained, the question arises whether it does indeed represent an independent subject category, or is just an instantiation of the very common partitive ABL. Given that Armenian is a pro-drop language, this ABL could be interpreted as depending on a Ø head, an implied indefinite quantifier. While further studies are necessary, native speaker grammaticality judgements suggest that the ABL does not pass subjecthood tests. Overall, the morphosyntactic alignment of Modern Eastern Armenian is an interesting example of NOM–ACC alignment with certain complications owing to the role definiteness and animacy play in the case assignment of O.⁴⁴ Given that syntactic function is expressed by a combination of morphological case marking and constituent order, however, no systematic ambiguity arises, nor is a different interpretation of this alignment possible.

9.6 Conclusions The development of morphosyntactic alignment in the history of Armenian is uncomplicated for the most part, as NOM–ACC dominated for most of its existence and in the majority of tenses. The Classical Armenian periphrastic perfect stands alone, showing tripartite alignment for at least the beginning of the attested history of this language, but is eventually ousted in favour of NOM–ACC alignment towards the end of the classical period. Its preliterary history and genesis has been the topic of a hundred-year-long debate, during which a great number of solutions internal and external to this language have been proposed. The approach advocated here, based on the well-attested, extensive, and prolific contact between Classical Armenian and the West Middle Iranian languages, chief among which Parthian, suggests that the tripartite alignment of the periphrastic perfect is the result of a grammaticalization process which started with the replication of the Parthian ERG–ABS aligned past tense. This participle-based tense, through pivot matching and polysemy copying, was replicated in Armenian with the participle in -eal and a GEN agent; the reanalysis of the original nominative object as accusative resulted in part from the large-scale isomorphy between those cases as well as from system pressure from the synthetic tenses. The introduction and later quasi-obligatorification of a fossilized 3.SG copula with Ø-agreement is an Armenian-internal development. The eventual loss of tripartite in favour of NOM–ACC alignment is caused once more by system pressure, since the identity of S and A prevails in the synthetic tenses, and even in the perfect the subject of intransitive verbs is marked as ⁴⁴ As indicated above, on the morphological level there is a split between the nominal (direct/neutral alignment) and pronominal (NOM–ACC alignment) system. The main reason to consider the alignment system on the whole NOM–ACC lies in the morphosyntax, since verb agreement is consistently with S=A and constituent order also clearly separates S=A from O.

ARMENIAN MORPHOSYNTACTIC ALIGNMENT IN DIACHRONY

299

NOM. Once this transition was concluded, the alignment pattern of Armenian has remained stable even though its verbal morphology and constituent order rules have undergone significant changes. While the prehistory and early stages of alignment in Armenian have been documented and investigated thoroughly, the later transition period between Classical and Middle Armenian, and thus between the two alignment types, remains a subject further enquiry into which should prove fruitful.

Acknowledgements The research on which most of this chapter is based was funded by the Arts and Humanities Research Council, for whose support I am very grateful. I also owe thanks to the anonymous reviewers for their keen eyes and constructive comments; all errors and omissions are, of course, mine.

Abbreviations of primary sources Ag. BBB Eł. HLS Kor. MKG MMiii PB

Ter-Mkrtcˇ‘ean and Kanayeanc‘(1909) Henning (1937) Thomson (1993) Durkin-Meisterernst (2006) Mat‘evosyan (1994) Sundermann (1981) Andreas and Henning (1934) Garsoïan (1984)

10 Areal and diachronic trends in argument flagging across Slavic Ilja A. Serzˇant, Bjo¨rn Wiemer, Eleni Buzˇarovska, Martina Ivanová, Maxim Makartsev, Stefan Savic´, Dmitri Sitchinava, Karolı´na Skwarska, and Mladen Uhlik

10.1 Introduction Flagging refers to the marking of the semantic and/or syntactic role of arguments by means of inflectional cases or adpositions including their combinations and morphologically intermediate subtypes (Haspelmath 2019). The more traditional terms are dependent marking or Case. In this pilot study, we examine areal and diachronic trends of argument flagging of ten modern Slavic languages: Bulgarian, Macedonian, Serbian, Slovenian (South Slavic), Czech, Slovak, Polish (West Slavic), Belarusian, Russian, and Ukrainian (East Slavic) as compared with Old Church Slavic, which is our proxy for Proto-Slavic, in addition to the comparative evidence from other ancient IndoEuropean languages. Our study is primarily explorative and aims at understanding how and why the flagging patterns of erstwhile dialects develop after the split from their common proto-language. The split of the Slavic languages is quite recent. It must have happened no earlier than 1,300 years ago. Their geographic spread has remained quite compact despite the fact that during approximately the sixth to ninth centuries the Slavic-speaking territory in Europe expanded at an astonishing speed (Nichols 1993, 2020). However, apart from the southwest movement of Ugric (Magyar) tribes into today’s Hungary, the persistence of the Daco-Romanian population (see today’s Romania) and rather minor-scale migrations of non-Slavic-speaking communities (e.g. on the Balkans), the Slavic-speaking territory has not been internally driven apart. Recency of the split as well as compactness of the Slavic territory must have been responsible for a considerable degree of inheritance and, thus, homogeneity in the flagging patterns of modern Slavic languages. Thus, all modern languages remained accusatively aligned as was Proto-Slavic and Proto-Indo-European

Ilja A. Serzˇant et al., Areal and diachronic trends in argument flagging across Slavic. In: Alignment and Alignment Change in the Indo-European Family. Edited by Eystein Dahl, Oxford University Press. © Ilja A. Serzˇant et al., (2022). DOI: 10.1093/oso/9780198857907.003.0010

ARGUMENT FL AGGING ACROSS SL AVIC

301

(PIE).¹ The case systems of most Slavic languages are very conservative. For these reasons and given the evidence we provide below (see especially Section 10.3.3), inheritance is an important factor determining homogeneity of flagging patterns across the modern Slavic languages. The compactness of the modern Slavic territory allowed for mutual contacts, which, in turn, contributed to the maintenance of the same inherited patterns across Slavic. Only two South Slavic languages are somewhat distinct: Bulgarian and Macedonian have lost morphological cases on nouns (reminiscent of neutral alignment). However, their pronouns retained three cases (nominative, accusative, and dative) and these two languages, distributively, show accusative alignment, too. The emergence of differential object indexing (term coined in Iemmolo 2011) via clitic doubling is likewise indicative of accusative alignment in these two languages, as is the subject indexing in all Slavic languages.² In addition to the inheritance factor, common innovations also contribute to homogeneity. Thus, the joint abandonment of typologically dispreferred flagging patterns of Proto-Slavic is also responsible for similarity. As a result of these innovations, modern Slavic languages—in contrast to Proto-Slavic—now predominantly rely on unmarked subjects and only on differentially marked objects. By contrast, Proto-Slavic (as well as PIE) marked both arguments of a transitive verb via dedicated nominative and accusative affixes across all NP types and in all inflectional classes (except neuter nouns).³ This morphologically redundant pattern is rare cross-linguistically and is functionally dispreferred (Dixon 1979, Handschuh 2014, Creissels 2018). The marked nominative system has been abandoned in most of the Slavic nominal inflectional classes (declensions) in the singular (except for the a-stems), while some plural declensions still retain marked nominatives (cumulatively expressing number as well) in some of the languages.⁴ Likewise, the Proto-Slavic flagging system of the direct object was also of a typologically rare type in that it involved no differentiation, i.e. all object types—including inanimate, indefinite, non-anaphoric, non-topical ones, etc.— were explicitly marked by accusative affixes across the board regardless of their ¹ Note that some researchers suggested active/inactive (Gamkrelidze and Ivanov 1984) and some even ergative alignment of PIE despite the fact that all ancient Indo-European languages such as Vedic Sanskrit, Avestan, Ancient Greek, Tocharian, or Old Church Slavic are strictly accusatively aligned in both flagging and indexing. We refrain here from a critical discussion of the spare and quite indirect and only morphological evidence at disposal (e.g. morphological nominative–accusative syncretism with neuter nouns). See, however, Chapter 2 in this volume for discussion and evaluation of these hypotheses. ² Bulgarian and Macedonian differ from each other with respect to some specific rules of object indexing (clitic doubling) and the obligatoriness with which they apply (Friedman 2008). Moreover, although object clitics are primarily hosted by verbs, other hosts in the Wackernagel position are possible too (Corbett 2006: 13). ³ The nominative affixes of neuter nouns were homonymous with the accusative affixes. ⁴ An exception was early Old Novgorodian with the marked nominative singular in both major declensional classes (o- and a-stems). However later, the marked nominative was abandoned in this variety as well.

302

ILJA A . SERŽ ANT ET AL.

likelihood of becoming an object. By contrast, modern Slavic languages adopted and expanded differential object marking (DOM) which also has been shown to be more efficient and cross-linguistically the preferred pattern (Sinnema¨ki 2014, Schmidtke-Bode and Levshina 2018: 527, Haspelmath 2021; see also more generally on DOM in Witzlack-Makarevich and Serzˇant 2018). In particular, most modern Slavic languages have animacy-based DOM systems such that animate nouns may employ the case marker that is homonymous for accusative and genitive (at least in the singular) while, in the singular, inanimate nouns bear no marker at all or, with neuter nouns, the marker is homonymous with the accusative and nominative (the a-declension is an exception). In the course of time, modern East and West Slavic languages expanded this DOM system onto the plural declension (to different degrees). Ukrainian and Polish DOM systems expanded even beyond animacy. Furthermore, a number of West and Southeast Macedonian dialects (South Slavic) in contact with Balkan Romance have developed an analytical pattern of DOM based on the so-called na-dative (formed by the preposition na ‘on’; see Asenova and Aleksova 2008, Buzˇarovska 2017). Thus, modern Slavic languages are uniform in adapting differential object flagging in various ways and abandoning the across-the-board object flagging of Proto-Slavic and PIE. This trend is likely to be conditioned by the universal dispreference for across-the-board object flagging (on which see Sinnema¨ki 2014, Haspelmath 2021).⁵ While the inheritance factor conditioned by the recency of the split of ProtoSlavic as well as the pressure towards more preferred flagging patterns are responsible for a considerable degree of homogeneity of the modern flagging patterns, one of the factors that must have been responsible for divergence is the distinct geographic locations of the modern Slavic languages, as we argue below. In what follows, we explore the interplay of the inheritance factor, the areal factor, and the local factor in the development of the modern flagging patterns. We proceed as follows. First, we present our sampling method and the database which builds on translations based on 46 verb meanings into each language (Section 10.2). Section 10.3 is the main part of the chapter and is devoted to the changes in the ratio of flagging alternations (Section 10.3.1), changes in transitivity prominence (Section 10.3.2), homogeneity of flagging across Slavic (Section 10.3.3), ratio of nominative marking of the subject(-like) argument (Section 10.3.4) and, finally, to establishing a big picture on argument flagging across Slavic in terms of language clusters (Section 10.3.5). Section 10.4 summarizes the results and presents conclusions. ⁵ In addition, West Slavic languages have developed differential subject marking in the plural, both for NPs and pronouns (Zieniukowa 1981, Laskowski 1986, Mindak 1990, Rappaport 2010, Žigo 2012). This is accompanied by differential subject indexing, inasmuch as verbs in the past tense plural adapt to differentially marked subject NPs according to general agreement rules on clause level. The different patterns are all based on the animacy hierarchy, or a hierarchy of inherent lexical content (Silverstein 1976).

ARGUMENT FL AGGING ACROSS SL AVIC

303

10.2 The database and the sampling method In order to explore the diachronic and areal trends in the evolution of flagging patterns of modern Slavic languages, we created a database comprising 11 Slavic languages, i.e. all major modern Slavic languages (Belarusian, Russian, Ukrainian (East Slavic), Czech, Polish, Slovak (West), Bulgarian, Macedonian, Serbian, Slovenian (South)) as well as Old Church Slavic. We use the latter as a proxy for Proto-Slavic, since it is the oldest attested stage of Slavic and comes close to Proto-Slavic. Our database was obtained from translating 46 verb meanings into each language. These meanings were disambiguated by sentences providing specific contexts.⁶ This was necessary in order to make sure that exactly the same reading of the verb meaning is entered into the database for all Slavic languages. The full list of verb meanings is found in (1): (1)

The 46 verb meanings represented in our sample ‘adversary float’, ‘adversary kill’, ‘damage’, ‘explain’, ‘help’, ‘leave behind’, ‘pull’, ‘see’, ‘tell sth.’, ‘threaten’, ‘demand’, ‘forbid’, ‘give’, ‘meet’, ‘play with’, ‘possess’, ‘resist’, ‘search for’, ‘serve’, ‘thank’, ‘obey sb.’, ‘defend’, ‘fear’, ‘like’, ‘listen’, ‘move’, ‘vomit’, ‘wait’, ‘avoid’, ‘bring forward’, ‘forgive’, ‘hate’, ‘name’, ‘play games’, ‘reach’, ‘beg request’, ‘feel pain in’, ‘follow’, ‘hear’, ‘look at’, ‘oppose’, ‘rule/govern’, ‘need’, ‘remind’, ‘think’, ‘disturb/hinder’

Our sample primarily involves verb meanings that populate the middle part of the transitivity prominence scale as found in typological works (Tsunoda 1985, Malchukov 2005, Haspelmath 2015, Say 2014, 2017, 2018: 563, 568, Say 2020). In Haspelmath (2015), transitivity prominence is a value from 0 to 1 that predicts the likelihood of a verb meaning to occur with the transitive construction, i.e. with the NOM–ACC flagging in accusative languages like Slavic, based on cross-linguistic evidence. It is computed as the proportion of languages in which the particular verb meaning is found with the transitive flagging pattern. For example, the upper part of the scale comprises such verbs as ‘break’ (transitivity prominence 1.0 in Haspelmath 2015: 143) or ‘kill’ (1.0). The transitivity prominence value 1.0 means that these verbs do not occur in constructions other than the transitive in the languages of the worldwide sample in Haspelmath (2015). There are only two verbs in our sample (1) that are very high on the transitivity scale, namely, ‘give’ (0.98) and ‘see’ (0.93). Otherwise we have aimed at excluding verbs with a high transitivity prominence from our sample because these verbs are extremely stable and thus may not reveal anything about inner-Slavic variation. ⁶ Similar method has been applied for the collection of the database of two-place predicates BivalTyp from 130 languages of Western Eurasia (www.bivaltyp.info/) in Say (2020).

304

ILJA A . SERŽ ANT ET AL.

Furthermore, we have also excluded the lower part of the transitivity prominence scale since these verbs are equally resistant to variation. They primarily occur as one-argument predicates and are also quite stable across languages, e.g. ‘run’ (0.05), ‘sit’ (0.05), or ‘jump’ (0). These verbs are invariably intransitive in all modern Slavic languages.⁷ Thus, the likelihood that the verbs from the upper and those from the lower end of the transitivity prominence scale would show any variation in argument flagging of Slavic is extremely low. Accordingly, most of the 46 verb meanings in (1) come from the middle part of the transitivity prominence scale; for example ‘search for’ (0.88), ‘name’ (0.80), ‘help’ (0.78) ‘like’ (0.78), ‘tell’ (0.78), ‘follow’ (0.74), ‘look at’ (0.73), ‘meet’ (0.70), ‘fear’ (0.53). The least transitivity-prominent verb we included is ‘play’ (0.10). We also included some verb meanings that are not treated in Haspelmath (2015), e.g. adversity impersonals with ‘(adversary) kill’ or ‘(adversary) float’. In order to make our comparison more rigorous we did not compare simple verb meanings across the Slavic languages but specific sentences that disambiguate the specific readings of these verbs. Furthermore, since Old Church Slavic does not have any native speakers, we started out by collecting Old Church Slavic sentences from the dictionary Kurz ([1966–1997] 2006) for each of the 46 preselected verb meanings. In the second step, we selected those Old Church Slavic sentences that represented best the meanings we were aiming at, and which had all arguments explicitly expressed. These sentences were translated into Russian with only slight adaptations. Subsequently, the 46 Russian sentences were translated into all ten modern languages, taking into account all possible close translational variants. Thus, we made sure that our database contains only comparable meanings across all languages including Old Church Slavic. In addition to native speakers, we also consulted relevant dictionaries and parallel corpora (the latter especially in cases of Belarusian and Ukrainian).⁸ Note that we did not include sentences with negated predicates into the sample, since in some Slavic languages (Polish, East Slavic as well as in Old Church Slavic) flagging is highly sensitive to polarity. Since each of the 46 verb meanings can sometimes be rendered by different predicates and/or by different flagging patterns (case/adposition frames), there are more entries than meanings. We ended up by having c. 80 entries per language. In

⁷ Internal-object verbs (traditionally figura etymologica) such as, for example, skocˇiť skok lit. ‘to jump a jump’, bezˇať beh lit. ‘to run a run’ in Slovak (or Czech) have no bearing on this. ⁸ Belarusian N-korpus (bnkorpus.info/index.html); Maria Shvedova, Ruprecht von Waldenfels, Sergiy Yarygin, Mikhail Kruk, Andriy Rysin, Vasyl Starko, Michał Woźniak (2017–2021): GRAC: General Regionally Annotated Corpus of Ukrainian. Electronic resource: Kyiv, Oslo, Jena (uacorpus.org); Russian National Corpus (ruscorpora.ru).

ARGUMENT FL AGGING ACROSS SL AVIC

305

total, the database contains 825 entries. Synonyms for Old Church Slavic were entered as well if the dictionary Kurz ([1966–1997] 2006) indicated them as such. Note that we did not take into account frequencies of different flagging patterns, which would have been too laborious for this study. The translators were advised to enter all translational variants into the database as long as they were acceptable to native speakers of the standard variety. A number of conventions have been made. Thus, we entered only verbal predicates unless there was a non-verbal predicate that could be considered the default, i.e. one of the most frequent choices like Russ. nuzˇno or nado for ‘need’. All entries were tagged for flagging of the main arguments, both for case and preposition (if applicable). In Macedonian and Bulgarian, the case was disambiguated by (clitic) pronouns since nouns no longer distinguish morphological cases. For example, Bulgarian and Macedonian show the dative case with pronouns and the new, prepositional dative marking na with nouns. Table 10.1 provides some examples from the database with the tagging.⁹ Table 10.1 Structure of the database

Mnogie ljudi slušali ego. ‘Many people listened to him.’ On zašcˇitil menja. ‘He defended me.’ V e˙tix vešcˇax vy nuzˇdaetes’. ‘You need these things.’

Lang.

Example

Predicate

Verb

subj

iobj

dobj

Czech

Mnozí lidé ho poslouchali.

poslouchat

Listen

nom

NA

acc

OCS

zaštititъ nъi (Kiev Missal 6a 13sq)

zaštititi

Defend

nom

NA

acc

Polish

Potrzeba wam tych rzeczy.

potrzeba

Need

dat

NA

gen

NA=non-applicable; OCS=Old Church Slavic

It was crucial for our method that the database contains entries on the same set of verbal meanings for each of the languages. Since the selection of these particular verbal meanings from the middle part of the transitivity–prominence scale was to some extent arbitrary, our data can only be meaningfully used for relative and not for absolute claims about Slavic languages. In this study, we only explore relative differences among the Slavic languages on the basis of our sample. By convention, we aligned all obligees of the necessity modals, experiencers of the experience predicates, inanimate causers of adversary ‘kill’ and ‘float’ and possessors in the predicative–possession construction as the first argument (tagged ⁹ We also introduced other tags that will not be relevant in this chapter.

306

ILJA A . SERŽ ANT ET AL.

as subj in the database in Serzˇant et al. 2021) and the possessum as the second one (tagged as dobject in the database). Note that the accusative of such objects was tagged as acc2 in contrast to the regular accusative (tagged as acc1) in a transitive clause with the nominative-marked first argument (subject). Furthermore, we excluded clausal arguments in order not to complicate matters too much (see the same convention in the valency database ValPal in Haspelmath and Hartmann 2015: 56). Another convention was to unify the spelling of the cognate prepositions. For example, Russian ot ‘from’, Bulgarian, Macedonian, and Polish od and Old Church Slavic otъ were uniformly rendered as ot in the database so that the automated methods would not treat these as different prepositions. The entire database is published online in Serzˇant et al. (2021) and is freely accessible.

10.3 Argument flagging in Slavic Although modern Slavic languages show a high degree of homogeneity in the flagging of the arguments of bivalent verbs (see Section 10.3.3), there are, of course, also differences. In the following two sections, we discuss the variation in the flagging of the object argument along two criteria: the ratio of flagging alternation (Section 10.3.1) and transitivity prominence (Section 10.3.2). Section 10.3.4 discusses the degree of variation of the subject-like argument and Section 10.3.5 presents an overall clustering analysis that takes into account both arguments.

10.3.1 Ratio of flagging alternation In this section, we examine the degree of available object flagging strategies per verb meaning. For example, the meaning ‘wait’ is rendered by the verb ˇcekati in Serbian, which may govern either the accusative or the preposition na:¹⁰ (2)

Serbian a. Oni ˇcekaju Zahar-a 3PL.NOM wait.3PL Zahar-ACC.SG¹¹ ‘They are waiting for Zahar.’

¹⁰ There is a slight difference between (2a) and (2b). Unlike the former, the latter example indicates that the subjects are waiting for Zahar to do a particular thing. For example, (2b) is not optimal in the context where the subject is waiting for Zahar to show up at the appointed place. Similar variation is also found in Slovenian (Žele 2006: 405). ¹¹ Slavic languages have animacy-based DOM systems such that animate nouns may employ the case marker that is homonymous for accusative and genitive (at least in the singular), while inanimate nouns employ markers that are homonymous for nominative and accusative. In what follows, we simply ignore these differences and gloss both types as ACC.

ARGUMENT FL AGGING ACROSS SL AVIC

307

b. Oni ˇcekaju na Zahar-a 3PL.NOM wait.3PL on Zahar-ACC.SG ‘They are waiting for Zahar.’ By contrast, Macedonian requires accusative marking (seen on pronouns): (3)

Macedonian Tie go ˇcekaat Zaharije 3PL 3SG.M.ACC wait.3PL Zahar[M] ‘They are waiting for Zahar.’

Thus, it can be said that ‘wait’ in Serbian has a higher ratio of ( flagging) alternation than the same verb in Macedonian. We measure this ratio by assigning 2 to Serbian and 1 to Macedonian for ‘wait’. Once this procedure has been applied to all verbs of the database, we measure the ratio of alternation for each Slavic language as the mean of the flagging–alternation ratios of all its verbs. Figure 10.1 presents the results. 1.4

1.2

1.0

0.8 Macedonian Polish

Belarusian Bulgarian Serbian Slovenian Ukrainian

Czech

Slovak

Russian

Church Slavic

Fig. 10.1 Ratios of (flagging) alternation

The variation ranges from 1.17 in Macedonian and Polish to 1.51 in Old Church Slavic. Thus, Old Church Slavic had the highest ratio of flagging variation of the object while all modern Slavic languages have been tending towards more rigid case government to varying degrees. The alternating government options of Old Church Slavic contributed to the semantic interpretation of the entire clause in a compositional way, very much in the spirit of construction-based syntax. We take this as indicative that flagging was more semantic in Old Church Slavic and other attested early stages of Slavic (Grkovic´-Major 2007, 2010, see also Bartula 1954, Xodova 1963). By contrast, rigid government is not susceptible to semantic nuances and flagging itself and carries primarily the syntactic function of marking the arguments of the verb. For example, the accusative vs (partitive) genitive alternation was more productive in the older layers of Slavic such as Old East Slavic (Borkovskij and Kuznecov [1963] 2006: 427–428, Krys’ko 2006, Malyševa 2008) and Old Church Slavic (Miklosich 1883: 473–476, Bartula 1969: 67), not only in token but also in type

308

ILJA A . SERŽ ANT ET AL.

frequency. In many contexts, the genitive is no longer possible, for example with verbs meaning ‘see’. In modern Slavic languages, in contrast to Old Church Slavic (4), the (partitive) genitive is no longer available:¹² (4)

Old Church Slavic (Blagova et al. 1994: 242; Euch. 1a 12) zьręšte bo zemlę nedvizˇimy see.NOM.PL.PTCP PTCL earth.F.GEN.SG motionless.GEN.SG.F ‘seeing (a part of ) the motionless earth’

A single exception is the mirative use of videti in Štokavian varieties:¹³ (5)

Vidi ti see.IMP.2SG 2SG.NOM ‘Look at her!’

nje 3SG.F.GEN (indicating surprise by the person’s behaviour)

However, in addition to its archaic flagging alternations, Old Church Slavic had already started replacing bare cases with prepositions—a process that expanded in modern languages, especially in the genealogically closely related Bulgarian and Macedonian. This optionality also contributes to the high ratio of (flagging) alternation of Old Church Slavic. For example, ‘fear’ (bojati sę) is found with the old (source) genitive but also with the new, prepositional government with otъ ‘from’ to become the only government in modern Bulgarian and Macedonian. Another innovation that contributed to the high ratio of alternation is the rise of the transitive possession predicate in Old Church Slavic (iměti ‘have’) alongside the ancient dative marking of the predicative possessor (see Section 10.3.2). Russian seems to be conservative in this regard, retaining a number of alternative flagging patterns. However, at the same time, Russian—similarly to other modern Slavic languages (Section 10.3.2)—also compensated for the loss of some of the alternations by introducing new alternations, primarily, based on the gradual expansion of the transitive pattern, i.e. of the accusative. Thus, Russian has introduced accusative marking alongside the old genitive marking in a number of verbs, e.g. with bojat’sja ‘fear’ which may either take the original genitive (e.g. bojat’sja ucˇitel’nicy ‘to fear the teacher (F.GEN)’) or the new accusative (bojat’sja ucˇitel’nicu ‘to fear the teacher (F.ACC)’, see Nesset and Kuznetsova 2015a, 2015b).

10.3.2 Transitivity prominence In (6), we define transitivity and transitive encoding for Slavic, following the typological definition of Haspelmath (2015: 136; see Haspelmath 2011) in (6): ¹² The partitive meaning of the object often yields the meaning of low degree of affectedness. ¹³ The genitive was also possible in Slovenian in the nineteenth century (see examples in Pleteršnik 1894–1895).

ARGUMENT FL AGGING ACROSS SL AVIC (6)

309

Typological, comparative definition of transitivity (Haspelmath 2015: 136): A verb is considered transitive if it contains an A and a P argument. A and P are defined as the arguments of a verb with at least two arguments that are coded like the ‘breaker’ and the ‘broken thing’ micro-roles of the ‘break’ verb.

Accordingly, we define transitivity in Slavic as follows, given that ‘break’ takes a NOM–ACC case frame in Slavic: (7)

Morphological definition of transitivity in Slavic adopted in this chapter: A verb is considered transitive if its subject argument is in the nominative and its object argument/one of its object arguments is in the accusative case.

Thus, this study does not take transitivity as a semantic notion in the sense of, among others, Hopper and Thompson (1980). Likewise, our definition does not take into account syntactic properties for reasons of feasibility, and only focuses on flagging. For example, not every accusative object can be promoted into the subject role under passivization, in which case there is a serious deviation from the transitive pattern. Conversely, some Slavic languages (particularly East Slavic and Polish) have a certain amount of bi- or trivalent verbs whose object argument (in the active voice) is exclusively or by default coded in the genitive (e.g. Russ. trebovat’ ‘demand’, Pol. unikac´ ‘avoid’) or in the instrumental (e.g. Russ. upravljat’ ‘administer, govern’, Pol. dowodzic´ (wojskiem) ‘conduct (an army)’), but which nevertheless behave syntactically like typical transitive verbs (on Polish, see Z ˙ elazko 1975: 13–79, Buttler 1976: 110–129, 163–169, Pisarkowa 1984: 95–97, Sawicki 1988: 25–31, 47–51, Holvoet 1991: §§7, 8, 10, Lesz-Duk 1995: 7–34).¹⁴ We disregard these differences here according to the definition in (7). In (7), we have defined transitivity for Slavic as crucially based on the accusative marking of the direct object and the nominative marking of the subject. Only these accusatives were tagged as acc1 in our database (Serzˇant et al. 2021). Accusativemarked arguments of the predicates with a non-canonical case marking (tagged as acc2) were thus excluded. This allows us to measure and compare transitivity prominence across Slavic languages. Note that we apply the term transitivity prominence to both verbs and languages. In both cases, the transitivity value is computed as the proportion between the number of transitive patterns and the total number of patterns available for the verb meaning or in the language. That is, transitivity prominence of a particular verb meaning is computed by dividing the number of transitive case assignments, i.e. of the accusatives, by the total number of ¹⁴ In other Slavic languages, the set of bi- and trivalent verbs with objects in the instrumental or genitive case (regardless of negation) has been radically reduced (unless it has not ceased to exist completely) or underlies severe restrictions. As a rule, these cases are replaced by the accusative; see Hausenblas (1958), Lamprecht et al. (1986), Gebauer (2007) on Czech, Grković-Major (2007) on Serbian, and Skwarska (2004). In other cases, a PP appeared instead; see, for instance, Buttler (1976: 134–137), Pisarkowa (1984: 101f., 105f.) on Polish.

