New Challenges in Typology: Transcending the Borders and Refining the Distinctions 9783110219067, 9783110219050

Citation preview

New Challenges in Typology

≥

Trends in Linguistics Studies and Monographs 217

Editors

Walter Bisang (main editor for this volume)

Hans Henrich Hock Werner Winter

Mouton de Gruyter Berlin · New York

New Challenges in Typology Transcending the Borders and Refining the Distinctions

edited by

Patience Epps Alexandre Arkhipov

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data is available at http://catalog.loc.gov.

ISBN 978-3-11-021905-0 ISSN 1861-4302 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” Copyright 2009 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Laufen. Printed in Germany.

Preface Greville Corbett

The first volume in this series was an innovation. By their skill and confidence in bringing together this second set of varied and interesting papers, the new editors have already established a tradition. As their introduction shows, a set of people tackling a range of languages with open minds will find interesting new things; and equally they will be confronted by some traditional issues. To take one instance: as the editors point out, typology means drawing distinctions. We make distinctions since that is what careful linguistic analysis is about. As typologists we also draw distinctions for each language we investigate, so that when we compare across languages we compare like with like. And this is a way to get into seriously deep issues quite quickly. New research on less-studied languages shows us that our analytical categories are not fully adequate. Part of the excitement of research is precisely in finding these new phenomena. Here the innovative meets the traditional: we face once again the situation in which new data lead us to a restatement of our categories, after which we have to reassess and perhaps reanalyse the original data that we were comparing with, preserving all we can of the original insights. This is potentially a virtuous circle, from better description, to clearer conceptual distinctions, to better description… up into better understanding. An important development in recent years, which impacts on the volume, has been the greater care devoted to constructing the samples for typological research. Scanning the library shelves to select an armful of thicker grammars is not innovative methodology. The best samples are controlled for genetic and areal balance, and are large enough to make statistically valid generalizations. Indeed, some of the language samples investigated are impressively large. This is where the issue of distinctions and categories raises its head again. When we come to language 497 in our careful sample, we need to know exactly why we are ticking the box to record that it has a particular construction or feature value. If we are to avoid the accusation of being ‘drive-by typologists’ we need to be able to demonstrate that we are using precisely the same criteria as for the preceding 496 languages. (Of course, ‘the author says so’ will not do, for too

vi

Greville Corbett

many reasons than there is space for here.) Even when we have carefully analysed our sample, a traditional rule of thumb is that before proposing a new typological generalization, and awaiting the fame and fortune which will inevitably follow, it is worth checking the situation in Kayardild and Archi, as well as other favourite generalization-breakers. The mention of these typological star languages leads naturally to the relation of fieldwork and typology. As the volume shows, this relation is arguably healthier than a while ago. The evident overlap in concerns between the domains is reflected in the papers. We also see it more widely in linguistic personnel: we find fieldworkers who make the occasional foray into typology, typologists who do some fieldwork, and some splendid colleagues who the fieldworkers believe are fieldworkers because they are always in the field, and who the typologists believe are typologists, because they are always publishing on typology. The contributors form a good sample of these different types of linguist, collaborating in this joint project. These are a few considerations you may have in mind as you enjoy the varied papers here. As you consider the typological issues, you may also reflect on the quality of the researchers involved, and what that means for the future of our field. The volume brings together innovative research and traditional issues in a collaborative volume. I commend it to you warmly. Greville G. Corbett Surrey Morphology Group University of Surrey

Contents

Preface Greville Corbett ....................................................................................

v

Contents ...............................................................................................

vii

List of contributors ................................................................................

ix

Introduction Patience Epps and Alexandre Arkhipov ...............................................

1

Part I.

Word and phrase structure

Patterns of clitic placement: Evidence from ‘mixed’ clitic systems Ana R. Luís ............................................................................................

11

Eton tonology and morphosyntax: A holistic typological approach Mark L. O. Van de Velde .......................................................................

35

Part II. Case, agreement, and localization A hierarchical indexation system: The example of Emerillon (Teko) Françoise Rose ......................................................................................

63

Where differential object marking and split plurality intersect: Evidence from Hup Patience Epps ........................................................................................

85

Syncretisms and neutralizations involving morphological case: Challenges for markedness theory Peter Arkadiev........................................................................................ 105 Towards a typology of ‘attachment’ markers: Evidence from East Caucasian languages Dmitry Ganenkov .................................................................................. 127 Part III. Tense, aspect, and desire Revisiting perfect pathways: Trends in the grammaticalization of periphrastic pasts Chad Howe............................................................................................. 151

viii

Contents

Individual-level meanings in the semantic domain of pluractionality Andrey Shluinsky ................................................................................... 175 The symbiosis of descriptive linguistics and typology: A case study of desideratives Olesya Khanina ..................................................................................... 199 Part IV. Clause structure and verbal derivation Comitative as a cross-linguistically valid category Alexandre Arkhipov................................................................................ 223 Towards a typology of labile verbs: Lability vs. derivation Alexander Letuchiy ................................................................................ 247 Towards the typology of raising: A functional approach Natalia Serdobolskaya .......................................................................... 269 Historical pathways in Northern Paiute verb formation Tim Thornes ........................................................................................... 295 Part V. Class struggle: Erasing borderlines Reference and predication in Movima Katharina Haude ................................................................................... 323 All typologies leak: Predicates of change in Lowland Chontal of Oaxaca Loretta O’Connor .................................................................................. 343 Multidimensional typology and Miraña class markers Frank Seifart .......................................................................................... 365 Part VI. New challenges in methodology Steps toward a grammar embedded in data Nicholas Thieberger............................................................................... 389 Index of languages ................................................................................. 409 Index of authors .................................................................................... 414 Index of subjects .................................................................................... 422

List of Contributors

Peter Arkadiev Department of Typology and Comparative Linguistics, Institute of Slavic Studies, Russian Academy of Sciences Leninsky Prospekt 32-A 117334 Moscow Russia [email protected] Alexandre Arkhipov Department of Theoretical and Applied Linguistics, Philological Faculty, Moscow State University Leninskie gory, 1 k. gum. fak., 953 119991 GSP-1 Moscow Russia [email protected] Greville Corbett Surrey Morphology Group, English (J1) Faculty of Arts and Human Sciences University of Surrey Guildford, Surrey, GU2 7XH Great Britain [email protected] Patience Epps Department of Linguistics, University of Texas at Austin 1 University Station B5100 Austin, Texas 78712 USA [email protected]

Dmitry Ganenkov Department of Caucasian Languages, Institute of Linguistics, Russian Academy of Sciences Bolshoy Kislovsky pereulok 1/12 125009 Moscow Russia [email protected] Katharina Haude Department of Linguistics, University of Cologne 50923 Cologne Germany [email protected] Chad Howe Department of Romance Languages, University of Georgia Gilbert Hall 370J Athens, Georgia 30602-1815 USA [email protected] Olesya Khanina Department of Linguistics, Max Planck Institute for Evolutionary Anthropology Deutscher Platz 6 04103 Leipzig Germany [email protected]

x

List of contributors

Alexander Letuchiy Russian State University for Humanities; Russian Language Institute, Russian Academy of Sciences Dekabristov ul. 38-134 127273 Moscow Russia [email protected] Ana R. Luís Faculdade de Letras Universidade de Coimbra Praça da Porta Férrea 3004-530 Coimbra Portugal [email protected] Loretta O’Connor DoBeS Documentation Initiative (University of Hamburg/ Max Planck Institute for Psycholinguistics) Schoolstraat 144 6581 BG Malden Netherlands [email protected] Françoise Rose Dynamique Du Langage (CNRS, Université Lyon 2) Institut des Sciences de l’Homme 14 avenue Berthelot 69363 Lyon Cedex 07 France [email protected]

Frank Seifart Department of Linguistics, Max Planck Institute for Evolutionary Anthropology Deutscher Platz 6 04103 Leipzig Germany [email protected] Natalia Serdobolskaya Moscow Municipal University for Psychology and Pedagogy Oktiabrskaya ul. 42-51 127018 Moscow Russia [email protected] Andrey Shluinsky Department of Linguistics, Max Planck Institute for Evolutionary Anthropology; Department of African Languages, Institute of Linguistics, Russian Academy of Sciences Deutscher Platz 6 04103 Leipzig Germany [email protected] Nicholas Thieberger Department of Linguistics, University of Melbourne (&Department of Linguistics, University of Hawai’i at Manoa) Vic 3010 Australia [email protected]

List of contributors Tim Thornes Department of Writing, University of Central Arkansas 201 Donaghey Avenue Conway, Arkansas 72035 USA [email protected]

xi

Mark Van de Velde Research Foundation-Flanders; Center for Grammar, Cognition and Typology, University of Antwerp Rodestraat 14, 2000 Antwerpen Belgium [email protected]

Introduction Patience Epps and Alexandre Arkhipov

Since its origins, linguistic typology has been concerned with classifying human languages according to their various properties. These properties become ever more varied in nature and seemingly infinite in number as new languages come into the broadening horizons of linguists, and the distinctions found between and within languages become ever more finegrained. In their effort, typologists continuously face the need to transcend all kinds of borders – political and geographical, empirical and theoretical – in order to gather the necessary data in a comparable format and to finally arrive at an integrated view of the phenomena under scrutiny. These are borders between countries and continents where languages are spoken; borders between the linguistic notions – such as noun and verb, root and affix, coordination and subordination – which under thorough examination often appear to be quite ephemeral; and, in a broader perspective, borders between alleged components of the language system like morphology and syntax, syntax and semantics, phonology and grammar. Perhaps the most difficult to confront are the well-entrenched borders separating theories, frameworks, schools and traditions in linguistics – including that between the search for universals and the interest in language-specific particulars. As the work presented here shows, however, there are few things as rewarding as an open-minded approach that is willing to look beyond traditional distinctions. Typology is concerned with similarities: defining universals and strong tendencies across languages. At the same time, it is concerned with differences, with identifying finer and finer distinctions in our understanding of linguistic categories and units as our knowledge of them increases. Meanings are sliced into sub-meanings; presumably uniform classes are divided into many smaller ones; alongside well-established, clear-cut types emerge unpredicted and mixed types; patterns formerly assumed to characterize languages as a whole appear to be combined in distinct constructions within one and the same language. Similarly, while typology seeks to define generalizations across languages, it also values the discovery and explanation of exceptions to those generalizations. Accordingly, as this volume demonstrates, linguistic typology and linguistic description are closely

2

Patience Epps and Alexandre Arkhipov

paired enterprises, each of which informs and enriches the other. Just as generalizations about human language rely on – and can be radically altered by – data from individual languages, so is an understanding of a particular language deepened immeasurably by contextualizing it within the framework of what we know about languages more generally.

1. The chapters of this volume The present volume continues the series opened with New Challenges in Typology: Broadening the Horizons and Redefining the Foundations (2007), edited by Matti Miestamo and Bernhard Wälchli. The book brings together seventeen articles by postdoctoral scholars whose thesis work involved comparative typological study or typologically oriented language description. The papers focus on the relevance of this earlier work to issues of broad theoretical interest. While the scope of individual studies varies from worldwide typological samples to family-internal typology to single language studies, all underscore the usefulness of typologically oriented investigation for our understanding of linguistic theory more generally. The research presented in this volume highlights many of the new and exciting dimensions in which the field of linguistics is currently being shaped. The chapters are organized thematically so as to guide the reader through the various linguistic phenomena addressed by the authors. Their range includes cliticization, case, agreement, spatial relations, TAM categories, lability, verbal derivation, nominal classification, grammaticalization, and the integration of data and analysis. The papers in the first section, ‘Word and phrase structure’, discuss issues related to the phonological and morphosyntactic structure of words and phrases. In her contribution, Ana R. Luís considers ‘special clitics’ (i.e. those characterized by an idiosyncratic distribution), and challenges the claim that these should all be defined as phrasal affixes. She shows that the clitic system in European Portuguese is effectively ‘mixed’: pronominal enclitics behave like verbal suffixes, triggering genuine allomorphy on their host and themselves undergoing genuine allomorphy, whereas pronominal proclitics are closer to phrasal affixes in that they are not necessarily adjacent to the verb and may take wide scope over verb phrases. A similarly mixed clitic system is shown to exist in unrelated Udi (East Caucasian). Mark Van de Velde offers a historical perspective on the phonological and morphosyntactic complexity of words and verb phrases in Eton, a

Introduction

3

Bantu language of Cameroon. In Eton, as in some other Niger-Congo languages, phonological and morphosyntactic structure have arisen interdependently. Prosodic constraints on stems have arguably led to the restructuring of derivational morphology and of the verb phrase. Likewise, Eton’s complex tonal system has been shaped by the same prosodic constraints, via the erosion of morphological material over time. The four chapters in the ‘Case, agreement, and localization’ section address questions related to the marking of arguments, including case marking and verbal agreement (indexation). The first two chapters focus on two previously undescribed Amazonian languages, and consider the challenges to existing theories posed by the phenomena observed in these languages. Françoise Rose analyzes the verbal agreement system of Emerillon (or Teko, a Tupí language of Amazonia). The most peculiar feature of the system is the selection of the obligatory person marker on transitive predicates; only one person marker is allowed, and its choice is determined by two hierarchies, the person hierarchy (1/2 > 3) and the grammatical relation hierarchy (S > O). Rose argues that such hierarchical systems constitute a type apart alongside nominative-accusative and ergative-absolutive, and that inverse systems are best analyzed as a subtype thereof. In her contribution, Patience Epps presents a discussion of splits in case and number marking in Hup (Nadahup family, Amazonia), and shows that these splits correspond to the animacy and definiteness of the referent in ways that are largely typologically consistent with systems of ‘differential object marking’ and split plurality in other languages of the world. However, their intersection yields a typologically unusual result: plural-marked nominals in Hup are obligatorily object marked, despite their less individuated status. Peter Arkadiev contributes a typological perspective on syncretisms and neutralizations in morphological case, and addresses the challenges these pose for our understanding of markedness. He considers the generalization that neutralizations of particular morphosyntactic features occur only in marked contexts with respect to other features, and examines exceptions relating to morphological case marking. For many such examples, he argues, the best explanations for the patterns we observe need appeal only to mechanisms of language change, rather than to abstract notions like markedness. Many languages have developed sophisticated systems for expressing fine-grained distinctions in the spatial domain, arousing considerable recent typological interest. Such marking of spatial relations is a prominent feature of East Caucasian languages, which tend to indicate such relations

4

Patience Epps and Alexandre Arkhipov

morphologically on nouns within systems of non-core case marking. Dmitry Ganenkov undertakes a family-internal typological study addressing a particular subdomain of spatial relations, the expression of ‘attachment’. It is typical of these languages to have two different markers that share roughly the domain of the English preposition on. As Ganenkov demonstrates, the distinction between these two markers is often more complicated than the generally assumed opposition of ‘horizontal surface’ vs. ‘vertical surface’. It may be based on a number of parameters of spatial configuration, including inherence of association between figure and ground, manner of attachment, closeness of contact, etc. The following section is devoted to the categories of ‘Tense, aspect, and desire’. In his chapter, Chad Howe approaches the variation in the use of the present perfect construction in Peninsular and Peruvian Spanish. His discussion indicates that previous claims suggesting distributions similar to that of the simple perfective (and thus presupposing grammaticalization paths analogous to the development of the passé composé in French) are overly simplistic; in fact, significant variation is found with respect to temporal and discourse features well known in typology, such as occurrence in sequenced narratives. A typological approach is thus shown to be revealing and fruitful for the study of evolution of verbal forms even within dialects of a single language. The paper by Andrey Shluinsky centers on one of the subdomains of aspect: pluractionality. In addition to the generally accepted event-internal and event-external uses of the pluractional (i.e. the iteration of components within an event vs. iteration of entire events), ‘individual-level’ pluractionality (that is, relating to a stable property) is demonstrated to be attested in cross-linguistically stable contexts. Five categories of ‘individual-level’ pluractionality are further distinguished: ‘property’, ‘qualitative’, ‘capacitive’, ‘generic’, and an additional option for predicates that are lexically individual-level, denoting a long-term state. In her contribution, Olesya Khanina addresses the symbiotic relationship between linguistic typology and description through the lens of desideratives, or expressions of desire. Drawing from a sample of over seventy languages, she discusses typological parameters of the realization of the desiderative category, focusing on the ways in which arguments are encoded in desiderative constructions. The case of desideratives illustrates that, where typologically relevant parameters are not widely addressed in grammatical descriptions, this leads to an impoverished typological understanding of the category in question, which may in turn fail to inform further descriptive work.

Introduction

5

The section ‘Clause structure and verbal derivation’ is composed of four chapters related to issues of argument structure, valency and complex predicates. Alexandre Arkhipov addresses the category of comitative, putting forward a language-independent definition. He claims that comitative should be treated as a type of construction used for pluralization of a participant, not as a standalone semantic role. He also demarkates the borders between comitative and several closely related categories, many of which have not always been distinguished from comitative. In his paper, Alexander Letuchiy offers a typologically based discussion of labile, or ambitransitive, verbs. He considers the parallels between lability and valency derivation, focusing in particular on the anticausative subtype of labile verbs. Lability, he argues, is nevertheless best understood not as a mechanism of valency change per se (by which one argument structure is derived from another), but rather as a strategy for combining two meanings and two argument structures in one verb. Natalia Serdobolskaya considers the phenomenon of raising, focusing on the expression of the subject of the embedded clause as an argument of the matrix verb (as in ‘I believe him to be a linguist’), in typological perspective. The conventional understanding of raising, she argues, is based on data from only a few languages, particularly English. However, phenomena resembling raising can in fact be identified in a wide range of languages, but few exhibit morphosyntactic properties identical to those of English raising. Furthermore, raising may – contrary to the received definition – be sensitive to the semantics of the matrix verb and/or to semantic and pragmatic properties of the noun phrase to be raised, and thus cannot be considered a purely syntactic phenomenon. These facts require a revised definition of raising. The paper by Tim Thornes zeros in on the historical puzzle of complexity in the Northern Paiute (Uto-Aztecan, northwest USA) verb. Despite the similarities between Northern Paiute’s two main strategies for forming complex predicates – the instrumental prefix construction and the secondary verb construction – these are shown to be synchronically distinct. Their differences can probably be attributed to different historical origins: the instrumental prefix construction derives from compounding, whereas the secondary verb construction may have developed from a subordinate form that lost its dependent morphology. Thornes’ discussion illustrates how fine-grained synchronic description and diachronic analysis may work together to inform our understanding of grammaticalization processes and of linguistic structures such as complex predication.

6

Patience Epps and Alexandre Arkhipov

The chapters of the following section, ‘Class struggle: erasing borderlines’, provide a vivid illustration of ‘transcending borders’ and showcase the fuzziness of word classes and the mixture of presumed language types in a single language. Katharina Haude’s contribution examines the functional and formal similarities between the classes of noun and verb in Movima, an isolate of lowland Bolivia. In Movima, nouns and verbs have similar syntactic identities: both can function as predicates, and both can refer to concrete entities. Furthermore, the blurred distinction between these two word classes may explain Movima’s unusual alignment system, by which arguments of transitive clauses are encoded either as syntactic arguments or like nominal possessors, depending on their relative position in a referential hierarchy. In the following chapter, Loretta O’Connor discusses predicates encoding change events in Lowland Chontal (isolate, Mexico). Pulling from a wide range of data, she argues that expressions of change in this language are strikingly inconsistent with established typologies of lexicalization and constructional patterns. She suggests that the formal patterns evident in Chontal’s expressions of change may be better understood from the perspective of typologies of use and of constructions, rather than typologies of languages. Frank Seifart offers a discussion of nominal classification in Miraña (Witotoan, northwest Amazonia), and observes that the Miraña system cannot be categorized according to the two most widely accepted parameters for a typology of noun classification systems: the type of morphosyntactic host (e.g. noun or numeral) with which the classifier occurs, and the presence or absence of agreement (by which noun classes are conventionally differentiated from classifiers). He suggests an alternative approach that makes use of more detailed parameters, and observes that a move away from typologizing nominal classification systems on the basis of a few general parameters may bring us to a more sophisticated understanding of classification more generally. We are particularly happy to include a section on ‘New challenges in methodology’. Meta-questions like those raised by Nicholas Thieberger in his contribution are indispensable for the science of linguistics to exist and develop. As Thieberger emphasizes, a lack of attention to the presentation of data and the distinction between data and analysis may ultimately compromise the results, making a researcher’s claims unverifiable or uninterpretable for future researchers. New digital technologies allow for a previously unthinkable embedding of the raw linguistic data in the published work. We hope that Thieberger’s chapter will encourage typologists and

Introduction

7

grammarians to experiment further with novel technologies, and also to share their methodological experience and expertise with their colleagues, so as to open doors to further innovations.

2. Technical notes The volume makes use of the following technical conventions. The glossing of examples conforms generally to the Leipzig Glossing Rules (see http://www.eva.mpg.de/lingua/resources/glossing-rules.php for details), although there is some variation in the abbreviations used from chapter to chapter. For each chapter, the abbreviations are listed in the endnotes. For chapters that refer to data from multiple languages, each set of examples is introduced with the name of the language and language family and genus, cited to the extent possible according to the names used in The World Atlas of Language Structures Online (http://wals.info/languoid). The source of the data is given together with the language name, except where different sources are indicated after each example in a set. Where no source is mentioned, the data should be understood as coming from the author’s own fieldwork or personal knowledge. To refer to works published in Russian, we used a uniform transliteration for titles and authors’ names. Where publications in other languages are cited, the transliterated versions of Russian authors’ names in brackets are preceded by the more traditional spellings found in these versions.

3. Acknowledgements We are very grateful to the many colleagues and others who made this volume possible. First, it is thanks to Matti Miestamo and Bernhard Wälchli that this and the preceding volume of New Challenges in Typology exist in the first place. We deeply appreciate their initiative in setting this in motion, and we thank them for passing the baton to us, and for all their helpful advice during the process. We also owe an enormous thanks to all those colleagues who reviewed the papers published here, and whose detailed and insightful comments have contributed vastly to the final outcome: Oliver Bond, Denis Creissels, Michael Cysouw, Östen Dahl, Alice Gaby, Jeff Good, Birgit Hellwig, Andreas Jäger, Timur Maisak, Jeanette Sakel, Dmitri Sitchinava, Hein Van der Voort, Christina Willis, Fernando Zúñiga, and a number of others who preferred to remain anonymous.

8

Patience Epps and Alexandre Arkhipov

We are very grateful to all the authors of the papers for their excellent contributions, for their hard work in fine-tuning them, and for their help in reviewing each others’ papers. A very big thanks goes to Greville Corbett for contributing the preface to the volume. Thanks also to Walter Bisang, the editor of Mouton’s ‘New Trends in Linguistics: Studies and Monographs’ series, for all his help with the volume’s preparation, and to Birgit Sievert and Monika Wendland from Mouton for their assistance in the publishing process. We are grateful as well to the University of Texas at Austin and Moscow State University for institutional and financial support. Finally, we very much appreciate the willingness of the Association for Linguistic Typology to foster this book project, both in its first instantiation in the work of Matti and Bernhard, and now in its second. Our thanks also go to the members of the Pāņini and Greenberg award juries for putting in motion the process that led to this book’s existence. We now join Matti and Bernhard in hoping that further volumes of New Challenges in Typology will come into being. We also hope that the cooperative work of international communities of postdoctoral scholars will continue to transcend borders of all kinds, while reconsidering and refining the distinctions that inform our understanding of human language.