310

ILJA A . SERŽ ANT ET AL.

flagging options for the same verbal meaning.¹⁵ For example, the meaning ‘obey sb.’ is found with two verbs in Bulgarian that differ in flagging: accusative (unmarked on nouns) in (8) and dative (marked with preposition na) in (9): (8)

(9)

Bulgarian I ti slušaj PTCL 2SG.NOM listen.IPFV ‘Obey your father!’

bašta si father REFL.POSS

Bulgarian I ti se podcˇinjavaj na PTCL 2SG.NOM REFL obey.IPFV on(=DAT)¹⁶ ‘Obey your father!’

bašta si father REFL.POSS

In this case, transitivity prominence of this verb meaning (0.5) is computed as the number of accusatives, i.e. 1, divided by the total number of flagging options for the object, i.e. 2. By contrast, transitivity prominence of ‘hate’ in Ukrainian is 1 (= 1/1), the only available object flagging here is accusative: (10) Ukrainian Vin nenavydyt’ tebe 3SG hate.IPFV.3SG 2SG.ACC ‘He hates you and your father.’

i CONJ

tvoho bat’ka 2POSS.SG.ACC father.ACC.SG

We compute transitivity prominence of each language by averaging the transitivity prominence of all verb meanings in the language. The results are presented in Figure 10.2.

0.6

0.5

0.4

0.3 Polish

Ukrainian Russian Belarusian

Church Slavic

Czech

Serbian

Slovenian Macedonian Slovak

Bulgarian

Fig. 10.2 Transitivity prominence of different Slavic languages ¹⁵ Note that the number of accusative assignments may be higher than 1 in those cases where a particular meaning is rendered by two or more different lexical verbs (close synonyms) in the language, each of which take the accusative. ¹⁶ Note that the dative case is seen on clitic pronouns in Bulgarian, but when it comes to nouns, the dative preposition na ‘on’ is used, e.g. podcˇinjavaj mu se (obey.IPFV 3SG.DAT.M REFL) ‘Obey him’.

ARGUMENT FL AGGING ACROSS SL AVIC

311

The transitivity prominence varies from 0.37 in Polish to 0.627 in Bulgarian. The higher the value, the higher the transitivity prominence of the language. Thus, Polish has the lowest transitivity prominence, Bulgarian the highest. Importantly, data collected independently from ours (Say 2020, map ‘transitivity ratios’) point to the same result (based on the Hemming distance): Polish patterns with East Slavic (Say (2017) has only data for Ukrainian and Russian) (Say 2017: 734) (Figure 10.3). Transitivity

0.40 0.45 0.50 0.55 0.60

Old Church Slavic

300 km 200 mi

Fig. 10.3 Transitivity degree of Slavic languages

Notably, the differences in the transitivity prominence are not randomly distributed. First, we observe genealogical clusters for South and East Slavic. All modern South Slavic languages, in addition to Slovak, occupy the highest ranges of transitivity prominence and thus pattern alike. Likewise, East Slavic languages also pattern all alike, but on the lowest edge of the transitivity scale (Figure 10.2).

312

ILJA A . SERŽ ANT ET AL.

By contrast, the West Slavic languages do not form a cluster here; they are most scattered: while Polish patterns with East Slavic, Slovak goes with South Slavic, and Czech takes an intermediate position on the scale. Thus, it is likely that, in addition to the genealogical factor, there is also the areal factor of language contact. This would explain why East Slavic and Polish form a cluster despite the fact that Polish belongs to a different subfamily. Likewise, this would support earlier claims about South Slavic impact on Slovak (Krajcˇovicˇ 1974, Nuorluoto 2010, Greenberg 2017: 178–179). Second, and more generally, there is a clear geographical trend from the less transitive and—given the evidence from Old Church Slavic—possibly more conservative Northeast (East Slavic plus Polish) to the more innovative Southwest starting from Czech to even stronger transitivity prominence in the South Slavic languages Bulgarian and Macedonian, see Figure 10.3.¹⁷ The comparison with Old Church Slavic shows that the high transitivity prominence of Bulgarian and Macedonian must be diachronically an innovation. South Slavic languages have increased their transitivity prominence in the course of their development (Gortan-Premk 1971: 159, Sobolev 2009, Grkovic´-Major 2010: 67). By contrast, East Slavic languages and Polish are more conservative in this regard and have moved towards even less transitivity than Old Church Slavic. Say (2017: 728) shows that there is a strong negative correlation between the number of cases and the transitivity prominence of the language such that more cases correlate with a lower transitivity prominence and vice versa. Note that nondirect objects can also be marked by various types of adpositions as well and thus still exhibit an intransitive pattern (Say 2014: 139). In any event, the demise of case systems may have a diachronic effect—possibly only temporarily—on the transitivity prominence of the language. Indeed, the rise of transitivity prominence in South Slavic is in some cases due to the loss of nominal case inflection in Bulgarian and Macedonian, which is, in turn, also areally motivated (Sandfeld 1926/1930, Joseph 2010). Consider the verb izbe˘gati ‘avoid’ in Old Church Slavic: (11)

Old Church Slavic (Kurz [1966–1997] 2006: I: 724; tit. ad Ps 141 Pog Bon.) jako da my navyknemь kymь obrazomь … izbe˘gati how PTCL 1PL learn.1PL which.INS way.INS avoid.INF zъla evil.GEN.SG ‘How are we to learn which way to avoid evil (things)?’

¹⁷ The map was created in R with the map.feature function (Moroz 2017).

ARGUMENT FL AGGING ACROSS SL AVIC

313

This verb takes the genitive in most of the modern Slavic languages (where it exists) as well. By contrast, Bulgarian, Macedonian, and Serbian take the accusative with this verb—see Bulgarian: (12)

(13)

Bulgarian Te go izbjagvat 3PL 3SG.ACC.M/N avoid.3PL ‘They avoid him/this man.’ Bulgarian Te izbjagvat tozi ˇcovek 3PL avoid.3PL DEM.ACC/NOM man.ACC/NOM ‘They avoid him/this man.’

In modern Eastern South Slavic (Balkan Slavic), the former genitive marking— originally required by the prefix iz-—is lost as the genitive case is entirely lost here. By contrast, the dative case is retained in the declension of pronouns. Accordingly, verbs that originally required dative flagging on (one of ) their objects (‘help’, ‘explain’, ‘give’, etc.) also retain the dative marking in Eastern South Slavic: (14)

Macedonian Taa mi pomogna so taa rabota 3SG.F 1SG.DAT help.AOR.3SG with DEM.SG.F work.SG.F ‘She helped me with this work.’

Finally, the expansion of the transitive pattern is also found in the accusative– dative syncretism in Southern Macedonian dialects (Buzˇarovska 2001, 2020). Note that the decrease in the number of morphological cases per language on nouns is not a genuinely Balkan Slavic phenomenon, it is due to a larger macroareal cline from East to West, see Figure 10.4 (Lazard 1998: 106–107, Iggesen 2013). Another factor contributing to the increase of the overall transitivity prominence particularly in South and West Slavic is the expansion of the transitive havetype predicative–possession constructions that is lexically based on an Aktionsart derivation from jęti ‘take’. Proto-Slavic—as well as its ancestor PIE—originally relied on the locational strategy to code predicative possession, which was based on the dative marking of the possessor and the existential verb ‘be’ (Meillet 1923, Isacˇenko 1974: 44–45). This strategy was also retained in Old Church Slavic (alongside the have-type with ime˘ti ‘have, possess’) but it entirely disappeared from all modern Slavic languages. All modern Slavic languages—except East Slavic—generalized the have-type (Safarewiczowa 1964, Grkovic´-Major 2007, Clancy 2010, 2020). East Slavic has modified the ancient possessive construction with the dative possessor to prepositional possessor marked by the preposition u

314

ILJA A . SERŽ ANT ET AL.

Values No morphological case-marking 2 cases 3 cases 4 cases 5 cases 6−7 cases 8−9 cases 10 or more cases

Fig. 10.4 Number of morphological cases in languages of Europe Source: based on Iggesen (2013).

‘at’ (15) in accordance with a more general tendency of East Slavic to expand this preposition in the functional domain of the dative (Serzˇant 2015b).¹⁸ (15)

Russian U menja est’ at 1SG.GEN COP.PRS ‘I have money.’

den’gi. money.NOM.PL

The distribution of the two different predicative–possession constructions in Slavic also follows the macro-areal cline seen in Figure 10.5. East Slavic, and especially, Russian follows the cline here. Similarly to the distribution of case loss (Figure 10.4), the trend towards the transitive have-type expands from Northeast (Finnic, Baltic, and East Slavic languages) to the West and South of Europe, as can be observed on Figure 10.5. Needless to say, this areal effect is the result of various and possibly independent language-contact situations. ¹⁸ Note that the PP u ‘at’ may also denote the affectee (sometimes misleadingly referred to as external possessor, see Serzˇant 2016) in East Slavic. Moreover, attested is such usage as in older Serbian (nowadays obsolete), e.g. (‘Srpska devojka’, a folk poem): U Milice duge trepavice (at Milica.GEN long.NOM.PL eyelash.NOM.PL) ‘Milica has long eyelashes’. Note that this need not be a possessive construction since its meaning is not to convey that *‘Milica possesses long eyelashes (somewhere)’ but rather to say that ‘her eyelashes are long’. Compare the difference in Russian: U Milicy est’ dlinnye resnicy which is a true possessive construction but which does not yield a meaningful sentence vs U Milicy dlinnye resnicy ‘Milica’s eyelashes are long’ (lit. ‘To Milica, eyelashes are long’). The latter is demonstrably syntactically different from the former (Serzˇant 2012: 378).

ARGUMENT FL AGGING ACROSS SL AVIC

315

Values Locational Genitive ‘Have’

Fig. 10.5 Predicative possession marking in Europe Source: based on Stassen (2013).

Interestingly, the locational strategy developed and became again obsolete in some Štokavian dialects (West South Slavic) due to competition with the ‘have’ verb (Piper et al. 2005: 146, Mrazovic´ 2009: 398)—see (16): (16)

Štokavian (Brozovic´ 2010: 1659; South Slavic; Lug, HerzegovinaNeretvian Canton, place 58) U njih ima ´cuko at 3PL.GEN have.3SG dog.NOM.SG ‘They have a dog.’

The loss of this strategy in this South Slavic variety reinforces our claim that an areal effect is at work here. We now turn to the split among West Slavic languages that range from the lowest transitivity prominence in Polish (0.37) via Czech (0.551) to Slovak (0.598). While Czech and Slovak have otherwise similar flagging patterns (Figure 10.7 in Section 10.3.5), one of the most striking differences between the two is their degrees of transitivity prominence. Many verbs with genitive object marking in Czech take the accusative case in Slovak, e.g. Cz. nechat predominantly with genitive ‘stop doing something’ (Hausenblas 1958: 175) vs Sk. nechať which takes the accusative with the same meaning. This is also true of some other verbs or

316

ILJA A . SERŽ ANT ET AL.

meanings which are not included in our database, e.g. Cz. dosahovat ‘arrive at a place’ with genitive vs the synonymous verb Sk. dosahovať with accusative.¹⁹ In other cases, the accusative is available alongside the original dative in Slovak, while only the dative case is found in Czech, e.g. Sk. nacˇúvať ‘listen’ (dat/acc), Sk. pocˇúvať ‘listen’ (acc) vs Cz. naslouchat (dat), Sk. napomáhať ‘bring forward, favour’ (dat/acc) vs Cz. napomáhat (dat). Thus, Slovak shows a more extensive spread of the accusative to the detriment of the dative and the genitive case when compared with Czech. Having said this, Czech also attests some expansion of transitivity prominence (Gebauer 2007: 331). Although verbs with the prefix do‘until’ originally assigned the genitive, some of these verbs substituted the genitive with the accusative (Stanislav 1973). The recessive nature of the genitive marking has been attested in Czech since the nineteenth century (Hausenblas 1958: 169). The expansion of the accusative flagging is also found in non-prefixed verbs both in Slovak and Czech, e.g. Cz. zˇádat/Sk. zˇiadať ‘request’, Cz. hledat/Slk. hl’adať ‘search’, Cz. ˇcekat/Slk. ˇcakať ‘wait’ (Trávnı´cˇek (1938) for Czech; Stanislav (1973) for Slovak). For all these verbs, the accusative is currently used instead of the former genitive. Still, the expansion of the transitivity pattern is more advanced in Slovak than in closely related Czech. Genitive object flagging is recessive in other Slavic languages as well. Thus, the verbs ˇcakati ‘wait’, zahtevati ‘request’ or iskati ‘search’ are only used with the accusative object flagging in present-day Slovenian, while the original flagging was the genitive as evidenced by the historical IMP corpus of Slovenian.²⁰ Old Church Slavic exclusively attests the genitive with these verb meanings. While the hotbed of the expansion of the transitive pattern is found in the Southwest of the Slavic area (Figure 10.4), this process is also found in the East Slavic languages, albeit to a minor degree. The genitive marking (alongside the accusative marking) is still widely encountered, e.g. with zˇdat’ ‘wait’, bojat’sja ‘fear’ (Nesset and Kuznetsova 2015a, 2015b) or iskat’ ‘search’ in Russian. Finally, in contrast to the other Slavic languages, the genitive marking is the almost only option in standard Polish with verbs like ‘wait’, ‘search’, or ‘listen’. Accusative flagging is quite rare in Polish with słuchac´ ‘listen’ or szukac´ ‘search for’ and may only be found in colloquial or regional speech, most frequently with pronouns.²¹ Thus, it does not come as a surprise that Polish clusters with East Slavic when it comes to transitivity prominence (Figure 10.2). This finding is supported by independent quantitative studies (Say 2014, 2017, 2018, 2020), which also find ¹⁹ Note that this verb is used in a different meaning in our database, i.e. ‘reach’, with a different argument coding pattern. ²⁰ http://nl.ijs.si/imp/index-en.html. ²¹ Our database does not allow for taking into account different frequencies. For this reason, the decision about whether or not to include the accusative option was to some extent subjective. Since accusative flagging is still quite rare for Pol. słuchać ‘listen’ (e.g. słuchać chorą ‘listen to the sick’ in a medical examination (Markowski 2002)), we did not include this option. By contrast, szukać ‘search for’ is listed as allowing both the genitive (the more frequent option) and the accusative (most frequently with pronouns, to be seen in the feminine declension).

ARGUMENT FL AGGING ACROSS SL AVIC

317

Russian and Polish to have very low transitivity prominence among the languages of Europe (Say 2014: 136, 138, 2018: 577, Say 2020). The similarity between East Slavic and Polish cannot be motivated genealogically since Polish belongs to the West Slavic branch. It is thus likely to assume that diverse contacts occurring over the last 600 years between Polish and East Slavic languages (i.e. after an earlier split between East and West Slavic) must have contributed to the retention of the original, genitive, or dative patterns; see Section 10.3.5. To summarize, the increase in transitivity prominence observed in many modern Slavic languages follows a geographical cline from Northeast (Belarusian, Russian, Ukrainian, Polish) to the South and the West (Bulgarian, Czech, Macedonian, Serbian, Slovak, Slovenian) with Balkan Slavic scoring the highest. Since the distribution of transitivity prominences in modern Slavic (Figure 10.3) is neither random nor solely driven by the degree of genealogical closeness, we shall assume that various language-contact situations are responsible for this cline. We have argued that the southward expansion of transitivity prominence might be reinforced by such areal clines as the loss of nominal case (Figure 10.4) and the expansion of the transitive possession construction (Figure 10.5) which shows similar geographic distribution. Moreover, the expansion of the transitive pattern in Slovak but also in Czech might also have been influenced by German, which equally shows relatively high transitivity prominence (Say 2014: 136). Importantly, and more generally, this cline independently established within Slavic is part of the larger European trend towards increasing transitivity prominence from Northeast to Southwest, established in Say (2014: 136) on the basis of 29 languages from geographical Europe (from Komi-Zyrian in the East thereof to Basque in its West) (see also Say 2017, 2018: 577). Furthermore, the diachronic evidence for Slavic—as observed in comparison with Old Church Slavic—suggests that Southwestern Slavic languages have increased the transitivity prominence while East Slavic and, especially, Polish decreased it, see Figure 10.2 and Figure 10.3. Other Indo-European languages likewise increased their transitivity prominence; see Say (2014: 136), who observes the same diachronic trend when comparing Modern Greek with Ancient Greek (see also Say 2017: 736, 743). This suggests that high transitivity prominence in the South (and West) of Europe is a local, areal property that was not brought by the Slavic languages moving into these regions but which rather itself heavily affected these languages upon arrival in the Balkans (see also Buzˇarovska 2001, 2020).

10.3.3 High homogeneity across the Slavic languages In this section, we test whether or not the differences among the Slavic languages illustrated so far are statistically significant. For testing the differences in the alternation ratio (Figure 10.1) and transitivity prominence (Figure 10.2), we ran the Poisson regression model with glm in R (R Core Team 2018) with

318

ILJA A . SERŽ ANT ET AL.

Polish as the baseline. Neither the alternation ratio nor the transitivity prominence reaches statistical significance. Even the distinction between Polish and Bulgarian in terms of transitivity prominence did not reach statistical significance (p = 0.079), the differences between Old Church Slavic and other languages were even less significant. Obviously, the primary reason for why our data do not reach statistical significance is that modern Slavic languages have not considerably diverged from Proto-Slavic (for which Old Church Slavic is our proxy); the entire family still remains quite homogeneous when it comes to argument flagging, and the differences are not sufficiently numerous. This is despite the fact that we purposefully aimed at those verbs that tend to show more variation cross-linguistically, i.e. those verbs that are likely to be more amenable to changes in argument coding (see Section 10.2). To illustrate the high homogeneity of our database, Table 10.2 provides the verbs and the number of object flagging patterns they show. Table 10.2 The number of object-marking patterns across all 46 verb meanings in different Slavic lgs verb meanings

number of patterns

‘damage’ (acc),a ‘explain’ (acc), ‘help (dat)’, ‘leave behind’ (acc), ‘pull’ (acc), ‘see’ (acc), ‘tell sth.’ (acc), ‘threaten (acc)’ ‘demand’, ‘forbid’, ‘give’, ‘meet’, ‘play with’, ‘possess’, ‘resist’, ‘search for’, ‘serve’, ‘thank’ ‘obey sb.’, ‘defend’, ‘fear’, ‘like’, ‘listen’, ‘move’, ‘vomit’, ‘wait’ ‘avoid’, ‘bring forward’, ‘forgive’, ‘hate’, ‘name’, ‘play games’, ‘reach’ ‘beg request’, ‘feel pain in’, ‘follow’, ‘hear’, ‘look at’, ‘oppose’, ‘rule/govern’ ‘need’, ‘remind’, ‘think’ ‘disturb/hinder’

1 2 3 4 5 6 8

a

While there are different meanings of ‘damage’ (with different case assignments), our database features only one particular meaning ‘destroy (e.g. a church)’.

Eight out of 46 verb meanings (the first row in Table 10.2) do not show any differences in object-marking pattern at all and are consistently transitive across all Slavic languages. Thus, more than 17 per cent of our verb meanings do not show any variation across Slavic at all. Another ten verb meanings (the second row) alternate only between two flagging options. For example, the possessee of ‘possess’ may either be coded by the accusative (have-strategy) or by the nominative (the locational strategy), but even this class does not significantly contribute to variation within Slavic. This is shown by the fact that the verb ‘possess’ is invariably transitive outside of East Slavic (see also example (16)). Likewise, ‘give’ allows the (partitive) genitive case alongside the default accusative marking of the direct

ARGUMENT FL AGGING ACROSS SL AVIC

319

object (the theme) only in East Slavic and Polish. Another type of alternation is found with those verbs that recently started allowing accusative instead of the former genitive (e.g. ‘search for’, ‘wait’ (third row), ‘listen’, ‘need’ (the last but one)). This variation, by contrast, affects many Slavic languages (Section 10.3.2). The high degree of homogeneity of Slavic even with less stable verb meanings is an important finding of this study. Finally, the high degree of diachronic stability and homogeneity of transitivity prominence is an important result per se, as it means that transitivity prominence has not changed above statistical significance despite a number of restructurings of the original Proto-Slavic flagging system—especially in Eastern South Slavic, which lost all six nominal cases found in Old Church Slavic and developed object indexing instead. We also mentioned the recency of the split as well as geographical compactness of modern Slavic as factors that contributed to homogeneity (Section 10.1). In addition, we may tentatively suggest that transitivity prominence in general is diachronically relatively stable, certainly more stable than case inflection that instantiated it (see also Say 2014: 160, 2017: 743).

10.3.4 Ratio of nominative marking of the subject-like argument In the previous sections (10.3.1–10.3.3), we have examined the variation in object flagging. In this section, we focus on the encoding of the subject or the subject-like argument. Recall from Section 10.2 that the subject-like arguments include experiencers of the experience predicates, obligees of the necessity modals, inanimate causers of adversary ‘kill’ and ‘float’ as well as the possessors of the predicative– possession predicates. Modern Slavic languages allow only a small subset of verbs that require a non-nominative flagging on the subject-like argument. However, there are some minor differences across Slavic languages too. First, South Slavic languages score the highest with respect to the average number of nominative flagging in the database. This is in concord with their general tendency to rely more on the transitive pattern (nominative–accusative) than other Slavic languages (recall Figure 10.2). Figure 10.6 illustrates this. We computed the ratio of nominative marking of subject-like arguments as the proportion between the number of the nominative marking (i.e. either 1 or 0) and the total of all other available marking options for the subject-like argument. This has been done for each particular verb meaning in each language. Subsequently, these values were averaged per language and put on a heat map, see Figure 10.6. With the disclaimer that the differences between the languages are not statistically significant, we may tentatively draw the following cautious conclusions. First, there is again a clear areal and genealogical trend here such that East Slavic shows the lowest ratio of nominative marking (differently in Say 2020, map ‘X ratios’). These languages have the highest number of case frames with a non-nominative

320

ILJA A . SERŽ ANT ET AL. Nominative subjects

0.88 0.90 0.92 0.94 0.96

Old Church Slavic

200 km 100 mi

Fig. 10.6 Nominative marking of subject-like argument across Slavic

flagging on the subject-like argument. By contrast, Bulgarian has the lowest number. The only two verb meanings that allow the non-nominative flagging of the subject-like argument in Bulgarian is bolja ‘ache, feel pain’ (the experiencer is marked by the accusative case) and ‘like’ (haresva). The latter may either be coded by the nominative or dative experiencer. Likewise, adversity impersonals with the causer in the instrumental case—illustrated here by an example from Ukrainian in (17)—are considerably less productive outside East Slavic; nor are we aware of any instances of these in Old Church Slavic, which seems to rely on the transitive pattern here (see example (19)). Adversity impersonals are active constructions with the causer marked with the instrumental and the patient with the accusative case (Babby 1994, Mustajoki and Kopotev 2005)—see (17): (17)

Ukrainian Joho vbylo blyskavkoju 3SG.M.ACC kill.PST.SG.N lightning.INS.SG ‘Lightning killed him.’

ARGUMENT FL AGGING ACROSS SL AVIC

321

Outside of East Slavic, adversity constructions with an instrumental causer occur only in Polish and are almost non-existent in Slovak; in South Slavic they do not occur at all (Schlund 2020: 41–42). Though, even in Polish, this construction with an explicit instrumental causer is much less frequent than in East Slavic. This is true in terms of token frequency, as research with parallel corpora shows (Schlund 2020: 47–48), but also in terms of type frequency (Schlund 2020). The two adversity meanings in our database (‘The boat got carried away by the wind’ and ‘Lightning killed him’) cannot be rendered by this construction in either West or South Slavic languages; instead, these attest the transitive pattern for these two meanings—see (18) from Slovak: (18)

Slovak Zabil ho blesk. kill.PST.3SG 3SG.M.ACC lightning.NOM.SG ‘Lightning killed him.’

Likewise, Old Church Slavic employs the transitive pattern, as the following example (19) shows, whereas Russian would have preferred the INS–ACC adversity impersonal for this meaning (20b). (19)

Old Church Slavic (Euch. 34a 5; Kurz ([1966–1997] 2006: I, 115) bliskъ izъmetъ zrakъ lightning.NOM.SG throw.AOR.3SG eye.ACC.SG ‘lightning struck out (his) eye’

The transitive construction (19) would be somewhat infelicitous in Russian (20a) while the adversity construction (20b) is idiomatic with this input: (20) Russian a. ? Molnija vybila emu glaz lightning.NOM.SG hit.PST.SG 3SG.M.DAT eye.ACC.SG ‘lightning struck out (his) eye’ b. Emu vybilo glaz molniej 3SG.M.DAT hit.PST.SG eye.ACC.SG lightning.INS.SG ‘lightning struck out (his) eye’ Given only the evidence ex negativo for Old Church Slavic we cannot be entirely sure that the adversity construction did not exist in this language. However, crucially, the fact that the transitive construction (19) was the only one attested with this lexical input can be taken as sufficiently indicative that Old Church Slavic did not have a productive adversity construction to the extent East Slavic languages do. Diachronically, the high ratio of nominative marking is a conservative feature in Slavic since Old Church Slavic does not attest non-nominative flagging even with the verb ‘to ache, feel pain’ (bole˘ti), which takes a non-nominative experiencer

322

ILJA A . SERŽ ANT ET AL.

in all modern Slavic languages. Instead, this verb requires the nominative marking on the experiencer and the locative marking on the body part (see Serzˇant and Bjarnadóttir (2014) on the diachronic account for Russian). This option is ungrammatical in the modern Slavic languages. It follows that modern South Slavic languages are conservative and modern East Slavic languages are innovative in this respect, West Slavic languages taking an intermediate position with Polish, again, being closest to East Slavic. Thus, our data do not seem to confirm an allegedly inherited nature of the socalled non-canonical subjects and earlier inactive alignment retained in modern Slavic languages (pace Barðdal et al. 2012, Barðdal and Smitherman 2013, see methodological criticism in Serzˇant 2015a). Quite to the contrary, our evidence suggests that there is a reverse trend from stronger nominative marking towards introducing more non-nominative subject-like arguments. This trend is particularly strong in East Slavic and in the Circum-Baltic area (Serzˇant 2015b), which two of the three East Slavic languages (Russian and Belarusian) as well as Polish belong to. Likewise, quantitative data on other Indo-European languages also suggest an increase of non-nominativeness from earlier languages to modern languages, see the figures on Modern Greek (5 to 6 per cent of all verbs in Say 2020) vs Ancient Greek (below 3 per cent). Moreover, we may conclude that the distribution of the nominativeness ratios is not random and that it is not determined genealogically since Old Church Slavic does not attest non-nominative subject-like arguments with our verb meanings. Accordingly, the areal trend is primarily driven by innovation and language contact here.

10.3.5 Argument-flagging clusters within Slavic In this section, we aggregate the variation. In order to reveal mutual influences in argument flagging, we have explored how Slavic languages cluster among each other with regard to flagging. While the previous sections focus on the flagging patterns of one argument (the object or the subject), in this section, we compare the entire case frames, i.e. flagging of the subject, object and, where applicable, indirect object. Table 10.3 illustrates the verb meaning ‘leave sth. to sb.’ in Slavic. It can be observed that, except for an alternative pattern in Czech and Slovak, all Slavic languages share the same pattern: NOM–DAT–ACC.²² They thus would cluster together if only this verb meaning were taken into consideration.