Part I. Word and phrase structure

Patterns of clitic placement: Evidence from ‘mixed’ clitic systems Ana R. Luís

1. Introduction1 A recurrent theme throughout the literature on clitics, and throughout this chapter, is the idea that the placement of ‘special clitics’ cannot be defined in terms of the normal rules of syntax. Special clitics are notorious for appearing in positions that are generally not available to words with a similar function (Zwicky 1977). For example, a) Second Position (or Wackernagel) clitics, such as Serbo-Croatian auxiliaries and pronouns, must appear either after the first word or after the first phrase in the sentence (Klavans 1980, Anderson 1992, 1996); b) verbal clitics in Macedonian and in various Romance languages are placed either before or after the verb depending on the morphosyntactic properties of the clause (Spencer 1992, Halpern 1998); and c) Tobler-Mussafia clitics in Old French or Bulgarian must appear before the verb, except where this would make them sentence-initial (Spencer 2001, Anderson 2000).2 Based on such observations, it is generally accepted that the idiosyncratic nature of clitic positioning constitutes an intrinsic property of ‘special clitics’. A number of attempts have been made in the literature to account for the distribution of special clitics in a unified way (e.g., Klavans 1980, Anderson 1992, Sadock 1991). One of the first proposals was formulated by Klavans (1980, 1985) who suggests a limited set of parameters for the definition of all the possible clitic positions. To capture the idiosyncratic distribution of special clitics, Klavans proposes that cliticization should be analysed as phrasal morphology. In Klavans’ terms, ‘special clitics’ should be regarded as phrasal affixes, i.e., affixes whose host is defined with respect to a phrasal domain (e.g., NP, N′, VP, V′, etc.). However, Klavans (1980) also highlights the fact that some special clitics appear systematically on the head of a given phrase and, hence, select the grammatical category of the host they attach to. Such special clitics, far from being promiscuous, differ from phrasal affixes in that they attach morphologically to a word rather than phonologically to a phrasal node. Klavans

12

Ana R. Luís

(1980) refers to the case of Spanish clitic pronouns and suggests that these are best analysed as head affixes (i.e., word-level affixes). Klavans’ insights on cliticization have gained wide acceptance in the literature and have been formalized in numerous studies within different theoretical frameworks. In particular, the claim that special clitics constitute phrasal affixes is further developed in Anderson (1992), Halpern (1995), Spencer (2001), Legendre (2001), and references therein, for phenomena such as Bulgarian or Serbo-Croatian pronominal and auxiliary clitics. Similarly, the idea that some ‘special clitics’ behave in effect like word-level affixes has found strong empirical support in the work of Zwicky and Pullum (1983), Monachesi (1999) and Miller and Sag (1997), among other. These authors have drawn attention to the fact that some clitics in English, French or Italian exhibit properties that are normally associated only with standard affixes (e.g., idiosyncratic morphophonological alternations). The primary purpose of this chapter is to demonstrate that the distinction between word-level affixation and phrasal affixation, originally formulated by Klavans (1980) in her theory of clitics, proves crucial in accounting for the European Portuguese clitic system (Luís 2004). In European Portuguese (EP), pronominal enclitics trigger genuine allomorphy on the host and undergo genuine allomorphy themselves, while pronominal proclitics can take wide scope over entire verb phrases and do not need to be immediately adjacent to the verb. So, in postverbal position, pronominal clitics behave like word-level affixes, whereas in preverbal position they exhibit the behaviour of phrasal affixes. The ability for affixal clitics to select either a morphological host or phrasal host is not exclusive to EP and can also be found in Udi (Harris 2002). As shown here, subject agreement clitics in this East-Caucasian (Nakh-Daghestanian) language show up either as phrasal affixes or as word-level affixes.3 Our empirical findings therefore weaken the recent claim, made in the clitic literature, that all special clitics constitute phrasal affixes (Anderson 1992, 1996, 2005). Anderson’s theory of cliticization effectively emphasizes the morphological nature of special clitics, in line with Klavans (1980). However, it departs quite substantially from Klavans’ proposal by defining special clitics uniformly as the “morphology of phrases” (i.e., as affixes that are positioned with respect to a phrasal domain). Anderson (2005) specifically addresses the case of both EP pronominal clitics and Udi person markers, arguing that they also behave like phrasal affixes. However, as our data show, Anderson’s monolithic definition of special clitics is empirically untenable, because it simply fails to take into account

Patterns of clitic placement

13

the observed grammatical differences between the word-level and phraselevel attachment of EP and Udi clitics. Ultimately, we argue that a more fine-grained approach to cliticization must be explored to accommodate mixed clitic systems.4 The structure of this chapter is as follows. Section 2 offers an outline of the empirical data on EP clitic placement. Section 3 sketches an informal typology of special clitics and argues that the clitic theory developed by Anderson (1992, 2005) fails to capture the empirical differences between the word-level and phrase-level attachment of special clitics. Section 4 briefly sketches an analysis of the EP clitic system within the theory of Generalized Paradigm Function Morphology (Spencer 2005, ms.). Section 5 summarizes our findings.

2. Evidence from European Portuguese 2.1. The affix-properties of clitic sequences In this section we show that clitic clusters in EP display a variety of morphophonological phenomena that are typical of word-level affixes. In particular, our data illustrates the presence of fusional morphology (cf. 1), syncretism (cf. 2), and non-productive allomorphy (cf. 3). Morphophonological fusion takes place between 3rd person accusative clitics and vowel-final dative clitics. In (1a), the 3rd person singular feminine accusative a is amalgamated with the 1st person singular dative me, giving rise to the portmanteau cluster ma. In (1b), the same accusative clitic is fused with the 2nd person singular dative te. These portmanteau clusters surface both postverbally (cf. 1a) and preverbally (cf. 1b).5 (Clitics are given in bold face here and throughout the chapter.) (1)

European Portuguese (Indo-European, Romance) a. A Maria ofereceu -ma. (*me-a) the M. gave -1SG.DAT-3SG.FEM.ACC ‘Maria gave it to me.’ b. Acho que não to dei (*te-o) think that not 2SG.DAT-3SG.MASC.ACC gave ‘I think that I didn’t give it to you.’

Other portmanteau forms which result from the obligatory fusion between clitics are illustrated below, in Table 1.

14

Ana R. Luís

Table 1. Portmanteau clusters in EP 1SG.DAT 2SG.DAT

3SG.MASC.ACC mo (*me-o) to (*te-o)

3SG.FEM.ACC ma (*me-a) ta (*te-a)

3PL.MASC.ACC mos (*me-os) tos (*te-os)

3PL.FEM.ACC mas (*me-as) tas (*te-as)

As for syncretism, we find syncretic clitic clusters when 3rd person dative clitics combine with 3rd person accusative clitics. This combination neutralizes the number features of the dative clitics. As illustrated in (2a), the portmanteau cluster lho can mean either ‘it to him/her’ or to ‘it to them’; and, in (2b), the cluster lhos can mean either ‘them to him/her’ or ‘them to them’. (2)

a. Compra -lho! (*lhe-o) buy -3SG/PL.DAT-3SG.MASC.ACC ‘Buy it for him/her/them.’ b. Quem lhos entregou? (*lhe-o) who 3SG/PL.DAT-3PL.MASC.ACC delivered ‘Who delivered them to him/her/them?’

The complete set of syncretic forms found in EP is illustrated in Table 2. Table 2. Syncretism inside clitic clusters 3DAT.SG 3DAT.PL

3SG.MASC.ACC

3SG.FEM.ACC

lho

lha

3PL.MASC.ACC lhos

3PL.FEM.ACC lhas

Finally, object pronouns generally exhibit morphophonological alternations when 3rd person accusative pronouns (i.e., o ‘him’, a ‘her’, os ‘them.MASC’ and as ‘them.FEM’) are preceded by the 1st or 2nd person plural dative pronouns nos ‘us’ or vos ‘you.pl’. This combination triggers ‘reciprocal’ allomorphy inside the cluster given that both clitics undergo shape alternations. As shown in (3), dative clitics undergo clitic-final consonant deletion and accusative clitics surface as l-initial allomorphs. (3)

a. Deu -no-lo. (*nos-o) gave -1PL.DAT-3SG.MASC.ACC ‘S/he gave it to us.’

Patterns of clitic placement

b. Ninguém vo-lo nobody 2PL.DAT-3SG.MASC.ACC ‘Nobody gave it to you.pl.’

15

deu. (*vos-o) gave

Table 3 illustrates the complete inventory of clitic clusters displaying ‘reciprocal’ allomorphy in EP. Table 3. Morphophonological alternations inside clitic clusters 1PL.DAT 2PL.DAT

3SG.MASC.ACC

3SG.FEM.ACC

3PL.MASC.ACC

no-lo (*nos-o) vo-lo (*vos-o)

no-la (*nos-a) vo-la (*vos-a)

no-los (*nos-os) vo-los (*vos-os)

3PL.MASC.ACC no-las (*nos-as) vo-las (*vos-as)

Thus far, we have shown that a wide range of idiosyncratic processes take place inside the clitic cluster, indicating that sequences of clitics behave like sequences of word-level affixes. 2.2. The morphological attachment of enclitic pronouns The first observation to make about EP enclitics is that they can occur either verb-finally, as illustrated in (4–8), or verb-internally between the stem and the future/conditional agreement endings, as in (10). In both positions, enclitics interact with the verb in a number of morphophonologically complex ways. More precisely, they trigger allomorphy on the verb (cf. 4) and undergo stem-induced allomorphy (cf. 6). In some cases, they both induce and undergo allomorphic variation within the same verb-enclitic sequence (cf. 8). Stem-allomorphy takes place when 1st and 2nd person plural enclitic pronouns nos ‘us’ and vos ‘you.pl’, respectively, are preceded by a 1st person plural verb form, regardless of the tense or aspect properties of the verb: (4)

a. Vêmo -nos amanhã. (*vêmos-nos) see.1PL -1PL.REFL tomorrow ‘We will see us tomorrow.’ b. Queremo -vos em casa cedo. (*queremos-vos) want.1PL -2PL.DAT in house early ‘We want you back home early.’

16

Ana R. Luís

Crucially, verb-final consonant deletion is restricted to the morphosyntactic context illustrated in (4). In contrast, if the enclitic pronouns nos and vos are preceded by a consonant-final verb that is not in the 1st person plural, as in (5a), then the verb-final consonant is preserved. Similarly, if a 1st person plural verb is followed by a consonant-initial enclitic other than the 3rd person accusative pronoun, no clitic allomorphy is observed, as illustrated in (5b). (5)

a casa? (*vende-nos) a. Vendes -nos sell.2SG.PRES -1PL.ACC the house ‘Would you sell us the house?’ b. Ouvimos -te aos gritos. (*ouvíamo-te) hear.2PL.IMPF -2SG.ACC at screams ‘We heard you scream.’

Enclitics also exhibit morphophonological variation when 3rd person accusative clitics combine with 3rd plural verb forms. In this context, accusative clitics, which by default are vowel-initial, surface as n-initial allomorphs. This generalization applies to both lexical verbs (6a) and auxiliaries (6b). On the contrary, other nasal-final verb forms, such as 3rd singular present indicative forms, as shown in (7), select the vowel-initial allomorph. Hence, n-allomorphs are sensitive to the morphosyntactic properties of the preceding verb rather than to the phonological form of the verb. (6)

(*lavam-o) a. As enfermeiras levam -no. The nurses take.3PL.PRES -3SG.MASC.ACC ‘The nurses take him.’ visto. (*tinham-as) b. As mulheres tinham -nas the women had.3PL.IMPF -3PL.FEM.ACC seen ‘The women had seen them.’

(7)

O professor tem -os visitado todos the teacher has 3PL.MASC.ACC seen every os dias. (*tem-nos) the days ‘The teacher has been visiting them every day.’

Reciprocal allomorphy takes place when a 3rd person accusative clitic is preceded by a verb ending in one of the following consonants: -s, -z or -r.

Patterns of clitic placement

17

In this context, the accusative clitic surfaces as an l-initial allomorph and the verb undergoes verb-final consonant deletion: (8)

Procuramo -lo searched.1PL.PRES -3SG.MASC.ACC o dia. (*procuramos-o) the day ‘(We) searched for him/her/it all day.’

todo all

Note that the selection of the l-allomorph is restricted to vowel-initial enclitics and their preceding verbal host. It cannot be triggered across word boundaries, as (9) illustrates: although the consonant-final noun lápis ‘pencil’ is followed by the vowel initial adjective azul ‘blue’, we see that consonant-deletion and l-epenthesis cannot apply. (9)

lápis azul (*lápi∅ lazul) ‘blue pencil’

Reciprocal allomorphy is also found when enclitic pronouns surface verb-internally between the verb stem and the future/conditional agreement marker, as shown in (10). (10) a. cantá∅ -lo sing.INF -3SG.MASC.ACC ‘(I) will sing it’ b. escrevê∅ -las write.INF -3.PL.FEM.ACC ‘(you) would write them’

-ei -1SG.FUT

(*cantar-o-ei)

-ías -2.SG.COND

(*escrever-o-ías)

Summing up, enclitic pronouns exhibit the following affixal properties: a) they must be immediately adjacent to the verb, b) they intervene between the stem and the future/conditional agreement markers, c) they trigger stem-allomorphy and d) undergo stem-induced allomorphy. As to c) and d), allomorphic alternations are determined by highly restricted contexts, such as the morphosyntactic properties of the verb (cf. 6) or the morphosyntactic properties of both the verb and the enclitic (cf. 4). The facts therefore show that enclitic pronouns interact morphologically with the verb in ways which indicate that they constitute word-level suffixes.

18

Ana R. Luís

2.3. The phrasal properties of proclitic pronouns 2.3.1. The separability between the proclitic and the verb Contrary to enclitics, pronominal proclitics exhibit none of the morphological properties surveyed in Section 2.2. Instead, they show clear-cut phrasal properties: a) they can have wide scope over two conjoined verb phrases (cf. 11), and b) they can be separated from the verb by intervening words (cf. 12). Wide scope over coordinated phrases is shown below: [ajudou (11) a. Apenas a minha mãe me only the my mother 1SG.ACC [helped e incentivou]. and encouraged] ‘Only my mother helped me and encouraged me.’ b. Acho que lhes [leram uma história e deram think that 3PL.DAT [read a story and gave um livro]. a book] ‘I think that they have read them a story and given them a book.’

In (11a), the proclitic me functions as the object of the coordinated verb phrase [ajudou e incentivou] ‘helped and encouraged’. Similarly, in (11b), the preverbal clitic lhes realizes the dative argument of the conjoined verb phrase [leram uma historia e deram um livro]. In (11b), the arguments are non-identical object complements: uma historia ‘a story’ is the complement of leram ‘read’ and um livro ‘a book’ is the complement of deram ‘give’. The ability to take wide scope across VPs constitutes the kind of data which lends support to the view that proclitics do not attach morphologically to the immediately adjacent verb. Prototypically word-level affixes cannot be shared by conjoined syntactic phrases (Luís and Sadler 2003). The phrasal properties of proclitics are further supported by the fact that proclitics allow certain words to intervene between them and the verb: ainda não disse. (12) a. Acho que ela lho think that she 3SG.DAT/3SG.MASC.ACC yet not told ‘I think that s/he hasn’t told it to him/her/them yet.’

Patterns of clitic placement

19

já compraste, … b. Embora eu saiba que a although I know that 3SG.ACC.FEM already bought ‘Although I know that you have already bought it, …’

In (12a), the proclitic and the verb are separated by the particles ainda ‘yet’ and não ‘not’, and in (12b) by the particle já ‘already’. Although there are quite severe restrictions on the lexical items that can intervene between the proclitic cluster and the verbal host, the data indicates quite clearly that proclitics – quite unlike enclitics – are syntactically transparent. 2.3.2. The proclitic contexts Unlike pronominal clitics in French, Italian or Spanish, the alternation between preverbal and postverbal placement in EP is not dependent on the finiteness of the verb but on a set of specific words and phrases in preverbal position. We refer to such elements as proclitic triggers (Martins 1994, Crysmann 2002, Luís 2004, Luís and Otoguro 2005). In (13), preverbal placement is triggered by the complementizer que ‘because’. (13) Acho que ele os think.1SG that he 3PL.MASC.ACC ‘I think he wants them.’

quer. wants

Proclisis is also sensitive to a number of other preverbal triggers, including adverbial particles (14a), wh-phrases (14b), quantified subjects (14c), relative pronouns (14d), among other. (14) a. As crianças também o viram. the children also 1SG.MASC.ACC see ‘The children saw him too.’ b. Quantos livros te compraram? how_many books 2SG.DAT buy ‘How many books did they buy you?’ c. Todos os irmãos lhe telefonaram. all the brothers 3SG.DAT phoned ‘All the brothers phone him/her.’

20

Ana R. Luís

d. A quem os to whom 2PL.MASC.ACC ‘Who did you give them to?’

entregaste? give

The contexts triggering proclitics further reinforce the syntactic transparency of preverbal pronominal clitics in EP. Previous syntactic analyses of EP cliticization have tried to identify configurational properties triggering proclisis, by placing triggers under functional nodes such as CP or IP or under functional projections such as NegP or FocP (Martins 1994). However, finding a common denominator for all proclitic contexts in the phrase structure has proven difficult. Some attempts have therefore been made at identifying natural semantic classes (e.g. downward monotone quantifiers; Crysmann 2002), but such classes also fail to encompass all proclitic triggers. Other studies have argued that clitic placement is largely driven by discourse-informational considerations (McConvell 1996), but it would be rather difficult to extend this intuition to subordinating complementizers or conjunctions. Overall, then, there appears to be no single configurational, semantic or discourse explanation for procliticization in EP. 2.4. Summary The puzzle posed by EP pronominal clitics seems to be the following: there is evidence indicating that enclitics in EP are syntactically opaque and attached to the verb in the morphology (Section 2.2). However, robust evidence strongly suggest that proclitics behave like syntactically visible units (Section 2.3): Proclitics a) license the interpolation of syntactic material (even if limited to Xº elements) and b) take wide scope over coordinated VPs (Section 2.3). Despite these marked differences, however, proclitics and enclitics are formally and semantically exactly identical (Section 2.1, examples 1 and 2). Assuming then that both enclitics and proclitics are instantiations of the same set of pronominal units, we are now left with the need to account for the syntactic transparency of the proclitic-verb combination, on the one hand, and the opaqueness of the verb-enclitic combination, on the other. In Section 4, adopting Klavans’ (1980) distinction between word-level and phrase-level affixation, we formulate an account in which preverbal clitics select a phrasal node in the syntax, while postverbal clitics select a verbal base in the morphology. Anticipating our conclusion somewhat, we argue

Patterns of clitic placement

21

that the difference between enclitics and proclitics is not one of kind, but one of domain of clitic placement. This hypothesis naturally accounts for both the formal similarities, and the distributional/scopal differences between enclitics and proclitics in EP.

3. The ‘mixed’ nature of clitic systems 3.1. Uniform clitic systems In word-level affixation, affixes are bound morphological units which need a morphological base to attach to. They can undergo idiosyncratic morphological variation of a kind normally associated with irregular inflection. Word-level affixes select a base belonging to a specific word class, while phrasal affixes select a specific phrasal domain. So, while word-level affixes combine with the host in the morphology, phrasal affixes attach to their host in the syntax (Klavans 1980, Anderson 1992). In uniform clitic systems, clitics attach either exclusively at the wordlevel or exclusively at the phrase level. Uniform word-level attachment has been proposed for enclitic and proclitic pronouns in French and Italian, by Monachesi (1999) and by Miller and Sag (1997). It is claimed that both proclitics and enclitics behave like verbal affixes given that they cannot take wide scope over conjoined verbs and must attach to a verb. Formally, these studies treat preverbal and postverbal clitics as positional variants with the same morphological status. Uniform phrase-level attachment is also well documented. In Serbo-Croatian, for example, several facts indicate that pronominal clitics and auxiliary clitics exhibit properties that are typical of ordinary affixes: a) they undergo non-productive morphophonological effects (e.g., the pronominal clitic je ‘FEM. SG. ACC’ surfaces as ju when followed by another clitic je), b) they are rigidly ordered inside the cluster and c) the order of clitics does not reflect general syntactic principles (Spencer 1992, Anderson 1996). However, despite such inflectional properties, they do not select the category of the host they attach to. As shown below, the clitic je is preceded by a determiner in (15a) and by a noun in (15b). (15) Serbo-Croatian (Indo-European, Slavic; Halpern 1995: 16) čovek voleo Mariju. a. Taj je that AUX.3SG man love.PPT Maria

22

Ana R. Luís

voleo b. Taj čovek je that man AUX.3SG love.PPT (a=b) ‘That man loved Maria.’

Mariju. Maria

The concept of phrasal affixation entails the idea that the position of clitics must be defined in phrasal terms. As a bona fide second position clitic, the auxiliary clitic surfaces in a specific phrasal position – either after the first word (cf. 15a) or after the first phrase (cf. 15b). It is precisely the ‘promiscuous’ nature of these affixes that forms the empirical basis of Klavans’ (1980) notion of phrasal affixation. 3.2. Mixed clitic placement in European Portuguese In EP, unlike in the above-mentioned Romance languages, there is a difference in status between the preverbal and postverbal attachment of pronominal clitics. In particular, whereas enclitics select a word-level host, proclitics select a phrasal domain. As phrasal affixes, EP proclitics select the VP domain. However, within this VP domain, they attach to the left of the Vº, which is immediately dominated by the VP. The high degree of locality between the proclitic and the verb is aimed at emphasizing the fact that EP proclitics are effectively Vº-oriented phrasal affixes. Although they can be separated from the verb by interpolated particles, proclitics in Modern Portuguese no longer allow whole phrases to interpolate (Martins 1994). We capture the restricted nature of interpolation by treating interpolated elements as non-projecting words which form with the verb a small (phrasal) construction, as shown in (16). The ‘small construction’ is dominated by the zero-level projection Vº (Toivonen 2003). (16) a.

b.

Crucially, proclitics attach to the left of the Vº immediately dominated by VP. Attachment to a phrasal node prevents proclitics from forming a mor-

Patterns of clitic placement

23

phologically cohering unit with the immediately adjacent verb. The representation in (16) also predicts a complete absence of non-productive allomorphic variation at the proclitic-host boundary. Enclitics, on the contrary, exhibit a number of properties which indicate that they combine morphologically with the verb: As shown in Section 2.2, a) they trigger stem allomorphy on the verb, b) undergo allomorphic variation, c) occur between the infinitival stem and the future/conditional agreement endings and d) can only attach to a verbal host. It is not obvious how these properties can be insightfully explained if enclitics are regarded as phrasal affixes. In effect, an overall phrasal affix approach, as proposed by Anderson (2005) for EP, fails to provide an insightful account of the distributional, scopal and morphophonological asymmetries between proclitics and enclitics.6 These differences can only be accommodated if enclitics are analyzed as verbal suffixes and if the verb-enclitic sequence is derived morphologically as an inflected verb form (Luís 2004). 3.3. Other ‘mixed’ clitic systems: person agreement makers in Udi The ability for clitics to be partly selective and partly promiscuous with respect to their host is not exclusive to EP. Person markers in Udi, an EastCaucasian (Nakh-Daghestanian) language, undergo morphological attachment in some contexts and phrasal attachment in others (Luís and Spencer 2005b). In what follows we survey the positions in which Udi person markers appear and highlight the mixed status of the Udi clitic system.7 If the clause contains a focused word or constituent, Udi clitics encliticise to it. In (17a) the person marker attaches to a negative particle, and in (17b) to a wh-phrase. Both the particle te and the wh-phrase are obligatorily in focus in Udi. The person marker can also attach to a focused argument, as in (17c–d), and to a nominal predicate (17e). If two different focused constituents co-occur within the same clause, precedence relations determine the correct output (Harris 2002). The data nicely shows that Udi person makers are sensitive to the information-structure of the clause. (17) Udi (East Caucasian, Lezgic) bụɣa-b-e p’ạ ačik’alšey a. nana-n te-ne mother-ERG NEG-3SG find-DO-AORII two toy.ABSL ‘Mother did not find two toys.’ (Harris 2002: 117)

24

Ana R. Luís

ayt-exa? b. mano muz-in-nu which language-INST-2SG word-SAY.PRES ‘What language are you speaking?’ (Harris 2002: 118) c. yaq’-a-ne ba-st’a road-DAT-3SG in-LV.PRES ‘ON THE ROAD he opens it.’ (Harris 2002: 3) d. merab-en ayt-ne ef-sa Merab-ERG word.ABSL-3SG keep-PRES ‘Merab keeps his WORD.’ (Harris 2002: 3) e. nana k’wa-ne mother.ABSL house.DAT-3SG ‘Mother is at the house.’ (Harris 2002: 121)

In some contexts, however, Udi person markers are attracted by the inflectional properties of the main verb. In particular, if the main verb is in the future II, the subjunctive I, the subjunctive II, or imperative, Udi clitics are positioned immediately after it: (18) a. boš-t’-al-le in-LV-FUTII-3SG ‘S/he will plant.’ (Harris 2002: 65) b. eɣ-a-q’un? come-SUBJVII-3PL ‘Will they come?’ (Harris 2002: 117) c. besp’-a-nan kill-IMP1-2PL ‘You kill [her].’ (Harris 2002: 117) In fact, these TMA properties take precedence over focused contexts. In (19), the person marker must follow the future II, rather than the whphrase: (19) šu-x k’al’k’al-a? who-DAT call-say-FUTII-3SG ‘Whom will she invite?’ (Harris 2002: 119) Verb-final placement, then, is triggered by the morphosyntactic properties of the verb. In this respect, Udi person markers ‘look inside’ the properties of the verb (Zwicky and Pullum 1983). Another common property of word-level affixes is the fact that neither words nor phrases can intervene

Patterns of clitic placement

25

between them and the morphological host. This generalization also holds for the person markers in (18). Udi person markers also undergo morphosyntactically conditioned allomorphy. As shown in (20), the 3rd singular marker -ne undergoes vowel elision after the conditional makers -a-y. Vowel elision in this context is obligatory both in the Nizh dialect (Timur Maisak, p.c.) and in the Vartashen/Oktomberi dialect. (20) a. e-y-a-y-n hither-LV-SUBJV-PAST-3SG ‘whoever may come’ (Harris 2002: 106) b. uk’-a-y-n say-SUBJVI-PAST-2SG ‘may you say’ (Harris 2002: 106) The word-level attachment of verb-final clitics is further reinforced by the data in (21), which shows that they can induce morphosyntactically conditioned stem allomorphy. In the Nizh dialect, as illustrated in (21a), the future suffix [-al] surfaces as [-o] if followed by the 1st singular person marker -zu. In (21b), on the contrary, the -al suffix is left unchanged. (21) a. bak-[o]-z (*bak-[al]-z) become-FUT-1SG ‘(I) will become’ (Timur Maisak, p.c.) b. bak-[al]-nu become-FUT-2SG ‘(you.sg) will become’ (Timur Maisak, p.c.) Finally, the fact that clitics surface inside the verb further reinforces our view that person markers form a morphologically cohering unit with the verb. As shown succinctly in (22), Udi person markers can either be inserted into a complex verbal stem between the incorporated element and the light verb or the verbal root, as in (22a), or they can appear inside a (diachronically complex) monomorphemic verbal root, as in (22b).8 (22) a. aš-ne-b-sa work-3SG-do-PRES ‘s/he works’ (Harris 2002: 122)

26

Ana R. Luís

b. a-z-q’-e receive1-1SG-receive2-AORII ‘(I) received’ (Harris 2002: 125) Both verb-internal positions are blocked by the same TAM properties that trigger verb-final placement in (18). Whereas the structures in (22) contain a verb-internal person marker, the corresponding future forms in (23) must surface with a verb-final one. As alluded to before, then, the verbal attachment of Udi person markers is dependent on the (inflectional) tense features of the verb. (23) a. aš-b-al-le work-do-FUTII-3SG ‘s/he will work’ (Harris 2002: 123) b. aq’-al-le receive1-FUTII-3SG ‘(I) received’ (Harris 2002: 126) Summing up, there is evidence showing that Udi person markers in (17) select a focused constituent. In line with Harris (2002) and Anderson (2005), we therefore take the view that Udi clitics, in those contexts, attach to a phrasal host. However, there is also evidence indicating that Udi clitics, in verb-final and verb-internal position, behave like word-level affixes. The presence of phenomena such as a) intermorphemic and intramorphemic placement (cf. 22), b) clitic allomorphy (cf. 20) and c) clitic induced stem allomorphy (cf. 21), can only receive an insightful account if person markers in (18) to (23) combine with the verb in the morphology. 3.4. Summary As alluded to before, the distinction between phrase-level affixes and word-level affixes, originally introduced by Klavans (1980), accounts for two major types of clitic system: on the one hand, the phrasal system found in languages like Serbo-Croatian, Kwakwala, Nganhcara or Ngiyambaa (Klavans 1980, Anderson 1996, 2005); on the other, the word-level system observed in most Romance languages and Macedonian, where highly selective clitics form a morphologically cohering unit with the word-level host (Monachesi 1999, Miller and Sag 1997, Spencer 1999).

Patterns of clitic placement

27

Mixed clitic systems constitute a third type. They are found in languages where clitics behave partly like phrase-level affixes and partly like word-level affixes. The marked feature of EP pronominal clitics and Udi person markers resides in their ability to select either a morphological host or a syntactic host, depending on inflectional properties of the verb or on syntactic properties of the clause. These mixed clitic systems have received much less attention in the clitic literature, but the empirical evidence so far provided in this chapter indicates that they do effectively exist. In recent work, Anderson (2005) explicitly argues that clitics in EP and Udi attach uniformly to a phrasal host. The obvious problem with this claim is that, in some contexts, clitics in both languages exhibit typically morphological properties that cannot be derived post-syntactically. Phenomena such as a) selectivity, b) adjacency, c) clitic allomorphy and d) stem allomorphy indicate that there is insufficient empirical evidence in favour of the phrasal attachment of verb-internal and verb-final clitics in EP and Udi.

4. An inflectional approach to mixed clitic systems The analysis proposed in this section treats EP pronominal clitics as inflectional formatives with the ability to attach either to a phrasal node or to a verbal stem. Our analysis is couched within an extended version of Stump’s (2001) theory of Paradigm Function Morphology (PFM), and is proposed in Luís and Spencer (2005a) and Spencer (2004, ms.). In Stump’s original PFM model, a paradigm function takes a pairing of a root and a set of features and delivers a fully inflected word form. In general, the paradigm function is defined in terms of a sequence of realization rules which add successive affixes to the root. We can think of the paradigm function as a set of assembly instructions for word forms. In PFM, the realization rules take the general form X ⇒ X-suffix/prefix-X, where the ‘X’ can be the lexeme’s root or any intermediate affixed form. In the Generalized Paradigm Function Morphology (GPFM), we separate out two distinct aspects of the realization rule, exponence (a definition of the form of each affix) and placement (what kind of stem it attaches to and in what direction). We code this idea by reformulating the notion of paradigm function. This is now a mapping from a pairing of the full representation of the lexeme and a set of features. The paradigm function defines the word form of a given lexeme corresponding to that feature set. In

28

Ana R. Luís

the extended model the paradigm function itself is split into four subfunctions, as in (24). (24) Generalized Paradigm Function: (i) Domain (ii) Host (iii) Exponence (iv) Linearization The modification of the paradigm function given in (24) allows us to collapse word-level affixation with phrasal affixation. Simplifying somewhat, in the case of word-level affixation: the Domain is the word and the Host is some appropriate stem or base; the subfunction Exponence realizes inflectional formatives and Linearization places the formative to the left or right of the host. In the case of phrasal affixation, the subfunction Domain determines the phrase or prosodic category with respect to which placement takes place and the Host specifies where within that domain attachment is made. The clitic cluster is defined under (24.iii) and linearized under (24.iv). A simple illustration of the difference between the two modes of attachment is given in (25) for word-level affixation (e.g., enclisis in EP) and in (26) for phrasal affixation (e.g., proclisis in EP): (25) PF for vê-me ‘s/he sees me’ PF () = Domain: Iº/Vº Host: verb stem Exponence: vê; me Linearization: vê-me (26) PF for me vê ‘s/he sees me’ PF() = Domain: VP Host: Vº/VP Exponence: vê; me Linearization: [me [Vº]…]VP

Patterns of clitic placement

29

Simplifying again, in both (25) and (26), the morphosyntactic features associated with the lexeme are given in curly brackets: 3rd person singular present tense and accusative 1st person singular pronominal object. The subfunction Exponence applies to this set of morphosyntactic features yielding the stem vê and the pronominal affix me. The derivation of enclitics and proclitics in (25) and (26) differs with respect to the subfunctions that are responsible for defining the Domain, the Host and the Linearization of the pronominal clitics. By default, the affix me is aligned to the right of the stem, as in (25). On the contrary, in proclitic contexts, the affix is positioned as a phrasal affix to the left of the Vº dominated by VP. Inside the morphological component, proclisis is induced by the markedness feature {Restricted: yes}, which is contained in the morphosyntactic feature set associated to the lexeme in (26) but not in (25) (for details, see Luís and Otoguro 2005).9 The analysis sketched above shows that the Generalized Paradigm Function Morphology model, with its extended notion of paradigm function, can capture the mixed properties of the EP clitic systems by allowing a single morphological formative to be linearized with respect to either a morphological host or a syntactic host.10

5. Conclusion A full contribution to typology depends largely on our ability to identify the rich variation of phenomena that are present in the languages of the world. In the process of identifying such phenomena, it is important not to assign identical labels to phenomena that are in fact distinct (Corbett 2005). In this chapter, we have tried to clarify the properties of mixed clitic systems in two genetically distinct languages, namely European Portuguese and Udi. In the literature, mixed clitic systems have gone largely unnoticed. However, we have shown that they effectively exist: pronominal clitics in EP and person markers in Udi select either a word-level host or a phrasal host. Our account of the data elaborates on the distinction between head affixes and phrasal-affixes, formulated originally by Klavans (1980, 1985). We have argued that the difference between enclitics and proclitics in EP, on the one hand, and between verb-final and focus-oriented clitics in Udi, on the other, must be analyzed as difference in status between the wordlevel and phrase-level attachment of the same set of pronominal affixes or person markers, respectively.

30

Ana R. Luís

Although we support the claim that clitics are inflectional formatives, throughout this chapter we have also taken the view that Anderson’s (2005) theory of clitics is empirically untenable. It is based on the underlying assumption that EP and Udi clitics attach uniformly to phrasal domains. The problem with this view, for EP and Udi, is that it fails to explain why postverbal clitics and preverbal clitics in EP differ so strongly with respect to phenomena such as stem allomorphy (cf. Section 2.2), scope and interpolation (cf. Section 2.3). The phrasal affix approach also leaves unexplained the fact that Udi clitics in verb-final position interact morphophonologically with the verb, unlike focus-oriented clitics (cf. Section 3.3).

Notes 1.

2. 3.

4.

5.

6.

I would like to express my deep gratitude to Andrew Spencer and Ryo Otoguro for suggestions during the various stages of this study. Thanks also to an anonymous reviewer, Pattie Epps and Timur Maisak for comments on an earlier draft. The work reported here was conducted with the support of the AHRB/British Academy, the Portuguese Foundation for Science and Technology (FCT) and the University of Coimbra. All errors are my own. A wider range of ‘special clitics’ can be found in Klavans (1980), Spencer (1992), and Halpern (1998). In this study, the term ‘phrasal affixation’ is not synonymous with ‘edge inflection’. We assume that edge inflections are morphologically associated with the host, even though they tend to appear at the edge of a phrase. Typical edge inflections include Bulgarian and Macedonian determiners and the English possessive ’s (Halpern 1995). For the sake of completeness, the claim that special clitics are phrasal affixes is not only problematic for languages with mixed clitic systems, but also for languages in which clitics attach exclusively to a word-level host (e.g., most Romance languages and Macedonian, among other). As alluded to before, the problem posed by Anderson’s theory resides in its inability to accommodate the word-level attachment of clitics. On this topic, see Bermúdez-Otero and Payne (forthcoming). Abbreviations in glossing: ABSL absolutive, ACC accusative, AOR aorist, AUX auxiliary, COND conditional, DAT dative, ERG ergative, FEM feminine, FUT future, IMP imperative, IMPF imperfect, INF infinitive, INST instrumental, LV light verb, MASC masculine, NEG negative, PAST past, PL plural, PPT past participle, PRES present, REFL reflexive, SG singular, SUBJV subjunctive Anderson’s phrasal approach is also committed to the view that the morphosyntactically triggered allomorphy surveyed in Section 2.3 is derived at the

Patterns of clitic placement

31

phrasal (or post-lexical) level. This view, which was formulated for EP by Vigário (2003), raises one serious problem: since enclitic allomorphy is triggered by highly specific grammatical contexts, it is not entirely obvious how phrasal phonology can have access to both categorical and morphosyntactic information (on this issue, see Luís 2006). 7. We illustrate the mixed properties of Udi person markers with data from two dialects: the Vartashen/Oktomberi dialect of Udi, documented in Harris (2002), and the Nizh dialect of Udi, provided to us by Timur Maisak (p.c.). We would like to thank Timur Maisak for drawing our attention to the morphophonological processes attested in the Nizh variety. 8. For a detailed description of both intermorphemic and intramorphemic clitic placement, see Harris (2002). On the diachronic origins of intramorphemic placement, see Harris (2002) and Luís and Spencer (2005b). 9. An analysis of the mixed placement of Udi person markers is sketched in Luís and Spencer (2005b). 10. For reasons of space, we leave the verb-internal placement of EP pronominal affixes unaddressed (cf. Section 2.2). However, in Luís (2004), we account for the fact that the pronominal suffix can precede the Future and Conditional suffixes by allowing the pronominal suffix to attach to an infinitival (morphemic) stem (Aronoff 1994). See also Luís and Spencer (2005a) for an analysis within Generalized Paradigm Function Morphology.

References Anderson, Stephen R. 1992 A-Morphous Morphology. Cambridge: Cambridge University Press. 1996 How to put your clitics in their place. The Linguistic Review 13: 165–191. 2005 Aspects of the Theory of Clitics. Oxford: Oxford University Press. Aronoff, Mark 1994 Morphology by Itself. Cambridge, Mass.: The MIT Press. Corbett, Greville G. 2005 The canonical approach in typology. In Linguistic Diversity and Language Theories, Zygmunt Frajzyngier, David Rood and Adam Hodges (eds.), 25–49. Amsterdam: John Benjamins. Crysmann, Berthold 2002 Constraint-based coanalysis. Ph.D. diss., Universität des Saarlandes & DFKI Gmbh.

32

Ana R. Luís

Bermúdez-Otero, Ricardo, and John Payne forthc. There are no clitics. In Morphology and its interfaces, Alexandra Galani, Glyn Hicks, and George Tsoulas (eds.). Amsterdam: John Benjamins. Halpern, Aaron 1995 On the Placement and Morphology of Clitics. Stanford: CSLI publications. 1998 Clitics. In The Handbook of Morphology, Andrew Spencer and Arnold M. Zwicky (eds.), 101–122. Oxford: Blackwell Publishers. Harris, Alice C. 2002 Endoclitics and the Origins of Udi Morphosyntax. Oxford: Oxford University Press. Klavans, Judith 1980 Some Problems in a Theory of Clitics. Bloomington: Indiana University Linguistics Club. 1985 The independence of syntax and phonology in cliticisation. Language 61: 95–120. Lapointe, Stephen 1980 A theory of grammatical agreement. Ph.D. diss., University of Massachusetts, Amherst. Legendre, Géraldine 2000 Positioning Romanian verbal clitics at PF: an Optimality-Theoretic analysis. In Clitics in Phonology, Morphology, and Syntax, Birgit Gerlach and Janet Grijzenhout (eds.), 219–254. Amsterdam: John Benjamins. Luís, Ana R. 2004 Clitics as morphology. Ph.D. diss., Department of Language and Linguistics, University of Essex. 2006 Para uma (re)definição da sufixação no Português Europeu: a adjunção prosódica do enclítico. Talk given at the Annual Meeting of the Associação Portuguesa de Linguística Portuguesa, October, University of Coimbra. Luís, Ana R., and Ryo Otoguro 2005 Morphological and syntactic well-formedness: The case of European Portuguese proclitics. In Proceedings of the LFG-2005 Conference, Miriam Butt and Tracy Holloway King (eds.), 253–270, Stanford, CA: CSLI Publications. Luís, Ana R., and Louisa Sadler 2003 Object clitic and marked morphology. In Empirical Isues in Formal Syntax and Semantics 4, C. Beyssade, O. Bonami, P. Hofherr and F. Corblin (eds.), 133–153. Paris: Presses de l’Université de ParisSorbonne.

Patterns of clitic placement

33

Luís, Ana R., and Andrew Spencer 2005a A paradigm function account of ‘mesoclisis’ in European Portuguese. In Yearbook of Morphology 2004, Geert Booij and Jaap van Marle (eds.), 177–228. Dordrecht: Kluwer. 2005b Udi clitics: A Generalized Paradigm Function Morphology approach. In Essex Research Reports in Linguistics 48, Ryo Otoguro, Gergana Popova, and Andrew Spencer (eds.), 35–47. Colchester: University of Essex. Martins, Ana Maria 1994 Clíticos na história do Português. Ph.D. diss., University of Lisbon. McConvell, Patrick 1996 The functions of split-Wackernagel clitic systems: Pronominal clitics in the Ngumpin languages (Pama-Nyungan family, Northern Australia). In Approaching Second Position Clitics and Related Phenomena, Aaron Halpern and Arnold Zwicky (eds.), 299–331. Stanford: CSLI. Miller, Philip, and Ivan A. Sag 1997 French clitic movement without clitics or movement. Natural Language and Linguistic Theory 15 (3): 573–639. Monachesi, Paola 1999 A Lexical Approach to Italian Cliticization. Stanford: CSLI. Sadock, Jerrold M. 1991 Autolexical Syntax: A Theory of Parallel Grammatical Representations. Chicago, University of Chicago Press. Spencer, Andrew 1992 Morphological Theory. Oxford: Basil Blackwell. 2001 Verbal clitics in Bulgarian: A Paradigm-Function approach. In Clitics in Phonology, Morphology and Syntax, Birgit Gerlach and Janet Grijzenhout (eds.), 355–386. Amsterdam/Philadelphia: John Benjamins. 2004 Generalized Paradigm Function Morphology – A synopsis. In York Papers in Linguistics, Alexandra Galani and Beck Sinar (eds.), (2): 93–106. ms. Generalized Paradigm Function Morphology, University of Essex. Available at http://privatewww.essex.ac.uk/~spena/papers/GPFM.pdf Stump, Gregory T. 2001 Inflectional Morphology. A Theory of Paradigm Structure. Cambridge: Cambridge University Press. Toivonen, Ida 2001 The phrase structure of non-projecting words. Ph.D. diss., Stanford University, CA.

34

Ana R. Luís

Vigário, Marina 2003 The Prosodic Word in European Portuguese. Berlin: Mouton de Gruyter. Zwicky, Arnold M. 1977 On Clitics. Indiana University Linguistics Club. Zwicky, Arnold M., and Geoffrey Pullum 1983 Cliticization vs. inflection: English n’t. Language 59 (3): 502–512.

Eton tonology and morphosyntax: A holistic typological approach Mark L. O. Van de Velde

1. Introduction1 This contribution shows how prosodic restrictions on the shape of verb stems have played a major role in structuring the morphosyntax and complex morphotonology of Eton, a Bantu language spoken in Cameroon. It proves to be impossible to provide an accurate description of the allomorphy of verbal suffixes and the behaviour of tonal morphemes without referring to two prosodic characteristics of stems, viz. initial prominence and a constraint specifying that stems can have maximally three (sometimes four) syllables. Moreover, it is argued, following Hyman (2004), that the same prosodic characteristics of stems are an essential ingredient of an explanation for the morphosyntactic differences between the Northwestern Bantu languages (such as Eton) and the other Bantu languages, and, more generally, among the languages of the Niger-Congo phylum. All this illustrates the usefulness of a holistic approach to language description, linguistic typology and reconstruction, in which explanations for one aspect of the grammar can be found at other levels of grammatical analysis. The discussion begins with a brief illustration of how systematic covariation between prosody and morphosyntax can explain the profound structural differences between languages of the Niger-Congo family (Section 2). It is then shown that Eton’s highly complex tone system is a result of the constant erosion of segmental material in the periphery of the word, i.e. everything before and after the prominent stem-initial syllable (Section 3). We will see that a fairly simple description of this tone system is possible if we take into account stem-initial prominence as well as the way in which ‘floating’ tones emerged. This leads to two interesting observations on tone in Eton. First, it is clear that syllables are the tone-bearing units in Eton, rather than moras or segments, but not all syllables behave the same way. Prominent syllables can accommodate two structural tones, one more than non-prominent syllables, irrespective of their segmental makeup. Second, tonal morphemes show the same degrees of morphological bonding

36

Mark L. O. Van de Velde

(word vs. clitic vs. affix) as morphemes with segmental form, even though they are clearly not prosodically independent. In Section 4, the chapter moves to a comparison of some morphosyntactic characteristics of Eton to those of the Central and Eastern Bantu languages, and explains the differences along the lines of the holistic approach outlined in Section 2. Three valency-changing operations are discussed: causative, applicative and passive. Finally, Section 5 briefly discusses what the study of Eton can (and cannot) contribute to the debate on the typological characteristics of the proto-language.