²² Bulgarian and Macedonian show the dative case with pronouns and the new, prepositional dative marking na with nouns.

ARGUMENT FL AGGING ACROSS SL AVIC

323

Table 10.3 Case frames and verbs for ‘leave sth to sb’ Belarusian Bulgarian Czech Czech Macedonian Polish Polish Russian Serbian Slovak Slovak Slovenian Ukrainian

pakinuc’ ostavja nechat nechat ostavi zostawić pozostawić ostavit’ ostaviti nechať nechať zapustiti zalyšyty

nom nom nom nom nom nom nom nom nom nom nom nom nom

dat dat dat pro dat dat dat dat dat dat pre dat dat

acc acc acc acc acc acc acc acc acc acc acc acc acc

In order to compute the mutual–similarity index for all pairs of languages with respect to all verb meanings, we counted the proportion of shared flagging patterns (Jaccard similarity), i.e. the number of shared flagging patterns for each language pair was divided by the total of the flagging patterns of the pair taken jointly (R function dist, Meyer and Buchta 2019). In the next step, with Jaccard similarity values as the input, we performed hierarchical clustering analysis via hclust (R Core Team 2018). The result is plotted in Figure 10.7. Similarity dendogram of Slavic languages

ukrainian

0.2

belarusian

maced

bulgarian

serbian

slovenian

0.3

slovak

0.4

czech

0.5

russian

0.6

polish

old_church

0.7

Fig. 10.7 Similarity dendrogram of Slavic languages

The dendrogram should be interpreted as follows. Those languages that form a close cluster—i.e. have the highest number of identical flagging patterns—are represented as branching sisters (e.g. Belarusian and Ukrainian). Languages outside

324

ILJA A . SERŽ ANT ET AL.

of the node have less intersecting flagging patterns, etc. Based on this, the following observations can be made. First, the clusters in Figure 10.7 mirror the degree of genealogical closeness, which means that genealogical relations are one of the main factors conditioning correlations in argument flagging. Thus, Slovenian clusters with Serbian, Bulgarian with Macedonian and both pairs form a common branch, thus corresponding to the genealogical and areal subbranch of South Slavic. Likewise, Czech clusters together with Slovak (both West Slavic), while Belarusian clusters with Ukrainian and both, somewhat more remotely, with Russian (East Slavic). However, there are also exceptions: Polish (West Slavic) and Old Church Slavic (South Slavic), both showing an external but close relationship to East Slavic. Indeed, the linguistic history of Polish is quite different from other West Slavic languages. It has long-standing mutual contacts with Russian, Belarusian, and Ukrainian (Moser 1998, Wiemer 2003, Grenoble 2015). It is thus not unexpected that argument flagging of Polish is influenced by East Slavic and vice versa. Second, it follows from Figure 10.7 that East Slavic and Polish are closest to Old Church Slavic when it comes to flagging. This can only mean that these languages are the most conservative Slavic languages regarding flagging, since any close genealogical relation with Old Church Slavic (which is a South Slavic language) can be safely excluded for either of them. For example, consider the object flagging strategies of ‘govern, rule’: Russian (pravit’), Belarusian (kiravac’), Ukrainian (pravyty) and Polish (kierowac´, rządzic´) require the old instrumental or (as with Pol. panowac´) the new prepositional flagging nad ‘above’. By contrast, the other two West Slavic languages (Czech and Slovak), in addition to the new prepositional government nad ‘above’, take the dative and no longer the instrumental case.²³ Interestingly, the instrumental government in Polish and in East Slavic is not a pattern borrowing. It represents the original Common Slavic government as is witnessed by Old Church Slavic, i.e. the instrumental case. There are more correlations between East Slavic, Polish, and Old Church Slavic of this kind. This suggests that—in addition to contact-induced convergence—Polish and East Slavic have retained the original flagging patterns more faithfully than other Slavic languages. Yet, the retention of the same properties in geographically neighbouring languages as opposed to non-retention in geographically more distant languages is likely to be motivated by language contact as well (Serzˇant, forthcoming (a), forthcoming (b)). Third, at the same time, despite long-standing contacts between Ukrainian and Polish (as well as between Belarusian and Polish) that appear so pervasively in the lexicon (Shevelov 1952, Richhardt 1957, Łesiów 1998), case and prepositional

²³ The original instrumental case is still attested for vládnuť in Slovak and vládnout in Czech with the meaning ‘be dominant, prevail, rule’ in the idiomatic expressions nevďak vládne svetom (Sk.), nevděk vládne světem (Cz.) ‘ingratitude rules the world’.

ARGUMENT FL AGGING ACROSS SL AVIC

325

government seem to be less affected here so that the distance in the genealogical relation between Polish and the two East Slavic languages plays out stronger in Figure 10.7—Polish clusters only at the node of East Slavic. Fourth, Belarusian and Ukrainian are more closely clustered with each other than with Russian, although all three languages belong to the same subbranch. This does not seem to be genealogically motivated since Belarusian is no more related to Ukrainian than to Russian, as is suggested, for example, by Shevelov’s genealogy (1953: 93) in Table 10.4. Table 10.4 Dialect continua that have formed modern East Slavic languages 11th c. 16th c.

Novgorod-Suzdal’ Polock-Rjazan’ Kiev-Polesie Galicia-Podolia Russian Belarusian Ukrainian

Source: Shevelov (1953: 93).

It is thus more likely that there is another factor at play in addition to the genealogical factor that have influenced a stronger mutual similarity of Ukrainian with Belarusian than with Russian. This factor is language contact. Ukrainian and Belarusian have introduced common innovations in flagging that are not found in Russian for the given meanings. For example, the object of the verb meaning ‘think’ may be coded either with accusative (only the question word in Belarusian što ‘what’ and Ukrainian šcˇo ‘what’) or, most frequently, with the preposition pro ‘about’. The use of this preposition is likely to be an innovation in these two languages as it is not found with this verb in Old Church Slavic. By contrast, Russian employs the preposition o ‘about’ here as the main option, while pro without a regional connotation is only possible to a limited extent with this verb in standard Russian.²⁴ Another common—only Belarusian–Ukrainian innovation—is the flagging of the experiencer of the verb meaning ‘ache’. Only these two languages allow dative marking in addition to prepositional flagging with u ‘at’ in Ukrainian (as in Russian): (21)

Belarusian Im balic’ se˙rca 3PL.DAT ache.PRS.3SG heart.NOM.SG ‘They have heartache.’

(22)

Ukrainian Jim bolyt’ serce 3PL.DAT ache.PRS.3SG heart.NOM.SG ‘They have heartache.’

²⁴ Thus, this preposition cannot be used in Russian to render the meaning of ‘think’ that we collected in our database, namely, ‘What do you think?’ without a strong regional connotation.

326

ILJA A . SERŽ ANT ET AL.

This is another common Belarusian and Ukrainian innovation. The possibility of there being also some influence of Baltic is an open question. Compare Lith. skaude˙ti ‘ache’, which has two flagging patterns with this verb meaning: either the same as in Belarusian and Ukrainian (see Serzˇant 2015b), or the body part is marked with accusative. In Old Church Slavic, this verb meaning (coded by the verb bole˘ti) requires the nominative case of the experiencer and the locative case of the affected body part (see also Serzˇant and Bjarnadóttir 2014). The Belarusian and Ukrainian pattern is thus clearly an innovation.

10.4 Conclusions In this pilot study, we have examined the variation in the flagging patterns across ten modern Slavic languages covering all three major Slavic branches: South, West, and East Slavic. Our goal was to examine and identify diachronic and areal trends that constrain the argument flagging patterns of modern Slavic. Before we recapitulate the main results, a note of caution is in order. Since the differences between the languages did not reach statistical significance—which in itself is an indication of high homogeneity in this domain across Slavic—our results are preliminary and should be corroborated on the basis of a considerably larger data set than ours (825 entries). First, by comparing flagging–alternation ratios across Slavic, we found that Slavic languages have consistently reduced the number of available case frames for each verb, Old Church Slavic attesting the highest flagging–alternation ratio. The reduction of the alternation ratio in favour of rigid, uniform government is particularly strong in Polish, Macedonian, and Bulgarian, whereas Slovak and Russian are less rigid with respect to object flagging.²⁵ While flagging alternations in Old Church Slavic were often related to slightly different meanings—compare, for example, the alternation between the accusative and the partitive genitive—this is different in Slovak. Here, the somewhat higher flagging–alternation ratio is not only due to the retention of the original Slavic patterns for each verb but is primarily due to the gradual expansion of the transitive pattern. The transitional stage, at which both, the original and the new, accusative flagging are possible, boosts the flagging–alternation ratio of this language. However, in contrast to many flagging alternations found in Old Church Slavic, the Slovak alternations are not semantically driven but result from the expansion of the transitive pattern. Second, when it comes to transitivity prominence, we have established an areal trend which splits Slavic languages into Northeast Slavic (Belarusian, Polish, Russian, Ukrainian) and Southwest Slavic (all other languages) such that the former ²⁵ Note that we did not take into account the genitive alternation conditioned by the negative polarity in languages such as Polish.

ARGUMENT FL AGGING ACROSS SL AVIC

327

group shows relatively low and the latter relatively high transitivity prominence. This trend has been established on large-scale cross-linguistic data that was collected independently from our study in Say (2014, 2017, 2018). Interestingly, this split is also seen in the ratio of nominative marking, albeit to a minor degree. Here, too, Slavic languages in the northeast have a lower ratio than the ones in the southwest. However, the diachronic underpinnings are different with these properties: while the high transitivity prominence of South Slavic is an innovation, the high nominativeness ratio of South Slavic is, conversely, an archaism. We have argued that the eastward decrease in transitivity prominence and in the ratio of nominative marking might be partly supported by such macro-areal clines as more morphological cases in the East of Eurasia as opposed to fewer or no cases in the West and the Southwest of it. Likewise, the macro-areal trend with the preference for the locational possessive strategy in the East and the preference for the have-strategy in the South and West must also have played a role here. Finally, we compared flagging patterns across Slavic languages in a cluster analysis based on Jaccard similarity in order to see how Slavic languages relate to each other when all arguments are taken into account at once. While the genealogical relations still largely determine similarities in argument flagging, the effect of language contact percolates here as well. Thus, Ukrainian and Belarusian cluster closer than any of them with Russian, although all three languages genealogically belong to the same subbranch. This is due to some innovations in flagging that these two languages share with each other but not with Russian. Since common innovations are hardly accidental in general, it is likely to assume that these innovations are due to more intense language contact between these two languages. Furthermore, we found that Polish is closer to East Slavic than to other West Slavic languages when it comes to flagging. Thus, Polish patterns with East Slavic with respect to its nominativeness ratio as well as with respect to transitivity prominence. This suggests that language contact must have played an important role here too. Indeed, historically, Polish had strong contacts with all three East Slavic languages. Finally, note that Polish patterns with East Slavic not so much with respect to innovations but rather with respect to the inherited patterns. It seems thus that language contact has a preserving, conservative effect here.

Acknowledgements We cordially thank Sergey Say, two anonymous reviewers, and the editor of the volume for their valuable comments on one of the previous versions of the chapter. We also thank Maria Shvedova for help with the Ukrainian data and Uladzimir Košcˇanka for help with the Belarusian data.

References Abrahamyan, Sergey. G. (2004). Hayocʹ lezu. Šarahyusutʹyun, Erewan: Luys. Adams, James N. (2013). Social Variation and the Latin Language. Cambridge: Cambridge University Press. Adjarian, Hrachia (1909). Classification des dialectes arméniens, Paris: Librairie Honore Champion. Ageno, Franca (1964). Il verbo nell’italiano antico. Milan: Ricciardi. Ageno, Franca (1978). ‘Infinito’, in Enciclopedia dantesca. Appendice. Roma: Istituto della Enciclopedia Italiana, 268–292. Ahmed, Tafseer (2010). ‘The unaccusativity/unergativity distinction in Urdu.’ Journal of South Asian Linguistics 3, 3–22. Aikhenvald, Alexandra Y., Robert M. W. Dixon, and Masayuki Onishi (eds) (2001). NonCanonical Marking of Subjects and Objects. Amsterdam: Benjamins. Alcina Franch, Juan and José Manuel Blecua (1975). Gramática española. Barcelona: Ariel. Aldridge, Edith (2017). ‘Intransitivity and the Development of Ergative Alignment’, in Jessica Coon, Diane Massam, and Lisa Demena Travis (eds), The Oxford Handbook of Ergativity. Oxford: Oxford University Press, 501–529. Allen, Cynthia L. (1995). Case Marking and Reanalysis. Oxford: Clarendon Press. Amberber, Mengistu (2005). ‘Differential subject marking in Amharic’, in Mengistu Amberber and Helen de Hoop (eds), Competition and Variation in Natural Languages: The Case for Case. Amsterdam: Elsevier, 295–315. Ambrazas, Vyautas (ed.) (2006). Lithuanian Grammar. Vilnius: Institute of the Lithuanian Language. Anand, Pranav and Andrew Nevins (2006). ‘The locus of ergative case assignment: Evidence from scope’, in Alana Johns, Diane Massam, and Juvenal Ndayiragije (eds), Ergativity: Studies in Natural Language and Linguistic Theory. Dordrecht: Springer, 2–26. Andreas, Friedrich C. and Walter B. Henning (1934). ‘Mitteliranische Manichaica aus Chinesisch-Turkestan III’, Sitzungsberichte der Preußischen Akademie der Wissenschaften, 848–912. Aristar, Anthony R. (1996). ‘The relationship between dative and locative: Kuryłowicz’s argument from a typological perspective’, Diachronica 13(2): 207–224. Arkadiev, Peter M. (2008). ‘Thematic roles, event structure, and argument encoding in semantically aligned languages’, in Mark Donohue and Søren Wichmann (eds), The Typology of Semantic Alignment. Oxford: Oxford University Press, 101–117. Asenova, Petya and Vassilka Aleksova (2008). ‘L’aspect balkanique de la nota accusativi personalis’, Zeitschrift fu¨r Balkanologie 44(1): 1–23. Avery, John (1872). ‘Contributions to the history of verb-inflection in Sanskrit’. Journal of the American Oriental Society 10: 219–324. Babby, Leonard H. (1994). ‘A theta-theoretic analysis of adversity impersonal sentences in Russian’, in Sergey Avrutin, Steven Franks, and Ljiljana Progovac (eds), Formal Approaches to Slavic Linguistics (FASL-2). The MIT Meeting 1993 (Michigan Slavic Materials, 36). Ann Arbor, MI: Michigan Slavic Publications, 25–67.

REFERENCES

329

Barðdal, Jóhanna (2001). ‘The perplexity of DAT–NOM verbs in Icelandic’, Nordic Journal of Linguistics 24: 47–70. Barðdal, Jóhanna (2004). ‘The semantics of the impersonal construction in Icelandic, German and Faroese: Beyond thematic roles’, in Werner Abraham (ed.), Focus on Germanic Typology (Studia Typologica 6). Berlin: Akademie Verlag, 105–137. Barðdal, Jóhanna (2006). ‘Construction-specific properties of syntactic subjects in Icelandic and German’, Cognitive Linguistics 17(1): 39–106. Barðdal, Jóhanna (2008). Productivity. Evidence from Case and Argument Structure in Icelandic. Amsterdam: Benjamins. Barðdal, Jóhanna and Thórhallur Eythórsson (2003). ‘The change that never happened: The story of oblique subjects’, Linguistics 39: 439–472. Barðdal, Jóhanna and Thórhallur Eythórsson (2009). ‘The origin of the oblique subject construction: An Indo-European comparison’, in Vit Bubenik, John Hewson, and Sarah Rose (eds), Grammatical change in Indo-European languages: papers presented at the workshop on Indo-European linguistics at the XVIIIth International Conference on Historical Linguistics, Montreal, 2007. Amsterdam: Benjamins, 179–193. Barðdal, Jóhanna and Thomas Smitherman (2013). ‘The quest for cognates: A reconstruction of oblique subject constructions in Proto-Indo-European’, Language Dynamics and Change 3(1): 28–67. Barðdal, Jóhanna, Eleonora Cattafi, Laura Bruno, and Serena Danesi (2020). ‘Nonnominative subjects in Latin and Ancient Greek: Applying the subject tests on early Indo-European material’. Preprint, February 2020, www.researchgate.net/publication/ 336146514_Non-Nominative_Subjects_in_Latin_and_Ancient_Greek_Applying_the_ Subject_Tests_on_Early_Indo-European_Material Barðdal, Jóhanna, Thomas Smitherman, Valgerður Bjarnadóttir, Serena Danesi, Gard B. Jenset, and Barbara McGillivray (2012). ‘Reconstructing constructional semantics: The dative subject construction in Old Norse-Icelandic, Latin, Ancient Greek, Old Russian and Old Lithuanian’, in Studies in Language 36(3): 511–547. Bartula, Czesław (1954). Związki czasownika z dopełnieniem w najstarszych zabytkach języka staro-cerkiewno-słowian´skiego. Wrocław: Ossolineum. Bartula, Czesław (1969). Składnia zdania pojedynczego w języku staro-cerkiewnosłowian´skim. Kraków: Uniwersytet Jagiellon´ski. Bassols de Climent, Mariano (1948). Sintaxis Histórica de la Lengua Latina, vol. II. Barcelona: Escuela de Filologı´a. Bauer, Anna H. (2014). Morphosyntax of the Noun Phrase in Hieroglyphic Luwian. Leiden: Brill. Bauer, Brigitte (2000). Archaic Syntax in Indo-European (Trends in Linguistics 125). Berlin: Mouton de Gruyter. Bayer, Josef (2004). ‘Non-nominative subjects in comparison’, in P. Bhaskararao and K. V. Subbarao (eds), Non-Nominative Subjects. Amsterdam: Benjamins, 49–76. Beavers, John (2011). ‘On affectedness’, Natural Language and Linguistic Theory 29(2): 335–370. Beavers, John and Andrew Koontz-Garboden (2017). ‘Result verbs, scalar change and the typology of motion verbs’, Language 93(4): 842–876. Beavers, John and Andrew Koontz-Garboden (2020). The Roots of Verbal Meaning. Oxford: Oxford University Press. Belletti, Adriana and Luigi Rizzi (1988). ‘Psych-verbs and θ-theory’, Natural Language and Linguistic Theory 6(3): 291–352.

330

REFERENCES

Benedetti, Marina (2002). ‘Radici, morfemi nominali e verbali: alla ricerca dell’inaccusatività indoeuropea’, Archivio Glottologico Italiano 87: 20–45. Bennett, Charles E. (1914). The Syntax of Early Latin, vol. II. Boston, MA: Allyn & Bacon. Bentley, Delia (2006). Split Intransitivity in Italian. Berlin: Mouton de Gruyter. Bentley, Delia (2010). ‘Principles of subject markedness in Romance’, Archivio Glottologico Italiano 95: 152–189. Benveniste, Émile. (1952). ‘La construction passive du parfait transitif ’, Bulletin de la Société Linguistique de Paris 48: 52–62. Benveniste, Émile. (1959). ‘Sur la phonétique et la syntaxe de l’arménien classique’, Bulletin de la Société Linguistique de Paris 54: 46–68. Benveniste, Émile (1962). ‘Les substantifs en -ant du Hittite’, Bulletin de la Societe de linguistique de Paris 57: 44–51. Berro, Ane and Ricardo Etxepare (2017). ‘Ergativity in Basque’, in Jessica Coon, Diane Massam, and Lisa Demena Travis (eds), The Oxford Handbook of Ergativity. Oxford: Oxford University Press, 782–830. Bertinetto, Pier Marco and Mario Squartini (1995). ‘An attempt at defining the class of “gradual completion verbs”’, in Piermarco Bertinetto, Valentina Bianchi, James Higginbotham, and Mario Squartini (eds), Temporal Reference, Aspect and Actionality. Vol. I: Semantic and Syntactic Perspectives, Turin: Rosenberg and Sellier, 11–26. Bhaskararao, Peri and Karumuri Venkata Subbarao (eds) (2004). Non-Nominative Subjects, 2 vols. Amsterdam: Benjamins. Bickel, Balthasar (2004). ‘The syntax of experiencers in the Himalayas’, in Peri Bhaskararao and Karumura V. Subbarao (eds), Non-nominative Subjects, vol. I. Amsterdam: Benjamins, 77–111. Bickel, Balthasar (2008). ‘On the scope of the referential hierarchy in the typology of grammatical relations’, in Greville G. Corbett and Michael Noonan (eds), Case and Grammatical Relations: Studies in Honor of Bernard Comrie. Amsterdam: Benjamins, 191–210. Bickel, Balthasar (2011). ‘Grammatical relations typology’, in Jae Jung Song (ed), The Oxford Handbook of Linguistic Typology. Oxford: Oxford University Press, 399–444. Bickel, Balthasar and Johanna Nichols (2009). ‘Case marking and alignment’, in Andrei Malchukov and Andrew Spencer (eds). The Oxford Handbook of Case. Oxford: Oxford University Press, 304–321. Bjarnadóttir, Valgerður (2014a). ‘Non-canonical case-marking on core arguments in Lithuanian’. Doctoral thesis, University of Stockholm. Bjarnadóttir, Valgerður (2014b). ‘Dialectal and diachronic distribution of case variation in Lithuanian pain-verb constructions’, Baltic Linguistics 5: 9–57. Blagova, Ė., R. M. Cejtlin, S. Gerodes et al. (1994). Staroslavjanskij slovar’ (po rukopisjam X–XI vekov). Moscow: Russkij Jazyk. Bogard, Sergio (1993). ‘El doble valor lexemático de los verbos psicológicos del español’, Signos 7: 47–65. Bolognesi, Giancarlo (1960). Le fonti dialettali degli imprestiti iranici in armeno. Milan: Società Editrice Vita e Pensiero. Borkovskij, Viktor I. and Petr S. Kuznecov ([1963] 2006). Istoricˇeskaja grammatika russkogo jayzka. Izd. 3-e, streotipnoe. Moscow: KomKniga. Bossong, Georg (1998). ‘Le marquage de l’expérient dans les langues d’Europe’, in Jack Feuillet (ed.), Actance et valence dans les langues de l’Europe. Berlin: Mouton de Gruyter, 259–294. Brekke, Magnar (1988). ‘The experiencer constraint’, Linguistic Inquiry 19: 169–181.

REFERENCES

331

Briscoe, John (2007). A Commentary on Livy 38-40. Oxford: Oxford University Press. Brozovic´, Dalibor (ed.) (2010). Materijal bosanskohercegovacˇkih punktova za opc´eslavenski lingvisticˇki atlas (OLA), građa, knjiga XXIX, Centar za leksikologiju i leksikografiju, Knjiga 1, Trec´i dio (Tramošnica 51, Grude 57, Lug 58). Sarajevo 2010: Akademija nauka i umjetnosti Bosne i Hercegovine. Brugmann, Karl (1891). ‘Zur Frage der Entstehung des grammatischen Geschlechtes’, Beitra¨ge zur Geschichte der deutschen Sprache und Literatur (PBB) 15: 523–531. Brugmann, Karl and Berthold Delbru¨ck (1893). Grundriss der vergleichenden Grammatik der indogermanischen Sprachen: kurzgefasste Darstellung der Geschichte desAltindischen, Altiranischen (Avestischen u. Altpersischen), Altarmenischen, Altgriechischen, Albanesischen, Lateinischen, Oskisch-Umbrischen, Altirischen, Gotischen, Althochdeutschen, Litauischen und Altkirchenslavischen. Bd 3, Vergleichende Syntax der indogermanischen Sprachen, T. 1 Strassburg: Tru¨bner. Brugmann, Karl and Berthold Delbru¨ck (1897). Grundriss der vergleichenden Grammatik der indogermanischen Sprachen: kurzgefasste Darstellung der Geschichte des Altindischen, Altiranischen (Avestischen u. Altpersischen), Altarmenischen, Altgriechischen, Albanesischen, Lateinischen, Oskisch-Umbrischen, Altirischen, Gotischen, Althochdeutschen, Litauischen und Altkirchenslavischen. Bd 4, Vergleichende Syntax der indogermanischen Sprachen, T. 2. Strassburg: Tru¨bner. Brugmann, Karl and Berthold Delbru¨ck (1900). Grundriss der vergleichenden Grammatik der indogermanischen Sprachen: kurzgefasste Darstellung der Geschichte des Altindischen, Altiranischen (Avestischen u. Altpersischen), Altarmenischen, Altgriechischen, Albanesischen, Lateinischen, Oskisch-Umbrischen, Altirischen, Gotischen, Althochdeutschen, Litauischen und Altkirchenslavischen. Bd 5, Vergleichende Syntax der indogermanischen Sprachen, T. 3 Strassburg: Tru¨bner. Bubenik, Vit (2016). ‘On the establishment of ergative during the Late Middle Indo-Aryan period’, in Dahl and Stron´ski (eds), 109–131. Butt, Miriam (2017). ‘Hindi/Urdu and related languages’, in Jessica Coon, Diane Massam, and Lisa Demena Travis (eds), The Oxford Handbook of Ergativity. Oxford: Oxford University Press, 807–831. Butt, Miriam and Ashwini Deo (2017). ‘Developments into and out of ergativity: IndoAryan diachrony’, in Jessica Coon, Diane Massam, and Lisa Demena Travis (eds), The Oxford Handbook of Ergativity. Oxford: Oxford University Press, 531–552. Butt, Miriam and Tikaram Poudel (2007). ‘Distribution of the ergative in Nepali’ [handout]. Butt, Miriam, Scott Grimm, and Tafseer Ahmed (2006). ‘Dative subjects’. Paper presented at the NWO/DFG Workshop on Optimal Sentence Processing. Nijmegen, June 2006. Buttler, Danuta (1976). Innowacje składniowe współczesnej polszczyzny (walencja wyrazów). Warsaw: Pan´stwowe Wydawnicwto Naukowe. Buzˇarovska, Eleni (2001). ‘Accusative and dative clitics in Southern Macedonian and Northern Greek dialects’, Balkanistica 14: 1–18. Buzˇarovska, Eleni (2017). ‘The contemporary use of DOM in South-Western Macedonian dialects’, Rhema 3: 65–87. Buzˇarovska, Eleni (2020). ‘The contact hypothesis revised: DOM in Macedonian dialects’, Journal of Language Contact 13: 57–95. 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.