2. Co-variation of prosody and morphosyntax in the Niger-Congo languages There have been several claims in the literature that there is systematic covariation between phonology, morphology and syntax in the languages of the world (see Plank 1998 for a historical overview). As far as I know none of these claims has been tested on a large and balanced sample of the world’s languages, and to me it seems unlikely that any of them would stand the test. However, a holistic typological perspective may be useful in explaining profound and systematic typological differences between languages of the same family by showing that a change in one part of the grammar triggered changes elsewhere. For instance, the two main language groups of the Austro-Asiatic family, Munda and Mon-Khmer, are structurally divergent on every level of linguistic analysis: Munda has a synthetic morphosyntactic structure, MonKhmer an analytical one; word order is generally head-last in Munda and head-first in Mon-Khmer; there are prefixes and suffixes in Munda, but only (rare) prefixes in Mon-Khmer; Munda is agglutinative, Mon-Khmer fusional; Munda has trochaic word stress, Mon-Khmer iambic, etc. Although the Munda languages (the Indian subcontinent) and the Mon-Khmer languages (South-East Asia) are spoken in different linguistic areas where their respective characteristics are common, Donegan and Stampe (2004) argue that areal influence cannot explain their profound divergences. Rather, they claim that Proto-Austro-Asiatic was typologically similar to contemporary Mon-Khmer languages and that Munda underwent an independent drift to synthetic and head-last morphosyntax due to a shift from rising (iambic) to falling (trochaic) rhythm. In their view, falling rhythm goes together with head-last syntax and suffixing because of the back-

Eton tonology and morphosyntax

37

grounding of heads relative to their dependents and of affixes relative to their stems.2 Hyman (2004) gives a similar account of the profound typological differences among languages of the Niger-Congo family. He points out a correlation between the prosodic characteristics of stems, the presence versus absence of derivational morphology and the structure of the verb phrase. Languages that have maximality constraints on verb stems tend to disallow (stacking of) valency-changing affixes as well as the occurrence of multiple unmarked objects with one verb, in contrast to languages without such maximality constraints. The reason is quite simple: if a verb stem can have maximally two or three syllables, there is no room to accommodate multiple derivational suffixes. The structural differences between Niger-Congo languages can be illustrated by comparing the two extremes: the Eastern Bantu languages versus the Kwa languages. Example (1) from Makwe, a Bantu language of Mozambique, shows a verb stem with three derivational suffixes. As far as the number of derivational suffixes is concerned, this is not an extreme example with respect to Bantu norms. The base verb is monotransitive and the causative and applicative each add a nuclear argument, respectively an agent and an instrument. The passive suffix promotes the instrument to subject position and deletes the expression of the agent. The causer is left unmentioned. The result of a passivized verb with four nuclear arguments is hardly translatable in English.3 Makwe (Niger-Congo, Bantu; Devos 2008: 239) (1) ǹnáandi unkukóméshéléwá wáana mu-nandi u-nku-kom-ish-il-iw-a wa-ana 3-branch III-PROG-hit-CAUS-APPL-PAS-FV 2-child ‘The branch is used to make somebody hit the children.’ Asked to provide a translation of this example (without the causative element) in Ewe, a Kwa language of Ghana, Nada Gbegble provided two alternatives (2a–b). Note that Ewe does not have a passive construction (nor does it have other valency-changing morphology). The applicative argument in (1) has to be introduced by means of a preposition-like relator (2a), or a combination of two serial verbs (2b). (The word àtílɔ̀à ‘the tree branch’ in the examples in (2) is a literal translation and would not normally be used in this context.)

38

Mark L. O. Van de Velde

Ewe (Niger-Congo, Kwa; Nada Gbegble p.c.) (2) a. Wóƒò ɖèvìàwó kplé àtílɔ̀à wó-ƒò ɖèvì-à-wó kplé àtí-lɔ̀-à 3PL-beat child-DEF-PL with tree-branch-DEF ‘They beat the children with the tree branch.’ b. àtílɔ̀à wózã tsɔ́ ƒò ɖèvíáwóè àtí-lɔ̀-à wó-zã tsɔ́ ƒò ɖèví-á-wó-è tree-branch-DEF 3PL-use take beat child-DEF-PL-FOC ‘The tree branch they use to beat the children.’ Hyman (2004) assumes that Proto-Niger-Congo must have been typologically similar to the contemporary Eastern Bantu languages in allowing stacking of derivational suffixes and thus creating long verb stems that license multiple objects. The evolution towards Niger-Congo languages of the Kwa-type – where verb stems are monosyllabic, verbal derivation is quasi-absent and serial verbs are used to avoid multiple objects – thus involves mutually reinforcing prosodic, morphological and syntactic changes. Verb stems that are subject to constraints on the maximum number of syllables cannot accommodate a high number of derivational suffixes, which in turn negatively affects their ability to license causee, beneficiary or instrumental arguments without the help of additional relators such as adpositions and serial verbs. As in Donegan and Stampe’s scenario for Austro-Asiatic, prosodic factors trigger changes at other levels of analysis. The impact of prosodic restrictions on morphology can be observed in a group of languages that is typologically (and geographically) between Kwa and Eastern Bantu, and that comprises the Northwestern Bantu languages (spoken in Cameroon, Gabon, Equatorial Guinea, CongoBrazzaville and parts of Congo-Kinshasa). Consider the Congolese Bantu language Tiene as a rather spectacular example (Hyman, to appear, citing data from Ellington 1977). Verb stems are subject to a number of prosodic constraints, summarized below. The impact of these constraints on verbal derivation is illustrated in (3–4). Prosodic constraints on stems in Tiene: – stem shape: CV, CVV, CVCV, CVVCV, CVCVCV – if CVCVCV – C2 must be coronal – C3 must be non-coronal – C2 & C3 must agree in nasality – V2 is predictable

Eton tonology and morphosyntax

39

Tiene (Niger-Congo, Bantu; Hyman, to appear) (3) causative suffix -esa. mat-a ‘go away’ maas-a ‘cause to go away’ b. lab-a ‘walk’ lasab-a ‘cause to walk’ c. dím-a ‘become extinguished’ díseb-ɛ ‘extinguish’ d. lɛ ‘eat’ lees-ɛ ‘feed’ In (3a), the consonant of the causative replaces the second consonant of the root in order to avoid a coronal consonant in C3 position. In (3b), the causative morpheme is infixed in order to avoid a non-coronal second consonant and a coronal third consonant. The same happens in (3c) where, additionally, the second root consonant is denasalized in order to obtain agreement in nasality between the second and third stem consonant. Tiene (Niger-Congo, Bantu; Hyman, to appear) (4) stative suffix -ek- / -ala. yaat-a ‘split’ yatak-a ‘be split’ b. kab-a ‘divide’ kalab-a ‘be divided’ The stative suffix exemplified in (4) has suppletive forms to ensure compliance with the prosodic constraints on stems. The variant with a velar consonant is suffixed to roots with a coronal second consonant (4a), whereas the -al- variant is infixed in roots with a non-coronal second consonant (4b). Hyman (2004) gives a less spectacular, but for our purposes more revealing example from Koyo, a Bantu language spoken in CongoBrazzaville. Koyo has a constraint that specifies that stems can be maximally of three syllables. The examples in (5–6) show the combinations of two verb roots tá ‘see’ and bar ‘bite’ with the causative -is- and reciprocal -in- suffix. Koyo (Niger-Congo, Bantu; Hyman 2004) (5) a. tá-a ‘see’ b. tá-s-a ‘show’ c. tá-n-a ‘see each other’ d. tá-s-an-a ‘show each other’ (6)

a. bar-a ‘bite’ b. bar-is-a ‘cause to bite’ c. bar-in-a ‘bite each other’

40

Mark L. O. Van de Velde

d. *bar-is-in-a ‘cause each other to bite’ e. *bar-in-is-a ‘cause to bite each other’ Here, it is shown that prosody dictates which morpheme combinations are possible and which are not. Two derivational VC-suffixes can be added to the CV stem in (5), but not to the CVC stem in (6). Many (perhaps all) of the Northwestern Bantu languages with maximality constraints have another characteristic of prosodic stems in common, viz. stem-initial prominence (see, e.g., Paulian 1975 for Küküya; Hyman 1989). This can be observed most easily in the phonotactic skewing of consonant phonemes. The inventory of consonants decreases as one moves away from the first stem consonant. Thus, Basaá has twenty-two consonant phonemes in C1 position, twelve in C2 position, six in C3 position and three in C4 position in stems (Hyman 2004: 81). In Eton, stem-initial prominence is manifested phonotactically, phonetically, phonologically and tonologically (see Section 3 for the latter). The occurrence of about half of the consonant phonemes is restricted to the onset of stem-initial syllables (C1). These are /ʤ/, /g͡b/ and /v/ as well as all voiceless consonants, i.e. /p, t, ʧ, k͡p, k, s/. As for vowels, some phonological oppositions are neutralized in prefixes (/ɔ/, /o/ and /e/ are absent and /ə/ and /ɛ/ are in complementary distribution), whereas in non-initial stem syllables only two structural vowels contrast, viz. |i| and |a|. In word-final position the latter is represented by a vowel identical to that of the first stem vowel, except if the first stem vowel is |i| or |u|. This can be illustrated by means of the agentive derivational suffix |-à| in (7). All Eton data are taken from Van de Velde (2008). (7)

a. kòm ‘make’, ŋ̀kòmò ‘creator’ b. sùm ‘labour’, ǹsùmà ‘labourer’

Phonetically, prominent syllables have an audibly longer onset consonant than non-prominent syllables. This consonant is not subject to the lenition rules that apply to non-C1 consonants in intervocalic position. The prominent syllable is underlined in the structural representations in (8), where it can be seen that a /d/ is realized as [r] in a non-prominent position and as [d] in a prominent position, in an otherwise identical phonological context (tone never has any influence on the realization of consonants). (8)

a. |à-H-dàŋ|
/àdâŋ/
[àdâŋ] ‘he crossed’ b. |kádá|
/kádá/
[kárá] ‘a crab’

Eton tonology and morphosyntax

41

The co-occurrence of maximality constraints and stem-initial prominence raises the question of whether there is a causal link between the two, i.e. whether stem-initial prominence provides some kind of centripetal force (that may incidentally also explain the attrition observed in verbal prefixes in Eton and other Northwestern Bantu languages; see below). It would be interesting to check, in a large Niger-Congo sample, whether there are correlations between stem-initial prominence (as manifested in phonotactic skewing, among other things), stem or word length, and analytic morphosyntax.4 In what follows, I assume that this general correlation is the case and that stem-initial prominence causes attrition in non-prominent syllables. Constant attrition may have led to the emergence of maximality constraints. From this perspective, I will first describe and explain the tone system of Eton (Section 3), and then discuss some morphosyntactic differences between Eton and the better-known Eastern Bantu languages (Section 4). Eton stems have maximally three syllables (four in the presence of the inflectional suffix -ə̀ŋgànà).5 As has been noted, the first syllable of every stem is prosodically prominent. All non-prominent syllables, i.e. those following and preceding stem-initial syllables, are possible targets of attrition. The resulting loss of segmental material creates a considerable number of floating tones, which need to attach to the available segmental material. We will see in Section 4 that Eton hardly allows stacking of derivational suffixes, but that some of the current suffixes are combinations of older suffixes. Syntactically, Eton allows multiple unmarked complements with one verb, but has alternative periphrastic constructions as well.

3. Eton tonology: floating tones as traces of lost segmental material 3.1. Introduction The Beti-Bulu-Fang languages have tone systems that at first sight look complex, even according to Bantu norms. Eton syllables can carry a high tone (á), a low tone (à), a falling tone (â), a rising tone (ǎ), or a downstepped high tone (ꜜá). Some monosyllabic morphemes can occur with any of these surface tones, depending on the context. The examples with the terminative quasi-auxiliary mà (underlined) in (9) illustrate this. (9)

a. àmá ꜜdí ‘He ate it (earlier today).’ b. àŋgábé màgà tìl ‘He stopped writing (some time ago).’

42

Mark L. O. Van de Velde

c. àŋgámâ cág ‘He crushed it (some time ago).’ d. àté mǎ tìl ‘He stops writing.’ e. béꜜmágá ꜜbólí ‘They left it (yesterday).’ Descriptions of the closely related language Ewondo have dealt with this complexity in different ways. Redden’s description (1978) appeared at a time when the theoretical study of tone was still young, and does not provide an analysis of the tone system. Instead, Redden provides entire paradigms for a number of TAM forms, enumerating the tone pattern on every verb form (evidently for a very restricted number of paradigms and situations). The most recent description is Essono (2000), whose tonal analysis can be illustrated by means of the first person singular Hodiernal past perfective form of the verb làd ‘sew’ (Essono 2000: 512). (10) |mə́-làt-àgà|
/mə̀lâd/ ‘I sewed’ The passage from the structural to the phonological form involves a number of rules, including two tonological operations that Essono calls hypertonemes. One hypertoneme that applies in example (10) is called ´D in Essono’s description. It replaces the low tone of the stem by a high tone in this TAM-form. Another hypertoneme replaces the high tone of subject prefix mə́- by a low tone (hypertoneme À). The low part of the falling tone on the verb stem in (10) is a trace of the suffix -àgà which loses its segmental form in the derivation. In a number of notes Essono remarks that this analysis does not work for structurally high verb stems (which surface with a high tone, rather than the predicted falling tone), nor for verbs that are followed by a complement. Moreover, on the basis of parallels with Eton I suspect that this analysis makes the wrong predictions for di- and trisyllabic verb stems as well. The result of this approach is a highly complex, non-concatenative morphology with an enormous number of arbitrary tone rules such as replace the lexical low tone of a monosyllabic verb by a high tone in the Hodiernal past perfective if the verb is not followed by a complement. A much more simple and insightful analysis is possible by taking into account the characteristics of prosodic stems discussed in the previous section. First, the tonal complexity of Eton is due to a high number of floating tones that arise as a result of the erosion of segmental material due to prosodic weakening of non-prominent syllables, a phenomenon that includes maximality constraints. Floating tones arise from concatenative segmental morphology and can therefore best be analysed as morphemes

Eton tonology and morphosyntax

43

(affixes, clitics, words) that combine with stems in a certain linear order (see Section 3.3). Second, the distinction between prominent and nonprominent syllables is as relevant for tone as it is for segmental phonology, as shown in the next section. 3.2. Prominent syllables Prominent syllables can accommodate two structural tones, against only one for non-prominent syllables, irrespective of syllable weight. This is true even in a marginal stem shape such as CV.CVC (the dot marks a syllable boundary), where the first, light but prominent syllable can carry two structural tones, whereas the second, heavy but non-prominent syllable can carry only one. The Consecutive suffix -H in (11) illustrates this. When it follows a low, prominent syllable it adds to the low tone, forming a rising tone pattern (11a). But when it follows a low, non-prominent syllable, it deletes the original low tone and takes its place, resulting in a high tone (11b). (11) a. |à-kɛ̀-H|
/àkə̌/ ‘and he goes’ */àkə́/ b. |à-bɛ̀glɛ̀-H|
/àbɛ̀glə́/ ‘and he carries it’ */àbɛ̀glə̌/ The same principles can be observed with the high tone of the locative preposition á, which copies itself to the right. If it finds a low, prominent syllable, it adds to the low tone forming a falling tone (12a). In contrast, a following low tone on a non prominent syllable is delinked and pushed to the right (12b).6 (12) a. |á d-ùmá|
/á dûmá/ ‘in the nest’ */á dúꜜmá/ b. |á mə̀-ndím|
/á mə́ꜜndím/ ‘in the water’ */á mə̂ndi ́m/ 3.3. Origin and behaviour of floating tones Floating tones show where historical deletion of segmental material has occurred. They can be morphemes, such as the nominalizing suffix -H in (13), or parts of morphemes that also have segmental form, such as the floating low tones in the indefinite future prefix LŋgáL- in (14). They can be either H(igh) or L(ow).

44

Mark L. O. Van de Velde

(13) /ʤàb/ ‘be long’ |à-ʤàb-H|
/àʤǎb/ ‘length, height’ (14) /ŋgə́ bə́ꜜŋgáꜜpám/ |ŋgɛ́ bə́-LŋgáL-pám-H| if II-IFUT-arrive-IFUT ‘if ever they will arrive’ This section shows how the origin of floating tones determines their behaviour. More specifically, it will be shown that morphological structure is remarkably preserved when segmental material is lost and only tones remain. Thus, understanding the prosodic mechanisms that lead to the emergence of floating tones helps describe and explain their behaviour. In order to prove this, we will first consider some examples of the behaviour of tones that are structurally linked to different types of morphemes. Eton has a rule that copies every structurally linked high tone followed by a boundary, as schematized in (15), where # symbolizes a word boundary, = a clitic boundary and - an affix boundary. Rules of high tone copy are not unusual in the Bantu languages (see e.g. Kisseberth and Odden 2003 for an overview of recurrent tone rules). (15) cv́# → cv́H# cv́= → cv́H= cv́- → cv́HThe situation that is of interest for our purposes is where the copied high tone is followed by a low, non-prominent syllable, because in that case the way the high tone attaches to the following syllable depends on the kind of boundary it has to cross. This will become clear in examples (16) to (18), where all prominent syllables are underlined. The morpheme í- in (16), called augment, is prefixed to nouns modified by demonstratives and relative clauses. Its high tone is copied to the right, where it crosses the prefix boundary and attaches to the gender 8 prefix bì-. The original low tone of this gender marker is deleted and replaced by the copied high tone. (16) a. /íbísíŋâ bí/ |í-bì-síŋà # bí| AU-8-cat VIII.DEM ‘these cats’

Eton tonology and morphosyntax

45

b. /íbílwâlì bí/ |í-bì-lɔ̀li ̀ # bí| AU-8-duck VIII.DEM ‘these ducks’ The examples in (16) also illustrate a rule that is similar to high tone copy, viz. high tone spread, which is responsible for the falling tone on the last syllable of /íbísíŋâ/ and the third syllable of /íbílwâlì/ in (16). High tone spread is a rule of perseverative tonal assimilation that results in a falling tone on a structurally low syllable that is preceded by a high syllable within the same word. Note that due to high tone spread falling tones can surface on non-prominent syllables. The underlying tone is low in such cases and most speakers perceive it as such, since they accept a low pronunciation in very slow and careful speech. Now consider the examples of connective (genitive) constructions in (17). The high tone of the connective proclitic is copied to the right, across a clitic boundary, where it replaces the original low tone of the prefixed gender 7 marker ì- of the modifying noun. The resulting tone pattern on ìsíŋâ ‘cat’ in (17a) is identical to that in (16a). The difference between an affix and clitic boundary becomes clear in (17b), where the stem of the modifying noun ìlwàlì ‘duck’ remains entirely low: a high tone that crosses a clitic boundary does not spread. (17) a. /bjǎ bísíŋâ/ |bj-ǎ # bí=ì-síŋà| 8-claw VIII.CON-7-cat ‘the claws of a cat’ b. /mə̀vùl mílwàlì/ |mə̀-vùl # mə́=ì-lɔ̀li ̀| 6-feather VI.CON=7-duck ‘the feathers of a duck’ Finally, a high tone copied across a word boundary does not replace the structurally low tone of a following non-prominent syllable, but ‘pushes’ it to the right. The result of this is visible only if the next syllable has a high tone, which is then downstepped, symbolized by a downward arrow. In (18) it is the high tone of the locative preposition á that copies to the right.

46

Mark L. O. Van de Velde

(18) a. /á mə́ꜜʤɔ́ŋ/ |á # mə̀-ʤɔ́ŋ| LOC 6-hole ‘in the holes’ b. /á mə́ʤɔ̀ŋ/ |á # mə̀-ʤɔ̀ŋ| LOC 6-clan ‘in the clans’ All this can be summarized by the schemes in (19). (The choice for CV syllable shapes in these schematic representations is arbitrary. Syllable shape does not matter here.) (19) cv́ # cv̀-→ cv́ # cv́Lcv́=cv̀- → cv́=cv́- (no tone spread) cv́-cv̀ → cv́-cv́- (tone spread) The three morphemes that trigger high tone spread in the previous examples (16–18) all have tonal allomorphs. That is, for all these morphemes there are contexts in which they lost their segmental substance and kept their high tone only. Interestingly, these tonal allomorphs behave exactly like the copied high tone from their segmental counterparts when they attach to a following syllable. Thus, even in the absence of segmental material, morphological structure is preserved. In (20), the tonal allomorph of the augment that is selected before vocalic gender prefixes replaces the low tone of the following prefix, and then spreads onto the stem syllable in (20b). (20) a. /ɛ́ʤɔ́ŋ dî/ |H-ɛ̀-ʤɔ́ŋ # dì| AU-5-hole V.DEM ‘this hole’ b. /ɛ́ʤɔ̂ŋ dí/ |H-ɛ̀-ʤɔ̀ŋ # dí| AU-5-clan V.DEM ‘this clan’ Likewise, the tonal allomorphs of the gender 8 (21a) and gender 6 (21b) connective morphemes behave exactly like a high tone that is structurally linked to a connective proclitic with segmental substance that copies itself

Eton tonology and morphosyntax

47

to the right: they produce replacement of the low tone of a following prefix, without subsequent tone spread. (21) a. /bjǎ bísíŋâ/ |bj-ǎ # H=bì-síŋà| 8-claw VIII.CON-8-cat ‘the claws of the cats’ b. /mə̀vùl bílwàlì/ |mə̀-vùl # H=bì-lɔ̀li ̀| 6-feather VI.CON=7-duck ‘the feathers of the ducks’ Finally, the tonal allomorph of the locative preposition that appears before vowel-initial words behaves as predicted for a tone that crosses a word boundary. It does not delete the low tone of a following prefix, but pushes it to the right, resulting in downstep if the following syllable is high. (22) a. /ɛ́ʤɔ̀ŋ/ |H # ɛ̀-ʤɔ̀ŋ| LOC 5-clan ‘in the clan’ b. /ɛ́ꜜʤɔ́ŋ/ |H ɛ̀-ʤɔ́ŋ| LOC 5-hole ‘in the hole’ Exactly the same is true for tonal morphemes that do not have segmental allomorphs. By analogy, these can be divided into affixes, clitics and even words. An example of a tonal proclitic is the connective morpheme of gender 3 H=, which in no circumstances has a segmental form. It behaves exactly as the tonal allomorphs of the gender 5 and 8 connectives in (21). There is one example of a purely tonal morpheme that behaves as a word with respect to the tonal attachment rules in (19), the so-called link tone H. When it attaches to a low, non-prominent syllable, it delinks the original low tone and pushes it to the right. That is, it behaves as a high tone that attaches across a boundary between two prosodically independent words (recall that attachment across a prefix or clitic boundary would result in the deletion of the following low tone), whereas, obviously, this tonal element is not prosodically independent. The link tone appears be-

48

Mark L. O. Van de Velde

tween an infinitive and a following (pro)noun (23) within the same clause, whether this (pro)noun is a complement of the infinitive (23a) or not (23b). (23) a. /mə̀tə́ twàgdɔ̀ mə́ꜜndím/ |mə̀-Ltɛ́ L-tɔ̀gdɔ̀ H mə̀-ndím| 1SG-PR INF-boil LT 6-water ‘I’m boiling water.’ b. /bə́ꜜtə́ ꜜkwál ɲ́ꜜʧámá/ H |bə́-Ltɛ́ L-kɔ́là ɴ̀-Sámá| II-PR INF-talk LT 3-group ‘They are all talking at the same time.’ Some speakers even allow the link tone to occur between an infinitive and a following (pro)noun if there is no syntactic relation between both, as between the relative verb form àté sòm and the second complement bìbùmá in (24). Other speakers reject this. As for its function, this element does not mark anything and therefore it qualifies as a pseudomorpheme, or, given its tonal behaviour, a pseudoword (see Idiatov 2005: 74 for this notion). Since infinitives are nominal-verbal forms it is possible that this link tone developed out of a connective morpheme. However, in that case we would have expected it to behave as a proclitic. (24) /àté pàd ímôd àté sòm bíbùmá/ H í-ɴ-òd |à-Ltɛ́ L-pàd I-PR INF-pick LT AU-1-person ‘She picks fruit for the hunter.’

H à-Ltɛ́ L-sòm bì-bùmá| I-PR INF-hunt LT 8-fruit

Examples of purely tonal affixes, both prefixes and suffixes, abound. The Hodiernal past perfective, for instance, is marked by a prefix H- in between the subject prefix and the verb stem. Thus, the underlying representation for the Ewondo example in (10), which is identical in Eton, is as follows: (25) /mə̀lâd/ |mə̀-H-làd| 1SG-PST-sew ‘I sewed.’ Comparing the analysis of (25) to that of (10), it is clear that we do not need a set of complex, arbitrary tonal replacement rules. The H- prefix

Eton tonology and morphosyntax

49

makes the right predictions in all contexts, including when the verb stem is high (26a) and/or disyllabic (26b). Moreover, it signals that the past perfective was formerly marked by a segmental prefix with high tone that lost its segmental substance. (26) a. /ɛ̀lád/ |ɛ̀-̀ H-lád| V-PST-be_sticky ‘It was sticky.’ b. /mə̀swâlbɔ̀/ |mə̀-H-sɔ̀lbà| 1SG-PST-hide_oneself ‘I hid.’ To summarize, the complex and seemingly non-concatenative surface tonology of Eton is the representation of simple, concatenative underlying morphology. That much of this morphology is tonal is due to attrition of the segmental form of non-root morphemes caused by maximality constraints and, more generally, the centripetal force of stem-initial prominence. Synchronically, stem-initial prominence determines the way floating tones attach to the available segmental material. Morphological structure is surprisingly well preserved in the process of attrition. As far as I know this has not been shown for other tone languages, which might be due to the rarity of the phenomenon, or, more likely, to the descriptive models used.

4. Morphosyntax This section discusses three valency changing or rearranging categories that are expressed by highly productive, stacking verbal suffixes in the Eastern Bantu languages and that are reconstructed as such in Proto-Bantu (see Meeussen 1967), viz. causative, passive and applicative. The Makwe example in (1) above contains an instance of each of these suffixes in one, complex verb stem. The aim of this section is to illustrate the effects of maximality constraints on this typical aspect of Bantu languages: loss and rearrangement of the proto-suffixes, loss of productivity of the derivational suffixes and gradual replacement by periphrastic constructions, and loss of the suffixes’ ability to stack. The role of maximality constraints is most relevant in a comparative perspective, but their influence can also be ob-

50

Mark L. O. Van de Velde

served synchronically, as will be shown with respect to the choice of allomorphs of the passive suffix. 4.1. Causatives Eton has two causative suffixes, -à and -là. Whether -à or -là is used is mainly determined by the form of the base. It is not entirely clear whether these suffixes are reflexes of the two causative suffixes *-ici and *-i reconstructed for Proto-Bantu (see Meeussen 1967, Bastin 1986 and Schadeberg 2003 for these reconstructions). Nzang Bie (to appear), following leads by Janssens (1993) and Bastin (1986), argues that -à is a direct reflex of *-i, whereas -là is either the result of morphological reanalysis of an unstable root-final *d as part of the causative suffix, or a reflex of the Proto-Bantu applicative suffix *-ed. The argument is plausible, but such historical analyses are notoriously hard to prove. The causative suffixes are not very productive in Eton. Instead, causation is mostly expressed periphrastically, involving the verb kwàm ‘make’ and a subordinated form of the lexical verb. This can be either a participial verb form, e.g. mə́pâmgì in (27), or a subjunctive (28). In the latter case the causative clause is introduced by the complementizer nâ ‘that’. (27) /ùkwàngì bùnì, ìnə̌ pə̂ ńtwi ́ mə̀ʧǐ, íꜜtə́ kwàm mə̀ci ̌ mə́pâmgi/̀ L |ù-kɔ̀n-Lgì bùnì ì-nɛ̌ pɛ̀ ńtwí =mə̀-ʧǐ| 2SG-be_ill-G [9]amoebiasis IX-RCOP also diarrhea I.CON=6-blood L |í-Ltɛ́ -kɔ̀m mə̀-ʧǐ mə́-pám-Lgì| VII-PR INF-make 6-blood VI-come_out-G ‘You (are) suffering from amoebiasis, which is also (known as) diarrhea of blood, which makes the blood come out.’ (28) /ǹwúdnə́ŋgáná útə̂ wɔ̂ wákwâm nâ ìbǎb ítɛ̂g/ H |ɴ̀-wúdnə́ŋgáná ú-tɛ̀ -wɔ̋ ú-à-H-kɔ̀m| 3-massage III-ANA AU-III.PPR III-SP-REL-make H |nâ ì-bǎb -í-tɛ̀g-L| CMP 7-asthma SB-VII-weaken-SB ‘It’s this massage that weakens the asthma.’

Eton tonology and morphosyntax

51

4.2. Passives Part of Westermann and Bryan’s (1952) characterization of the Kwa languages is that they lack passive voice. This is opposed to Eastern Bantu languages, where passivization is highly productive. Eton is somewhere in between. It does have a passive suffix -bàn, but the passive is rarely used. The usual strategy for leaving the agent unmentioned is to use an impersonal construction with a third person plural subject prefix (29). (29) /bə́ꜜmá ꜜbó bɔ̂ bájàsálə̂m/ L H -bó bɔ̀ bájàsálə̂m| |bə́-H-mà-H II-PST-TMN-NF INF-chase LT PL retailer ‘They chased away the retailers.’ Moreover, -bàn can also have a reflexive (30) or a potential (31) reading with certain verbs. (30) /mə̀kɛ̂ŋbàn/ |mə̀-H-kɛ̀ŋ-bàn| 1SG-PST-shave-SF ‘I shaved.’ (31) /ùwònò àtə́ pùmbàn ávól/ H |ùwònò à-Ltɛ́ L-pùm-bàn à-vól| groundnut I-PR INF-uproot-SF LT 3-quick ‘Groundnuts harvest easily.’ The form of the passive suffix is remarkable from a Bantu perspective. It is not a reflex of the suffix reconstructed for Proto-Bantu, viz. *-o/*-ibo. Instead, it is a combination of an element -b (possibly a reflex of the passive/middle suffix *-(a)b(e) suggested as a reconstruction for Northwestern Bantu, Schadeberg 2003: 78) and the original associative/reciprocal suffix -àn. Depending on the form of the verb stem, these two historical components can be used by themselves to mark the passive. The form of the passive suffix is -àn after CVC.CV stems, in order to avoid a succession of three consonants (32), and -b in front of the inflectional suffix ə̀ŋgàn(à), in order not to violate the maximality constraints (33). (32) /bàglà/ ‘conserve’; /bàgl-àn/ ‘be conserved’

52

Mark L. O. Van de Velde

(33) /kódô/ ‘leave’; /àtán úkód-b-ə̂ŋgàn/ ‘The village was left.’ This can be compared to the contemporary reciprocal suffix -ːnì, which is historically a combination of the original suffix -àn and (anticausative?) -ì. The vowel length symbol points out that the vowel of a CV-root is lengthened before the reciprocal suffix (34) and that tone spread is blocked (35). (34) /ʤù/ ‘forgive’; /ʤùùnì/ ‘forgive each other’ (35) /wúd/ ‘massage’; /wúdnì/ ‘massage each other’ *wúdnî The reciprocal suffix has the same form as the passive suffix after viz. -àn (36). This ambiguity between a passive and a reciprocal reading, which arose through prosodic constraints, may further diminish the frequency of use of the passive.

CVC.CV-stems,

(36) a. /kòglò/ ‘bite’ /kòglàn/ ‘bite each other’ or ‘be bitten’ b. /ɲwàmlɔ̀/ ‘tickle’ /ɲwàmlàn/ ‘tickle each other’ or ‘be tickled’ The -àn allomorph of the passive and reciprocal suffixes makes it prosodically possible to add these suffixes to the causative -là suffix. This is the only type of suffix stacking found in Eton. (37) a. /ʤɔ́ŋ/ ‘be hot’; /ʤɔ́ŋ-lɔ̂/ ‘heat’; /ʤɔ́ŋ-l-ân/ ‘be heated/heat e.o.’ b. /jɛ́gi/̀ ‘learn’; /jɛ́g-lâ/ ‘teach’; /jɛ́g-l-ân/ ‘be taught/teach e.o.’ The vocalic causative suffix -à, on the other hand, is deleted in the passive, so that there is no formal difference between the passive of the underived verb and that of the derived causative in these cases (38b). Note that prosodic constraints predict the deletion of the vowel of the causative suffix (CV.CV.CVC is not a possible syllable scheme for Eton verb stems), but they do not explain or predict the total elision of its low tone, which could have become floating (as happened in other suffixes that lost their segmental substance). Alternatively, causative -à could have left a trace in blocking high tone spread (as does the lost à in the reciprocal suffix -ːnì). The absence of a floating low tone or high tone spread blocking could

Eton tonology and morphosyntax

53

point to a dispreference for suffix stacking, even where prosodic constraints allow it. (38) a. /ɲúŋ/ ‘drink’; /ɲúŋâ/ ‘make drink, quench’ b. /ɲúŋbân/ ‘be drunk’ or ‘be quenched’ c. *ɲûŋbàn, *ɲúŋbàn 4.3. Applicatives The last derivational category I discuss is the applicative, reconstructed in Proto-Bantu as *-ed. The Proto-Bantu applicative suffix was lost in Eton (or perhaps reinterpreted as a causative suffix, see above). Contrary to the causative no new applicative suffix arose, which makes Eton very unBantu-like from a morphological point of view. Benefactives – to name only this type of applicatives – can be introduced by means of the preposition ású H=, which originates in a connective construction with the head noun àsú ‘face’ and the gender 3 connective proclitic H=. (39) /àkɔ́m dɔ̂ ású ɲə̂/ ású |à-H-kɔ̀m-H dɔ̋ I-PST-do-NF 5.PPR for ‘He did it for her.’

=ɲɛ̋|

H

III.CON=I.PPR

However, benefactive (40) and circumstantial (41) complements licensed by an applicative suffix in other Bantu languages are normally expressed without any additional marking in Eton. That is, the loss of morphology does not negatively affect the ability of verbs to license multiple unmarked complements in this case. In example (41) the speaker explains that if one digs up the root of a plant called wogzo-wogzo in order to make medicine with it, one has to be careful to dig up only a single root. Otherwise, a child will die because of your actions. (40) /báꜜjám bɔ̂ bi ̀di/́ |bə́-à-jám bɔ̋ bì-dí| II-SP-prepare II.PPR 8-food ‘They prepare food for them.’

54

Mark L. O. Van de Velde

(41) /ŋgə́ úꜜɲɔ́ŋ ŋ́kǎŋ ḿꜜpɛ́bə̂, ùjɛ̀mgi ̀ nâ i ́ꜜzá môd àwú ꜜwɔ́/ |ŋgɛ́ H-ù-ɲɔ̀ŋ-H ɴ̀-kǎŋ m̀ -pɛ́bɛ̀ ù-jɛ̀m-Lgì nâ ízǎ| if 2SG-take-CS 3-root III-other 2SG-know-G CMP sb_else’s |ɴ-òd à-à-wú wɔ̌| 1-person I-SP-die 2SG.PPR ‘If you take another root, you must know that somebody else’s child will die because of you.’

5. Diachronic remarks If it can be shown that there is some degree of systematic co-variation between levels of grammatical structure (prosody, morphology, syntax) within languages of particular families or even in languages generally, the explanation for it will be diachronic. Therefore, systematic co-variation between daughter languages naturally leads to questions about the typological characteristics of the protolanguage. Hyman (2004) argues that Proto-Niger-Congo must have been similar to the contemporary Eastern Bantu languages, which allow strings of derivational suffixes to attach to verb stems thanks to the absence of maximality constraints. The changes towards the Kwa prototype were triggered by increasingly stronger maximality constraints on stems and can still be observed in a number of Benue-Congo languages, including Northwestern Bantu languages such as Eton, which are considered to be both typologically and diachronically somewhere halfway between the proto-situation and the Kwa type. Indeed, the previous sections show that Eton can be characterized as an intermediate type between the Kwa and the Eastern Bantu prototypes: – stems have maximally three (sometimes four) syllables, as compared to one (or two) in languages of the Kwa type and no maximal number in the Eastern Bantu type – in contrast to the Kwa prototype there is derivational morphology, but it is less productive than in the Eastern Bantu type and suffix combinations are marginal – syntactically, a verb can license multiple unmarked complements, but there are also periphrastic or prepositional alternatives

Güldemann (2007) argues, contra Hyman (2004), that Proto-Bantu and Proto-Niger-Congo must have resembled the Northwestern rather than the Eastern Bantu languages, pointing out, among other things, that Proto-

Eton tonology and morphosyntax

55

Bantu originates in the so-called Macro-Sudan belt, an area where languages from different families are typologically similar to the Northwestern Bantu languages. The evolution toward the Eastern Bantu and the Kwa prototypes would then involve respectively a weakening/loss and a strengthening of the maximality constraints. The advantage of this scenario is that maximality constraints are already present in the proto-language, so that their emergence does not need to be accounted for. The question I would like to briefly address in this section is this: In the absence of a large and careful comparative study, what can the study of one language add to this debate? The following observations are relevant. First, in Eton there are about as many floating tones in the pre-stem domain as within the stem, mostly in verbal inflectional morphemes. Judging by the number of structurally floating tones in prefixes it can be concluded that the pre-stem domain is at least as prone to attrition as are stem-final syllables. Maximality constraints on stems cannot explain the loss of segmental material in prefixes. Therefore, it seems that stem-initial prominence constitutes a centripetal force that triggers attrition in non-prominent syllables and that maximality constraints are merely a symptom of this. Second, if Eton were evolving from an Eastern Bantu to a Kwa-type grammar, we might expect to find two types of traces of productive suffix stacking: (i) existing suffix stacking where prosodic constraints allow it and (ii) tonal allomorphs of suffixes where stacking of segmental suffixes is prosodically impossible. However, the only more or less productive suffix stacking in Eton is the causative-passive combination described in Section 4.2. Moreover, none of the verbal derivational suffixes has a floating tone allomorph that appears in circumstances where a segmental allomorph would have violated maximality constraints. The presence of such tonal allomorphs would provide evidence for the hypothesis that Eton lost or is losing the ability to stack multiple derivational affixes. Their actual absence does not prove the opposite, however. The loss of the applicative suffix could be interpreted as a symptom of morphological change in the direction of the Kwa prototype under the pressure of maximality constraints. But there may be other explanations for this loss. The functions of an optional applicative7 are typically (i) marking discourse prominence of non-nuclear terms and (ii) making non-nuclear terms accessible to passivization, relativization, agreement, immediate postverbal position, etc. The second function, which amounts to giving ‘object properties’ to non-nuclear terms, is fundamentally a manifestation of the first function: passives are a means to ensure topic continuity, object agreement goes together with high topicality and the immediate postverbal

56

Mark L. O. Van de Velde

position often has a special discourse function. The reason that Eton does not have applicatives, nor objects clearly definable in terms of a converging set of object diagnostics, can therefore be sought in the fact that it marks discourse prominence differently from Eastern Bantu languages, viz. by means of discourse particles, emphatic pronouns and fronting (see Van de Velde to appear). Eton shares this feature with many other western Niger-Congo languages. Third, there are no indications that the grammar of Eton is evolving towards the Kwa prototype. We saw evidence of erosion, but there is also evidence of renewal. For instance, there is an interesting example of an auxiliary that is apparently in the process of morphologization (i.e. an evolution towards stronger morphological bonding), leading to a gradual loss of syllabic prominence, viz. the present tense auxiliary Ltə́ illustrated in (42). This auxiliary is a grammaticalization of the resultative verb tɛ́lə̂ ‘stand, be upright’. In Southern dialects of Eton, which are closer to Ewondo, this auxiliary has a present progressive reading, which contrasts with a simple present marked by the prefix à- (also the present tense prefix in Ewondo). The initial floating low tone in the auxiliary is a trace of the original tense prefix. The L- infinitive prefix is the trace of a nominal gender marker. The grammaticalization of Ltə́ must be relatively recent, because none of the closely related Beti-Bulu-Fang languages has it. (42) |bə́-Ltɛ́ L-dí|
/bə́ꜜtə́ ꜜdí/ ‘they are eating’ In the dialects I recorded, this form clearly consists of a prominent syllable, judging on its tonal behaviour and its initial voiceless consonant. Also, morphosyntactically it is an auxiliary rather than a tense prefix, since pronominals and adverbs can stand between it and a following lexical verb. However, my consultants reported that some speakers pronounce this morpheme as [rə́], with a voiced and lenited (and hence non-prominent) onset consonant. Clauses I produced in which this variant was followed by a pronominal or an adverb were consistently rejected, showing that the [rə́] pronunciation implies morphological reinterpretation as a prefix. This and similar exemples of renewal shows that the existence of attrition due to maximality constraints does not have to lead to a Kwa-type grammar. Eton does not provide clear evidence for either of the two diachronic scenarios. Rather, we need more comparative evidence based on more descriptions of individual languages that pay due attention to the role of prosody in shaping other aspects of the grammar.

Eton tonology and morphosyntax

57

6. Conclusion It was shown that two logically unrelated aspects of the grammar of Eton, the complex tone system and the relatively analytic morphosyntax (as compared to Eastern Bantu), can be attributed to the same underlying organizing principle: stem-initial prominence and the ensuing prosodic constraints on stems. In the case of the tone system, recognizing the importance of stem-initial prominence is essential for a description of the system. It is well known that the constant erosion of segmental material is responsible for the high number of floating tones in the languages of the region, but as far as I know it has never been suggested in the analysis of other languages that morphological structure is so well preserved in the process. We saw in Section 3 that with regard to their behaviour according to tone rules, tonal morphemes can be affixes, clitics or even words. This reflects their origin, at least in the cases for which we know their origin. In the other cases, the analysis of their tonal behaviour can provide evidence for their reconstruction. Thus, the holistic perspective on variation in the Niger-Congo phylum suggested in Hyman (2004) proves to be a useful framework for describing and explaining the grammar of Eton. Whether it is a valid explanation for the typological differences found in the Niger-Congo languages needs to be proved by means of a large comparative-historical study. On the basis of the available evidence it seems to me that it is valid, and that it therefore makes sense for typologists to look for systematic co-variation between levels of grammar.

Notes 1. 2.

3.

I would like to thank Pattie Epps, Larry Hyman, Dmitry Idiatov and two anonymous reviewers for their useful comments. It falls outside of the scope of this chapter to evaluate Donegan and Stampe’s proposal for the Austro-Asiatic languages. Its value depends to a large extent on whether a relation between rhythm and the order of meaningful elements can be proved to exist in a large and balanced sample. Note that their analysis rests on a number of generalizations, the validity and usefulness of which have been denied in the typological literature, such as consistent head serialization (e.g. Dryer 1992) and agglutination (Haspelmath, to appear). The following abbreviations are used: 1, 2, … overt marker of gender 1, 2, …; I, II, … prefix of agreement pattern I, II, …; APPL applicative, ANA anaphoric

58

4.

5.

6.

7.

Mark L. O. Van de Velde marker, AU augment, CAUS causative, CMP complementizer, CON connective proclitic, CS consecutive, DEF definite, FOC focus, FV final vowel (a multifunctional TAM morpheme), G inflectional suffix -Lgì, IFUT indefinite future, INF infinitive, LT link tone, NF non-final form, PAS passive, PL plural, PPR personal pronoun, PR present, PROG progressive, PST past, RCOP relative form of the copula, RECIP reciprocal, REL relative verb marker, SB subjunctive, SF suffix, SG singular, SP present (southern dialects), TMN terminative. There are other grammatical features that might correlate with these. For instance, within the Niger-Congo phylum long, complex verb stems can be found in the easternmost languages (Eastern Bantu) and in some of the westernmost languages (of the Atlantic family). These two groups also provide the only non-tonal Niger-Congo languages. Languages that are tonal in these groups often have accent-like tone systems in which tone is privative (high versus zero or low versus zero), whereas as far as I know tonal oppositions are equipollent in languages with maximality constraints. As for the typical noun class systems of the Niger-Congo languages, Creissels et al. (2008: 116) remark, “this particular type of nominal classification is mainly bound to a particular phylum – rather than a particular area – and within the area occupied by Niger-Congo languages, its most typical representatives are found among Atlantic languages and Bantu languages, i.e. in two areas very distant from each other”. Perhaps we can add where overt gender markers and agreement markers are not subject to attrition. This morpheme allows a stem to take an extra syllable in many of the Bantu languages with maximality constraints. Possibly it is (or was until recently) an enclitic, rather than a suffix. It was stated earlier that every first stem syllable is prominent in Eton. Note, however, that there are some mismatches between prosodic stems and morphological stems. A rather trivial one is that the onset of a prominent syllable can be a prefix if the morphological stem is vowel initial, as in (12a). Other examples can be found in borrowings and proper names. The initial syllable of the borrowing mə̀twâ ‘car’ (from Pidgin English motor or French moteur) is formally identical to the gender 6 prefix mə̀-. Morphologically it must be analysed as part of the stem, but prosodically it behaves as a prefix, as the following example shows: /mə̀twâ/ ‘the car’, /bɔ̀ mə̀twâ/ ‘the cars’, /á mə́twâ/ ‘in the car’, not */á mə̂twâ/. Optional applicatives are those for which there is an alternative, oblique instantiation for the applicative object, as opposed to obligatory applicatives, which are the only means available to express a certain type of term.