332

REFERENCES

Calboli, Gualtiero (1962). ‘Sulla costruzione impersonale Dicitur eo tempore matrem Pausaniae vixisse (Nep. 4.5.3)’, in Gualtiero Calboli Studi Grammaticali. Bologna: Zanichelli, 3–115. Campbell, Dennis R. M. (2008). ‘Split ergativity in Hurrian’, Zeitschrift fu¨r Assyriologie 98: 262–294. Cano Aguilar, Rafael (1981). Estructuras sintácticas transitivas en el español actual. Madrid: Gredos. Cardona, George (1974). ‘Pāṇini’s kārakas: Agency, animation and identity’, Journal of Indian Philosophy 2: 231–306. Cardona, George (1976). ‘Subject in Sanskrit’, in Manindra K. Verma (ed.), The Notion of Subject in South Asian Languages (South Asian Studies Publication Series, 2). Madison, WI: University of Wisconsin, 1–38. Cardona, George (1978). ‘Relations between causatives and passives in Indo-Iranian’, in Braj B. Kachru (ed.), Linguistics in the Seventies: Forum Lectures Presented at the 1978 Linguistic Institute, special issue of Studies in the Linguistic Sciences, 8(2), 1–42. Carling, Gerd (2017). DiACL—Diachronic Atlas of Comparative Linguistics Online. https:// diacl.ht.lu.se/. Carling, Gerd and Chundra Cathcart (2021). ‘Evolutionary dynamics of Indo-European alignment patterns’, Diachronica 38(3): 358–412. Carruba, Onofrio (1992). ‘Le notazioni dell’agente animato delle lingue anatoliche (e l’ergativo)’, in Onofrio Carruba (ed.), Per una grammatica ittita/Towards a Hittite Grammar. Pavia: Italian University Press, 61–98. Cennamo, Michela (1993). The Reanalysis of Reflexives: A Diachronic Perspective. Naples: Liguori. Cennamo, Michela (1997). ‘Passive and impersonal constructions’, in Martin Maiden and Mair Parry (eds), Dialects of Italy. London: Routledge, 145–161. Cennamo, Michela (1998). ‘The loss of the voice dimension between Late Latin and Early Romance’, in Monika S. Schmid, Jennifer R. Austin, and Dieter Stein (eds), Historical Linguistics 1997: Selected Papers from the 13th International Conference on Historical Linguistics, Du¨sseldorf, August 1997. Amsterdam: Benjamins, 7–100. Cennamo, Michela (2000). ‘Patterns of active syntax in late Latin pleonastic reflexives’, in John Charles Smith and Delia Bentley (eds), Historical Linguistics 1995: Selected Papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995. Amsterdam: Benjamins, 33–55. Cennamo, Michela (2001a). ‘On the reorganization of voice distinctions and grammatical relations in Late Latin’, in Claude Moussy (ed.), De lingua latina novae questiones: Actes du Xeme Colloque International de Linguistique Latine. Leuven: Peeters, 51–65. Cennamo, Michela (2001b). ‘L’extended accusative e le nozioni di voce e relazione grammaticale nel latino tardo e medievale’, in Valeria Viparelli (ed.), Ricerche Linguistiche tra antico e Moderno. Naples: Liguori, 3–27. Cennamo, Michela (2003). ‘(In)transitivity and object marking: Some current issues’, in Giuliana Fiorentino (ed.), Romance Objects: Transitivity in the Romance Languages. Berlin: Mouton de Gruyter, 49–194. Cennamo, Michela (2005). ‘Passive auxiliaries in Late Latin’, in Sándor Kiss, Luca Mondin and Giampaolo Salvi (eds), Latin et Langues Romanes. Études de Linguistique Offertes à József Herman à l’Occasion de son 80ème anniversaire. Tu¨bingen: Niemeyer, 177–194.

REFERENCES

333

Cennamo, Michela (2009). ‘Argument structure and alignment variations and changes in Late Latin’, in Jóhanna Barðdal and Shobana Chelliah (eds), The Role of Semantics and Pragmatics in the Development of Case. Amsterdam: Benjamins, 307–346. Cennamo, Michela (2010). ‘Control and argument marking in Latin’. Paper presented at the workshop ‘Subject and transitivity in Indo-European and beyond: A diachronic typological perspective’, 43rd Annual Meeting of the Societas Linguistica Europaea, Vilnius, 2–3 September. Cennamo, Michela (2011). ‘Impersonal constructions and accusative subjects in Late Latin’, in Andrej Malchukov and Anna Siewierska (eds), Impersonal Constructions: A CrossLinguistic Perspective. Amsterdam: Benjamins, 169–188. Cennamo, Michela (2016). ‘Voice’, in Adam Ledgeway and Martin Maiden (eds), The Oxford Guide to the Romance Languages. Oxford: Oxford University Press, 967–980. Cennamo, Michela (2020). ‘The actualization of new voice patterns in Romance. Persistence in diversity’, in Bridget Drinka (ed.), Historical Linguistics 2017: Selected Papers. Amsterdam: Benjamins, 109–142. Cennamo, Michela, Thórhallur Eythórsson, and Jóhanna Barðdal (2015). ‘Semantic and (morpho)syntactic constraints on anticausativization: Evidence from Latin and Old Norse-Icelandic’, Linguistics 53(4): 677–729. Chahinian, Talar and Anny Balakian (2016). ‘Language in Armenian American communities: Western Armenian and efforts for preservation’, International Journal of the Sociology of Language 237: 37–57. CHD = Gu¨terbock, Hans Gustav, Harry A. Hoffner, Jr, and Theo P. J. van den Hout (1989–). The Hittite Dictionary of the Oriental Institute of the University of Chicago. Chicago: Oriental Institute of the University of Chicago. Clackson, James (1994). The Linguistic Relationship between Armenian and Greek. Oxford: Blackwell. Clancy, Steven J. (2010). The Chain of Being and Having in Slavic. Amsterdam: Benjamins. Clancy, Steven J. (2020). ‘Be and have: General semantics, grammaticalization, and alignment’, in Marc L. Greenberg (ed.), Encyclopedia of Slavic Languages and Linguistics Online. Brill: Leiden, https://referenceworks.brillonline.com/browse/encyclopedia-ofslavic-languages-and-linguistics-online (accessed 17/05/21). Coghill, Eleanor (2016). The Rise and Fall of Ergativity in Aramaic. Cycles of Alignment Change. Oxford: Oxford University Press. Cole, Peter, Wayne Harbert, Gabriella Hermon, and Shikaripur N. Sridhar (1980). ‘The acquisition of subjecthood’, Language 56(4): 719–743. Collinge, Neville (1978). ‘Restructuring of noun cases in syntax: Why “anti-” will not do’, in Werner Abraham (ed.), Valence, Case and Grammatical Relations. Amsterdam: Benjamins, 617–633. Comrie, Bernard (1978). ‘Ergativity’, in Winfred P. Lehmann (ed.), Syntactic Typology: Studies in the Phenomenology of Language. Sussex: Harvester Press, 329–394. Comrie, Bernard (1985). ‘Causative verb-formation and other verb-deriving morphology’, in Timothy Shopen (ed.), Language Typology and Syntactic Description. Vol. III: Grammatical Categories and the Lexicon. Cambridge: Cambridge University Press, 301–348. Comrie, Bernard (1989). Language Universals and Linguistic Typology (2nd ed.). London: Blackwell. Conti, Luz (2010). ‘Análisis del dativo en construcciones impersonales: Los conceptos de sujeto y de semisujeto en griego antiguo’, Emerita 78(2): 249–273.

334

REFERENCES

Conti, Luz and Silvia Luraghi (2014). ‘The Ancient Greek partitive genitive in typological perspective’, in Silvia Luraghi and Tuomas Huumo (eds), Partitive Cases and Related Categories. Berlin: De Gruyter, 443–476. Coon, Jessica (2013). Aspects of Split Ergativity. Oxford: Oxford University Press. Coon, Jessica, Diane Massam, and Lisa Demena Travis (eds) (2017). The Oxford Handbook of Ergativity. Oxford: Oxford University Press. Corbett, Greville G. (2000). Number. Cambridge: Cambridge University Press. Corbett, Greville G. (2006). Agreement. Cambridge: Cambridge University Press. Creissels, Denis (2018). ‘The obligatory coding principle in diachronic perspective’, in Cristofaro and Zúñiga (eds), 59–109. Cristofaro, Sonia and Fernando Zúñiga (eds) (2018). Typological Hierarchies in Synchrony and Diachrony. Amsterdam: Benjamins. Croft, William (1993). ‘Case marking and the semantics of mental verbs’, in James Pustejovsky (ed.), Semantics and the Lexicon. Dordrecht: Kluwer Academic, 55–72. Cuervo, Marı´a Cristina (2010). ‘Some dative subjects are born, some are made’, in Claudia Borgonovo, Manuel Español-Echevarrı´a, and Philippe Prévost (eds), Selected Proceedings of the 12th Hispanic Linguistics Symposium. Somerville, MA: Cascadilla Proceedings Project, 26–37. Cuzzolin, Pierluigi and Gerd Haverling (2009). ‘Syntax, sociolinguistics, and literary genres’, in Philip Baldi and Pierluigi Cuzzolin (eds), New Perspectives on Historical Latin Syntax, vol.1: Syntax of the Sentence. Berlin: De Gruyter, 19–64. Dahl, Eystein (2009). ‘Some semantic and pragmatic aspects of object alternation in Archaic Vedic’, in Jóhanna Barðdal and Shobana Chelliah (eds), The Role of Semantics and Pragmatics in the Development of Case (Studies in Language Companion Series 108). Amsterdam: Benjamins, 23–55. Dahl, Eystein (2010). Time, Tense and Aspect in Early Vedic Grammar. Exploring Inflectional Semantics in the Rigveda. Leiden: Brill. Dahl, Eystein (2014). ‘Partitive subjects and objects in Indo-Iranian and beyond’, in Silvia Luraghi and Tuomas Huumo (eds), Partitive Cases and Related Categories. Berlin: De Gruyter, 417–441. Dahl, Eystein (2015). ‘Toward a formal model of semantic change: A neo-Reichenbachian approach to the development of the Vedic past tense system’, Lingua Posnaniensis 57(1): 41–76. Dahl, Eystein (2016). ‘The origin and development of the Old Indo-Aryan predicated -tá construction’, in Eystein Dahl and Krzysztof Stron´ski (eds), Indo-Aryan Ergativity in Typological and Diachronic Perspective. Amsterdam: Benjamins, 63–110. Dahl, Eystein (2020). ‘Come, rain or shine: The etymology and lexical emantics of some precipitation verbs in Indo-European’, in H. Bichlmeier, O. Šefcˇ´ı k, and R. Sukacˇ (eds), Etymologus. Festschrift for Václav Blazˇek. Hamburg: Baar, 131–141. Dahl, Eystein (2021a). ‘Pathways to split ergativity: The rise of ergative alignment in Anatolian and Indo-Aryan’, Diachronica 38(3): 413–456. Dahl, Eystein (2021b). ‘The verb agreement patterns of neuter plural subjects in Homeric Greek’, in Matteo Tarsi (ed.), Studies in General and Historical Linguistics Offered to Jón Axel Harðarson on the Occasion of His 65th Birthday. Innsbruck: Innsbrucker Beitra¨ge zur Sprachwissenschaft, 89–108. Dahl, Eystein and Chiara Fedriani (2012). ‘The argument structure of experience: Experiential constructions in Early Vedic, Homeric Greek and Early Latin’, Transactions of the Philological Society 110(3): 342–362. Dahl, Eystein and Krzysztof Stron´ski (eds) (2016). Indo-Aryan Ergativity in Typological and Diachronic Perspective. Amsterdam: Benjamins.

REFERENCES

335

Danesi, Serena (2014). ‘Accusative subjects in Avestan: “Errors” or non-canonically marked arguments?’, Indo-Iranian Journal 57(3): 223–260. Dardano, Paola (2006). Die hethitischen Tontafelkataloge aus Hattusa (CTH 276–282). [Studien zu den Bog˘azko¨y-Texten 47]. Wiesbaden: Harrassowitz. Dardano, Paola (2010). ‘Zur anatolischen Morphosyntax: das Suffix -(a)nt- und seine Bildungen’, inAygu¨l Su¨el (ed.), Acts of the VIIth International Congress of Hittitology, Çorum, August 25–31, 2008. Ankara, 173–188. Dardano, Paola (2013). ‘L’allineamento sintattico delle lingue indoeuropee dell’Anatolia: vecchi problemi e nuove proposte alla luce di unarecente pubblicazione’, Orientalia 82(2): 29–67. Dardano, Paola (2018). ‘Zur Subjektmarkierung im Hethitischen: syntaktische und semantische Fragen’, in Elisabeth Rieken, Ulrich Geupel, and Theresa Maria Roth (eds), Morphosyntaktische Kategorien in Sprachgeschichte und Forschung. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 21. bis 23. September 2015 in Marburg. Wiesbaden: Reichert Verlag, 39–62. Deeters, Gerhard (1927). Armenisch und Su¨dkaukasisch. Leipzig: Verlag der Asia Major. Delbru¨ck, Berthold (1874). Das altindische Verbum aus den Hymnen des Ṛigveda seinem Baue nach dargestellt. Halle a. S.: Verlag der Buchhandlung des Waisenhauses. Delbru¨ck, Berthold (1888). Altindische Syntax. Halle a. S.: Verlag der Buchhandlung des Waisenhauses. De Melo, Wolfgang (ed. and trans.) (2012). The Little Carthaginian. Pseudolus. The Rope. Cambridge, MA: Harvard University Press. De Smet, Hendrik (2009). ‘Analysing reanalysis’, Lingua 119: 1728–1755. De Smet, Hendrik (2012). ‘The course of actualization’, Language 88(3): 601–633. Dixon, Robert M. W. (1976). ‘Rapporteur’s summary [topic B: ergative, locative, and instrumental case inflections]’, in R. M. W. Dixon (ed.), Grammatical Categories in Australian languages. Canberra: Australian Institute of Aboriginal Studies, 411–414. Dixon, Robert M. W. (1979). ‘Ergativity’, Language 55: 59–138. Dixon, Robert M. W. (1994). Ergativity. Cambridge: Cambridge University Press. Dixon, Robert M. W. and Alexandra Y. Aikhenvald (eds) (2000a). Changing Valency: Case Studies in Transitivity. Cambridge: Cambridge University Press. Dixon, Robert M. W. and Alexandra Y. Aikhenvald (2000b). ‘Introduction’, in Robert M. W. Dixon and Alexandra Y. Aikhenvald (eds), Changing Valency: Case Studies in Transitivity. Cambridge: Cambridge University Press, 1–29. Di Tullio, Ángela. (1996). Verbos psicológicos en español. Signo & Seña 5: 219–238. Di Tullio, Ángela (1998). ‘A lternancia acusativo-dativo en verbos psicológicos del español’, in Giovanni Ruffino (ed.), Atti del XXI Congresso di Lingu¨´ıstica e Filologia Romanza. Tu¨bingen: Max Niemeyer, 255–260. Di Tullio, Ángela (2004). ‘Los verbos psicológicos y la estatividad: realizaciones del español’, Cuadernos de Lingu¨´ıstica del Instituto Ortega y Gasset 11: 23–43. Djahukian, Gevorg (2003). ‘Notes on some lexical correspondences between Armenian and the Kartvelian Languages’, Iran and the Caucasus 7(1–2): 191–192. Donohue, Mark (2008). ‘Semantic alignment systems: What’s what, and what’s not’, in Mark Donohue and Søren Wichmann (eds), The Typology of Semantic Alignment. Oxford: Oxford University Press, 24–75. Donohue, Mark and Søren Wichmann (eds) (2008). The Typology of Semantic Alignment. Oxford: Oxford University Press. Dowty, David (1979). Word Meaning and Montague Grammar. Dordrecht: Reidel.

336

REFERENCES

Dowty, David (1991). ‘Thematic proto-roles and argument selection’, Language 67(3): 547–619. Draeger, Anton (1878). Historische Syntax der Lateinischen Sprache. Leipzig: Teubner. Dum-Tragut, Jasmine (2009). Armenian: Modern Eastern Armenian, Amsterdam: Benjamins. Durie, Mark (1985). ‘Control and decontrol in Acehnese’, Australian Journal of Linguistics 5: 43–54. Durkin-Meisterernst, Desmond (2006). The Hymns to the Living Soul. Middle Persian and Parthian Texts in the Turfan Collection. Turnhout: Brepols. Durkin-Meisterernst, Desmond (2014). Grammatik des Westmitteliranischen (Parthisch und Mittelpersisch). Vienna: Verlag der Österreichischen Akademie der Wissenschaften. Edelman, D. (Joy) I. and Leila R. Dodykhudova (2009). ‘The Pamir languages’, in Gernot Windfuhr (ed.), The Iranian Languages. London: Routledge, 773–786. Egerland, Verner (2010a). ‘Frasi subordinate all’infinito’, in G. Salvi and L. Renzi (eds), Grammatica dell’italiano antico. Bologna: Il Mulino, 817–835, 856–879. Egerland, Verner (2010b). ‘Infinito’, in R. Simone (ed.), Enciclopedia dell’Italiano, www. treccani.it/enciclopedia/infinito_(Enciclopedia_dell’Italiano). Eichner, Heiner (1985). ‘Das Problem des Ansatzes eines urindogermanischen Numerus “Kollektiv” (“Komprehensiv”)’, in Bernfried Schlerath (ed.), Grammatische Kategorien. Funktion und Geschichte. Wiesbaden: Reichert, 134–169. Elvira, Javier (2011). ‘Constructions of uncontrolled state or event: The increase in productivity of a new argument structure in Old Spanish’, Constructions and Frames 3: 184–207. Endzelīns, Janis (1923). Lettische Grammatik. Heidelberg: Carl Winter. Eriksen, Pål K., Seppo Kittila¨, and Leena Kolehmainen (2015). ‘The world is raining. Metereological predicates and their subjects in a typological perspective’, in M.-L. Helasvuo and T. Huumo (eds), Subjects in Constructions: Canonical and Non-Canonical. Amsterdam: Benjamins, 205–228. Ernout, Alfred (1909). ‘Recherches sur l’emploi du passif latin à l’èpoque républicaine’, Mémoires de la Société de Linguistique de Paris 15: 273–333. Ernout, Alfred and Thomas François (1964). Syntaxe Latine (2nd ed.). Paris: Klincksieck. Evans, Nicholas (2004). ‘Experiencer objects in Iawaidjan languages (Australia)’, in Peri Bhashaskararao and Karumuri V. Subbarao (eds), Non-Nominative Subjects, vol. I. Amsterdam: Benjamins, 169–192. Eythórsson, Thórhallur and Jóhanna Barðdal (2005). ‘Oblique subjects: A common Germanic inheritance’, Language 81: 824–881. Fabrizio, Claudia (2015). ‘Il caso curioso dell’infinito soggetto in latino’, in G. Marotta and Francesco Rovai (eds), Ancient Langages between Variation and Norm. Studi e Saggi Linguistici (special issue), 53: 401–418. Fabrizio, Claudia (2017). ‘«Par che m’acoglia lo vostro innamorare». Sull’infinito soggetto in italiano antico’, in Paolo D’Achille and Maria Grossmann (eds), Per la storia della formazione delle parole in italiano: un nuovo corpus in rete (MIDIA) e nuove prospettive di studio. Firenze, Franco Cesati, 201–219. Fabrizio, Claudia (2018). ‘On the distribution of subject infinitives in Latin and Homeric Greek’, Studi e Saggi Linguistici 61: 61–95. Fabrizio, Claudia (forthcoming). Alignment and Subjecthood in Latin. Leiden: Brill. Falk, Yehuda (2006). Subjects and Universal Grammar. An Explanatory Theory. Cambridge: CUP.

REFERENCES

337

Fedriani, Chiara (2009). ‘The “behaviour-before-coding” principle: Further evidence from Latin’, Archivio Glottologico Italiano 94: 156–184. Fedriani, Chiara (2013). ‘The me pudet construction in the history of Latin: Why and how fast non-canonical subjects come and go’, in Ilja Serzˇant and Leonid Kulikov (eds), The Diachronic Typology of Non-Canonical Subjects. Amsterdam: Benjamins, 203–229. Fedriani, Chiara (2014). Experiential Constructions in Latin: A Synchronic and Diachronic Study. Leiden: Brill. Fellner, Hannes A. and Laura Grestenberger (2018). ‘Die Reflexe der *-nt- unf *-mh1 noPartizipien im Hethitischen und Tocharischen’, in Elisabeth Rieken (ed.), 100 Jahre Entzifferung des Hethitischen Morphosyntaktische Kategorie in Sprachgeschichte und Forschung. Akten der Arbeitstagung der Indogermanischen Gesellschaft vom 21). bis 23 September 2015 in Marburg. Wiesbaden: Reichert, 63–82. Feltenius, Leif (1977). Intransitivizations in Latin. Uppsala: Almqvist and Wiksell. Fernández Soriano, Olga (1999). ‘Two types of impersonal sentences in Spanish: Locative and dative subjects’, Syntax 2: 101–140. Fischer, Olga (2013). ‘An inquiry into unidirectionality as a foundational element of grammaticalization: On the role played by analogy and the synchronic grammar system in processes of language change’, Studies in Language 37(3): 515–533. Fischer, Olga C. M. and Frederike C. van der Leek (1983). ‘The demise of the Old English impersonal construction’, Journal of Linguistics 19: 337–368. Flobert, P. (1975). Les Verbes déponents latins des origines à Charlemagne. Paris: Les Belles Lettres. Foley, William A. and Robert D. Van Valin, Jr. (1984). Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Folgar, Carlos (1993). Diacronı´a de los objetos directo e indirecto (del latı´n al castellano medieval). Verba, Anexo 37. Santiago de Compostela: Universidade de Santiago de Compostela. Fortson, Benjamin W. (2004). Indo-European Language and Culture: An Introduction. Malden, MA: Wiley-Blackwell. Fortson, Benjamin W. (2010). Indo-European Language and Culture: An Introduction. Second Edition. Malden, MA: Wiley-Blackwell. Franco, Josep M. (2000). ‘Agreement as a continuum: The case of Spanish pronominal clitics’, in Frits Beukema and Marcel den Dikken (eds), Clitic Phenomena in European Languages. Amsterdam: Benjamins, 147–189. Frantikova, Dita (2015). ‘Hittite participles—Adjectives or verbs?’, Journal of IndoEuropean Studies 43: 178–200. Friedman, Victor A. (2008). ‘Balkan object reduplication in areal and dialectological perspective’, in Dalina Kallulli and Liliane Tasmowski (eds), Clitic Doubling in the Balkan Languages. Amsterdam: Benjamins, 35–63. Friedman, Victor A. and Brian D. Joseph (2018). ‘Non-nominative and depersonalized subjects in the Balkans’, in Jóhanna Barðdal, Na’ama Pat-El, and Stephen M. Carey (eds), Non-Canonically Case-Marked Subjects. The Reykjavı´k-Eyjafjallajo¨kull Papers. Amsterdam: Benjamins, 24–53. Frotscher, Michael (2013). ‘Das hethitische -ant-Partizip und seine indogermanischen Grundlagen. Semantik, Morphologie, Syntax’. Ph.D. dissertation, Università degli Studi di Verona. Gamkrelidze, Thomas V. and Vjacˇeslav V. Ivanov (1984). Indoevropejskij jazyk i indoevropejcy: Rekonstrukcija i istoriko-tipologicˇeskij analiz prajazyka i protokul’tury (2 vols.). Tbilisi.

338

REFERENCES

Gamkrelidze, Thomas V. and Vjacheslav V. Ivanov (1995). Indo-European and the IndoEuropeans: A Reconstruction and Historical Analysis of a Proto-Language and a ProtoCulture [trans. Johanna Nichols of Indoevropejskij jazyk i Indoevropejcy. Rekonstrukcija i istoriko-tipologicˇeskij analiz prajazyka i protokul’tury, 1984]. Berlin: Mouton de Gruyter. Ganenkov, Dmitry (2006). ‘Experiencer coding in Nakh-Daghestanian’, in Leonid Kulikov, Andrej Malchukov, and Peter de Swart (eds), Case, Valency and Transitivity. Amsterdam: Benjamins, 179–202. Garcı´a-Miguel, José Ma . (1991). ‘La duplicación de complemento directo e indirecto como concordancia’, Verba 18: 375–410. Garcı´a-Miguel, José Ma . (1995). Transitividad y complementación preposicional en español. Verba, Anexo 40. Santiago de Compostela: Universidade de Santiago de Compostela. Garcı´a-Miguel, José Ma . (2015). ‘Variable coding and object alignment in Spanish: A corpus-based approach’, Folia Linguistica 49(1): 205–256. Garrett, Andrew (1990a). ‘The origin of NP split ergativity’, Language 66(2): 261–296. Garrett, Andrew (1990b). ‘The syntax of Anatolian subject clitics’. Ph.D. diss. Harvard University. Garrett, Andrew (1996). ‘Wackernagel’s law and unaccusativity in Hittite’, in Aaron L. Halpern and Arnold M. Zwicky (eds), Approaching Second: Second Position Clitics and Related Phenomena. Stanford, CA: CSLI Publications, 85–133. Garsoïan, Nina G. (1984). Buzandaran Patmutʹiwnkʹ [The Epic Histories], Delmar, NY: Caravan Books. Gebauer, Jan (2007). Historická mluvnice jazyka ˇceského. Dı´l IV. Skladba. Praha: Academia. Geldner, Karl Friedrich (trans.). (1951). Der Rig-Veda aus dem Sanskrit ins Deutsche u¨bersetzt und mit einem laufenden Kommentar versehen, 3 vols. (Harvard Oriental Series, 33–35) [repr. 2003 in one volume (Harvard Oriental Series, 63)] Cambridge, MA: Harvard University, Department of Sanskrit and Indian Studies. Gertz, Janet Elaine (1982). ‘The nominative–accusative neuter plural in Anatolian’. Ph.D. dissertation, Yale University. Gianollo, Chiara (2005). ‘Middle voice in Latin and the phenomenon of split intransitivity’, in Gualtiero Calboli et al. (eds), Lingua Latina! Proceedings of the XII Colloquium on Latin Linguistics. Roma: Herder, 97–109. Gianollo, Chiara (2014). ‘Labile verbs in Late Latin’, Linguistics 52(4): 945–1002 [special issue ‘Typology of Labile Verbs: Focus on Diachrony’, guest ed. Leonid Kulikov and Nikolaos Lavidas]. Gildea, Spike and Fernando Zuñiga (2016). ‘Referential hierarchies: A new look at some historical and typological patterns’, Linguistics 54(3): 483–529. Gippert, Jost (2005). ‘Das Armenische—eine indogermanische Sprache im kaukasischen Areal’, in Gerhard Meiser and Olav Hackstein (eds), Sprachkontakt und Sprachwandel: Akten der XI. Fachtagung der Indogermanischen Gesellschaft, 17.–23. September 2000, Halle an der Saale. Wiesbaden: Dr. Ludwig Reichert, 139–160. Givón, Talmy (1984). Syntax: A Functional–Typological Introduction, vol. I. Amsterdam: Benjamins. Godel, Robert (1975). An Introduction to the Study of Classical Armenian. Wiesbaden: Dr. Ludwig Reichert. Goedegebuure, Petra (2010). ‘The alignment of Hattian: an active language with an ergative base’, in Leonid E. Kogan, Natalia Koslova, Sergey Loesov, and Serguei Tishchenko (eds), Language in the Ancient Near East (Proceedings of the 53e Reoncontre Assyriologiquew Internationale, vol. (1). Winona Lake, IN: Eisenbrauns, 949–981.

REFERENCES

339

Goedegebuure, Petra (2012). ‘Split-ergativity in Hittite’, Zeitschrift fu¨r Assyriologie und vorderasiatische Archa¨ologie 102(2): 270–303. Goedegebuure, Petra (2015). ‘The rise of split-ergativity (or rather split-accusativity) in Hittite’ [handout from the workshop Language, Variation and Change, 8 June]. Goedegebuure, Petra (2018). ‘The packagers -ant- and -a-, and the origin of split-ergativity in Hittite (and Lycian)’, in David M. Goldstein, Stephanie W. Jamison, and Brent Vine (eds), Proceedings of the 29th Annual UCLA Indo-European Conference. Bremen: Hempen, 77–115. Goetze, Albrecht (1951). ‘The Hittite words for “year” and the seasons and for “night” and “‘day”’, Language 27: 467–476. Goldberg, Sander M. and Gesine Manuwald (eds) (2018). Fragmentary Republican Latin. Ennius. Dramatic Fragments Minor Works Edited and Translated by Sander M. Goldberg and Gesine Manuwald. Cambridge, MA: Harvard University Press. Gonda, Jan (1951). Remarks on the Sanskrit Passive. Leiden: Brill. Gortan-Premk, Darinka (1971). Akuzativne sintagme bez predloga u srpskohrvatskom jeziku (Biblioteka juznoslovenskog filologa, Nova serija 2). Beograd: Institut za Srpskohrvatski Jezik. Grassmann, Hermann (1873). Wo¨rterbuch zum Rig-Veda. Leipzig: Brockhaus [repr. 1964, Wiesbaden: Harrassowitz]. Greenberg, Marc L. (2017). ‘Slavs as Migrants: Mapping prehistoric language variation’, in René Genis, Eric de Haard, and Radovan Lucˇic´ (eds), Definitely Perfect. Festschrift for Janneke Kalsbeek (Pegasus Oost-Europeese Studies 29). Amsterdam: Pegasus, 169–183. Grenoble, Lenore A. (2015). ‘Language contact in the East Slavic contact zone’, Balkanistica 28: 225–250. Greppin, John A. C. (1996). ‘New data on the Hurro-Urartian substratum in Armenian’, Historische Sprachforschung 109(1): 40–44. Greppin, John A. C. and Igor M. Diakonoff (1991). ‘Some effects of the Hurro-Urartian people and their languages upon the earliest Armenians’, Journal of the American Oriental Society 111(4): 720–730. Greppin, John A. C. and Amalya A. Khachaturian (1986). A Handbook of Armenian Dialectology, Delmar, NY: Caravan. Grimshaw, Jane (1990), Argument Structure. Cambridge, MA: MIT Press. Grkovic´-Major, Jasmina (2007). ‘Razvoj sintaksicˇke tranzitivnosti’, Zbornik Matice Srpske za Slavistiku 71/72: 417–433. Grkovic´-Major, Jasmina (2010). ‘The role of syntactic transitivity in the development of Slavic syntactic structures’, in Bjo¨rn Hansen and Jasmina Grkovic´-Major (eds), Diachronic Slavonic Syntax. Gradual Changes in Focus (Wiener Slawistischer Almanach, Sonderband 74). Munich: Kubon & Sagner, 63–74. Gutiérrez-Bravo, Rodrigo (2005). ‘A reinterpretation of quirky subjects and related phenomena in Spanish’, in Jean-Pierre Y. Montreuil and Chiyo Nishida (eds), New Perspectives on Romance Linguistics, Vol. I: Morphology, Syntax, Semantics, and Pragmatics. Selected Papers from the 35th Linguistic Symposium on Romance Languages (LSRL), Austin, Texas, 2005. Amsterdam: Benjamins, 127–142. Haig, Geoffrey (2008). Alignment Change in Iranian Languages. A Construction Grammar Approach. Berlin: Mouton de Gruyter. Haig, Geoffrey (2017). ‘Deconstructing Iranian ergativity’, in Jessica Coon, Diane Massam, and Lisa Demena Travis (eds), The Oxford Handbook of Ergativity. Oxford: Oxford University Press, 465–500.

340

REFERENCES

Handschuh, Corinna (2014). A Typology of Marked-S Languages. Berlin: Language Science Press. Harris, Alice C. (1985). Diachronic Syntax: The Kartvelian Case. New York: Academic Press. Harris, Alice C. (1990). ‘A lignment typology and diachronic change’, in Winfred P. Lehmann (ed.), Language Typology 1987: Systematic Balance in Language: Papers from the Linguistic Typology Symposium, Berkeley, 1–3 Dec 1987. Amsterdam: Benjamins, 67-90. Harris, Alice C. and Lyle Campbell (1995). Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin (1990). ‘The grammaticization of passive morphology’, Studies in Language 14(1): 25–71. Haspelmath, Martin (1999). ‘Explaining article-possessor complementarity: Economic motivation in noun phrase syntax’, Language 75(2): 227–243. Haspelmath, Martin (2001). ‘Non-canonical marking of core arguments in European languages’, in Alexandra Aikhenvald, Robert M. W Dixon, and Masayuki Onishi (eds), Non-Canonical Marking of Subjects and Objects. Amsterdam: Benjamins, 53–83. Haspelmath, Martin (2010). ‘The behaviour-before-coding principle in syntactic change’, in Franck Floricic (ed.), Essais de typologie et de linguistique générale. Mélanges offerts à Denis Creissels. Lyon: ENS, 541–554. Haspelmath, Martin (2011). ‘On S, A, P, T and R as comparative concepts for alignment typology’, Linguistic Typology 15(3): 535–689. Haspelmath, Martin (2015). ‘Transitivity prominence’, in Andrej L. Malchukov and Bernard Comrie (eds), Valency Classes in the World’s Languages, vol. I: Introducing the Framework, and Case Studies from Africa and Eurasia (Comparative Handbooks of Linguistics 1/1). Berlin: De Gruyter, 131–147. Haspelmath, Martin (2016). ‘Universals of causative and anticausative verb formation and the spontaneity scale’, Lingua Posnaniensis 58(2): 33–63. Haspelmath, Martin (2019). ‘Indexing and flagging, and head and dependent marking’, Te Reo, the Journal of the Linguistic Society of New Zealand 62: 93–115. Haspelmath, Martin (2021). ‘Role-reference associations and the explanation of argument coding splits’, Linguistics 59(1): 123–174. Haspelmath, Martin and Iren Hartmann (2015). ‘Comparing verbal valency across languages’, in Andrej L. Malchukov and Bernard Comrie (eds), Valency Classes in the World’s Languages, vol. I: Introducing the Framework, and Case Studies from Africa and Eurasia (Comparative Handbooks of Linguistics 1/1). Berlin: De Gruyter, 41–71. Hasselbach, Rebecca (2013). Case in Semitic: Roles, Relations, and Reconstruction. Oxford: Oxford University Press. Hatcher, Anna Granville (1942). Reflexive Verbs: Latin, Old French, Modern French. Baltimore: John Hopkins Press. Hausenblas, Karel (1958). Vývoj prˇedmětového genitivu v ˇceštině. Praha: Nakladatelstvı´ Československé akademie věd. Hay, Jennifer, Cristopher Kennedy, and Beth Levin (1999). ‘Scalar structure underlies telicity in degree achievements’, in T. Matthews and D. Strolovitch (eds), Proceedings of Semantic and Linguistic Theory IX. Ithaca, NY: CLC Publications, Cornell University, 127–144. HED = Puhvel, Jaan (1984–). Hittite Etymological Dictionary. Berlin: Mouton de Gruyter. HEG = Tischler, Johann (1983–). Hethitisches Etymologisches Glossar. Innsbruck: Institut fu¨r Sprachwissenschaft.

REFERENCES

341

Heine, Bernd (2012). ‘On polysemy copying and grammaticalization in language contact’, in C. Chamoreau and I. Léglise (eds), Dynamics of Contact-Induced Language Change. Berlin: De Gruyter, 125–166. Heine, Bernd and Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd and Tania Kuteva (2005). Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Henning, Walter B. (1937). Ein manicha¨isches Bet- und Beichtbuch. Berlin: Akademie der Wissenschaften. Herman, Iószef (1991). ‘On the grammatical subject in Late Latin’, in Robert Coleman (ed.), New Studies in Latin Linguistics. Amsterdam: Benjamins, 415–425. Hettrich, Heinrich (1990). Der Agens in passivischen Sa¨tzen altindogermanischer Sprachen, Go¨ttingen: Vandenhoeck & Ruprecht. Hewson, John and Vit Bubenik (2006). From Case to Adposition: The Development of Configurational Syntax in Indo-European Languages. Amsterdam: Benjamins. Hock, Hans H. (1986). ‘Voice, mood, and the gerundive (kṛtya) in Sanskrit’, Indologica Taurinensia 13: 81–102. Hock, Hans H. (1990). ‘Oblique subjects in Sanskrit?’, in Manindra K. Verma and Karuvannur P. Mohanan (eds), Experiencer Subjects in South Asian Languages. Stanford, CA: CSLI Publications, 119–139. Hock, Hans H. (1991). ‘Possessive agents in Sanskrit?’, in Hans H. Hock (ed.), Studies in Sanskrit Syntax. Delhi: Motilal Banarsidass, 55–69. Hock, Hans H. (2016). ‘Pāṇini’s language: Real or not?’, in Andrew Miles Byrd, Jessica DeLisi, and Mark Wenthe (eds), Tavet Tat Satyam: Studies in Honor of Jared S. Klein on the Occasion of His Seventieth Birthday. Ann Arbor, MI: Beech Stave Press, 101–112. Hock, Hans H. (2019). ‘Anticausative and passive in Vedic: Which way reanalysis?’, in Lars Heltoft, Iván Igartua, Brian D. Joseph, Kirsten Jeppesen Kragh, and Lene Schøsler (eds), Perspectives on Language Structure and Language Change: Studies in Honor of Henning Andersen. Amsterdam: Benjamins, 181–191. Hoffner, Harry A., Jr. (1978). Review of: J. Friedrich and A. Kammenhuber, Hethitisches Wo¨rterbuch. Zweite, vo¨llig neubearbeitete Auflage auf der Grundlage der edierten hethitischen Texte, Lieferung 1, Heidelberg 1975). Bibliotheca Orientalis 35: 242–246. Hoffner, Harry A. and H. Craig Melchert (2008). A Grammar of the Hittite Language. Part I: Reference Grammar. Winona Lake: Eisenbrauns. Holisky, Dee Ann (1987). ‘The intransitive subject in Tsova-Tush.’ In Lingua 71, 103–132. Holst, Jan Henrik (2009). Armenische Studien. Wiesbaden: Harrassowitz. Holton, Gary (2008). ‘The rise and fall of semantic alignment in North Halmahera, Indonesia’, in Mark Donohue and Søren Wichmann (eds), The Typology of Semantic Alignment. Oxford: Oxford University Press, 252–293. Holvoet, Axel (1991). Transitivity and Clause Structure in Polish. A Study in Case Marking. Warsaw: SOW. Holvoet, Axel (2013). ‘Obliqueness, quasi-subjects and transitivity in Baltic and Slavonic’, in Ilja A. Serzˇant and Leonid Kulikov (eds), Diachronic Typology of Non-Canonical Subjects. Amsterdam: Benjamins, 257–282. Holvoet, Axel and Nicole Nau (eds) (2014). Grammatical Relations and their NonCanonical Encoding in Baltic. Amsterdam: Benjamins. Holvoet, Axel and Nicole Nau (eds) (2015). Voice and Argument Structure in Baltic. Amsterdam: Benjamins.

342

REFERENCES

Holvoet, Axel and Nicole Nau (eds) (2016). Argument Realization in Baltic. Amsterdam: Benjamins. Hook, Peter Edwin (1990). ‘Experiencers in South Asian languages: A gallery’, in Verma and Mohanan (eds), 319–334. Hopper, Paul J. and Sandra A. Thompson (1980). ‘Transitivity in grammar and discourse’, Language 56(2): 251–299. Hopper, Paul J. and Elizabeth C. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Hrozný, Bedrˇich (1915). ‘Die Lo¨sung des hethitischen Problems’, Mitteilungen der Deutschen Orient-Gesellschaft 56: 17–50. Hrozný, Bedrˇich (1917). Die Sprache der Hethiter, ihr Bau und ihre Zugeho¨rigkeit sum indogermanischen Sprachstamm. Leipzig: J. C. Hinrichs’sche Buchhandlung. Hrozný, Bedrˇich (1919). Hethitische Keilschrifttexte aus Boghazko¨i, in Umschrift, mit Übersetzung und Kommentar. Leipzig: J. C. Hinrichs’sche Buchhandlung. Iemmolo, Giorgio (2011). ‘Towards a typological study of differential object marking and differential object indexation’ (PhD Dissertation, University of Pavia). Iggesen, Oliver A. (2013). ‘Number of cases’, in Matthew S. Dryer and Martin Haspelmath, (eds), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology, http://wals.info/chapter/49 (accessed 13/05/20). Inglese, Guglielmo (2020). The Hittite Middle Voice. Leiden: Brill. Inglese, Guglielmo and Silvia Luraghi (2020). ‘The Hittite periphrastic perfect’, in Robert Crellin and Thomas Ju¨gel (eds), Perfects in Indo-European languages and beyond. Amsterdam: Benjamins, 377–410. Isacˇenko, Alexander V. (1974). ‘On “have” and “be” languages: A typological sketch’, in Michael Flier (ed), Slavic Forum: Essays in Linguistics and Literature. The Hague: Mouton, 43–77. Jackendoff, Ray (1976). ‘Toward an explanatory semantic representation’, Linguistic Inquiry 7: 89–150. Jamison, Stephanie (1979). ‘The case of the agent in Indo-European’, Die Sprache 25(2): 129–143. Jamison, Stephanie W. and Joel P. Brereton (trans.) (2014). The Rigveda: The Earliest Religious Poetry of India, 3 vols. Oxford: Oxford University Press. Janda, Laura A., Anna Endresen, Julia Kuznetsova, Olga Lyashevskaya, Anastasia Makarova, Tore Nesset, and Svetlana Sokolova (2013). Why Russian Aspectual Prefixes Aren’t Empty. Prefixes as Verb Classifiers. Bloomington, IN: Slavica. Janic, Katarzyna and Charlotte Hemmings (2021). ‘A lignment shift as functional markedness reversal’, Journal of Historical Linguistics 11(2): 299–341. Janko, Richard (2011). ‘πρῶτo´ν τε καὶ ὕστατoν αἰὲν ἀείδειν. Relative chronology and the literary history of the early Greek epos’, in Øyvind Andersen and Dag Trygve T. Haug (eds), Relative Chronology in Early Greek Poetry. Cambridge: Cambridge University Press, 20–43. Jensen, Hans (1959). Altarmenische Grammatik. Heidelberg: Carl Winter. Jespersen, Otto (1927). A Modern English Grammar on Historical Principles, Vol. III. London: Allen and Unwin. Joseph, Brian (2010). ‘Language contact in the Balkans’, in Raymond Hickey (ed.), The Handbook of Language Contact. Malden, MA: Wiley-Blackwell, 618–633. Josephson, Folke (2004). ‘Semantics and typology of Hittite -ant’, in James Clackson and Birgit A. Olsen (eds), Indo-European Word Formation. Proceedings of the Conference Held at the University of Copenhagen, October 20th–22nd 2000. Copenhagen: Museum Tusculanum, 91–118.

REFERENCES

343

Ju¨gel, Thomas (2015). Die Entwicklung der Ergativkonstruktion im Alt- und Mitteliranischen. Eine korpusbasierte Untersuchung zu Kasus, Kongruenz und Satzbau, Wiesbaden: Harrassowitz. Karst, Josef (1901). Historische Grammatik des Kilikisch-Armenischen. Strassbourg: Tru¨bner. Kassian, Alexei and Ilya Yakubovich (2007). ‘Muršili II’s prayer to Telipinu (CTH 377)’, in Detlev Groddek and Marina Zorman (eds), Tabularia Hethaeorum—Hethitologische Beitra¨ge Silvin Košak zum 65. Geburtstag. Wiesbaden: Harrassowitz, 423–454. Kayne, Richard (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Keenan, Edward (1976). ‘Towards a universal definition of subject’, in Charles Li (ed.), Subject and Topic. New York: Academic Press, 303–322. Kemmer, Suzanne (1993). The Middle Voice. Amsterdam: Benjamins. Kerleroux, Françoise (1990). ‘Du mode d’existence de l’infinitif substantivé en français contemporain’, Cahiers de Grammaire 15: 55–99. Keydana, Go¨tz (2013). Infinitive im Rgveda: Formen, Funktion, Diachronie. Leiden: Brill. Keydana, Go¨tz (2018). ‘The syntax of Proto-Indo-European’, in Jared Klein, Brian Joseph, and Matthias Fritz (eds), Handbook of Comparative and Historical Indo-European Linguistics. Berlin: De Gruyter, 2195–2228. Kim, Ronald L. (2014). ‘A tale of two suffixes: *-h2 -, *-ih2 -, and the evolution of feminine gender in Indo-European’, in Sergio Neri and Roland Schuhmann (eds), 115–136. Kimball, Sarah E. (1999). Hittite Historical Phonology. Innsbruck: Innsbrucker Beitra¨ge zur Sprachwissenschaft. Kirk, Geoffrey S. (1985). ‘Homer’, in Patricia E. Easterling and Bernard M. W. Knox (eds), The Cambridge History of Classical Literature. Cambridge: Cambridge University Press, 42–91. Klaiman, Miriam H. (1991). Grammatical Voice. Cambridge: Cambridge University Press. Klamer, Marian (2008). ‘Differential marking of intransitive subjects in Kambera (Austronesian)’, in Helen de Hoop and Peter de Swrt (eds), Differential Subject Marking. Dordrecht: Springer, 281–299. Klein, Jared S (2017). ‘The syntax of Classical Armenian’, in Jared Klein, Brian Joseph and Matthias Fritz (eds), Handbook of Comparative and Historical Indo-European Linguistics, vol. II. Berlin: De Gruyter, 1097–1115. Klimov, Georgy A. (1974). ‘On the character of languages of active typology’, Linguistics 12(131): 11–26. Klingenschmitt, Gert (1982). Das Altarmenische Verbum, Wiesbaden: Dr. Ludwig Reichert. Kloekhorst, Alwin (2008). Etymological Dictionary of the Hittite Inherited Lexicon. Leiden: Brill. Knudtzon, Jørgen A. (1902). Die zwei Arzawa-Briefe. Die a¨ltesten Urkunden in indogermanischer Sprache. Leipzig: J. C. Hinrichs’sche Buchhandlung. Ko¨lligan, Daniel (2013). ‘Non-canonical case marking: Genitive subjects in Classical Armenian’, in Ilja A. Serzˇant and Leonid Kulikov (eds), The Diachronic Typology of Non-Canonical Subjects. Amsterdam: Benjamins, 73–90. Ko¨lligan, Daniel (2020). ‘The perfect in Classical Armenian’, in Robert Crellin and Thomas Ju¨gel (eds), Perfects in Indo-European Languages and Beyond. Amsterdam: Benjamins, 351–376. Koontz-Garboden, Andrew (2005). ‘On the typology of state/change of state alternations’, Yearbook of Morphology: 83–117. Krajcˇovicˇ, Rudolf (1974). Slovencˇina a slovanské jazyky, vol. I: Praslovanská genéza slovencˇiny. Bratislava: SPN.

344

REFERENCES

Krishnamurti, Bhadriraju (2003). The Dravidian Languages. Cambridge: Cambridge University Press. Krys’ko, Vadim B. (2006). Istoricˇeskij sintaksis russkogo jazyka. Ob”ekt i perexodnost’. 2-e izdanie, ispravlennoe i dopolnennoe. Moscow: Azbukovnik. Kulikov, Leonid (1996). ‘Vedic -ya-presents: Semantics and the placement of stress’, in Wolfgang Meid (ed.), Sprache und Kultur der Indogermanen: Akten der X. Fachtagung der Indogermanischen Gesellschaft Innsbruck, 22.–28. September 1996. Innsbruck: Institut fu¨r Sprachwissenschaft, 341–350. Kulikov, Leonid (2001). The Vedic ya-presents. Universiteit Leiden Proefschrift. Kulikov, Leonid (2011a). ‘Drifting between passive and anticausative: True and alleged accent shifts in the history of Vedic -ya-presents’, Вoпpocы языкoвoгo poдcтвa/Journal of Language Relationship 6: 185–215 [also containing discussion by other scholars and a reply by Kulikov]. Kulikov, Leonid (2011b). ‘Passive to anticausative through impersonalization: The case of Vedic and Indo-European’, in Andrej Malchukov and Anna Siwierska (eds), Impersonal Constructions: A Cross-Linguistic Perspective. Amsterdam: Benjamins, 229–254. Kulikov, Leonid (2012). The Vedic -ya-presents: Passives and intransitivity in Old IndoAryan. Amsterdam: Rodopi. Kulikov, Leonid (2013). ‘Language vs. grammatical tradition in Ancient India: How real was Pāṇinian Sanskrit? Evidence from the history of late Sanskrit passives and pseudopassives’, Folia Linguistica Historica 34: 59–91. Kuryłowicz, Jerzy (1952). L’accentuation des langues indo-europeénnes (Polska Akademia Umiejętnos´ci: Prace komisji językowej, 37.) Kraków: Nakład Polskiej Akademii Umiejętnos´ci. Kuryłowicz, Jerzy (1964). The Inflectional Categories of Indo-European. Heidelberg: Winter. Kuryłowicz, Jerzy (1965). ‘The evolution of grammatical categories’, Diogenes 13(51): 55–71. Kurz, Josef ([1966–1997] 2006). Slovar’ staroslavjanskogo jazyka, vols. 1–4. St. Petersburg: Sankt-Peterburgskij gosudarstvennyj universitet [reprint of the original (1966– 1997). Slovnı´k jazyka staroslověnského. Lexicon linguae palaeoslovenicae, 4 vols. Prague: Academia, Euroslavica]. Kurzová, Helena (1993). From Indo-European to Latin. The Evolution of a Morphosyntactic Type. Amsterdam: Benjamins. La Fauci, Nunzio (1994). Objects and Subjects in the Formation of Romance Morphosyntax. Translated by Carol Rosen. Bloomington, IN: Indiana University Linguistics Club Publications [orig. pub. 1988 in Italian]. Lakoff, George (1977). ‘Linguistic gestalts’, CLS 13: 238–287. Lamprecht, Arnošt, Dušan Šlosar, and Jaroslav Bauer (1986). Historická mluvnice ˇceštiny. Prague. Landau, Idan (2010). The Locative Syntax of Experiencers. Cambridge, MA: MIT Press. Langacker, Ronald W. (1991). Concept, Image, and Symbol. The Cognitive basis of Grammar. Belin: Mouton de Gruyter. Lapesa, Rafael. (1968). ‘Evolución sintáctica y forma lingu¨´ıstica interior en español’, in Antonio Quilis, Ramón B. Carril, and Margarita Cantarero (eds), Actas del XI Congreso Internacional de Lingu¨´ıstica y Filologı´a Románicas (Madrid, 1–9 septiembre 1965). Madrid: Revista de Filologı´a Española 1, 131–150. Laroche, Emmanuel (1962). ‘Un “ergative” en indo-europeen d’A sie Mineure’, Bulletin de la Societe de linguistique de Paris 57: 23–43.

REFERENCES

345

Laroche, Emmanuel (1971). Catalogue des textes hittites. Paris: Klincksieck. Laskowski, Roman (1986). ‘The development of the category of gender in the Slavic languages’, in Dieter Kastowsky and Alexander Szwedek (eds), Linguistics across Historical and Geographical Boundaries, vol. I: Linguistic Theory and Historical Linguistics. Berlin, New York: Mouton De Gruyter, 459–472. Lavine, J. (2016). ‘Variable argument realization in Lithuanian impersonals’, in A. Holvoet and N. Nau (eds), Argument Realization in Baltic. Amsterdam: Benjamins, 107–135. Lazard, Gilbert (1998). ‘Définition des actants dans les langues européennes’, in Jack Feuillet (ed.), Actance et valence dans les langues de l’Europe. Berlin: Mouton de Gruyter, 11–146. Lazzeroni, Romano (2002a). ‘Ruoli tematici e genere grammaticale: un aspetto della morfosintassi indoeuropea?’, Archivio Glottologico Italiano 87: 1–17. Lazzeroni, Romano (2002b). ‘Il nome greco del sogno e il neutro indoeuropeo’, Archivio Glottologico Italiano 87: 145–162. Lazzeroni, Romano (2009). ‘Deissi e tempo grammaticale. Il caso dell’imperativo indoeuropeo’, Archivio Glottologico Italiano 94: 226–234. Lazzeroni, Romano (2013). ‘Fra ruoli semantici e ruoli pragmatici’: il cosiddetto “schema pindarico” nel greco antico’, Archivio Glottologico Italiano 98: 1–15. Ledo-Lemos, Francisco (2003). Femininum Genus. A Study on the Origins of the IndoEuropean Feminine Grammatical Gender. Munich: Lincom Europa. Legate, Julie Ann (2014). ‘Split ergativity based on nminal type’, in Lingua 148, 183–212. Lehmann, Christian (1988). ‘Predicate classes and participation’, in Studies in General Comparative Linguistics. Ko¨ln: AKUP, 33–77. Lehmann, Christian (1991). ‘Predicate classes and participation’, in Hansjakob Seiler and Waldfried Premper (eds), Partizipation: das sprachliche Erfassen von Sachverhalten. Tu¨bingen: Gunter Narr Verlag, 183–239. Lehmann, Christian (2015). Thoughts on grammaticalisation. 3rd edition. Berlin: Language Science Press. Lehmann, Thomas (1998). ‘Old Tamil’, in Sanford B. Steever (ed.), The Dravidian Languages. London: Routledge, 75–99. Lehmann, Winfred P. (1989a). ‘Problems in Proto-Indo-European grammar: Residues from Pre-Indo-European’, General Linguistics 29: 228–246. Lehmann, Winfred P. (1989b). ‘Earlier stages of Proto-Indo-European’, in Karin Heller, Oswald Panagl, and Johann Tischler (eds), Indogermanica Europaea: Festschrift fu¨r Wolfgang Meid zum 60. Geburtstag. Graz: Institut fu¨r Sprachwissenschaft der Universita¨t Graz, 109–131. Łesiów, Michał (1998). ‘The Polish and Ukrainian languages: A mutually beneficial relationship’, Harvard Ukrainian Studies 22: 393–406. Lesz-Duk, Maria (1995). Funkcje składniowe rzeczowników w dopełniaczu w historii polszczyzny. Częstochowa: Wydawnictwo WSP. Leumann, Manu, Johann B. Hofmann, and Anton Szantyr (1965). Lateinische Grammatik. Bd. II Syntax und Stilistik. Munich: Beck. Levin, Beth and Malka Rappaport Hovav (1995). Unaccusativity. At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT Press. Levin, Beth and Malka Rappaport Hovav (2005). Argument Realization. Cambridge: Cambridge University Press. Levin, Beth and Bonnie Krejci (2019). ‘Talking about the weather: Two construals of precipitation events in English’, Glossa 4(1): 1–29. Lewis, Charlton D. and Charles Short (1942). A Latin Dictionary. Oxford: Oxford University Press

346

REFERENCES

Lightfoot, David (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lindsay, Wallace M. (1895). A Short Historical Latin Grammar. Oxford: Clarendon Press. Lindsay, Wallace M. (1907). Syntax of Plautus. Oxford: Parker. LIV2 = Lexikon der indogermanischen Verben, 2nd ed., revised and edited by Martin Ku¨mmel and Helmut Rix. Wiesbaden: Reichert, 2001. Lo¨fstedt, Einar (1911). Philologischer Kommentar zur Peregrinatio Aetheriae. Uppsala: Almqvist & Wiksell. Lo¨fstedt, Einar (1936). Vermischte Studien. Lund: Gleerup. Lo¨fstedt, Einar (1942). Syntactica, vol. II. Lund: Gleerup. Lohmann, Johannes (1937). ‘Ist das idg. Perfektum nominalen Ursprungs?’, Zeitschrift fu¨r vergleichende Sprachforschung 64: 42–61. Lubotsky, Alexander (1997). A Ṛgvedic Word Concordance, 2 parts (American Oriental Series, 82–83). New Haven, CT: American Oriental Society. Luraghi, Silvia (1993). ‘La modificazione nominale nelle lingue anatoliche’, Archivio Glottologico Italiano 78(2): 144–166. Luraghi, Silvia (1995). ‘Prototypicality and agenthood in Indo-European’, in Henning Andersen (ed.), Historical Linguistics 1993: Selected Papers from the 11th International Conference on Historical Linguistics. Los Angeles: Benjamins, 259–268. Luraghi, Silvia (1997). Hittite [Languages of the World/Materials 114]. Munich: Lincom Europa. Luraghi, Silvia (2008). ‘Possessive constructions in Anatolian, Hurrian, Urartean, and Armenian as evidence for language contact’, in Billie K. Collins, Mary R. Bachvarova, and Ian C. Rutherford (eds), Anatolian Interfaces. Oxford: Oxbow Press, 147–155. Luraghi, Silvia (2009a). ‘The origin of the feminine gender in PIE: An old problem in a new perspective’, in Vit Bubenik, John Hewson, and Sarah Rose (eds), Grammatical Change in Indo-European Languages. Amsterdam: Benjamins, 3–13. Luraghi, Silvia (2009b). ‘Indo-European nominal classification: From abstract to feminine’, in Stephanie W. Jamison, H. Craig Melchert, and Brent Vine (eds), Proceedings of the 20th Annual UCLA Indo-European Conference. Bremen: Hempen, 115–131. Luraghi, Silvia (2010). ‘Experiencer predicates in Hittite’, in Ronald Kim, Norbert Oettinger, Elisabeth Rieken, and Michael Weiss (eds), Ex Anatolian Lux: Anatolian and Indo-European Studies in Honor of H. Craig Melchert. Ann Arbor, MI: Beech Stave Press, 248–264. Luraghi, Silvia (2011). ‘The origin of the Proto-Indo-European gender system: Typological considerations’, Folia Linguistica 45(2): 435–464. Luraghi, Silvia (2015). ‘Instrument and cause in the Indo-European languages and in ProtoIndo-European’, Indo-European Linguistics and Classical Philology 19: 603–618. Luraghi, Silvia (2017). ‘The syntax of Anatolian: The simple sentence’, in Jared Klein, Brian Joseph, and Matthias Fritz (eds), Handbook of Comparative and Historical IndoEuropean Linguistics (Handbooks of Linguistics and Communication Science 41/1). Berlin: De Gruyter, 274–290. Lyons, Christopher (1999). Definiteness. Cambridge: Cambridge University Press. Lyons, John (1969). Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. Lyutikova, Ekaterina and Andrei Sideltsev (2021). ‘Voice neutrality in Hittite infinitives: A restructuring analysis’, Journal of Historical Syntax 5: 1–70. MacDonell, Arthur Anthony (1916). A Vedic Grammar for Students. Oxford: Clarendon.