Eton tonology and morphosyntax

59

References Bastin, Yvonne 1986 Les suffixes causatifs dans les langues bantoues. Africana Linguistica 10: 55–145. Creissels, Denis, Gerrit J. Dimmendaal, Zygmunt Frajzyngier, and Christa König 2008 Africa as a morphosyntactic area. In A Linguistic Geography of Africa, Bernd Heine and Derek Nurse (eds.), 86–150. Cambridge. Cambridge University Press. Devos, Maud 2008 A Grammar of Makwe. München: LINCOM Europa. Donegan, Patricia, and David Stampe 2004 Rhythm and the synthetic drift of Munda. In The Yearbook of South Asian Languages and Linguistics 2004, Rajendra Singh (ed.), 3–36. Dryer, Matthew S. 1992 The Greenbergian word order correlations. Language 68: 81–138. Ellington, John 1977 Aspects of the Tiene language. Ph.D. diss., University of Wisconsin, Madison. Essono, Jean-J. Marie 2000 L’ewondo. Yaoundé: Presses de l’Université Catholique de l’Afrique Centrale. Güldemann, Tom 2007 Bantu in its macro-areal context of Africa and implications for the early typology of Bantu and Niger-Congo. Paper read at the International Conference on Bantu Languages, Gothenburg: October 4–6 2007. Haspelmath, Martin to appear An empirical test of the Agglutination Hypothesis. To appear in Universals of Language Today, Sergio Scalise, Elisabetta Magni, and Antonietta Bisetto (eds.). Berlin: Springer. Accessed at http://email.eva.mpg.de/~haspelmt/publist.html on 28.09.2008. Hyman, Larry M. 1989 Accent in Bantu: an appraisal. Studies in the Linguistic Sciences 19 (2): 115–134. 2004 How to become a Kwa verb. Journal of West African Languages 30 (2): 69–88. to appear Affixation by place of articulation: the strange case of Tiene. To appear in the Proceedings of the Rara and Rarissima Conference, MPI Leipzig. Accessed at http://www.linguistics.berkeley.edu/ people/person_detail.php?person=19 on 16.01.2007.

60

Mark L. O. Van de Velde

Idiatov, Dmitry 2005 The exceptional morphology of Tura numerals and restrictors: endoclitics, infixes and pseudowords. Journal of African Languages and Linguistics 26 (1): 31–78. Janssens, Baudouin 1993 Doubles reflexes consonantiques: quatre études sur le bantou de zone A (bubi, nen, bafia, ewondo). Ph.D. diss., Université Libre de Bruxelles. Kisseberth, Charles, and David Odden 2003 Tone. In The Bantu Languages, Derek Nurse and Gérard Philippson (eds.), 59–70. London: Routledge. Meeussen, Achilles Emile 1967 Bantu Grammatical Reconstructions. Africana Linguistica 3: 80–122. Nzang Bie, Yolande to appear La dérivation causative dans les langues du groupe A70. To appear in Africana Linguistica 14. Paulian, Christiane 1975 Le kukuya. Langue Teke du Congo. Phonologie, classes nominales. Paris. SELAF. Plank, Frans 1998 The co-variation of phonology with morphology and syntax: A hopeful history. Linguistic Typology 2–2:195–230. Redden, James 1979 A Descriptive Grammar of Ewondo. Carbondale: Department of Linguistics, Southern Illinois University. Schadeberg, Thilo 2003 Derivation. In The Bantu Languages, Derek Nurse and Gérard Philippson (eds.), 71–89. London: Routledge. Van de Velde, Mark L. O. 2008 A Grammar of Eton. (Mouton Grammar Library 46.) Berlin: Mouton de Gruyter. to appear The syntax of verb complements and the loss of the applicative in Eton (A71). Proceedings of the International Conference on Bantu Languages, Gothenburg. Köln, Rüdiger Köppe. Westermann, Dietrich, and Margaret A. Bryan 1952 Languages of West Africa. Handbook of African languages, Part II. London: Oxford University Press.

Part II. Case, agreement, and localization

A hierarchical indexation system: The example of Emerillon (Teko) Françoise Rose

1. Introduction This chapter deals with a particularly challenging facet of the Emerillon language: the person indexation system. Its most remarkable feature is the selection of the obligatory person index on transitive predicates out of two series. Transitive predicates allow only one person index, which is selected according to the relative position of the two arguments on both the person hierarchy 1/2 > 3 and the grammatical roles hierarchy A > P. I will argue that this system should be classified as a distinct type of indexation system, here termed ‘hierarchical’. Emerillon is a language spoken in French Guyana by about 400 speakers who call it Teko. It belongs to the Tupí-Guaraní family (e.g. Rodrigues 1984–1985), itself part of the Tupí stock. The Tupí-Guaraní family consists of about 40 languages spread out through Brazil and neighbouring countries and “is noted for a high degree of lexical and morphological similarity among its member languages in spite of their extensive geographical separation” (Jensen 1999). Person-indexation, often called cross-referencing, is one of the major themes of Tupí-Guaraní morphosyntax, which has been studied both within particular languages (Seki 1990 for Kamaiurá; Rodrigues 1990 for Tupínamba; Leite 1990 for Tapirapé; Harrison 1994 for Guajajara inter alia) and for the whole family (Jensen 1990; Payne 1994). As far as the Emerillon language is concerned, it was little described before my own work (Rose 2003a, 2008, to appear). However, the question of person indexation in Emerillon had been presented before in an article on Emerillon parts of speech (Couchili, Maurel, and Queixalós 2002). The present work goes further into the description of the indexation system, and takes a typological perspective. The chapter first presents the Emerillon person indexation system on verbs (Section 2). It then discusses its characterization as a hierarchical system (Section 3) and rejects the hypothesis that it is an inverse system. After considering these two points, I address the issue of the alignment

64

Françoise Rose

type to which Emerillon belongs. This study focuses specifically on the verb morphology and does not deal with nominal predication.1

2. Indexation in Emerillon The person markers used on verbs are divided in two main sets, often called Set I and Set II in the Tupí-Guaraní literature (see for instance Jensen 1998). Table 1. Emerillon sets of person indexes 1 SG 2 SG 1 INCL 1 EXCL 2 PL 3 indeterminate

Set I aeɾesi- 2 oɾopeoza-

Set II edenõde-/kõdeoɾe- / oɾonepe- / penei- 2 / ∅zo- / poɾo- 3

Set II indexes are also used beyond the verb morphology to refer to the possessor of nouns (1) and the object of postpositions (2).4 (1)

kob-nẽ nõde-kuɾu-o COP-CONTRAST 1INCL.II-manioc_beer-CONT ‘But our traditional drink (manioc beer) still exists.’

(2)

kob aʔe-kom-ãhã nõde-pe DEM-PL-only 1INCL.II-for COP ‘These things only exist for us.’

I will now turn to the specific distribution of person markers on intransitive and transitive verbs. An important preliminary point is that all verb forms obligatorily carry a person index, regardless of whether arguments are expressed as full nominal phrases or not.

Emerillon hierarchical indexation system

65

2.1. Intransitive verbs On intransitive verbs, only Set I is found, referring to S. (3)

si-mãnõ-taɾ 1INCL.I-die-FUT ‘We are going to die.’

2.2. Transitive verbs On transitive verbs, Set I is used for A and Set II for P. Table 2. Distribution of person indexes on transitive verbs Set I Set II

A x

P x

The interesting point about indexation on transitive verbs is that only one slot is normally available for person indexes. Whether the A or the P is marked depends on two different hierarchies. These hierarchies will be introduced when relevant in the following presentation of the three transitive scenarios: local (involving speech act participants only), non-local (involving third persons only) and mixed (involving a speech act participant and a third person). 2.2.1. Mixed scenarios Whenever a speech act participant and a third person interact, the person hierarchy is called into play. Speech act participants are higher than the third person on this scale. 1/2 > 3 Figure 1. Person hierarchy

The participant higher on the hierarchy is marked on the verb, whether A or P. In both examples below, a first person inclusive and a third person are involved. The first person being higher in the hierarchy, it is in both cases

66

Françoise Rose

the one to fill the person index slot, with a Set I index when it is A (4), and with a Set II index when it is P (5). (4)

nõde-baʔekwəɾ-a-we si-kuwa-gatu 1INCL.II-story-REF-also 1INCL.I-know-well ‘And we also know our story very well.’

(5)

apam-a-nẽ nõde-apisi-tanẽ stranger-a-CONTRAST 1INCL.II-massacre-DESID ‘The strangers wanted to massacre us.’

Likewise, in both examples below, a second person singular and a third person are involved. The second person being higher in the hierarchy, it is the one to be indexed on the verb, with a Set I index when it is A (6), and with a Set II index when it is P (7). (6)

mama-nẽ eɾe-ʤika ! Mum-CONTRAST 2SG.I-kill ‘You killed Mum!’

(7)

e-ʔu-taɾ-eʔe zawaɾ 2SG.II-eat-FUT-INTENS jaguar ‘The jaguar is for sure going to eat me.’

In all of these cases, the indexing of a speech act participant on the verb, whether as A or P, indicates indirectly that the other participant, i.e. P or A, is to be interpreted as a third person. 2.2.2. Non-local scenarios The person hierarchy does not specify any hierarchy between different third persons nor among speech act participants. These distinctions are informed by the grammatical roles hierarchy, at work whenever the person hierarchy is not relevant, i.e. between two third persons or two speech act participants.

Emerillon hierarchical indexation system

67

A > P Figure 2. Grammatical roles hierarchy

When a third person acts on another third person, the grammatical roles hierarchy requires the A to be systematically marked (with a Set I index), whatever the arguments may refer to. (8)

o-wɨ o-zika 3.COREF-mother 3.I-kill ‘He killed his mother.’

(9)

patu-pope o-iɲuŋ pot-in 3.I-put ‘She puts them (the sweet potatoes) in the pot.’

(10) aɾakapusa-uhu o-mõduɾ-oŋ gun-big 3.I-send-PL.S ‘Guns were shooting bullets.’ (11) o-peɾo-peɾog e-iba 3.I-RED-lick 3.II-pet ‘His dog licks his face.’

bal bullet

∅-owa 3.II-face

The o- prefix of Set I (marking either Ss or As) here indicates indirectly that P is also a third person: if P was a speech act participant, it – instead of A – would have to occupy the index slot due to its higher position on the person hierarchy. A consequence of these two hierarchies is that the i- prefix of Set II never occurs on verbs.5 It is nevertheless used in nominal phrases (as a possessive marker) and in postpositional phrases. 2.2.3. Local scenario 2→1 In Emerillon, the local scenarios (i.e. when the two participants are speechact participants, either 2→1 or 1→2) all show the A indexed on the verb, regardless of the person of the participants. The two local scenarios are therefore described as following the grammatical roles hierarchy A > P already mentioned for the non-local scenario.

68

Françoise Rose

Table 3 presents the four instances in which a second person (singular or plural) acts on a first person (singular or exclusive). In each case, A is marked with Set I on the verb, and the verb is followed by a pronoun (the meaning of which is discussed below). Table 3. 2→1 examples in Emerillon6 A P 2SG 1SG

Examples eɾe-nupã eɾeɲ 2SG.I-hit 2SG ‘You (SG) hit me.’ 2SG 1EXCL eɾe-nupã oɾone-kom 2SG.I-hit 1EXCL.II-PL ‘You (SG) hit us.’ 2PL 1SG pe-nupã peɲ 2PL.I-hit 2 PL ‘You (PL) hit me.’ 2PL 1EXCL pe-nupã oɾone-kom 2PL.I-hit 1EXCL.II-PL ‘You (PL) hit us.’

The presence of the independent pronominal form is necessary to disambiguate the 2→1 configuration from the configuration in which a second person acts on a third person, as in (12). (12) eɾe-nupã 2SG.I-hit ‘You (SG) hit him/her/it/them.’ Because, when 2→1, A is systematically marked on the verb by a Set I prefix, we would expect the pronoun following the verb to refer to P, but this is not always the case. In the second and fourth line of Table 3, oɾonekom, the normal free pronoun for first person exclusive, refers explicitly to P, as expected. Surprisingly, in the first and third line, the independent forms eɾeɲ and peɲ, used only in this specific scenario, are linked to second person pronominal forms rather than to first persons (Proto-TupíGuaraní pronouns for second person singular and plural are reconstructed as *eré/ene and *pe…ẽ, Jensen 1998). The system seems a priori illogical in synchrony: in order to refer to a second person A and a first person P, two markers for second person are used. The oddity of the marking of

Emerillon hierarchical indexation system

69

2SG/PL→1SG can be explained as a residue of a former 1 > 2 hierarchy, substituted for by the grammatical roles hierarchy A > P.7 2.2.4. Local scenario 1→2 When a first person acts on a second person, Emerillon uses two different structures according to the number of P. oɾo- is used when a first person (singular or plural) acts on a second person singular (13), and a-poɾo- when a first person (singular or plural) acts on a second person plural (14). (13) oɾo-ʔu-taɾ 1EXCL.I-eat-FUT a) ‘I will eat you.’ (and also: ‘We will eat you.’) b) ‘We will eat it/her/him/them.’ (14) a-poɾo-nupã-taɾ 1SG.I-INDET.II-hit-FUT ‘I/we will hit you all.’ (Lit. ‘I hit people’; spoken in a narrative by a father furious at his misbehaving daughters) The use of oɾo- in (13) is ambiguous. As a Set I marker for first person exclusive (cf. Table 1), it can refer to scenario (b), where a first person exclusive acts on a third person, but it can also refer to the local scenario (a), i.e. a first person (singular or exclusive) acts on a second person singular. Accordingly, when a transitive verb is prefixed with a Set I first person exclusive marker, this marker refers to a first person A, and P is inferred. In the absence of any NP referring to P, only context can solve the ambiguity between a second and a third person P. Interestingly, with a second person P, oɾo- neutralizes the number opposition of A. Indexation in (14) is made up of two morphemes: a- first person singular of Set I (here neutralized in terms of number), referring to an A, and poɾo- usually referring to a generic human P (cf. Table 1). The opaque encoding of the local scenarios (due to substitution of forms or semantic neutralization) has been discussed in comparative and historical perspective and with reference to politeness (in keeping with Brown and Levinson 1987) in earlier work (Rose 2003a, 2003b). In summary, my analysis of the Emerillon local scenarios is as follows:

70

Françoise Rose

– For 2→1, A takes priority in filling the index slot on the verb. The second person marker, following the verb, refers to P either directly (oɾonekom), or indirectly (eɾeɲ, peɲ). – For 1→2SG, oɾo- is analyzed as an A marker. – For 1→2PL, A is marked with Set I (a-), whereas P is marked with a human generic object (poɾo-).

My conclusion is that there is no obvious person hierarchy between the two speech act participants. The various local scenarios are better explained by the grammatical roles hierarchy, which also applies to non-local scenarios, than with any possible person hierarchy. Indeed, in all cases, A is indexed on the verb with Set I, while P is either unexpressed or expressed in a rather marginal way, i.e. as an ‘incorporated’ generic noun, or as a pronoun following the verb. The main points of the indexation system on Emerillon verbs are repeated below. – On intransitive verbs, S is marked with Set I. – On transitive verbs, either A or P is marked (with Sets I or II, respectively) according to their relative position on a person hierarchy or on a grammatical roles hierarchy. The 1/2 > 3 person hierarchy operates when only one speech act participant is involved, and the A > P grammatical hierarchy operates elsewhere.

The following section focuses on the use and organization of the different hierarchies operating in the Emerillon transitive constructions. For the sake of the coming discussion, we note that the indexation system presented below is quite comparable to the indexation system of the independent clauses in other Tupí-Guaraní languages, although differences can be noted in the local scenarios (cf. Rose 2007). In Emerillon, the same system also applies in dependent clauses.

3. Characterization of the hierarchical indexation system on transitive verbs The terms ‘person hierarchy’ and ‘grammatical roles hierarchy’ used here correspond roughly to other designations that emerged in line with Silverstein’s pioneering work on hierarchies of features (Silverstein 1976). The author highlighted the role of semantic properties of nominals on casemarking and agreement (more specifically in the domain of ergative or

Emerillon hierarchical indexation system

71

split-ergative systems). A recent label that subsumes all aspects (semantic, referential, discursive) of these hierarchies is the ‘Indexability Hierarchy’ (Bickel and Nichols 2007: 224–227). The two hierarchies used in Emerillon could logically be justified in terms of saliency, with the most salient participant being put forward. It is nevertheless important to assert that this system is completely grammaticalized: whatever the characteristics of the participants are, what counts in the system is the grammatical persons and the grammatical roles. While the person hierarchy attested in most Tupí-Guaraní languages is usually presented as 1 > 2 > 3, the Emerillon data point clearly at only a 1/2 > 3 hierarchy. This reorganization of the person hierarchy (cf. endnote 7), involving the neutralization of the hierarchy between first and second persons, can probably be attributed to politeness conventions, just like the substitution patterns detailed in the local scenarios. This is tightly correlated with the fact that languages tend to disfavour transparent marking of first and second person combinations (Heath 1998). This assertion is confirmed by the fact that both in Emerillon (1/2 > 3) and in the hypothetical Proto-Tupí-Guaraní (1 > 2 > 3), the marking is very clear and systematic when only one speech act participant is involved, but less so when both first and second person are involved (Montserrat and Soares 1983). This difference in treatment is common in languages involving a person hierarchy. A similar example is given by Gildea (1994) for the inverse system of Caribe. DeLancey (ms.) proposes the deictic nature of inverse and hierarchical systems as an explanation for the fact that these systems give a special status to speech act participants. Likewise, in typological perspective, first and second person are independent within the hierarchy, and their relative order fluctuates from one language to the other (Silverstein 1976, DeLancey 1981). This may help account for the fact that the Emerillon language succeeded in reorganizing the Proto-Tupí-Guaraní hierarchy concerning specifically the local scenario, neutralizing the hierarchy between the speech act participants. The grammatical roles hierarchy came into play to compensate for this change. On transitive verbs in Emerillon, the correct index is selected according to the relative position of the two arguments on both the person hierarchy 1/2 > 3 and the grammatical roles hierarchy A > P. Two hierarchies are thus involved, and Couchili, Maurel, and Queixalós (2002) propose to order them as follows:

72

Françoise Rose

1/2 > 3 person hierarchy

>

A > P hierarchy

Figure 3. Hierarchy of hierarchies

However, the authors consider that the person hierarchy is 2 > 1 > 3, and therefore the grammatical roles hierarchy is used only when both participants are third persons. The present analysis describes the person hierarchy as 1/2 > 3 and extends the scope of the grammatical roles hierarchy to the local scenarios: it is also outside the scope of the person hierarchy, and thus obeys the grammatical roles hierarchy. On the whole, the grammatical roles hierarchy becomes relevant when, and only when, the person hierarchy is not. I argue that this type of indexation system on transitive verbs should be classified as ‘hierarchical’, as the primary organizing pattern is the notion of hierarchy.8 A first explicit definition of hierarchical systems considered on a par with neutral, accusative, ergative, stative-active and three-way systems, is Nichols (1992). Section 5 will discuss further developments of this definition, in relation to the question of alignment type: Access to inflectional slots for subject and/or object is based on person, number, and/or animacy rather than (or no less than) on syntactic relations. (Nichols 1992: 66)

Languages identified as displaying a hierarchical system are Cree, Tepehua, Mixe, Nunggubuyu, Kiowa (Nichols 1992), Tangut and some TibetoBurman languages (DeLancey 2001). Since the existence of the person hierarchy in Tupí-Guaraní languages has led to the interpretation of these agreement systems as inverse systems, I now turn to this interesting question.

4. Against the inverse hypothesis It was Doris Payne (1994) who first formulated the hypothesis that TupíGuaraní languages could be described as having an inverse system. The author follows Givón’s definition of the inverse, based on the notion of an ‘inherent topicality hierarchy’ 1 > 2 > 3 (Givón 1994). If 1→2/3 or 2→3, the action flows in the natural direction (A is more topical than P). This flow is considered to be direct. If the action flows the other way around (when P is more topical than A), the flow is considered to be inverse. Fur-

Emerillon hierarchical indexation system

73

thermore, a canonical inverse language is a language that expresses the inverse direction with morphosyntactic devices in a transitive construction. Examples from Fox, an Algonquian language are given below, where -aa is a direct marker (15), and -ek an inverse one (16). Fox (Algic, Algonquian; Comrie 1981: 122) (15) ne-waapam-aa-wa 1SG-see-DIRECT-3 ‘I see him.’ (16) ne-waapam-ek-wa 1SG-see-INVERSE-3 ‘He sees me.’ Payne (1994) applies this definition to the Tupí-Guaraní system, organized around a 1 > 2 > 3 hierarchy. Direct situations are marked with Set I for A. They include some mixed scenarios (when A is superior to P on the person hierarchy) and the non-local scenario, since Set I is also used when two third persons are involved. Inverse situations are marked with Set II for P, and consist of the remaining mixed scenarios (when P is superior to A on the person hierarchy). The encoding of the local scenarios is not explicitly classified as direct or inverse.9 The author’s main argument for the inverse analysis is that the relational r- prefix, found in co-occurrence with Set II, be analysed as an inverse marker in Tupí-Guaraní languages. There would be no specific marker for direct. Following this analysis, the Wayampi example (17) below would be a direct construction, with a Set I index for first person A and no index for P. Example (18) would be an inverse construction, with no index for A but a Set II index referring to first person P, and an overt inverse marker r-. This hypothesis could likewise be applied to Emerillon, since the indexation system on independent verbs diverges with the rest of the family only in local scenarios. Wayampi (Tupian, Tupí-Guaraní; Payne 1994: 314–315) (17) namu r-a’y jĩ a-juka tinamou LK-immature only 1SG.I-kill ‘I killed just a little tinamou.’ (18) e-r-aty-pa e-ke pe e-r-uwy 1SG.II-LK-cover-COMPLE 1SG.II-sleep in 1SG.II-LK-blood ‘My blood completely covered me in my dream.’