REFERENCES

347

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 typologicalsemiotic perspective’, Lingua 120: 1610–1636. McGregor, William B. (2017). ‘Grammaticalization of ergative case marking’, in Jessica Coon, Diane Massam, and Lisa Demena Travis (eds) The Oxford Handbook of Ergativity. Oxford: Oxford University Press, 448–464. Magnani, Marco. (2019). ‘Non-canonical case marking on subjects in Russian and Lithuanian. An interface approach’, in Evolutionary Linguistic Theory 1(2), 175–196. Mahajan, Anoop K. (2004). ‘On the origin of non-nominative subjects’, in P. Bhaskararao and K. V Subbarao (eds), Non-Nominative Subjects. Amsterdam: Benjamins, 283–299. Malchukov, Andrei (2005). ‘Case pattern splits, verb types, and construction competition’, in Mengistu Amberber and Helen de Hoop (eds), Competition and Variation in Natural Languages. London: Elsevier, 73–117. Malchukov, Andrei (2008). ‘Split intransitives, experiencer objects and “transimpersonal constructions”: (Re-)establishing the connection’, in Mark Donohue and Søren Wichmann (eds), The Typology of Semantic Alignment. Oxford: Oxford University Press, 76–100. Malchukov, Andrei and Akio Ogawa (2011). ‘Towards a typology of impersonal constructions’, in Andrei Malchukov and Anna Siewierska (eds), Impersonal Constructions: A Cross-Linguistic Perspective. Amsterdam: Benjamins, 19–56. Maldonado, Ricardo (1999). A media voz. Problemas conceptuales del clı´tico se. México: Universidad Nacional Autónoma de México. Maldonado, Ricardo (2009). ‘Middle as a basic voice system’, in Lilian Guerrero, Sergio Ibáñez Cerda, and Valeria A. Belloro (eds), Studies in Role and Reference Grammar. México: Universidad Nacional Autónoma de México, 69–109. Malyševa, Anna V. (2008). ‘“Variativnost” genitiva i akkuzativa pri glagolach s obšcˇim znacˇeniem “berecˇ”, “zašcˇišcˇat”‘v russkix letopisjax, Russkij Jazyk v Naucˇnom Osvešcˇenii 1(15): 79–97. Markowski, Andrzej (2002). Nowy słownik poprawnej polszczyzny PWN. Wyd. 2., 2 dodr. Warszawa: Wydawn. Naukowe PWN. Mat‘evosyan, A. S. (1994). Varkʹ Mesrop Maštocʹi. Erewan: Hayastan. Matasovic´, Ranko (2004). Gender in Indo-European. Heidelberg: Winter. Mathiassen, Terje (1996). A Short Grammar of Lithuanian. Columbus, OH: Slavica. Mathieu, Eric (2006). ‘Quirky subjects in Old French’, Studia Linguistica 60(3): 282–312. Matras, Yaron (2009). Language Contact. Cambridge: Cambridge University Press. Matras, Yaron and Jeanette Sakel (2007). ‘Investigating the mechanisms of pattern replication in language convergence’, Studies in Language 31(4): 829–865. Matzinger, Joachim (2005). Untersuchungen zum altarmenischen Nomen: die Flexion des Substantivs. Dettelbach: Ro¨ll. Meier-Bru¨gger, Michael (2000). Indogermanische Sprachwissenschaft, 7. vo¨llig neubearbeitete Auflage der fru¨heren Darstellung von Hans Krahe, unter Mitarbeit von Matthias Fritz und Manfred Mayrhofer. Berlin: de Gruyter. Meillet, Antoine (1899–1900). ‘Recherches sur la syntaxe comparée de l’arménien I’, Mémoires de la Société Linguistique de Paris 11(6): 369–388. Meillet, Antoine ([1921] 1948). Linguistique historique et linguistique générale. Paris: Champion.

348

REFERENCES

Meillet, Antoine (1923). ‘Le dèveloppement du verbe “avoir”’, in Jacob Wackernagel (ed.), Antidoron. Festschrift Jacob Wackernagel zur Vollendung des 70. Lebensjahres am 11. Dezember 1923. Go¨ttingen: Vandenhoeck & Ruprecht, 9–13. Meillet, Antoine (1936). Esquisse d’une grammaire comparée de l’arménien classique, Vienna: Imprimerie des PP. Mékhitaristes. Meillet, Antoine and Joseph Vendryes (1924). Traitè de Grammaire Comparée des Langues Classiques. Paris: Champion. Melchert, H. Craig (1977). ‘Ablative and instrumental in Hittite’ (Ph.D. dissertation, Harvard University). Melchert, H. Craig (1994). Anatolian Historical Phonology. Amsterdam: Rodopi. Melchert, H. Craig (2000). ‘Tocharian plurals in -nt- and related phenomena’, Journal of Tocharian and Indo-European Studies 9: 53–75. Melchert, H. Craig (2003). ‘Language’, in H. Craig Melchert (ed.), The Luwians. Leiden: Brill, 170–210. Melchert, H. Craig (2005). ‘The problem of the Luvian influence on Hittite: When and how much?’, in Gerhard Meiser and Olav Hackstein (eds), Sprachkontakt un Sprachwandel: Akten der XI. Fachtagung der Indogermanischen Gesellschaft, 17.–23. September 2000, Halle an der Saale. Wiesbaden: Reichert, 445–460. Melchert, H. Craig (2008). ‘Palaic’, in Roger D. Woodard (ed.), The Ancient Languages of Asia Minor. Cambridge: Cambridge University Press, 40–45. Melchert, H. Craig (2011a). ‘The PIE Collective Plural and the “ rule’, in Thomas Krisch and Thomas Lindtner (eds), Indogermanistik und Linguistik im Dialog. Akten der XIII. Fachtagung der Indogermanischen Gesellschaft vom 21. bis 27. September 2008 in Salzburg. Wiesbaden: Reichert, 395–400. Melchert, H. Craig (2011b). ‘The problem of the ergative case in Hittite’, in Michèle Fruyt, Michel Mazoyer, and Dennis Pardee (eds), Grammatical Case in the Languages of the Middle East and Europe. Acts of the International Colloquium Variations, concurrence et evolution des cas dans divers domaines linguistiques Paris 2–4 April 2007. Chicago: Oriental Institute, 161–167. Melchert, H. Craig (2012). ‘Genitive case and possessive adjective in Anatolian’, in Vincenzo Orioles (ed.), Per Roberto Gusmani. Studi in ricordo. Linguistica storica e teorica. Udine: Forum, 273–286. Melchert, H. Craig (2014). ‘PIE *-eh2 as an “individualizing” suffix and the feminine gender’, in Neri and Schuhmann (eds), 257–271. Melchert, H. Craig (2016). ‘The case of the agent in Anatolian and Indo-European’, in Dieter Gunkel, Joshua T. Katz, Brent Vine, and Michael Weiss (eds), Sahasram Ati Srajas: IndoIranian and Indo-European Studies in Honor of Stephanie W. Jamison. Ann Arbor, MI: Beech Stave Press, 239–249. Melchert, H. Craig (2017a). ‘Anatolian’, in Mate Kapovic´ (ed.), The Indo-European Languages (2nd ed.). London/New York: Routledge, 171–204. Melchert, H. Craig (2017b). ‘The source(s) of Indo-European participles in *-e/ont-’, in Claire Le Feuvre, Daniel Petit, and Georges-Jean Pinault (eds), Verbal Adjectives and Participles in Indo-European Languages. Proceedings of the Conference of the Society for Indo-European Studies (Indogermanische Gesellschaft), Paris, 24th to 26th September 2014. Bremen: Hempen, 203–206. Melchert, H. Craig and Norbert Oettinger (2009). ‘Ablativ und Instrumental im Hethitischen und Indogermanischen. Ein Beitrag zur relativen Chronologie’, Incontri Linguistici 32: 53–73.

REFERENCES

349

Melis, Chantal (1999a). ‘Variación sintáctica con los verbos de emoción’, Español actual 71: 49–62. Melis, Chantal (1999b). ‘Los verbos placer y pesar en la Edad Media: la expresión “impersonal” de las emociones’, in Fulvia Colombo Airoldi (ed.), El Centro de Lingu¨´ıstica Hispánica y la lengua española. Volumen conmemorativo del 30 aniversario de su fundación. México: Universidad Nacional Autónoma de México, 87–105. Melis, Chantal (2019). ‘Los causativos emocionales del español. Un estudio aspectual’, Anuario de Letras. Lingu¨´ıstica y Filologı´a 7 (1): 105–156. Melis, Chantal and Marcela Flores (2013). ‘On the historical expansion of non-canonically marked “subjects” in Spanish’, in Serzˇant and Kulikov (eds), 163–184. Melis, Chantal and Marcela Flores (2018). ‘The dative experiencer of Spanish gustar’, in Miriam Bouzouita, Ioanna Sitaridou, and Enrique Pato (eds), Studies in Historical IberoRomance Morpho-Syntax. Amsterdam: Benjamins, 123–147. Melis, Chantal, Marcela Flores, and Sergio Bogard (2003). ‘La historia del español: propuesta de un tercer perı´odo evolutivo’, Nueva Revista de Filologı´a Hispánica 51: 1–56. Merlan, Francesca (1985). ‘Split intransitivity: Functional oppositions in intransitive inflection’, in Johanna Nichols and Anthony C. Woodbury (eds), Grammar Inside and Outside the Clause. Cambridge: Cambridge University Press, 324–362. Meyer, David and Christian Buchta (2019). proxy: Distance and Similarity Measures. R package version 0.4-23. https://CRAN.R-project.org/package=proxy (accessed 18/05/21). Meyer, David, Achim Zeileis, and Kurt Hornik (2017). vcd: Visualizing Categorical Data. R package version 1.4–4. Meyer, Robin (2014). ‘Remodelling the historical morphology of the Classical Armenian -eal participle’, Banber Matenadarani 21: 385–398. Meyer, Robin (2016). ‘Morphosyntactic Alignment, Pattern Replication, and the Classical Armenian Periphrastic Perfect’, in Stephanie W. Jamison, H. Craig Melchert, and Brent Vine (eds), Proceedings of the 26th Annual UCLA Indo-European Conference: Los Angeles, October 24th and 25th, 2014, Bremen: Hempen, 117–133. Meyer, Robin (2017). ‘Iranian–Armenian language contact in and before the 5th century ce: An investigation into pattern replication and societal multilingualism’ (D.Phil. thesis, University of Oxford). Meyer, Robin (2022). ‘A lignment change and changing alignments: Armenian syntax and the first “death” of Parthian’, in Michele Bianconi, Marta Capano, Domenica Romagno, and Francesco Rovai (eds), Ancient Indo-European Languages between Linguistics and Philology. Contact, Variation, and Reconstruction. Leiden: Brill. Meyer, Robin (forthcoming). ‘Languages in contact: Armenian and Iranian’, in A. Orengo and I. Tinti (eds), Armenian Linguistics (Handbuch der Orientalistik 23(2)). Leiden: Brill. Michaelis, Laura (1993). ‘On deviant case-marking in Latin’, in Robert D. Van Valin, Jr (ed.), Advances in Role and Reference Grammar. Amsterdam: Benjamins, 311–373. Miklosich, Franz (1883). Vergleichende Grammatik der slavischen Sprachen. Syntax. Vierter Band. Wien: Wilhelm Braumu¨ller. Mindak, Jolanta (1990). Językowa kategoria z˙ywotnos´ci w polszczyz´nie i słowian´szczyz´nie na tle innych języków ´swiata. Próba ujęcia typologicznego. Wrocław: Ossolineum. Mithun, Marianne (1991). ‘Active/agentive case marking and its motivations’, Language 67(3): 510–546.

350

REFERENCES

Mithun, Marianne (2005). ‘Ergativity and language contact on the Oregon Coast: Alsea, Siuslaw and Coos’ in Andrew K. Simpson (ed.), Proceedings of the Annual Meeting of the Berkeley Linguistics Society 26. Special Session on Syntax and Semantics of the Indigenous Languages of the Americas. Berkeley, CA: University of California Press, 77–95. Mithun, Marianne (2008). ‘The emergence of agentive systems in core argument marking’, in Mark Donohue and Søren Wichmann (eds), The Typology of Semantic Alignment. Oxford: Oxford University Press, 297–333. Mohanan, Tara (1994). Argument Structure in Hindi. Stanford, CA: CSLI Publications. Montaut, Annie (2004). ‘Oblique main arguments in Hindi as localizing predications’, in Bhaskararao and Subbarao (eds), vol. II, 33–56. Montaut, Annie (2013). ‘The rise of non-canonical subjects and semantic alignments in Hindi’, in Ilja Serzˇant and Leonid Kulikov (eds), The Diachronic Typology of NonCanonical Subjects. Amsterdam: Benjamins, 91–117. Mørck, Endre (2005). ‘Morphological developments from Old Nordic to Early Modern Nordic: Inflection and word formation’, in Gun Widmark (ed.), The Nordic Languages: An International Handbook of the History of the North Germanic Languages, vol 2. Berlin: De Gruyter, 1213–1221. Moravcik, Edith A. (1978). On the distribution of ergative and accusative patterns. Lingua 45 (3–4): 233–279. Moreno Cabrera, Juan Carlos (1984). ‘La diátesis anticausativa. Ensayo de sintaxis general’, Revista Española de Lingu¨´ıstica 14: 21–43. Moro, Andrea (2010). Breve storia del verbo essere. Milan: Adelphi. Moroz, George (2017). lingtypology: easy mapping for Linguistic Typology. https://cran.rproject.org/package=lingtypology (accessed 18/05/21). Moser, Michael (1998). Die polnische, ukrainische und weißrussische Interferenzschicht im russischen Satzbau des 16. und 17. Jahrhunderts. Frankfurt: Lang. Mrazovic´, Pavica (2009). Gramatika srpskog jezika za strance. Sremski Karlovci, Novi Sad: Izdavacˇka knjizˇarnica Zorana Stojanovic´a. Mustajoki, Arto and Mikhail V. Kopotev (2005). ‘Lodku uneslo vetrom: uslovija i konteksty upotreblenija russkoj ‘stixijnoj’ konstrukcii’, Russian Linguistics 29(1): 1–38. Mu¨th, Angelika (2014). ‘Indefiniteness, animacy and object marking: A quantitative study based on the Classical Armenian Gospel translation’ (Ph.D. dissertation, University of Oslo). Næss, Åshild (2007). Prototypical Transitivity. Amsterdam: Benjamins. Næss, Åshild (2009). ‘Varieties of dative’, in Andrei Malchukov and Andrew Spencer (eds), The Oxford Handbook of Case. Oxford: Oxford University Press, 572–580. Napoli, Maria (2009). ‘Impersonal passivization and agentivity in Latin’, in S. Jamison, C. Melchert, and B. Vine (eds), Proceedings of the 20th Annual UCLA Indo-European Conference, October 31st and November 1st 2008. Bremen: Hempen, 163–178. Napoli, Maria (2010). ‘How impersonal is the Latin impersonal passive?’, in Peter Anreiter and Manfred Kienpointner (eds), Latin Linguistics Today: Proceedings of 15th International Colloqium on Latin Linguistics. Innsbruck: Innsbrucker Beitra¨ge zur Sprachwissenschaft, 163–178. Napoli, Maria (2013). ‘Semantic constraints on the Latin impersonal passives’, in van Gelderen, Cennamo, and Barðdal (eds), 373–403. Narbona Jiménez, Antonio (1989). Sintaxis española. Nuevos y viejos enfoques. Barcelona: Ariel.

REFERENCES

351

Neri, Sergio and Roland Schumann (eds) (2014). Studies in the Collective and Feminine in Indo-European from a Diachronic and Typological Perspective. Leiden: Brill. Nesset, Tore and Julia Kuznetsova (2015a). ‘In which case are Russians afraid? Bojat’sja with genitive and accusative objects’, Journal of Slavic Linguistics 23(2): 255–283. Nesset, Tore and Julia Kuznetsova (2015b). ‘Constructions and language change: From genitive to accusative objects in Russian’, Diachronica 32(3): 365–396. Neu, Erich (1969). Review of: Kastner, Wolfgang. 1967. Die griechischen Adjektive zweier Endungen auf –OΣ. Indogermanische Forschungen 74: 235–241. Neu, Erich (1989). ‘Zum Alter der personifizierenden -ant- Bildung des Hethitischen’, Historische Sprachforschung 102: 1–15. Neue, Friedrich and Christian Wagener (1985). Formenlehre der lateinischen Sprache, vols. III/IV. Hildesheim: Georg Olms. Neumann, Gu¨nter (1982). ‘Die Konstruktionen mit Adjectiva genitivalia in den luwischen Sprachen’, in Erich Neu (ed.), Ingestigationes philologicae et comparativae. Wiesbaden: Harrassowitz, 149–161. Newman, John (1996). Give. A Cognitive Linguistic Study. Berlin: Mouton de Gruyter. Nichols, Johanna (1986). ‘Head-marking and dependent-marking grammar’, Language 62: 56–84. Nichols, Johanna (1992). Linguistic Diversity in Space and Time. Chicago: The University of Chicago Press. Nichols, Johanna (1993). ‘The linguistic geography of the Slavic expansion’, in Robert A. Maguire and Alan Timberlake (eds), American Contributions to the Eleventh International Congress of Slavists. Bratislava, August–September 1993. Literature, Linguistics, Poetics. Columbus OH: Slavica, 377–391. Nichols, Johanna (2008). ‘Why are stative-active languages rare in Eurasia? A typological perspective on split-subject marking’, in Mark Donohue and Søren Wichmann (eds), The Typology of Semantic Alignment. Oxford: Oxford University Press,119–139. Nichols, Johanna (2020). ‘Slavic homeland and migrations’, in Marc L. Greenberg (ed.), Encyclopedia of Slavic Languages and Linguistics online. Brill: Leiden. Nuorluoto, Jussi (2010). ‘Central Slovak and Kajkavian structural convergences: A tentative survey’, Slovo. Journal of Slavic Languages and Literatures 50: 37–45. Oberlies, Thomas (2001). Pāli. A Grammar of the Language of the Theravāda Tipiṭaka. Berlin: De Gruyter. Oberlies, Thomas (2003). A Grammar of Epic Sanskrit. Berlin: De Gruyter. Oettinger, Norbert (2001). ‘Neue Gedanken u¨ber das nt-Suffix’, in Onofrio Carruba and Wolfgang Meid (eds), Anatolisch und Indogermanisch. Anatolico e Indoeuropeo. Akten des Kolloquiums der Indogermanischen Gesellschaft Pavia, 22.–25. September 1998. Innsbruck: Innsbrucker Beitra¨ge zur Sprachwissenschaft, 301–315. Onishi, Masayuki (2001). ‘Non-canonically marked subjects and objects: Parameters and properties’, in Aikhenvald, Dixon, and Onishi (eds), 1–51. Ormazabal, Javier and Juan Romero (2013). ‘Object clitics, agreement and dialectal variation’, Probus. International Journal of Latin and Romance Linguistics 25: 301–344. Orqueda, Verónica (2019). Reflexives in Vedic. Leiden: Brill. Ostrowski, Manfred (1985). ‘Zur Entstehung und Entwicklung des indogermanischen Neutrums’, in Benfried Schlerath (ed.), Grammatische Kategorien, Funktion und Geschichte. Wiesbaden: Harrassowitz, 313–323. Palancar, Enrique L. (2002). The Origin of Agent Markers. Berlin: Akademie Verlag. Palmer, Frank R. (1994). Grammatical Roles and Relations. Cambridge: Cambridge University Press.

352

REFERENCES

Pāṇini. 4th c. BCE (?). Aṣṭādhyāyī, ed. and trans. Otto Bo¨htlingk (1887). Paˆṇini’s Grammatik. Leipzig: Haessel [repr. 1971, Hildesheim: Olms]. Papoyan, Artašes and Khachik Badikyan (2003). Žamanakakicʹ hayocʹ lezvi šarahyusutʹyun, Erewan: Erewani hamalsarani hratarakcˇʹutʹyun. Patri, Sylvain (2007). L’alignement syntaxique dans les langues indo-européennes d’A natolie. Wiesbaden: Harrassowitz. Payne, John R. (1979). ‘Transitivity and intransitivity in the Iranian languages of the U.S.S.R’, in Paul R. Clyne, William F. Hanks, and Carol L. Hofbauer (eds), The Elements: A Parasession on Linguistic Units and Levels, April 20–21, 1979, including Papers from the Conference on Non-Slavic Languages of the USSR, April 18, 1979. Chicago: Linguistic Society, 436–447. Payne, John R. (1980). ‘The decay of ergativity in Pamir languages’, Lingua 51(2–3): 147–186. Pensado, Carmen (1986), ‘Inversion de marquage et perte du système casuel en ancien français’, Zeitschrift fu¨r ro¨manische Philologie 102: 271–296. Pérez Vázquez, Maria Enriqueta (2007). El infinitivo y su sujeto en español. Bologna: Gedit. Pesetsky, David (1995). Zero Syntax: Experiencers and Cascades. Cambridge, MA: MIT Press. Peterson, John (1998). Grammatical Relations in Pāli and the Emergence of Ergativity in Indo-Aryan. Munich: LINCOM. Piccini, Silvia (2008). ‘Traces of non-nominative alignment in Lithuanian: The impersonal constructions in Indo-European perspective’, Baltistica 43:3: 437–461. Pieroni, Silvia (2000). ‘Agents in Latin impersonal passives’, Mnemosyne 45(2): 288–301. Pinault, Georges-Jean (2008). Chrestomathie Tokharienne. Leuven: Peeters. Pinault, Georges-Jean (2011). ‘Sur l’histoire des cas en Tokharien’, in Michèle Fruyt, Michel Mazoyer, and Dennis Pardee (eds), Grammatical Case in the Languages of the Middle East and Europe. Acts of the International Colloquium ‘ Variations, concurrence et évolution des cas dans divers domaines linguistiques’, Paris, 2–4 April 2007, Chicago: Oriental Institute, 383–398. Pinkster, Harm (1988). Latin Syntax and Semantics. London: Routledge. Pinkster, Harm (1991). Sintassi e Semantica Latina. Turin: Rosenberg & Sellier. Pinkster, Harm (1992). ‘The Latin impersonal passive’, Mnemosyne 56: 159–177. Pinkster, Harm (2015). The Oxford Latin Syntax. Volume I. The Simple Clause. Oxford: Oxford University Press. Piper Predrag, Ivana Antonic´, Vladislava Ruzˇic´, Sreto Tanasic´, Ljudmila Popovic´, and Branko Tošovic´ (2005). Sintaksa savremenoga srpskog jezika: prosta recˇenica. Belgrade: Institut za srpski jezik SANU, Beogradska knjiga, Matica srpska. Pirejko, Liya A. (1966). ‘Talyšskij Jazyk’, in V. V. Vinogradow (ed.), Jazyki Narodov CCCR. Indoevropejskie jazyki. Moscow: Institute of Linguistics, Moscow Academy of Sciences of the USSR, 302–322. Pisarkowa, Krystyna (1984). Historia składni języka polskiego. Wrocław: Ossolineum. Plank, Frans (1985). ‘The extended accusative/restricted nominative in perspective’, in Frans Plank (ed.), Relational Typology. Berlin: De Gruyter, 269–310. Plank, Frans (1995). ‘Ergativity’, in Joachim Jacobs et al. (eds), Syntax: An International Handbook of Contemporary Research, vol. II. Berlin: Walter de Gruyter, 1184–1199. Pleteršnik, Maks (1894–1895). Slovensko-nemški slovar. Ljubljana: Knezoškofijstvo. Polinski, Maria (2016). Deconstructing Ergativity: Two Types of Ergative Languages and Their Features. Oxford: Oxford University Press.

REFERENCES

353

Pooth, Roland, Peter Alexander Kerkhof, Leonid Kulikov, and Jóhanna Barðdal (2019). ‘The origin of non-canonical case marking of subjects in Proto-Indo-European. Accusative, ergative, or semantic alignment’, Indogermanische Forschungen 124: 245–264. Pooth, Roland and Verónica Orqueda (2021). ‘A lignment change and the emergence of the thematic conjugation from Proto-Indo-European to Indo-European: A wedding of hypotheses’, Transactions of the Philological Society 119(2): 107–151 Prins, Anna (1997). Hittite Neuter Singular—Neuter Plural. Some Evidence for a Connection. Leiden: Research School CNWS. Radden, Gu¨nter and Kovecses, Zoltán (1999). ‘Metonymy: Developing a cognitive linguistic view’, Cognitive Linguistics 9(1): 37–78. Ramchand, Gillian (2008). Verb Meaning and the Lexicon. A First-Phase Syntax. Cambridge: Cambridge University Press. Ramchand, Gillian (2019). ‘Event structure and verbal decomposition’, in Robert Triswell (ed.), The Oxford Handbook of Event Structure. Oxford: Oxford University Press, 315– 341. Rappaport Hovav, Malka (2008). ‘Lexicalized meaning and the internal structure of events’, in Susan Rothstein (ed.), Theoretical and Crosslinguistic Approaches to the Semantics of Aspect. Amsterdam: Benjamins, 13–42. Rappaport Hovav, Malka (2014). ‘Building scalar changes’, in Artemis Alexiadou, Hagit Borer, and Florian Scha¨fer (eds), The Syntax of Roots and the Roots of Syntax. Oxford: Oxford University Press, 259–281 Rappaport, Gilbert C. (2010). ‘The grammaticalization of the category masculine personal in West Slavic’, in Bjo¨rn Hansen and Jasmina Grkovic´-Major (eds), Diachronic Slavonic Syntax: Gradual Changes in Focus (Wiener Slawistischer Almanach, Sonderband 74). Munich: Kubon & Sagner, 169–180. R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, www.R-project.org/ (accessed 18/05/21). R Core Team (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL www.R-project.org/. Reichelt, Hans (1909). Awestisches Elementarbuch. Heidelberg: Winter. Reid, Nicholas (2000). ‘Complex verb collocations in Ngan’gityemerri: A non-derivational strategy for encoding valency alternations’, in Dixon and Aikhenvald (eds), 333–359. Richhardt, Rosemarie (1957). Polnische Lehnwo¨rter im Ukrainischen. Wiesbaden: Harrasowitz. Rieken, Elisabeth (2006). ‘Zum hethitisch-luwischen Sprachkontakt in historischer Zeit’, Altorientalische Forschungen 33: 271–285. Rieken, Elisabeth (2017). ‘Das hethithische partizip: zur schittstelle von syntax un semantics’, in Claire Le Feuvre, Daniel Petit, and Georges-Jean Pinault (eds), Verbal Adjectives and Participles in Indo-European Languages. Proceedings of the Conference of the Society for Indo-European Studies (Indogermanische Gesellschaft), Paris, 24th to 26th September 2014. Bremen: Hempen, 391–403. Rijkhoff, Jan (2002). The Noun Phrase. Oxford: Oxford University Press. Ringe, Don (2006). From Proto-Indo-European to Proto-Germanic (A Linguistic History of English, 1). Oxford: Oxford University Press. Rivero, Maria Luisa (2004). ‘Spanish quirky subjects: Person restrictions and the person– case constraint’, Linguistic Inquiry 35: 494–502.

354

REFERENCES

Rizza, Alfredo (2009). ‘Ergatività in anatolico. Alcune premesse’, Atti del Sodalizio Glottologico Milanese 1–2 (n.s.): 250–256. Rizza, Alfredo (2010). ‘Contributi allo studio dell’ergatività in anatolico: basi teoreticotipologiche. Sopra alcune recenti pubblicazioni’, Atti del Sodalizio Glottologico Milanese 3 (n.s.): 144–162. Rizza, Alfredo (2014). “‘Ergativo” negli studi sulle lingue del ramo indoeuropeo anatolico’, in Vincenzo Orioles, Raffaella Bombi, and Marica Brazzo (eds), Metalinguaggio. Storia e statuto dei costrutti della linguistica. Rome: Il calamo, 271–289. Roberts, Ian (2010). ‘Grammaticalization, the clausal hierarchy, and semantic bleaching’, in Traugott and Trousdale (eds), 45–73. Roberts, John R. (2001). ‘Impersonal constructions in Amele’, in A. Aikhenvald, R. M. W. Dixon, and M. Onishi (eds), Non-Canonical Marking of Subjects and Objects. Amsterdam: Benjamins, 201–250. Ronconi, Alessandro (1968 [1948]). Il verbo latino. Principi di sintassi storica. Bologna: Zanichelli. Rosén, Haiim (1970). ‘Uterum dolet und verwandtes’, Folia Linguistica 4: 135–147. Rosén, Haiim (1992). ‘On some types of so-called “impersonality” and verbal valency in Indo-European’, in Beekes and Witerberg (eds), 383–390. Rovai, Francesco (2005). ‘L’estensione dell’accusativo in tardo latino e medievale’, Archivio Glottologico Italiano 90: 54–89. Rovai, Francesco (2007a). ‘Manifestazioni di sub-sistemi tipologici attivi in latino’, Archivio Glottologico italiano 92: 51–65. Rovai, Francesco (2007b). ‘Tratti attivi in latino. Il caso del genere’ (Ph.D. dissertation, University of Pisa). Rovai, Francesco (2012). Sistemi di codifica argomentale. Tipologia ed evoluzione. Pisa: Pacini 2012. Rovai, Francesco (2014). ‘Case marking in absolute constructions: Further evidence for a semantically based alignment in Late Latin’, Journal of Latin Linguistics 13: 115–143. Ruwet, Nicolas (1994). ‘Être ou ne pas être un verb de sentiment’, Langue française 103, 45–55. Safarewiczowa, Halina (1964). Obocznos´´c “ja imeju”i “u menja est” w języku rosyjskim dzis´ i dawniej. Wrocław: Ossolineum. Sandfeld, Kristian (1926/1930). Linguistique balkanique. Paris: Klincksieck [1st ed. 1926: Balkanfilologien. Copenhagen: Luno]. Sawicki, Lea (1988). Verb–Valency in Contemporary Polish. Tu¨bingen: Narr. Say, Sergey (2014). ‘Bivalent verb classes in the Languages of Europe: A quantitative typological study’, Language Dynamics and Change 4(1): 116–166. Say, Sergey (2017). ‘Lexicˇeskoe raspredelenie perexodnyx i neperexodnyx dvuxmestnyx glagolov v indoevropejskix jazykax: kvantitativno-tipologicˇeskoe issledovanie’, in Nikolay N. Kazanskij (ed.), Indoevropejskoe jazykoznanie i klassicˇeskaja filologija XXI. St. Petersburg: Nauka, 714–747. Say, Sergey (2018). ‘Markirovanie aktantov dvuxmestnyx predikatov: Predvaritel’nye itogi tipologicˇeskogo issledovanija’, in Sergey Say (ed.), Valentnostnye klassy dvuxmestnyx predikatov v raznostrukturnyx jazykax. St. Petersburg: ILI RAN, 557–616. Say, Sergey (ed.) (2020). BivalTyp: Typological Database of Bivalent Verbs and Their Encoding Frames. St. Petersburg: Institute for Linguistic Studies, RAS, www.bivaltyp.info (accessed 25/05/21).

REFERENCES

355

Sayeed, Ollie and Bert Vaux (2017). ‘The evolution of Armenian’, in Jared Klein, Brian Joseph, and Matthias Fritz (eds), Handbook of Comparative and Historical IndoEuropean Linguistics, vol. II. Berlin: De Gruyter, 1146–1167. Schlund, Katrin (2020). ‘Active transitive impersonals in Slavic and beyond: A parallel corpus analysis’, Russian Linguistics 44: 39–58. Schmidt, Johannes (1889). Die Pluralbildungen der indogermanischen Neutra. Weimar: Bo¨hlau. Schmidt, Karl-Horst (1962). ‘Zum umschriebenen Perfekt in indogermanischen Sprachen’, Indogermanische Forschungen 67(3): 225–236. Schmidt, Karl-Horst (1963). ‘Zum Agens beim Passiv’, Indogermanische Forschungen 68(1): 1–12. Schmidtke-Bode, Karsten and Natalia Levshina (2018). ‘Reassessing scale effects on differential case marking: Methodological, conceptual and theoretical issues in the quest for a universal’, in Ilja A. Serzˇant and Alena Witzlack-Makarevich (eds), Diachrony of Differential Argument Marking. Berlin: Language Science Press, 509–537. Schmitt, Ru¨diger (1983). ‘Iranisches Lehngut im Armenischen’, Revue des Études Arméniennes 17: 73–112. Schmitt, Ru¨diger (2007). Grammatik des Klassisch-Armenischen mit sprachvergleichenden Erla¨uterungen. Innsbruck: Innsbrucker Beitra¨ge zur Sprachwisschenschaft. Seefranz-Montag, Ariane von (1984). ‘Subjectless constructions and syntactic change’, in Jacek Fisiak (ed.), Historical Syntax. Berlin: Mouton, 521–553. Senn, Alfred (1957). Handbuch der Litauischen Sprache. Band II. Texte und Glossar. Heidelberg: Winter. Serzˇant, Ilja A. (2012). ‘The so-called possessive perfect in North Russian and the CircumBaltic area: A diachronic and areal approach’ Lingua 122: 356–385. Serzˇant, Ilja A. (2013). ‘Rise of canonical objecthood with the Lithuanian verbs of pain’, Baltic Linguistics 4: 187–211. Serzˇant, Ilja A. (2015a). ‘An approach to syntactic reconstruction’, in Carlotta Vitti (ed.), Perspectives on Historical Syntax. Amsterdam: Benjamins, 117–154. Serzˇant, Ilja A. (2015b). ‘Dative experiencer constructions as a Circum-Baltic isogloss’, in Peter Arkadiev, Axel Holvoet, and Bjo¨rn Wiemer (eds), Contemporary Approaches to Baltic Linguistics. Berlin: De Gruyter, 325–348. Serzˇant, Ilja A. (2016). ‘External possession and constructions that may have it’, Sprachtypologie und Universalienforschung STUF 69(1): 131–169. Serzˇant, Ilja A. (forthcoming (a)). ‘The Circum-Baltic area: An overview’, in Jan Fellerer and Neil Bermel (eds), Oxford Guides to the World’s Languages: The Slavonic Languages. Oxford: Oxford University Press. Serzˇant, Ilja A. (forthcoming (b)). ‘The dynamics of Slavic morphosyntax is primarily determined by the geographic location and contact configuration’, Scando-Slavica. Serzˇant, Ilja A. and Valgerður Bjarnadóttir (2014). ‘Verbalization and non-canonical case marking of some irregular verbs in *-ē- in Baltic and Russian’, in Tatjana Civjan, Marija Zavjalova, and Artūras Judzˇentis (eds), Baltai ir slavai: dvasinių kultūrų sankritos/Бaлты и cлaвянe: пepeceчeния духoвных культуp. Vilnius: Versme˙s, 218– 242. Serzˇant, Ilja A. and Leonid Kulikov (eds) (2013). The Diachronic Typology of NonCanonical Subjects. Amsterdam: Benjamins. Serzˇant Ilja A., Bjo¨rn Wiemer, Eleni Buzˇarovska, Martina Ivanová, Maxim Makartsev, Stefan Savic´, Dmitri Sitchinava, Karolı´na Skwarska, and Mladen Uhlik (2021). Dataset for

356

REFERENCES

the paper ‘Areal and Diachronic Trends in Argument Flagging across Slavic’. Zenodo, http://doi.org/10.5281/zenodo.4749564 (accessed 18/05/21). Shatskov, Andrej (2011). ‘ЭРΓAТИВHЫЙ ⊓AДЕЖ В XЕТТCКOM ЯЗЫКЕ/The ergative case in Hittite (in Russian with an English abstract)’, Acta Linguistica Petropolitana 7(1): 143–152. Shevelov, Yury (1952). ‘The problem of Ukrainian–Polish linguistic relations from the tenth to the fourteenth century’, WORD 8(4): 329–349. Shevelov, Yury (1953). ‘Problems in the Formation of Belorussian’, supplement to WORD 9, suppl. 1, 1–100. Shibatani, Masayoshi (1985). ‘Passives and related constructions: A prototype analysis’, Language 61: 821–848. Shibatani, Masayoshi (1994). ‘Voice’, in R. E. Asher (ed.), The Encyclopedia of Language and Linguistics. Oxford: Pergamon Press, 4398–4942. Siewierska, Anna (1984). The Passive: A Comparative Linguistic Analysis. London: Croom Helm. Siewierska, Anna (2008). ‘Introduction: Impersonalization from a subject-centred vs. agent centred perspective’, Transactions of the Philological Society 106(2): 115–137 [special issue ‘Impersonal Constructions in Grammatical Theory’, ed. A. Siewiweska]. Sigurðsson, Halldór Ármann (2002). ‘To be an oblique subject: Russian vs. Icelandic’, Natural Language and Linguistic Theory 20: 691–724. Sigurðsson, Halldór Ármann (2004). ‘Icelandic non-nominative subjects’, in Bhaskararao and Subbarao (eds), vol. II, 137–159. Sihler, Andrew L. (1995). New Comparative Grammar of Greek and Latin. New York: Oxford University Press. Silverstein, Michael (1976). ‘Hierarchy of features and ergativity’, in Robert M. W. Dixon (ed.), Grammatical Categories in Australian Languages. Atlantic Highlands, NJ: Humanities Press, 112–171. Simone, Raffaele (2003). ‘Maṣdar, ‘ismu al-marrati et la frontière verbe/nom’, in Cécile Brion and Eric Castagne (eds), Nom et verbe: catégorisation et référence. Acts du Colloque International de Reims 2001. Reims: Presse Universitaires de Reims, 227–249. Sinnema¨ki, Kaius (2014). ‘A typological perspective on differential object marking’, Linguistics 52(2): 281–313. Skjærvø, Prods O. (1983). ‘Case in inscriptional Middle Persian, inscriptional Parthian and the Pahlavi Psalter’, Studia Iranica 12(1): 69–94, 151–181. Skjærvø, Prods O. (2009). ‘Old Iranian’, in Gernot Windfuhr (ed.), The Iranian Languages. London: Routledge, 43–195. Skwarska, Karolina (2004). ‘Konkurence genitivu a akuzativu s tranzitivnı´mi slovesy v cˇeštině, ruštině, polštině a slovinštině’ (Ph.D. thesis, Charles University, Prague). Skytte, Gunver and Giampaolo Salvi (1991). ‘L’infinito come testa del sintagma nominale (infinito con l’articolo o altro determinante)’, in Lorenzo Renzi, Anna Cardinaletti, and Giampaolo Salvi (eds), Grande grammatica italiana di consultazione, II. Bologna: Il Mulino, 559–569. Smith, Herbert Weir (1956). Greek Grammar (rev. Gordon M. Messing). Cambridge, MA: Harvard University Press. Sobolev, Andrej N. (2009). ‘From synthetic to analytic case: Variation in South-Slavic dialects’, in Andrej Malchukov and Andrew Spencer (eds), Handbook of Case. Oxford: Oxford University Press, 716–729. Speijer, Jacob S. (1886). Sanskrit syntax. Leiden: Brill [repr. 1998, Delhi: Motilal Banarsidass].

REFERENCES

357

Speijer, Jacob S. (1896). Vedische und Sanskrit-Syntax. Strasbourg: Tru¨bner. Stanislav, Ján (1973). Dejiny slovenského jazyka IV. Syntax 1. Bratislava: Vydavatel’stvo Slovenskej akadémie vied. Starke, Frank (1977). Die Funktionen der dimensionalen Kasus und Ortsadverbien im Hethitischen (Studien zu den Bogazko¨y-Texten 22). Wiesbaden: Harrassowitz. Stassen, Leon (2013). ‘Predicative possession’, in Matthew S. Dryer and Martin Haspelmath (eds), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology, http://wals.info/chapter/117 (accessed 13/05/20). Stempel, Reinhard (1983). Die infiniten Verbalformen des Armenischen. Frankfurt a.M.: Peter Lang. Sundermann, W. (1981). Mitteliranische manicha¨ische Texte kirchengeschichtlichen Inhalts, Berlin: Akademie-Verlag. Suñer, Margarita (1988). ‘The role of agreement in clitic doubled constructions’, Natural Language and Linguistic Theory 6: 391–434. Svennung, Josef (1935). Untersuchungen zur Palladius und zur lateinischen Fach und Folkssprache. Uppsala: Almkvist and Wiksell. Sweetser, Eve S. (1990). From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: Cambridge University Press. Talmy, Leonard (1985). ‘Lexicalization patterns: Semantic structure in lexical forms’, in Timothy Shopen (ed.), Language Typology and Syntactic Description, Vol. III: Grammatical Categories and the Lexicon. Cambridge: Cambridge University Press, 57–149. Taracha, Piotr (1995). ‘Zum Stand der hattischen Studien’, in Onofrio Carruba, Mauro Giorgieri, and Clelia Mora (eds), Atti del II Congresso Internazionale di Hittitologia (Studia Mediterranea 9). Pavia: Iuculano, 351–358. Taracha, Piotr (1998). ‘Neues zur Sprache und Kultur der Harder’, Orientalistische Literaturzeitung 93: 9–18. Taylor, John R. (1989). ‘Possessive genitives in English’, Linguistics 27: 663–686. Tchekhoff, Claude (1978). “Le double cas-sujet des inanimées: un archaïsme de la syntaxe hittite?”, in: BSLP73 (1978) 225–241. Teffeteller, Annette (2015). ‘Anatolian morphosyntax. Inheritance and innovation’, in Carlotta Viti (ed.), Perspectives on Historical Syntax. Philadelphia: Benjamins, 155–184. Ter-Mkrtcˇ’ean, G. and S. Kanayeanc (1909). Agat‘angełay Patmut‘iwn Hayoc‘,Tiflis. Thomas, Werner (1952). Die tocharischen Verbaladjektive auf -l: eine syntaktische Untersuchung, Berlin: Akademie-Verlag. Thomson, Robert W. (1993). History of Vardan and the Armenian War. Delmar, NY: Caravan Books. Tikkanen, Karin (2011). A Sabellian Case Grammar. Heidelberg: Winter. Timberlake, Alan (1977). ‘Reanalysis and actualization in syntactic change’, in Charles N. Li (ed.), Mechamisms of Syntactic Change. Austin: University of Texas Press, 140–177. Traugott, Elizabeth Closs (2008). ‘The grammaticalization of NP of NP constructions’, in Alexander Bergs and Gabriele Diewald (eds), Constructions and Language Change. Berlin: Mouton de Gruyter, 21–43. Traugott, Elizabeth Closs and Graeme Trousdale (eds) (2010). Gradience, Gradualness and Grammaticalization. Amsterdam: Benjamins. Traugott, Elizabeth Closs and Graeme Trousdale (2010). ‘Gradience, gradualness and grammaticalization: How do they intersect?’, in Traugott and Trousdale (eds), 19–44. Trávnı´cˇek, František (1938). ‘Záporový genitiv v cˇeštině’, Slovo a slovesnost 4: 129–138. Tsunoda, Tasaku (1985). ‘Remarks on transitivity’, Journal of Linguistics 21(2): 385–396.

358

REFERENCES

Tsunoda, Tasaku (2004). ‘Issues in case-marking’, in Peri Bhaskararao and Karumuri V. Subbarao (eds), Non-Nominative Subjects. Amsterdam: Benjamins, 197–208. Tumanyan, E. G. (1974). ‘Sušcˇestvuet li èrgativ v indoevropejskix jazykax?’, in Luigi Heilmann (ed.), Proceedings of the Eleventh International Congress of Linguists, BolognaFlorence, Aug. 28–Sept. 2, 1972. Bologna: Il Mulino, 957–961. Uhlenbeck, Christianus Cornelius, (1901). ‘Agens und Patiens im Kasussystem der indogermanischen Sprachen’, Indogermanische Forschungen 12: 170–171. Uhlenbeck, Christianus Cornelius (1907). ‘Zu Seite 209ff. dieses Bandes’, Zeitschrift fu¨r Vergleichende Sprachforschung 41: 400. Va¨a¨na¨nen, Veikko (1966). Le latin vulgaire des inscriptions pompéiennes, 3rd ed. Berlin: Akademie Verlag. Valério, Miguel. 2009. Palaic fulāsinanza: One Anatolian Suffix, Two Possible Explanations. Journal of Indo-European Studies, 37(2), pp. 421–429. van Gelderen, Elly, Michela Cennamo, and Jóhanna Barðdal (eds) (2013). Argument Structure in Flux. The Naples–Capri Papers. Amsterdam: Benjamins. Van Valin, Robert D., Jr. (1990). ‘Semantic parameters of split intransitivity’, Language 66 (2): 221–260. Van Valin, Robert D., Jr. (1993). ‘A synopsis of role and reference grammar’, in Robert D. Van Valin, Jr. (ed.), Advances in Role and Reference Grammar. Amsterdam: Benjamins, 1–166. Van Valin, Robert D., Jr. (2005). Exploring the Syntax–Semantics Interface. Cambridge: Cambridge University Press. Van Valin, Robert D., Jr. and Randy J. Lapolla. 1997. Syntax. Structure, Meaning and Function. Cambridge: Cambridge University Press. Van Valin, Robert D., Jr. and David Wilkins (1996). ‘The case for “effector”: Case roles, agents, and agency revisited’, in Masayoshi Shibatani and Sandra Thompson (eds), Grammatical Constructions: Their Form and Meaning. Oxford: Clarendon, 289–322. Vanséveren, Sylvie (1997). ‘Questions sur l’infinitif arménien’, Annual of Armenian Linguistics 18: 23–37. Vanvolsem, Serge (1981). L’infinito sostantivato in italiano. Florence: Accademia della Crusca. Vázquez Rozas, Victoria (2006). ‘Gustar-type verbs’, in J. Clancy Clements and Jiyoung Yoon (eds), Functional Approaches to Spanish Syntax: Lexical Semantics, Discourse and Transitivity. New York: Palgrave Macmillan, 80–114. Vázquez Rozas, Victoria and Elena Rivas (2007). ‘Un análisis construccionista de la diacronı´a de gustar’, in Iraide Ibarrexte-Antuñano, Carlos Inchaurralde, and Jesús M. Sánchez-Garcı´a (eds), Language, Mind, and the Lexicon. Frankfurt: Peter Lang, 143–164. Vendler, Zeno (1967). Linguistics in Philosophy. Ithaca, NY: Cornell University Press. Verma, Manindra and Karuvannur P. Mohanan (eds) (1990). Experiencer Subjects in South Asian Languages. Stanford, CA: CSLI Publications. Verstraete, Jean-Christophe (2011). ‘Impersonal constructions in Umpithamu and the Lamalamic languages’, in Andrej Malchukov and Anna Siewierska (eds), Impersonal Constructions: A Cross-Linguistic Perspective. Amsterdam: Benjamins, 607–626. Villari, Cristina and Leonel Menacho López (2015). ‘La voz media en el quechua de Huaylas: las funciones semánticas del sufijo -ku’, Aporte Santiaguino 8(2): 300–318. Vincent, Nigel (2016). ‘Continuity and change from Latin to Romance’, in James N. Adams and Nigel Vincent (eds), Early and Late Latin. Continuity or Change? Cambridge: Cambridge University Press, 1–13.

REFERENCES

359

Viti, Carlotta (2014). ‘Reconstructing syntactic variation in Proto-Indo-European’, IndoEuropean Linguistics 2: 73–111. Viti, Carlotta (2016). ‘The areal distribution of argument marking of Indo-European experience predicates’, Journal of Indo-European Studies 44(1–2): 1–84. Vogt, Hans (1930). ‘Les deux thèmes verbaux de l’arménien classique. Étude sur la traduction des Évangiles’, Norsk Tidsskrift for Sprogvidenskap 4: 129–145. Vogt, Hans (1935). ‘Les formes nominales du verbe arménien’, Norsk Tidsskrift for Språkvidenskap 8: 5–70. Vogt, Hans (1938). ‘Arménien et Caucasique du Sud’, Norsk Tidsskrift for Sprogvidenskap 9: 321–338. Wackernagel, Jacob (1928). Vorlesungen u¨ber Syntax. Basel: Birkhau¨ser e Ciel. Walsh, Michael 1987. ‘The impersonal verb construction in Australian languages’, in Ross Steele and Terry Threadgold (eds), Language Topics: Essays in Honour of Michael Halliday. Amsterdam: Benjamins: 425–438. Watkins, Calvert (2001). ‘An Indo-European linguistic area and its characteristics: Ancient Anatolia. Areal diffusion as a challenge to the comparative method?’, in Alexandra Y. Aikhenvald and Robert M. W. Dixon (eds), Areal Diffusion and Genetic Inheritance. Oxford: Oxford University Press, 44–63. Weiss, Michael (2009). Outline of the Historical and Comparative Grammar of Latin. Ann Arbor, MI: Beech Stave Press. Weitenberg, Jos J. S. (1987). ‘Proto-Indo-European nominal classification and Old Hittite’, Mu¨nchener Studien zur Sprachwissenschaft 48: 213–230. Weitenberg, Jos J. S. (1995). ‘Sigmatization and thematization in Hittite’, in Theo van den Hout and Johan de Roos (eds), Studio historiae ardens. Ancient Near Eastern studies presented to Philo H.J Houwink ten Cate on the occasion of his 65th birthday. Publication de l’Institut Historique et Archéologique Néerlandais de Stamboul 74. Leiden, 333–344. Weitenberg, Jos J. S. (2017). ‘The dialectology of Armenian’, in Jared Klein, Brian Joseph, and Matthias Fritz (eds), Handbook of Comparative and Historical Indo-European Linguistics, vol. II. Berlin: De Gruyter, 1132–1146. West, Martin L. (2011a). ‘Towards a chronology of early Greek epic’, in Øyvind Andersen and Dag T. T. Haug (eds), Relative Chronology in Early Greek Poetry. Cambridge: Cambridge University Press, 224–241. West, Martin L. (2011b). Old Avestan Syntax and Stylistics. Berlin: De Gruyter. Whitman, John and Yuko Yanagida (2012). ‘The formal syntax of alignment change’, in Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds), Parameter Theory and Linguistic Change. Oxford: Oxford University Press, 177–195. Whitney, William Dwight (1879). Sanskrit grammar including both the classical language, and the older dialects, of Veda and Brahmana. Leipzig: Breitkopf & Ha¨rtel [2nd ed. 1889, repr. 1955, Cambridge, MA: Harvard University Press]. Wiemer, Bjo¨rn (2003). ‘Dialect and language contacts on the territory of the Grand Duchy of Lithuania from the 15th century until 1939’, in Kurt Braunmu¨ller and Gisella Ferraresi (eds), Aspects of Multilingualism in European Language History (Hamburg Studies on Multilingualism 2). Amsterdam: Benjamins, 105–143. Wiemer, Bjo¨rn and Valgerður Bjarnadóttir (2014). ‘On the non-canonical marking of the highest-ranking and argument in Lithuanian and Icelandic’, in Axel Holvoet and Nicole Nau (eds), Grammatical Relations and Their Non-Canonical Encoding in Baltic. Amsterdam: Benjamins, 301–361 Wilhelm, Gernot (1994). Medizinische Omina aus Ḫattuša in akkadischer Sprache. Wiesbaden: Harrassowitz.

360

REFERENCES

Willi, Andreas (2018). Origins of the Greek Verb. Cambridge: Cambridge University Press. Witzel, Michael (1989). ‘Tracing the Vedic dialects’, in Colette Caillat (ed.), Dialectes dans les langues indo-aryennes. Paris: College de France, Institut de Civilisation Indienne, 97–265. Witzel, Michael (1995). ‘Early Indian history: Linguistic and textual parameters’, in George Erdosy (ed.), Language, Material Culture and Ethnicity: The Indo-Aryans of Ancient South Asia. Berlin: de Gruyter, 85–125. Witzlack-Makarevich, Alena and Ilja A. Serzˇant (2018). ‘Differential argument marking: Patterns of variation’, in Ilja A. Serzˇant and Alena Witzlack-Makarevich (eds), Diachrony of Differential Argument Marking (Studies in Diversity Linguistics). Berlin: Language Science Press, 1–40, http://langsci-press.org/ (accessed 18/05/21). Woodcock, Eric Charles (1959). A New Latin Syntax. London: Methuen. Xodova, Kapitolina I. (1963). Sistema padezˇej staroslavjanskogo jazyka. Moscow: Nauka. Yakubovich, Ilya (2008). ‘The origin of the Luvian possessive adjectives’, in Karlene JonesBley, Martin E. Huld, Angela Della Volpe, and Miriam Robbins Dexter (eds), Proceedings of the Nineteenth Annual UCLA Indo-European Conference. Washington, DC: Institute for the Study of Man, 193–217. Yakubovich, Ilya (2010). Sociolinguistics of the Luvian Language (Brill’s Studies in IndoEuropean Language & Linguistics 2). Leiden: Brill. Yanagida, Yuko and John Whitman (2009). ‘A lignment and word order in Old Japanese’, Journal of East Asian Linguistics 18: 101–144. Zamboni, Alberto (1997). ‘Temi e problemi della transizione’, Rivista italiana di dialettologia 21: 9–71. Zamboni, Alberto (2002). ‘Sistemi casuali e orientamenti tipologici: la dimensione romanza’, Quaderni dell’Istituto di Fonetica e Dialettologia 4: 190–237. Zeilfelder, Suzanne (2001). Archaismus und Ausgliederung. Studien zur sprachlichen Stellung des Hethitischen. Heidelberg: Winter. Z ˙ elazko, Kazimierz (1975). Czasowniki przechodnie o składni wielorakiej w języku polskim. Wrocław: Ossolineum. Žele, Andreja (2006). ‘Valency in standard Slovenian with special reference to the verb’, Slavisticˇna revija 54(1): 401–414. Zieniukowa, Jadwiga (1981). Rodzaj męski osobowy we współczesnych językach zachodniosłowian´skich. Wrocław: Ossolineum. Žigo, Pavol (2012). Analógie a anomálie v substantı´vnej deklinácii slovanských jazykov. Bratislava: Veda. Zimmer, Stefan (1985). ‘Die Gebrauchsweisen des adnominalen Genetivs im Tocharischen’, in B. Schlerath (ed), Grammatische Kategorien, Funktion und Geschichte, Wiesbaden: Dr. Ludwig Reichert, 560–574. Zuñiga, Fernando (2018). ‘The diachrony of morphosyntactic alignment’, Language and Linguistics Compass 12: 1–21. Zúñiga, Fernando and Seppo Kittila¨ (2019). Grammatical Voice. Cambridge: Cambridge University Press.