74

Françoise Rose

However, a canonical inverse language is a language that expresses the inverse direction with an explicit morphosyntactic device. The relational morpheme that Payne suggests is an inverse marker in Tupí-Guaraní languages (the r- relational) is found throughout the family on a lexically defined subclass of transitive roots, but as well on certain intransitive roots, and some nouns and postpositions, when preceded with a Set II index or a full object/possessor (see for example the second occurrence of r- in example 18). My first counter-argument is that the syntactic distribution of r- is greater by far than is the function of the inverse category. It could reasonably be glossed as an ‘inverse marker’ only on transitive roots. This analysis would also not account for the unique function it serves throughout its distribution: the linking of a root with an immediately preceding complement (be it the P of a verb, the possessor or genitive of a noun or the object of a postposition). My second counter-argument is that r- is found only with certain lexically determined roots. Its absence with various other roots (in 5 and 7 for instance) is evidence that it is not essential to the system. Moreover, when present, it is redundant with the presence of Set II indexes, which are sufficient to refer to the whole participant scenario. In languages like Fox (illustrated in 15 and 16), the inverse marker is obviously necessary to sort out which of the two participants is A and which is P. I therefore consider that no morpheme can be interpreted as an inverse marker in Tupí-Guaraní languages. As a consequence, to make the inverse analysis tenable for this language family, one would have to be willing to accept that the distribution of person markers into two sets is in itself enough to constitute an inverse system. This assumption would be possible with a purely functional definition of inverse, like that of Givón (1994) or Klaiman (1991), which consider any situation where P is more topical than A, but A is still topical, to be inverse. This is in fact consistent with what T. Payne describes as “special verb agreement markers for inverse situations”, citing data from Wayampi, a close relative of Emerillon (Payne 1997). Therefore, if the Tupí-Guaraní indexation system, and the Emerillon system in particular, were to be described as an inverse system, it would not be canonical in relation to prototypical inverse systems (most notably of the Algonquian languages). First, direction of action is not expressed by a specific morpheme, since no morpheme can be interpreted as an inverse marker. Second, it is not a complete system, since the non-local scenario and the local scenarios are not involved; it is limited to the mixed scenarios, since a hierarchy would be lacking for the other scenarios. Therefore, although one can easily identify a function related to inverse in Tupí-

Emerillon hierarchical indexation system

75

Guaraní languages, there are insufficient grounds on which to consider this an inverse system. Positing person and grammatical role hierarchies is sufficient to explain the indexation system. This argument actually boils down to the confrontation between two different approaches. In the functional approach, inverse is to be found whenever P is more topical than A, but A is still topical (Givón 1994). In the syntactic approach, hierarchical indexation and the marking of direction (inverse) are independent (DeLancey 2001), as discussed by Zúñiga: Hierarchical alignment and direction are logically independent features that can, but need not, co-occur. A particular language may display verbal morphology that can be meaningfully described with the concept of hierarchical alignment alone, without there being direction marking. Similarly, the morphosyntax of a certain language may (i) be adequately described with a simple SAO model but (ii) allow for additional direction marking if there is no hierarchical alignment. (Zúñiga 2006: 28)

The so-called inverse systems are thus considered as a particular type of hierarchical systems (Nichols 1992, Siewierska 2004), or for DeLancey, both are expressions of deictic orientation (DeLancey 2001). In line with Heath’s argument (Heath 1998), I stand against the extension of the ‘inverse’ terminology to forms lacking an inverse marker, arguing that it actually undermines the usefulness of such a term. The Emerillon indexation system on transitive verbs is thus a plain hierarchical system.

5. Characterizing the alignment system of Emerillon Whatever the definition of alignment (either the way the arguments of a transitive predicate align with the unique argument of an intransitive verb, or the way grammatical relations map to grammatical roles), the hierarchical indexation system presented above can not be reduced to an alignment system. The relative position of arguments on the hierarchies does not assign them their grammatical roles. The system focuses on the relative saliency of the participants: We are used to thinking of verb agreement as tied to grammatical relations: a common claim about the typology of verb agreement is that if a language has verb agreement it will index the subject; some languages index both subject and object, and a rare handful index only objects. However, there are languages in which indexation of arguments in the verb reflects not gram-

76

Françoise Rose matical relations, but the person hierarchy. In these languages a verb will always agree with a SAP argument, regardless of its grammatical role. (DeLancey 2001)

This is where I want to draw a neat distinction between hierarchical indexation and hierarchical alignment. Hierarchical indexation systems simply rule access to inflectional slots on transitive predicates. Within the same languages, alignment can still be marked in other morphosyntactic domains (such as nominal morphology or syntax). Moreover, alignment is also concerned with intransitive predication. This explains why hierarchical systems can co-occur with other alignment patterns, as Nichols states: “most hierarchical languages also have an identifiable accusative, neutral, or stative-active component. We may speak of hierarchical languages, like stative-active languages, as admitting various base alignments, at least in theory.” (Nichols 1992: 68). As far as hierarchical alignment is concerned, we deal with languages in which the hierarchies constrain the assignment of grammatical relations to S, A and P. A notable case is when the indexability hierarchy governs the access to syntactic function, as in Kiowa (Zúñiga 2006: 65). For this reason, my position is that a system such as Emerillon’s is to be seen simply as an indexation system, rather than as an alignment type. The scope of hierarchies does not go beyond the selection of the argument to be encoded on the transitive verb. It does not influence the access to grammatical roles and even less so to grammatical relations. Consequently, the Emerillon alignment type has to be considered separately from its indexation system. Characterization of the alignment type of a specific language should take into account several parameters, including indexation, but also case, constituent order and syntax, since different types of alignment may co-occur in a language, depending on the parameter and/or the specific construction observed. Since arguments in Emerillon are not marked for case and their order is relatively flexible, the argumentation will be based on the indexation pattern described above and on syntactic phenomena. Even though the Emerillon indexation system is based upon the notion of hierarchies, grammatical roles are not completely excluded from it. Remember that the selection of the argument to be indexed on the transitive verb depends on its position on the hierarchies, yet the morphological paradigms of the indexes correlate with grammatical roles. Set I encodes A, while Set II encodes P. Since the unique argument of an intransitive verb is also encoded with Set I, it aligns with the A argument of a transitive

Emerillon hierarchical indexation system

77

verb. This forms a nominative-accusative system, where S and A are encoded likewise, and differently from P, as shown in Table 4. Table 4. Distribution of person indexes on verbs

Set I Set II

S x

A x

P x

Syntactic phenomena confirm the subject category postulated when looking at the distribution of the morphological paradigms. Some agreement and coreference patterns are indeed associated with a subject category comprising S of intransitive verbs and A of transitive verbs. Two examples are given below. First, the plural clitic -(o)ŋ agrees exclusively with the unique argument of a verbal intransitive predicate (19) and the A argument of a transitive verb (20), but not with the P argument of a transitive verb (21). Within the Tupí-Guaraní family, this criterion is specific to Emerillon, since the family does not offer a unified way of marking number. ikiʔɨ (19) o-paʔam-oŋ 3.I-get_up-PL.S then ‘Then they got up.’

(20) baipuɾi o-pɨhɨg-oŋ tapir 3.I-catch-PL.S ‘They caught a tapir.’ (21) *a-ɨkɨdʒ-oŋ 1SG.I-catch-PL.S ‘I caught them.’ Second, a criterion widely used by scholars of Tupí-Guaraní languages to define the subject grammatical relation is the use of an o- index for third person possessors or objects of postpositions specifically triggered by coreference with the subject, be it of an intransitive (22) or a transitive verb (23). (22) o-ho o-wɨ-kotɨ 3.I-go 3.COREF-mother-to ‘He is going to his (own) mother.’

78

Françoise Rose

o-mõde (23) bokal-a-pe o-akaŋ jar-REF-in 3.COREF-head 3.I-put ‘He put his (own) head in the jar.’

As a consequence, the coreferential index o- is in opposition with the Set II (i.e. non-coreferential) third person marker i-, both as a possessor (compare examples 23 and 24) and as an object of postposition (compare 25 and 26). zawaɾ o-wuɾ (24) i-(dʒ)akaŋ 3.II-head dog 3.I-go_up ‘The dog went up on his head.’

(25) o-iɲuŋ o-wɨb 3.I-put 3.COREF-under ‘She put it under herself.’ (26) o-bowɨg i-wɨb 3.I-put_firewood 3.II-under ‘They put more firewood under it.’ In brief, person index paradigms and syntactic patterns concur in the analysis of the Emerillon alignment type as nominative-accusative. Alignment is therefore completely independent of the hierarchical indexation system. Of course, in all hierarchical indexation systems, a device is necessary to disambiguate between A and P. In some languages, it is done through the existence of direct/inverse morphology. In Emerillon, this is done through the existence of two distinct person index sets. This is how this indexation system, though not organized so as to primarily mark such and/or such grammatical roles, nevertheless does mark grammatical roles and therefore participates in the general alignment type of the language, which is basically nominative-accusative.

6. Conclusion To give a brief summary of the Emerillon person indexation system on the verbs, intransitive verbs take a Set I index, while transitive verbs allow only one person index, from Set I for their subject or from Set II for their object. The correct index is selected according to the relative position of the two arguments on both the person hierarchy 1/2 > 3 and the grammati-

Emerillon hierarchical indexation system

79

cal roles hierarchy A > P. Thus, hierarchy is the primary organizing pattern of the Emerillon agreement system on transitive verbs. Since no morpheme can be considered to be marking direction of action (as normal or reversed, i.e. direct or inverse), the system is simply labelled ‘hierarchical’. A broader picture of the language (including syntax) shows a nominativeaccusative alignment, where the A of a transitive verb aligns with the S of an intransitive verb. The following table sums up the characterization of indexation and alignment in Emerillon. Table 5. Indexation and alignment in Emerillon Indexation on transitive verbs:

hierarchical with no direction marking

Morphological alignment: Syntactic alignment:

nominative-accusative nominative-accusative

It is clear from this table that, of the three possible implications of the indexability hierarchy in a language (access to marking slots, marking of direction, access to syntactic functions; Zúñiga 2006: 27, from Nichols 2002: 66), Emerillon shows only the first: the hierarchy only serves to select the argument indexed on the transitive verb. The analysis of the hierarchical indexation system of Emerillon, in comparison with other languages, leads to three theoretical assertions. First, ‘hierarchical’ systems should be considered alongside major indexation systems. Second, inverse systems should be considered as a subtype of hierarchical systems more generally, rather than the other way around. And finally, indexation should be considered independent of alignment.

Notes 1.

2. 3. 4.

I would like to thank Francesc Queixalós, Andrej Kibrik and Scott DeLancey for their helpful remarks on initial steps of this investigation, as well as Denis Creissels and Katharina Haude for more recent discussions. si- and i- are realized se- and e- before the vowel i. zo- is used on nouns, poɾo- on verbs to mark a generic human object ‘people’. The abbreviations used in this chapter are the following: I, II Set I, Set II; A most agent-like argument of a transitive clause; COMPL completive; CONT continuous; CONTRAST contrastive; COP existential copula; COREF coreferential; DEM demonstrative; DESID desiderative; EXCL exclusive; FUT future; INCL inclusive; INDET indeterminate; INTENS intensive ; INTER interrogative; LK linker; P most patient-like argument of a transitive clause; PL plural; PL.S plural of

80

5. 6.

7.

8.

9.

Françoise Rose subject; PRO pronoun; RED reduplication; REF referential; RELN relational; S unique argument of an intransitive clause; SG singular. The Set II prefix i- is actually found on some gerundives, as a residue of former absolutive marking on verbs in dependent clauses (Rose, ms). The configuration where a first person inclusive interacts with some other speech act participant (i.e. second person or first person singular), although logically possible, is never attested in my corpus. The overt expression of such a configuration is probably dispreferred due to the overlapping of the two referents. The peculiar use of the independent second person markers in 2→1SG scenarios can probably be explained as a hypothetical residue of a Proto-TupíGuaraní marker for A. Since the hierarchy in Proto-Tupí-Guaraní is considered to be 1 > 2 > 3 (Montserrat and Soares 1983, Jensen 1998), in the same 2→1 scenario, the first person P, being higher in the hierarchy, is procliticized to the verb, and the second person A is then expressed with an independent pronoun following the verb. For a detailed account of this diachronic hypothesis, see Rose (2003b and 2007). This chapter deals with person indexation on verbs only. However, in a study also taking into account nominal predicates (including among them descriptive words), the system could be considered as displaying split intransitivity, since different types of intransitive predicates take different person marking. Intransitive verbs take Set I prefixes while nominal predicates take Set II prefixes. In the analysis of Tupí-Guaraní languages where descriptive words are classified as verbs (such as Kakumasu 1986, Leite 1990, Jensen 1998, Seki 2000), the split intransitivity analysis holds within the verbal indexation system. For a detailed account of the debate about which part of speech the descriptive words belong to, see Meira (2006). Payne (1994) only mentions that the local scenario where a first person acts on a second person is traditionally described as using so-called portmanteau morphemes (considered to belong to a separate paradigm called Set IV, see for example Jensen 1998: 522). However, it would be easy to integrate the other local scenario participates in the putative direct/inverse system: when a second person acts on a first person (as in 10 from Tupínambá), the first person P is marked on the verb with a Set II clitic. This scenario illustrates both the inverse function and the so-called inverse morphology.

Emerillon hierarchical indexation system

81

References Bickel, Balthasar, and Johanna Nichols 2007 Inflectional morphology. In Language Typology and Syntactic Description. Vol. III, Timothy Shopen (ed.), 169–240. Cambridge: Cambridge University Press. Brown, Penelope, and Stephen C. Levinson 1987 Politeness. Cambridge: Cambridge University Press. Comrie, Bernard 1981 Language Universals and Linguistic Typology. Oxford: Basil Blackwell. Couchili, Ti’iwan, Didier Maurel, and Francesc Queixalós 2002 Classes de lexèmes en émérillon. Amerindia 26/27: 173–208. DeLancey, Scott ms. Deixis, Topicality, and the Inverse. Available at http://www.uoregon. edu/~delancey/papers/inverse.html. 1981 An interpretation of split ergativity and related patterns. Language 57: 626–657. 2001 Lectures on Functional Syntax. Class given at the LSA Summer Institute, University of California at Santa Barbara. Available at http://www.uoregon.edu/~delancey/sb/fs.html. Gildea, Spike 1994 Semantic and pragmatic inverse: ‘inverse alignment’ and ‘inverse voice’ in Carib of Surinam. In Voice and Inversion, Talmy Givón (ed.), 187–230. Amsterdam/Philadelphia: John Benjamins. Givón, Talmy 1994 The pragmatics of de-transitive voice: functional and typological aspects of inversion. In Voice and Inversion, Talmy Givón (ed.), 3– 44. Amsterdam/Philadelphia: John Benjamins. Harrison, Carl 1984 Verb prominence, verb initialness, ergativity and typological disharmony in Guajajara. In Handbook of Amazonian Languages. Vol. I, Desmond Derbyshire and Geoffrey Pullum (eds.), 407–439. Berlin: Mouton de Gruyter. Heath, Jeffrey 1998 Pragmatic skewing in 1–2 pronominal combinations in Native American Languages. International Journal of American Linguistics 64 (2): 83–104. Jensen, Cheryl 1990 Cross-referencing changes in some Tupí-Guaraní languages. In Amazonian Linguistics: Studies in Lowland South American Languages, Doris Payne (ed.), 117–158. Austin: University of Texas Press.

82

Françoise Rose 1998

Comparative Tupí-Guaraní Morpho-syntax. In Handbook of Amazonian Languages. Vol. IV, Desmond Derbyshire and Geoffrey Pullum (eds.), 489–619. Berlin: Mouton de Gruyter. 1999 Tupí-Guaraní. In The Amazonian Languages, R. M. W. Dixon and Alexandra Y. Aikhenvald (eds.), 125–163. Cambridge: Cambridge University Press. Kakumasu, James 1986 Urubu-Kaapor. In Handbook of Amazonian Languages. Vol. I, Desmond Derbyshire and Geoffrey Pullum (eds.), 326–403. Berlin: Mouton de Gruyter. Klaiman, M. H. 1991 Grammatical Voice. (Cambridge Studies in Linguistics 59). Cambridge: Cambridge University Press. Leite, Yonne 1990 Para uma tipologia ativa do Tapirapé, os clíticos referenciais de pessoa. Cadernos de Estudos Linguisticos 18: 37–56. Meira, Sérgio 2006 Mawe Stative Verbs and Predicate Possession. In Guaraní y MawetíTupí-Guaraní. Estudios Históricos y Descriptivos sobre una Familia Lingüística de América del Sur, Wolf Dietrich and Haralambos Symeonidis (eds.), 47–68. Berlin: LIT Verlag. Montserrat, Ruth, and Marília Faco Soares 1983 Hierarquia referencial em línguas Tupí. Ensaios de Lingüística 9: 164–187. Nichols, Johanna 1992 Linguistic Diversity in Space and Time. Chicago/London: University of Chicago Press. Payne, Doris L. 1994 The Tupí-Guaraní inverse. In Voice: Form and Function, Barbara Fox and Paul Hopper (eds.), 313–340. (Typological Studies in Language 27). Amsterdam/Philadelphia: John Benjamins. Payne, Thomas E. 1997 Describing Morphosyntax: a Guide for Field Linguists. Cambridge: Cambridge University Press. Queixalós, Francesc 2001 Le suffixe référentiant en émérillon. In Des Noms et des Verbes en Tupí-Guaraní: État de la Question, Francesc Queixalós (ed.), 115– 132. (LINCOM Studies in Native American Languages 37) München: Lincom Europa. Rodrigues, Aryon Dall’Igna 1984–1985 Relações internas na família lingüística Tupí–Guaraní. Revista de Antropologia 27/28: 33–53.

Emerillon hierarchical indexation system 1990

83

You and I = neither you nor I: the personal system of Tupínambá. In Amazonian Linguistics: Studies in Lowland South American Languages, Doris Payne (ed.), 393–405. Austin: University of Texas Press. Rose, Françoise 2003a Morphosyntaxe de l’émérillon, une langue tupí-guaraní de Guyane française. Ph.D. diss., Linguistics Department, Université Lumière Lyon II, France. 2003b Le marquage des personnes en émérillon (tupí-guaraní): un système d’accord hiérarchique. Faits de Langues 21: 107–120. 2007 As particularidades da língua Emerillon entre as línguas TupíGuaraní: O caso da marcação de pessoa. In Línguas e Culturas Tupí, Aryon Rodrigues and Ana Suelly Cabral (eds.), 325–332. Campinas: Curt Nimuendajú. 2008 A typological overview of Emerillon, a Tupí-Guaraní language from French Guiana. Linguistic Typology 12 (3): 431–460. ms. The origins of serialization: the case of Emerillon. to appear Grammaire de l’émérillon (teko), langue tupí-guaraní de Guyane française. (Langues et Sociétés d’Amérique traditionnelle). Louvain: Peeters. Seki, Lucy 1990 Kamaiurá (Tupí-Guaraní) as an active-stative language. In Amazonian Linguistics: Studies in Lowland South American Languages, Doris Payne (ed.), 367–391. Austin: University of Texas Press. Siewierska, Anna 2004 Person. (Cambridge Textbooks in Linguistics). Cambridge: Cambridge University Press. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, R. M. W. Dixon (ed.), 112–171. Canberra: Australian Institute of Aboriginal Studies. Zúñiga, Fernando 2006 Deixis and Alignment. Inverse Systems in Indigenous Languages of the Americas. (Typological Studies in Language 70). Amsterdam/Philadelphia: John Benjamins.

Where differential object marking and split plurality intersect: Evidence from Hup Patience Epps

1. Introduction1 The relevance of the Animacy Hierarchy to a range of linguistic structures is cross-linguistically familiar. Many languages treat nouns differently according to their position on the hierarchy, which is typically summarized as follows: pronouns 1/2 > 3 > kin, proper names > human > animate > inanimate (e.g., Silverstein 1976; Comrie 1989; Croft 1991) – although, as many scholars have pointed out, there are actually at least three parameters (or sub-hierarchies) involved: animacy proper (human > animal > inanimate), definiteness (personal pronoun > proper name > definite noun phrase > indefinite specific noun phrase > non-specific noun phrase), and person: 1/2 > 3. While it is thus difficult to reduce the Animacy Hierarchy to any single parameter, as noted by Comrie (1989: 195), it is widely observed that more general variables informing this ranking are the individuation of entities and their prominence or salience to speakers (e.g., Bossong 1991; Lazard 2001: 879; de Swart 2007). As an organizing principle in the variable treatment of nominals, the Animacy Hierarchy is relevant to a range of grammatical areas both within and across languages. In particular, these include ergative/absolutive splits, object case marking, and number marking, such that morphological marking (or a particular alignment pattern) occurs with those nominals toward the left end of the hierarchy, but is absent or realized differently on those nominals toward the right end. While the treatment of nominals according to the hierarchy itself is highly consistent across languages, considerable variation exists in the actual realization of the grammatical patterns involved. For example, languages differ as to where on the hierarchy they place the cut-off point between more and less (or differently) marked nominals, and as to whether morphological marking is optional or obligatory at different points along the hierarchy. Similarly, a given language may exhibit animacy effects in

86

Patience Epps

one or in several grammatical subsystems (such as number and/or case marking). Given this variability, if we take a situation in which animacy effects are reflected in multiple grammatical subsystems for a single language, there is no assurance that the parameters informing the treatment of different nominals will be identical for the different subsystems, even within that language. This potential mis-match suggests that the points at which grammatical subsystems informed by the Animacy Hierarchy intersect within a language could reveal some intriguing and unexpected twists – grammatical idiosyncrasies that challenge or complicate typological generalizations about the broader patterns. Yet while there is considerable discussion in the literature regarding the role of the Animacy Hierarchy in the patterning of individual subsystems, little attention has been devoted to what happens when such subsystems overlap. This paper presents a case of just such an intersection and its outcome. In Hup, a Nadahup (Makú)2 language of the northwest Amazonian Vaupés region, the Animacy Hierarchy informs both object marking and number marking: where relevant, each is obligatory on pronouns and specific human referents, optional on animal referents, and normally absent on inanimates. However, at the point where these two systems overlap – the treatment of plural-marked object nominals – we find a typologically unusual phenomenon: plural-marked nouns in Hup receive obligatory object marking, regardless of the animacy of the nominal referent. Thus, contrary to cross-linguistic generalizations that object-marked nouns will be relatively more individuated or discrete than unmarked nouns in such variable systems, in Hup morphologically marked plurality takes priority over considerations of animacy or definiteness/specificity for the purposes of case marking. The following sections turn first to a discussion of object marking and number marking as independent phenomena in Hup, and the role of the Animacy Hierarchy in determining their respective realizations. The paper then addresses the intersection of these patterns and the typological implications of this intersection.