Index Tables and figures are indicated by an italic t and f following the page number. ablative case 58, 74t, 89t, 95t, 104t, 123, 131–2, 141, 145 n.8, 147, 149–50, 190, 193, 195, 196, 197, 208, 211 absolutive case 8, 53, 66 accusative alignment see nominative-accusative alignment accusative case 5, 7, 8, 9, 10 n.7, 11, 12, 16, 17, 18, 22, 27, 28, 29, 30, 43, 45, 46 n.30, 53, 55, 57, 58, 59, 61, 62, 66, 67, 68, 69, 71, 74, 76, 77, 78, 79, 80, 81t, 82, 85, 87, 88, 89t, 95, 96, 100, 101, 102, 104, 106, 109, 111, 112, 113t, 115, 123, 127, 130, 143, 188, 190, 193, 195, 196, 197, 199, 201, 203, 204, 207, 208, 209, 210, 213, 214, 221, 228, 234 n.12, 235, 237, 248, 250, 251, 252, 253, 256, 258, 259, 261, 262, 264, 298, 301, 302, 306, 307, 308, 309, 310, 313, 315, 316, 318, 319, 320, 325, 326 active/active-inactive alignment see semantic alignment Acehnese 251 active voice 19, 56, 57, 58, 78 n.16, 168, 184, 188–217, 250, 253, 255, 257, 258, 280, 281, 288, 289, 290, 309, 320 affectedness 20, 189, 192t, 207, 210, 214, 216, 226, 246, 247, 249, 250, 251, 252, 253, 254, 257 n.8, 269, 270, 270, 274, 275, 308 n.12, 326 agent defocusing 20, 192, 195, 202, 210, 213, 214 agent/agentivity19, 20, 36, 58 n.40, 70, 76 n.12, 94, 98, 107, 128, 130, 132, 138, 142, 149, 150, 166, 167, 171, 172, 173, 175, 177–9, 180, 186, 190 n.4, 192, 193, 195, 201–2, 211, 212, 213, 214, 217, 247, 249, 251–3, 260, 269, 274, 279, 280, 281, 282, 285, 286, 287, 288, 289, 290, 295, 296, 298 agentive 7 n.5, 9, 21, 53, 57, 58, 59, 65, 94, 98, 132, 134, 138, 144, 145, 179, 188, 222 n.2, 229, 246, 250, 253, 258, 279, 287, 289, 296 agreement 4, 6–7, 10 n.9, 13, 15, 17, 18, 19, 25, 26, 27, 30–5, 37, 38, 37–44, 46, 50, 56, 59–62, 66, 67, 68, 71, 72, 74–5, 82, 83, 85–6, 87, 89–90, 91, 93, 94, 95–6, 97, 101–2, 103, 104, 105, 110, 114, 115, 120, 128, 130 n.7, 131, 137, 139, 141, 146, 150, 151, 165t, 200, 204, 213, 214, 239 n.15, 250, 251, 254, 255, 256 n.4, 258,

259, 260, 261, 265, 267, 271, 272, 273, 274, 275, 280, 281, 284, 286 n.14, 287 n.22, 289, 291, 293, 296, 298, 301, 302 n.5, 309, 319 Albanian 103 n.35 allative case 57, 58, 95t, 109t Amele 215, 217, 259 analogical extension 2, 11, 13, 15, 16, 72, 73, 107, 115, 116, 145, 148, 256 n.4, 258, 266, 268, 269, 270, 285, 288–90 Anatolian 17, 18, 25, 28, 42, 43, 50, 59, 95, 124, 125, 126, 127, 137, 139, 139, 145–7, 149, 151 Ancient Greek 1, 15, 17, 18, 24, 25, 26, 27, 28, 30, 31, 33, 34, 35, 38, 39, 40, 41, 43, 44, 46, 47, 48ft, 49, 50, 59, 60, 62, 64, 73, 80–7, 88, 91, 92, 93, 97, 103, 110, 112, 114, 115, 117, 119, 120, 137, 143, 199, 235 n.13, 253, 254, 255, 277, 287, 288, 289, 301 n.1, 317, 322 Ancient Macedonian 103 n.35 animacy 5 n.3, 6 n.4, 8, 22, 35, 53, 54, 55, 60, 61, 62, 67, 95t, 98, 99, 125, 126, 128, 128, 130, 138, 142, 143, 144, 145, 151, 188 n.1, 189, 193, 197, 209, 216, 228, 229, 235, 237, 238, 247, 249, 260, 262, 263, 266, 267, 271, 273, 285, 296, 298, 301, 302, 305, 306 n.11, 319 anticausative 1, 19, 94, 166–187, 189 n.3 antipassive 1, 13, 57, 58 Aramaic 2, 40 areal 22, 23, 103 n.37, 124, 147, 149, 151, 216, 300, 302, 303, 312, 313, 314–15, 317, 319, 322, 324, 326, 327 Armenian 1, 10, 11, 15, 18, 21, 22, 64, 73, 80 n.18, 103–8, 114, 115, 116, 117, 119, 120, 181, 187, 254, 277–99 Austronesian 1, 14, 215, 217 Avestan 41–42, 60, 92, 94, 287 n.22, 301 Baltic 17, 108, 111, 255, 314, 326 Basque 8, 317 Batsbi 66 Belarusian 22, 23, 300, 303, 304, 307f, 310f, 317, 322, 323, 324, 325, 326, 327 biargumental see bivalent Biblical Hebrew 40, 197

362

INDE X

bivalent 1, 3, 7 n.5, 9, 10, 13, 16, 17, 20, 23, 26,27, 30 n.8, 33, 35, 36, 37 n.16, 44, 45, 46, 47, 48ft, 49, 50, 51, 55, 56, 57, 61, 62, 68, 70, 71, 72, 75, 76, 79, 80, 82, 84, 85, 87, 88, 89, 90, 91, 94, 96, 97, 98, 99, 100, 102, 103, 104, 105, 106, 107, 108 n.43, 109, 112, 113, 114, 115, 117, 123, 126, 127, 128, 130, 131, 132, 133, 137, 138, 140, 141 n.22, 144, 150, 152t, 154t, 156t, 158t, 160t, 162t, 164t, 165t, 167, 188, 189, 190, 199, 201, 203, 205, 207, 210, 212, 215, 216, 221, 222, 223, 225, 226, 228, 229, 231, 233, 237, 238, 240, 246, 247, 248, 250, 255, 256, 258, 259, 260, 261, 262, 263, 264, 265, 266, 268 n.17, 271, 273, 279, 280, 281, 286, 287 n.19, 288, 293, 295, 296, 301, 303, 306, 308, 309, 312, 313, 314, 316, 317, 318, 319, 320, 321, 326 borrowing 13, 22, 72, 271, 277, 288, 292, 298, 324 Bulgarian 22, 23, 300, 301, 303, 305, 306, 307f, 308, 310, 311, 312, 313, 317, 318, 320, 322 n.22, 323ft, 324, 326 case marking 4, 5, 6 n.4, 7, 8, 9, 10 n.9, 11, 15, 16, 17, 18, 19, 22, 23, 25, 27–8, 30, 34, 44, 45, 50, 53, 54, 55, 56, 59, 62, 66, 67, 68, 72, 75, 79, 80, 81, 87, 90, 94, 104, 107, 108, 112, 113t, 114, 115, 150, 197, 234 n.12, 239 n.15, 250, 252, 264, 287, 298, 300–27 Caucasian 66, 148, 285, 288 causative 9, 21, 68, 127, 141 n.22, 166, 180, 197, 233, 235 n.13, 247, 248, 261–5, 266, 267, 268, 270, 271, 272, 274 Central Pomo 7, 8, 251 Classical Greek see Ancient Greek control (morphosyntactic) 19, 75, 77 n.14, 84, 87, 90 n.25, 91, 94, 100, 103, 105, 106, 112, 117, 210, 167, 175, 176, 177, 179, 180, 186 control (semantic) 19, 188, 189–91, 192, 201, 205, 207, 214, 216, 217, 226, 227, 249, 250, 251, 252, 253, 265, 267, 274 converb 19, 92, 112, 167, 175–6, 177, 179, 180, 186, 282, 287 n.19 Coosan 148 cross-reference see agreement Czech 22, 300, 303, 304 n.7, 305t, 307f, 309 n.14, 310f, 312, 315, 316, 317, 322, 323ft, 324 Dānesfāni 293 n.34 dative case 9, 21, 45, 57, 74t, 76–7, 81t, 85–6, 87, 105t, 95t, 100, 101, 104t, 109t, 111, 112, 113t, 188, 189, 190, 195, 203, 207–10, 214, 217, 235–7, 247, 248, 250, 251–3, 254, 255, 256–7, 258, 261, 263 n.12, 264–5, 266–74, 275, 301,

302, 305, 308, 310, 313, 314, 316–17, 320, 322 n.22, 324, 325 dative-locative see dative dependent-marking 6, 18, 255, 300 detransitivizing 2, 56, 92 differential marking 5 n.3, 6 n.4, 22, 59, 104, 222, 239–42, 263, 264, 280, 301, 302, 306 n.11 direct coding 9, 230 n.9 ditransitive see trivalent double-oblique alignment 3, 10, 11, 65 Dravidian 2, 257 n.7 Dutch 258 Dyirbal 54, 57 English 144, 247, 256 Epic Sanskrit 14, 15 n.15 ergative 2, 3–7, 8, 11, 13, 14, 15, 17, 18, 19, 21, 22, 24, 25, 50, 52–6, 58, 59, 65, 66, 67, 69, 79, 80, 93, 94, 98, 102, 103, 107, 115, 116, 123, 124, 129, 130–3, 135, 137–43, 144, 145, 146, 147, 148, 149, 150, 151, 165t, 258, 277, 281, 288, 317 n.1 experiencer predicates 8, 9, 10 n.8, 20, 21, 76–7, 86, 91, 92, 93, 100–1, 112, 189, 190, 194, 195, 203, 204, 205, 209, 210, 212, 214, 214–15, 216, 221, 234–9, 242, 246–75, 305, 319, 320, 321–2, 325, 326 extended accusative 7, 16, 73 Faliscan 73 Finnic 314 flagging see case marking fluid-S alignment see semantic alignment Folopa 65 n.1 Gatha Avestan see Avestan gender motion 126, 129, 133, 141, 143, 146, 150 genitive case 9, 10 n.7, 11, 58, 74t, 76, 81t, 85, 86, 87, 89t, 91, 92, 93, 94, 95t, 104t, 106, 107, 108, 109t, 111, 113t, 128, 129, 142, 143, 144, 149, 204, 251, 256, 267, 285, 287, 290, 302, 306 n.11, 307–8, 309, 313, 315, 316, 317, 318, 319, 326 Germanic 12, 37, 43, 255, 256, 258 Gothic 37–8, 40, 258 grammaticalization 9 n.6, 14, 43, 86, 98, 115, 116, 117 n.51, 124, 143, 145, 147, 148, 150, 294, 298 Hattian 147 head-marking 6, 18, 254 Hellenistic Greek see Ancient Greek High German 258 Hindi/Urdu 5–7, 11 n.12, 67, 94, 251, 293 n.34

INDE X Hittite 1, 17, 18, 19, 24, 25, 26, 28, 30, 31, 33, 35, 36, 37, 38, 40, 41, 42, 43, 44–6, 49, 50–1, 56, 57, 60, 61, 62, 64, 73, 92, 95–103, 105 n.40, 110, 114, 115, 116, 117, 118, 120, 123–45, 146, 147, 148, 149–51, 216, 254, 255, 257 n.22 Homeric Greek see Ancient Greek Hurrian 103 n.37, 133, 147, 147, 277 n.1 Icelandic 215, 216, 251, 252 Ilgar 259, 260 n.11 impersonal 1, 9, 16, 19, 20, 34, 113, 171, 188, 189–95, 198, 199–205, 206–10, 212, 213, 214, 215, 216, 217, 251, 256, 267, 272, 273, 274, 275, 304, 320, 321 inanimate see animacy indexation see agreement individuation 19, 61, 125, 126, 136, 138, 139, 141, 144, 145, 151, 189 Indo-Aryan, Indo-Aryan languages 2, 4, 6, 14–5, 87, 94, 116, 147, 150, 183f, 215, 217, 257, 285 Indo-Iranian 17, 19, 25, 28, 37, 41, 42, 43, 44, 59, 87, 94, 181 infinitive 20, 75 n.11, 84, 85, 91, 94, 100, 103, 105 n.40, 112, 117, 119, 120, 196, 199, 202, 203, 221–42, 265, 272, 285 n.13 instrumental case 58 n.40, 89t, 94, 95t, 104t, 109t, 112 n.47, 113t, 132, 147–8, 295, 296, 309, 320–1, 324 intransitive see monovalent Inuktitut 14, 58 inverse coding 9, 21, 94, 148, 262 Italian 12, 20, 221, 222, 231–42, 256, 258, 259, 262 Italic 17, 25, 28, 37, 73, 79 Iwaidjan 214, 217 Kartvelian 103 n.37, 277 n.1, 288 Lamalamic languages 214 Latin 1, 9–10, 12, 15, 16, 17, 18, 19, 20, 21, 23, 24, 25, 26, 28, 29, 30t, 31, 33, 37, 38, 40, 44, 46, 47, 48ft, 49, 50, 60, 62, 64, 64, 69, 71, 73–80, 81, 82, 83, 85, 87, 88, 91, 92, 112, 114, 115, 116, 117, 118, 120, 137, 188–217, 218, 221, 222–31, 233, 234, 235, 237, 239t, 242, 254, 256, 257, 258, 265, 267, 270, 287 n.22 Latvian 287 Lithuanian 1, 8, 17, 18, 64, 66, 73, 108–14, 115, 116, 117, 118, 120, 215, 216, 287, 289 locative 89t, 104t, 109t, 112 n.47, 113t, 252 n.3, 322, 326 Luwian 18, 43, 124, 127, 146, 147, 149, 151

363

Lycian 18, 124, 127, 146, 149, 151 Lydian 149 Macedonian 22, 300, 301, 302, 303, 305, 306, 307, 308, 310f, 312, 313, 317, 322 n.22, 323ft, 324, 326 Manipuri 65 n.1 marked nominative 107, 108, 301 mediopassive see middle/mediopassive middle/mediopassive 1, 19, 21, 56, 57, 58, 71, 77, 78, 80, 116, 166, 168 n.2, 169, 171, 181, 182, 183f, 189, 190, 207, 247, 248, 250, 253–4, 255, 257, 258, 261, 263 n.12, 265, 266, 268, 269, 270, 273, 274, 275 mixed alignment 4, 6, 15, 17, 18, 34, 64, 66, 67, 68, 72, 75, 83, 84, 90, 94, 114, 115, 120 monotransitive see monovalent monovalent 3, 9, 10, 13, 16, 17, 20, 30 n.8, 36, 37 n.16, 45, 47, 48ft, 49, 51, 55, 57, 62, 69, 70, 71, 72, 75, 76 n.12, 78, 79, 80, 82, 84, 86, 87, 88, 89, 90, 91, 92, 94, 96, 97, 98, 99, 100, 101, 102, 103, 105, 106, 107, 108, 109, 111, 113, 114, 115, 117, 123, 127, 129, 141, 152t, 154t, 156t, 158t, 160t, 162t, 164t, 170 n.5, 171, 175 n.11, 177, 181, 182, 189, 190, 195, 196, 197, 199, 201, 202, 203, 205, 206t, 208, 210, 211, 214, 215, 216, 222, 226, 227t, 230, 240, 246, 250, 258, 259, 265, 270, 271, 274, 275, 279, 280, 281, 286, 288, 289, 293, 295, 298, 304 Muna, Malayo-Polynesian Murrinh-Patha 214–15, 217 Mycenaean see Ancient Greek Nepali 6–7, 15 neutral alignment 3, 5 n.3, 10, 12, 15, 16, 17, 18, 20, 28, 30, 34, 44, 52f, 53, 55, 56, 54, 65, 66, 68, 74, 75, 80, 81, 83, 84, 87, 90, 94, 97, 102, 103, 114, 115, 234, 242, 280, 296, 298 n.44, 301 Ngan’gityemerri 26 n.11 nominative alignment see nominative-accusative alignment nominative case 8, 9, 11, 12, 17, 18, 22, 27–32, 43, 46 n.30, 53, 54 n.38, 55, 57, 59, 62, 66, 67, 68, 69, 71, 74, 75, 76, 77, 78, 79, 81t, 82, 85, 86, 87, 88, 89, 92, 94, 95, 96, 101, 102, 104, 107, 109, 111 n.46, 112, 115, 123, 126–7, 128, 129, 130, 135, 137, 141, 143, 145, 146, 148, 150, 152t, 154t, 156t, 158t, 160t, 162t, 164t, 188 n.1, 190, 229, 230, 237, 250, 251, 253, 254, 255, 256, 258, 259, 267, 269, 271, 272, 274, 298, 299, 301, 302, 306, 309, 318, 319, 320, 321, 322, 326, 327

364

INDE X

nominative-accusative alignment 2–5, 6, 7, 8, 9, 11, 13, 14, 15, 16, 17, 18, 20, 21, 22, 24, 25, 27, 28, 30, 32, 33, 34, 42, 43, 44, 50–2, 53, 54, 55, 56, 62, 64, 65–6, 67, 68, 71, 74–5, 80, 81, 82, 83, 84, 87, 88, 90, 93, 94, 95, 97, 103, 104, 106, 107, 108, 109, 110, 114, 115, 116, 117, 118, 119, 124, 130, 145, 217, 223, 228, 229, 230, 247, 250, 258, 259, 260, 274, 275, 277, 278, 298, 300, 301, 303, 319 non-canonical argument 8–9, 17, 18, 19, 57, 58 n.40, 64, 76, 77, 80, 84, 86, 87, 91, 93, 94, 100–1, 102, 106–7, 108, 111, 112, 113t, 114, 117, 118, 119, 120, 128, 150, 188–217, 226, 224, 235, 248, 251, 252, 254, 255, 256, 257, 258, 261, 265, 267, 269, 292 n.32, 309, 319, 320, 321, 322 non-nominative see non-canonical North Halmaheran 259, 260 North-West Caucasian 148 Norwegian 12 oblique argument see non-canonical argument Old Church Slavic (OCS) 22, 300, 301 n.1, 303, 304, 305, 306, 307, 308, 311f, 312, 313, 316, 317, 318, 319, 320, 321, 322, 323f, 324, 325, 326 one-place see monovalent Oscan 73 Pahlavi see West Middle Iranian Palaic 43, 146 Pāli 14, 15 n.15, 94 Pama–Nyungan languages 148, 209, 217 Pāṇini 167, 170 n.5, 173, 175, 177, 178, 179, 180, 186 Parthian see West Middle Iranian passive 1, 14, 19, 20, 80, 92, 93, 94, 106, 107, 113, 114, 116, 147, 150, 166–87, 188, 192, 193, 195, 207, 210, 212, 213, 214, 216, 217, 222, 223, 224, 226, 227, 232, 233, 258, 280, 281, 287 n.19, 288, 289, 290, 297 pattern replication see borrowing Phrygian 103 n.35 plurative 137, 138 Polish 22, 23, 300, 302, 303, 304, 305t, 306, 307, 309, 310f, 311, 312, 315, 316, 317, 318, 319, 321, 322, 323ft, 324, 325, 326, 327 Polynesian 148 polysemy copying see borrowing pro-drop 32, 76, 82, 90, 96, 104, 110, 201, 273, 281, 295, 298 Punjabi 67 n.4

Quechua 253 reanalysis 13, 58, 72, 142, 147, 148, 150, 151, 221, 237, 259, 260, 271, 275, 293, 294, 298 reflexive 85 n.19, 87, 105–6, 112, 170, 177, 179 n.15, 190, 207, 248, 252 n.2, 257, 258, 261, 265, 266, 269, 274 Rushani 11 Russian 22, 23, 138, 300, 303, 304, 306, 307, 308, 310f, 311, 314, 316, 317, 321, 322, 323ft, 324, 325, 326, 327 Sabellic 79 semantic alignment 2, 3, 7–9, 10, 16, 17, 20, 24, 25, 50, 56, 57, 59, 62, 64, 65, 66, 79, 80, 98, 99, 102, 103, 114, 124, 145, 146, 147, 150, 151, 188, 214, 215, 216, 217, 222, 227, 228, 229, 230, 241, 242, 247, 248, 250, 254, 255, 257 n.7, 258, 259, 260, 261, 270, 271, 274, 275, 301 n.1 Semitic 2 Serbian 22, 300, 303, 306, 307, 309 n.14, 310f, 313, 314 n.18, 317, 323ft, 324 singulative 128, 138 Slavic 22, 23, 255, 300–27 Slovak 22, 23, 300, 303, 304 n.7, 307f, 310f, 311, 312, 315, 316, 317, 321, 322, 323ft, 324, 326 Slovenian 22, 300, 303, 306 n.10, 307f, 308 n.13, 310f, 316, 317, 323ft, 324 split alignment 4, 5 nn.3–4, 6, 11, 13, 14, 15, 17, 18, 21, 22, 24, 28, 30, 34, 44, 49, 50, 52, 53, 54, 55, 56, 62, 66–8, 74, 81, 83, 84, 87, 94, 97, 98, 99, 103, 108, 114, 115, 116, 120, 123, 124, 125, 126, 127, 129, 139, 142, 145, 147, 150, 151, 231, 277, 278, 284, 287, 294, 298 n.44 split intransitive alignment see semantic alignment split-S alignment see semantic alignment stative-active alignment see semantic alignment subject properties 18, 20, 64, 72, 78 n.16, 87, 91, 94, 99, 100, 117, 118, 120, 221, 222, 230, 233, 252, 256 n.4, 265 n.13, 298 subjecthood see subject properties Tabasaran 251 Talyši 293 n.34 Tobelo 259 Tokharian 287, 289 transimpersonal 1, 9, 10, 21, 259, 260, 270, 271, 273, 274, 275 transitive see bivalent transitivity prominence 13, 16, 23, 302, 303, 304, 305, 306, 308–17, 318, 319, 326–7

INDE X tripartite alignment 3, 6 n.4, 10, 11, 21, 22, 52f, 64, 65, 106, 107, 108, 116, 119, 277, 281, 282, 284, 292, 293, 294, 298 trivalent 1, 3, 30 n.8, 36, 44, 53, 55, 56, 57, 72 n.9, 113, 211, 264, 279, 309 Tsova-Tush see Batsbi Tunica 217 two-place see bivalent Ukrainian 22, 23, 300, 302, 303, 304, 307f, 310, 311, 317, 320, 323, 324, 325, 326, 327 Umbrian 73 Umpithamu 214 unaccusative 8, 16, 20, 35, 36, 70, 71, 86, 87, 94, 98, 99, 101, 102, 103, 188 n.1, 229, 262 unergative 7, 8, 16, 70, 71, 94, 102, 103, 141 n.22, 222, 227, 229

365

Urartean 2, 103 n.37, 277 n.1 Vedic Sanskrit 1, 15, 17, 18, 19, 24, 25, 26, 27, 30, 31, 33, 35, 37, 38, 40, 41, 44, 46, 47, 48ft, 49, 50, 62, 64, 66, 67, 68, 73, 87–94, 105 n.40, 114, 115, 116, 117, 118, 120, 166–87, 235 n.13, 255, 287, 301 n.1 vocative case 59, 74t, 81t, 89t, 95t, 125t, 129 volitionality 8, 19, 58, 98, 99, 189, 191, 192 n.5, 207, 210, 214, 215, 217, 222 n.2, 249, 253, 254 Wangkumara 69 Waray 214–15, 217 West Middle Iranian 22, 41 n.24, 116, 277, 278, 288, 290, 291–3, 298 Yazgulyami 11 Young Avestan see Avestan

O X F O R D ST U D I E S I N D I A C H R O N I C A N D H IST O R I C A L L I N GU IST I C S General editors Adam Ledgeway and Ian Roberts, University of Cambridge Advisory editors Cynthia L. Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge PUBLISHED 1 From Latin to Romance Morphosyntactic Typology and Change Adam Ledgeway 2 Parameter Theory and Linguistic Change Edited by Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar 3 Case in Semitic Roles, Relations, and Reconstruction Rebecca Hasselbach 4 The Boundaries of Pure Morphology Diachronic and Synchronic Perspectives Edited by Silvio Cruschina, Martin Maiden, and John Charles Smith 5 The History of Negation in the Languages of Europe and the Mediterranean Volume I: Case Studies Edited by David Willis, Christopher Lucas, and Anne Breitbarth 6 Constructionalization and Constructional Changes Elizabeth Closs Traugott and Graeme Trousdale 7 Word Order in Old Italian Cecilia Poletto 8 Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent

9 Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli 10 Vowel Length from Latin to Romance Michele Loporcaro 11 The Evolution of Functional Left Peripheries in Hungarian Syntax Edited by Katalin É. Kiss 12 Syntactic Reconstruction and Proto-Germanic George Walkden 13 The History of Low German Negation Anne Breitbarth 14 Arabic Indefinites, Interrogatives, and Negators A Linguistic History of Western Dialects David Wilmsen 15 Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden 16 Syllable and Segment in Latin Ranjan Sen 17 Participles in Rigvedic Sanskrit The Syntax and Semantics of Adjectival Verb Forms John J. Lowe 18 Verb Movement and Clause Structure in Old Romanian Virginia Hill and Gabriela Alboiu 19 The Syntax of Old Romanian Edited by Gabriela Pană Dindelegan 20 Grammaticalization and the Rise of Configurationality in Indo-Aryan Uta Reino¨hl 21 The Rise and Fall of Ergativity in Aramaic Cycles of Alignment Change Eleanor Coghill

22 Portuguese Relative Clauses in Synchrony and Diachrony Adriana Cardoso 23 Micro-change and Macro-change in Diachronic Syntax Edited by Eric Mathieu and Robert Truswell 24 The Development of Latin Clause Structure A Study of the Extended Verb Phrase Lieven Danckaert 25 Transitive Nouns and Adjectives Evidence from Early Indo-Aryan John. J. Lowe 26 Quantitative Historical Linguistics A Corpus Framework Gard B. Jenset and Barbara McGillivray 27 Gender from Latin to Romance History, Geography, Typology Michele Loporcaro 28 Clause Structure and Word Order in the History of German Edited by Agnes Ja¨ger, Gisella Ferraresi, and Helmut Weiß 29 Word Order Change Edited by Ana Maria Martins and Adriana Cardoso 30 Arabic Historical Dialectology Linguistic and Sociolinguistic Approaches Edited by Clive Holes 31 Grammaticalization from a Typological Perspective Edited by Heiko Narrog and Bernd Heine 32 Negation and Nonveridicality in the History of Greek Katerina Chatzopoulou 33 Indefinites between Latin and Romance Chiara Gianollo

34 Verb Second in Medieval Romance Sam Wolfe 35 Referential Null Subjects in Early English Kristian A. Rusten 36 Word Order and Parameter Change in Romanian A Comparative Romance Perspective Alexandru Nicolae 37 Cycles in Language Change Edited by Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen 38 Palatal Sound Change in the Romance Languages Diachronic and Synchronic Perspectives André Zampaulo 39 Dative External Possessors in Early English Cynthia L. Allen 40 The History of Negation in the Languages of Europe and the Mediterranean Volume II: Patterns and Processes Anne Breitbarth, Christopher Lucas, and David Willis 41 Variation and Change in Gallo-Romance Grammar Edited by Sam Wolfe and Martin Maiden 42 Phonetic Causes of Sound Change The Palatalization and Assibilation of Obstruents Daniel Recasens 43 Syntactic Features and the Limits of Syntactic Change Edited by Jóhannes Gı´sli Jónsson and Thórhallur Eythórsson 44 Romance Object Clitics Microvariation and Linguistic Change Diego Pescarini 45 The Diachrony of Differential Object Marking in Romanian Virginia Hill and Alexandru Mardale

46 Noun-based Constructions in the History of Portuguese and Spanish Patrı´cia Amaral and Manuel Delicado Cantero 47 Syntactic Change in French Sam Wolfe 48 Periphrasis and Inflexion in Diachrony A View from Romance Edited by Adam Ledgeway, John Charles Smith, and Nigel Vincent 49 Functional Heads Across Time Syntactic Reanalysis and Change Edited by Barbara Egedi and Veronika Hegedűs 50 Alignment and Alignment Change in the Indo-European Family Edited by Eystein Dahl IN PREPARATION Redevelopment of Case Systems in Indo-Aryan Miriam Butt Classical Portuguese Grammar and History Charlotte Galves, Aroldo de Andrade, Christiane Namiuti, and Maria Clara Paixão de Sousa Morphological Borrowing Francesco Gardani Nominal Expressions and Language Change From Early Latin to Modern Romance Giuliana Giusti A Study in Grammatical Change The Modern Greek Weak Subject Pronoun τος and its Implications for Language Change and Structure Brian D. Joseph Reconstructing Pre-Islamic Arabic Dialects Alexander Magidow Iranian Syntax in Classical Armenian Robin Meyer Arabic and the Case against Linearity in Historical Linguistics Jonathan Owens