2. Differential object marking and its realization in Hup The variable treatment of object nominals depending on their location on the Animacy Hierarchy is known as differential object marking (abbreviated as DOM; cf. Bossong 1985a, etc.). In such a system, those nominals

Differential object marking and split plurality

87

that are more highly ranked on the hierarchy will be more (or at least no less) marked than those below them. The principal parameters involved in their placement along the hierarchy are animacy and definiteness/specificity (c.f. Comrie 1989; Croft 1991; Lazard 2001; etc.), or in Bossong’s (1991: 159) terms, ‘inherence’ (humanness/animacy and discreteness) and ‘reference’ (individuation and discourse-related definiteness); in addition, in Bossong’s (1991) framework, ‘constituence’ (whereby object nominals that are more integrated with the verbal predicate are unmarked) may also play a role. Why differential object marking exists has been the topic of considerable discussion in the linguistic literature. A primary function is clearly the resolution of potential ambiguity between subjects and objects.3 Studies such as Zeevat and Jäger (2002) and Øvrelid (2004) indicate that subjects are much more likely than objects to be animate, definite, and pronominal (cf. Haspelmath 2005a: 8; de Swart 2007: 81). Given this tendency, as Comrie (1977: 9) observes, “instances where confusion will be particularly likely will be where one has either indefinite... and/or inanimate subjects, or where one has definite... and/or animate direct objects”. In a differential system, marking thus falls on those objects that share features with prototypical subjects (cf. Bossong 1991: 162). However, the avoidance of subject/object ambiguity is not the whole story: Many objects receive morphological marking even in cases where ambiguity is resolved by other means, such as through the discourse or pragmatic context (cf. Aissen 2003: 437; Haspelmath 2005a: 9; etc.). Accordingly, DOM often functions to mark prominent objects (i.e. those objects that are animate, discrete, and/or definite/specific; see de Swart 2007) generally, regardless of ambiguity. This in turn can be explained as “maximizing distinctiveness with minimal effort, or minimizing confusion with maximal economy” (Haspelmath 2005a: 9); in other words, while the avoidance of ambiguity may be the primary motivation for a differential system, the convention of marking certain types of nouns has become systematized and routinized. Thus speakers effectively avoid all or most cases of potential ambiguity without having to engage in a case-by-case evaluation of whether ambiguity actually exists. We return to this issue in Section 4 below. Differential object marking is cross-linguistically widespread. It occurs, for example, in many branches of the Indo-European, Semitic, FinnoUgric, and Turkic families (Bossong 1998), in languages of Australia and the Americas (Bossong 1983, 1985b), and elsewhere. As with Animacy effects in other grammatical subsystems, however, it is subject to consider-

88

Patience Epps

able variation across languages. Typological approaches to DOM (Lazard 2001; Bossong 1991; etc.) observe that some languages privilege definiteness/specificity (or ‘reference’, in Bossong’s terminology) in the marking of nominals, while in others animacy (‘inherence’) takes priority. Languages differ in which nouns are marked (e.g., pronouns, human nouns, or animate nouns), and in whether marking is optional or obligatory. Just a few examples of this variation are evident in Hindi, which marks nouns only when the referent is both specific and animate (Blake 2001: 120), in Sinhalese, where optional object marking is limited to animate-referring objects, and in Hebrew, in which object marking is obligatory but limited to definite objects (Aissen 2003: 436). Even within the Vaupés region, where Hup is spoken and DOM is a general feature (cf. Aikhenvald 2002; Zúñiga 2007; Stenzel 2008), we find that in the Eastern Tucanoan languages definiteness plays a particularly important role (such that definite inanimates and even mass nouns may be case-marked; cf. Stenzel 2008), while the Nadahup languages in the region (Hup and Yuhup; see below and Ospina 2002: 139–147) share animacy as the primary variable but differ in still other ways. In Hup, differential object marking makes two primary distinctions between nouns with respect to the Animacy Hierarchy: humans vs. animals and animates vs. inanimates (see Epps 2008: 165–177 for further details). Definiteness and specificity also play a role, but one which is less central than animacy. The following discussion briefly outlines the patterning of DOM in Hup. The set of Hup nouns that is most highly ranked according to the Animacy Hierarchy is obligatorily marked with the object suffix -ǎn. This applies to pronouns and demonstratives, as in (1) and (2):4 (1)

́ -ay ʔəg-náʔ-ãw hɨ ́d-ǎn drink-lose_senses-FLR-INCH 3PL-OBJ ‘While drunk, he spoke to them.’

(2)

́ ́ -ãý ʔãh hipãh yúw-ǎn-ãh 1SG know-DYNM that.ITG-OBJ-DECL ‘I know that one (story).’

tɨh 3SG

ʔɨ ́d-ɨ ́h speak-DECL

Obligatory object marking also applies to kin terms (example 3) and personal names – even those referring to inanimates, such as the canoe in (4). However, proper names referring to places cannot receive object marking (in the unlikely event that they occur as objects).

Differential object marking and split plurality

89

(3)

nɨ ̌ tæ̃́ h-ǎn mæh-yɨʔ-tæ̌ n… ʔãh 1SG.POSS son-OBJ kill-TEL-COND 1SG d’oʔ-d’ɔh-yɨʔ-tæ̃́ ʔ-æ̃́ y ʔũhníy take-rot-TEL-CNTRFCT-DYNM maybe ‘If they killed my son, I would maybe make things bad (for them).’

(4)

ʔãh d’ó-óh hǎt-ǎn Alligator-OBJ 1SG take-DECL ‘I took Alligator (a canoe).’

Nouns having human referents are also obligatorily marked in almost all cases, including those in which the human noun is indefinite but specific (example 6): (5)

děh hɔn-yóʔ… yub=ʔãý -ǎn, cipó_vine=woman-OBJ water vomit-SEQ tɨh d’oʔ-macã-́ b’ay-áh 3SG take-gain_consciousness-AGAIN-DECL ‘Having (ritually) vomited water, he created cipó vine woman.’

(6)

ʔayǔp=ʔĩh-ǎn ʔãh kéy-éy, one=MSC-OBJ 1SG see-DYNM ‘I saw a man in the forest.’

j’ǔg-an forest-DIR

The only exception to the obligatory object marking of human nouns occurs in the relatively rare cases where the referent is non-specific, provided the identity of subject and object are clear from the discourse context. This is the case in (7a), which can be compared with the definite and specific referent in (7b): (7)

tæ̃ hʔín túk-úy a) wãʔ́ vulture wife want-DYNM ‘Vulture wants a wife.’ (i.e., he wants to get married) b) wãʔ́ tɨh=tæ̃ hʔín-ǎn túk-úy vulture 3SG=wife-OBJ want-DYNM ‘Vulture wants his wife.’ (e.g., they have separated)

In contrast to this obligatorily marked set of Hup nouns, object marking on nouns referring to animals (which are located at an intermediate point

90

Patience Epps

along the Animacy Hierarchy) is optional, even when the noun is both definite and specific. This optionality is illustrated by the following examples, which both come from the same text and refer to the same frog: (8)

hɔhɔ́h=mah tɨh ʔey-yɔhɔ́y-ɔ́h frog=REP 3SG call-search-DECL ‘He was calling and searching for the frog, it’s said.’

(9)

tɨh ʔéy-cud-ʔũhníy tɨnɨ ̌h cápu-ǎn=yɨ ́ʔ 3SG.POSS toad(Port.)-OBJ=TEL 3SG call-INFR-maybe ‘He’s apparently calling for his (pet) frog.’

While examples like these suggest that the factors governing the object marking of animal nouns are subtle and not easily defined, the choice to use the marker or not appears to be largely discourse-driven, and dependent in particular on the topicality of the referent. On the other hand, mythical animal protagonists in stories are almost always case-marked, probably because of their human-like attributes of agency and volitionality, as well as their capacity for speech and rational thought, and the fact that their names function like personal names. The last category of Hup nouns, as defined by DOM patterns, is that of inanimates. For these nouns, object marking is ungrammatical (in the singular; see below). This is apparently without exception, and without regard to the noun’s definiteness or specificity: hɨd bɨʔ-pɨ ́d-ɨ ́h, (10) yɨkán mɔ̌y(*-ǎn) over_there house(*-OBJ) 3PL make-DIST-DECL ‘There they built a house, (it was) big!’

póg ! big

Thus far, differential object marking in Hup corresponds neatly to the Animacy Hierarchy, and exhibits typologically commonplace behavior. As noted above, however, this is not the full story; we complete the picture of Hup DOM in Section 4 below.

3. Split plurality and its realization in Hup As with object marking, the morphological indication of number in many languages applies differently to nouns according to their position on the Animacy Hierarchy. This phenomenon is typically labelled ‘split plurality’

Differential object marking and split plurality

91

(Smith-Stark 1974; Corbett 2000). As Comrie (1989: 189) observes, the correlation between animacy and number may reflect a “greater human concern with entities of higher animacy as individuals, therefore countable, while entities of lower animacy are more readily perceived as an indeterminate mass”. This tendency, according to Haspelmath (2005b: 8), may lead speakers to prefer the use of available plural markers when they refer to multiple animate entities; the frequent adherence to this norm can then give rise to routinization and eventual obligatoriness in language structure. Split plurality, like differential object marking, is cross-linguistically common, but also varies in its realization across languages. This variation corresponds to many of the same parameters discussed above for DOM – where on the hierarchy the cut-off occurs, optional vs. obligatory marking, etc. To illustrate, Slave (Athabaskan) groups humans and dogs together as distinct from other animals and inanimates with respect to number marking, while Marind (Irian Jaya) observes an animate vs. inanimate distinction (Corbett 2000: 57, 59, after Rice 1989 and Drabbe 1955). Similarly, within the Vaupés region itself we find widespread split plurality but variable systems; for example, in Eastern Tucanoan languages plurality splits divide humans and higher-level animates from low-level animates and inanimates (Wanano; Stenzel 2004: 4.3.1–2) and animates from inanimates (Tucano: Ramirez 1997: 205–209), while in the Nadahup languages we find distinctions between human, animal, and inanimate nouns (Hup; see below), animates and inanimates (Yuhup; Ospina 2002: 258–262), and humans and non-humans (Dâw; Martins 2004: 401), together with further differences in marking conventions. In Hup, split plurality carves up the set of nouns along the Animacy Hierarchy in much the same way as does differential object marking; however, the parameters for marking within the two subsystems are not identical. Plural number in Hup is indicated by the enclitic =d’ǝh (which can also have a collective function in some contexts); when the plural marker occurs together with the object case marker they take the fused form =n’ǎn. The following discussion details the patterning of split plurality in Hup (see Epps 2008: 192–197 for further details). Like object marking, plural marking is in general obligatory in Hup for pronouns, demonstratives, kin terms, and human nouns. Pronouns have a lexical distinction between singular and plural for first, second, and third persons (e.g., ʔãh ‘I’ and ʔɨn ‘we’). Demonstratives always take the plural suffix if they stand in for or modify an animate noun, as in (11), while a separate collective demonstrative form is used for plural inanimates and mass nouns (12):

92

Patience Epps

nɨh-nɨ ́h-ay=pog’, páy-ay (11) ʔǝ̌g=wag nɨ-d’ǝ̌h drink=day this-PL be_like-NEG-INCH=EMPH1 bad-INCH ‘On drinking days, these (people) don’t do like this, (it’s) no good.’

(12) nɨ-n’ɨ ̌h j’ɨ ̌k this-NMLZ smoke ‘(all) this smoke’ Human nouns and referential kin terms must also, in most cases, be marked as plural: ʔǝg-náʔ-ãý ! (13) tiyǐʔ=d’ǝh-ǝw-ǝ́c man=PL-FLR-EXCL2 drink-lose_senses-DYNM ‘Only the men got drunk!’ tukáno=tæ̃́ h=d’ǝh NEG:R Tucano(Pt)=offspring=PL ‘There are no Tucano children.’

(14) nǽ

pǎ̃ NEG:EX

The only exception to this generalization occurs in cases where the human noun or kin term is indefinite and non-specific, as is also true for object marking (see above). Even these cases, however, require anaphoric agreement with a plural pronoun, as in (15) (whereas unmarked multiple inanimates may be referred to via a singular pronoun): (15) húp dǝ̌b, hɨd bɨ ́ʔ-ɨ ́h person many 3PL work-DECL ‘Lots of people worked.’ In contrast to human nouns, nouns referring to animals are optionally marked for number, which again is consistent with the conventions of DOM in Hup. The speaker included the plural suffix in example (16): (16) j’ám-ãṕ děh mí-ĩt,́ hɔ̌p̃ wæ̌ d=n’ǎn DST.CNTR-DEP water creek-OBL fish eat=PL.OBJ ʔãh nɔ́-ɔ̃p, yɔ̌k=d’ǝh 1SG say-DEP otter=PL ‘I’m talking about those that eat fish in the river, the otters.’

Differential object marking and split plurality

93

In the following example, on the other hand, the speaker did not use the plural suffix in (17a), but did in (17b) – although the two utterances come from the same text, refer to the same entities, and are almost identical. (17) a) b’ǒy ʔɨn kǝk-ʔay-nɨ ̌ŋ ! traira 1PL pull-VENT-COOP ‘Let’s go fish for traira together!’ ʔɨn kǝk-nɨ ̌ŋ-ay ! b) b’ǒy=n’ǎn traira=PL.OBJ 1PL pull-COOP-INCH ‘Let’s fish for traira together!’

Exactly what informs the choice to leave animal entities marked or unmarked for number is thus not obvious, but marking is in general preferred when the animal referent is more topical (and thus relatively specific, although it frequently occurs on nouns that refer to a generic group of animal entities – such fish or rats to be caught, as in 17 and in 19 below). In (17), the unmarked example (17a) comes from the first mention of the traira fish in the text, while subsequent mentions (such as 17b) are marked fairly consistently as plural. For nouns referring to inanimate entities, number marking is extremely rare in natural discourse; however, it is not ungrammatical (except in the case of mass nouns), as illustrated by the elicited example in (18). This marginal acceptability of number marking on inanimates constitutes the primary difference between the conventions governing object and number marking in Hup, although in actual discourse the two pattern in very similar ways. (18) mɔ̌y=d’ǝh ní-ĩý house=PL be-DYNM ‘There are some houses.’

4. The intersection of differential object marking and split plurality As we have seen, the Animacy Hierarchy informs two distinct areas of Hup’s grammar – the marking of object case and number – in ways that are similar but not quite identical. In such a situation, what will the intersection of the subsystems look like? How will animacy effects be realized if and when the two grammatical structures overlap?

94

Patience Epps

In Hup, this intersection occurs when plural-marked nouns appear as objects. In this context, we find that object marking on these nouns is absolutely obligatory, regardless of their animacy. While obligatory marking is expected in the Hup system for those entities that are highly ranked on the Animacy Hierarchy – pronouns, demonstratives, kin terms, and human nouns – it is much less clear why it should apply to nouns with animal and inanimate referents, for which object marking is otherwise optional or obligatorily absent. Examples (19–20) illustrate this phenomenon for nouns referring to animals, on which object marking is optional when number marking is absent. In these examples, however, the presence of the plural marker without the object marker is ungrammatical, although both may be dropped and a plural interpretation maintained. (As noted above, the combination of the plural and object markers yields the fused form =n’ǎn). tɨh hãʔ-ʔě-h (19) bǐʔ=n’ǎn=mah yúp rat=PL.OBJ=REP that.ITG 3SG search_with_hands-PERF-DECL ‘He searched out rats with his hands, it’s said.’ hɨd, hãý -an, (20) cǎy… cǎy=n’ǎn=mah centipede centipede=PL.OBJ=REP 3PL um-DIR hǝb-kæd-yóʔ, yuʔ-yóʔ=mah hɨd ʔǝg-pó-tíh ! dry-dry_in_heat-SEQ burn-SEQ=REP 3PL drink-EMPH1-EMPH2 ‘Centipedes… having dried centipedes out, in the whatchamacallit, having burned them (to ashes and mixed them with water), they drink them, it’s said!’

Examples (21–22) illustrate the same phenomenon for inanimate nouns: here object marking is obligatorily present when the number marker occurs, but obligatorily absent when it does not: (21) ʔãh́ cug’æ̌ t=n’ǎn pũhũt-d’ǝh-hi-yɨ ́ʔ-ɨ ́y 1SG leaf/paper=PL.OBJ blow-send-descend-TEL-DYNM ‘I blew the papers down.’ (22) pɨhɨ ́t=n’ǎn ʔãh yum-té-h banana=PL.OBJ 1SG plant-FUT-DECL ‘I’ll plant those bananas.’

Differential object marking and split plurality

95

Given that number marking rarely occurs on inanimates in natural discourse, in my corpus most examples like those above derive from elicitation contexts. Consultants nevertheless report that marked forms like those in (21–22) are not only grammatical but indicate a specific group of referents; in (22), for example, the speaker has a particular bunch of bananas in mind, whereas without the number/object marking the sentence would simply mean ‘I’ll plant bananas’. It is interesting to note that while the presence of number marking entails the presence of object marking for Hup nominals, the reverse does not appear to be the case. In example (23), the indefinite and non-specific noun ‘person’ is best understood as semantically plural but – as expected (compare 15 above, for example) – it does not receive number marking. It is nevertheless marked with the object suffix, which functions here to clarify that it is indeed an object, since the subject of the clause is also nonspecific and is in fact dropped. Examples like this one suggest that the motivations for object marking are generally more complex than are those for plural marking, and that in particular cases the need to disambiguate subject and object may take precedence over other considerations. tǝ́w-ǝ́y, húp-ǎn dóh-óy (23) húp-ǎn person-OBJ scold-DYNM person-OBJ curse-DYNM ‘(Some people) scold people, cast curses on people.’

As the examples above illustrate, the presence of number-marking morphology on nouns overrides all considerations of animacy in Hup for the purposes of differential object marking. Yet for plurality to take priority in a DOM system is typologically highly unusual, and appears to be almost without parallel in other languages for which discussions of DOM are available. This can be attributed to the fact that individuation and discreteness are important parameters informing DOM in general (cf. de Swart 2007; Lazard 2001: 879; Bossong 1991; etc.). As Hopper and Thompson (1980: 292) have likewise observed, the presence of object marking registers a relatively high degree of individuation of the nominal entity, and definiteness corresponds closely to the individuation of the object – as does animacy, as the discussion of split plurality above indicates. It is therefore less likely that subject/object ambiguity would arise where an object is plural, since plural entities are almost by definition less individuated than are singular entities, and are thus less likely to be interpreted as subjects. There is little doubt that differential object marking relates to the individuation of entitites in Hup, just as it does cross-linguistically. In addition

96

Patience Epps

to the link between object marking and specificity of referents, a further indication lies in the fact that object marking in Hup patterns similarly to the participation of nouns in the ‘bound’ or obligatorily compounded construction, which in many cases encodes inalienable possession. As I have argued elsewhere (Epps 2007, 2008: 256–263), one function of the ‘bound’ construction in Hup is to individuate an entity from a mass of potential referents. The parallels between this construction and object marking in Hup are illustrated in examples such as (7) above, where the definite and specific referent ‘his wife’ is marked both by the (normally obligatory) third person pronoun tɨh, indicating inalienable possession, as well as by the object suffix, whereas both are absent when ‘wife’ is indefinite and non-specific. How then can we explain the unusual feature of Hup’s differential object marking system, by which plural marking of nouns takes priority over considerations of animacy in object marking? The answer undoubtedly lies in the intersection of the two systems – DOM and split plurality – which are both informed by the Animacy Hierarchy but conform to slightly different parameters. In Hup, the vast majority of nouns that are candidates for plural marking are also candidates for object marking: that is, they refer to animates and are specific and/or topical. Because the parameters governing the two subsystems are so similar, speakers can – most of the time – safely assume that where plural marking is appropriate, object marking will be likewise. Thus, over time, the frequency of their co-occurrence has led to its systematicization, such that plural marking on object nominals actually entails object marking. This nicely illustrates Haspelmath’s (2005a: 9) point that “DOM is about maximizing distinctiveness with minimal effort”: Because plural-marked nouns are prime candidates for object-marking, the expenditure of minimal effort on the part of speakers leads these nouns to be object marked by default, thereby avoiding a case-by-case evaluation of their actual animacy or the likelihood of their confusion with subjects.5 This is hardly different from other languages’ propensity to case-mark all animate object nominals, for example, even where ambiguity between subject and object is not at issue, as discussed above. In Hup, this default generalization has extended to plural marking because of the similarities between the split plurality and DOM systems; the unusual nature of the outcome – that the presence of plural marking overrides considerations of animacy in DOM – is due primarily to the mis-match between the two systems, namely that inanimate nouns may take number marking (although they rarely do), but by themselves cannot be object-marked.6

Differential object marking and split plurality

97

As we have seen, the interaction between number and case marking in Hup gives its differential object marking system a typologically unusual twist. Yet Hup is certainly not the only language to exhibit animacy effects in multiple areas of its grammar. While there is relatively little discussion in the literature of how these animacy effects play out when different grammatical subsystems come together, there is nevertheless some indication that Hup is not unique in exhibiting an interaction between number and other areas of grammar in which the Animacy Hierarchy plays a role.7 Exactly what this interaction is and how it may be generalized across languages is unclear, however. Comrie, for example, makes the following observation: One parameter which... participates in a rather ambivalent interaction with animacy is number. We are not prepared to make any generalizations as to whether number raises or lowers the animacy of a noun phrase [with respect to object marking]... and certainly there is a fair amount of evidence where number is relevant in either direction, suggesting that over all it is randomly, rather than significantly, relevant. (Comrie 1989: 188)

Comrie contrasts the example of Russian, where plural nominals are more likely to take a special animate accusative ending, with that of Polish, where a plural noun phrase is less likely to take the animate accusative ending. Other cases in which plural number appears to win out over singular include the nominative/ergative splits found in several Australian languages (Filimonova 2005: 100–104). For example, Giramay (a dialect of Dyirbal) exhibits a split distinguishing first and second person dual and plural pronouns (which follow a nominative-accusative pattern), first and second person singular and third person singular/dual/plural (which follow a ‘contrastive’ pattern, such that subjects, agents, and patients are all marked differently), and all other nouns (which follow an ergative-absolutive pattern; Filimonova 2005: 104, after Dixon 1972: 50). Similarly, inverse systems in many Algonquian languages rank plural first and second person verbal suffixes higher than their singular counterparts (Zúñiga 2008). As Filimonova observes, Silverstein (in his classic [1976] paper dealing with ergative splits and animacy) states that the marking of some nominals (particularly pronouns) with number features can result in their being more highly ranked within split systems, as appears to be the case in languages like Giramay. This is clearly analogous to what we see in the differential object marking of nominals generally in Hup. Yet Filimonova (2005: 105),

98

Patience Epps

like Comrie, notes that this phenomenon is not consistent across languages, and that in many cases it is singular nouns that are elevated over plurals. While more conclusive cross-linguistic generalizations about the interaction between number and other animacy-informed grammatical subsystems must await more data from a broader sample of languages, the Hup case shows that a clear motivation can in fact be identified for the ranking of plural nouns over singular nouns in systems such as differential object marking: Where number and object marking are both informed by the Animacy Hierarchy, number marking may provide a default basis for object marking, allowing for a more efficient DOM system. At the same time, the fact that object marking tends to register a relatively high degree of individuation of the nominal entity will encourage the ranking of singular nouns over plurals. Thus there appear to be at least two competing patterns of interaction between number and animacy, which may account for some of the cross-linguistic variation that Comrie and others have observed.

5. Conclusion While the identification of cross-linguistic tendencies is the bread and butter of typology, it is also widely recognized that identifying and explaining the exceptions is a priority if we are to truly understand how language works and why the generalizations exist in the first place. One well-known source of explanations for such linguistic irregularities is found in transitional areas of various types – points of intersection between languages, linguistic areas, historical phases, etc. For example, historical transitions between grammatical categories or subsystems can give rise to typologically idiosyncratic features, as in the case of the unusual tripartite or contrastive case marking found in some Iranian languages (such that S, A, and P arguments are all marked differently), which may be understood as an intermediate phase in a shift from an ergative to an accusative alignment pattern (Filimonova 2005: 87–93).8 Similarly, languages that are geographically sandwiched between two distinct language families or contact zones may come to share features with both neighbours through language contact, leading to unusual combinations of characteristics – such as the deviations in word order patterns in languages of northern Iran that are located between Turkic and Semitic groups (Stilo 2005). This paper suggests that another point of intersection – that between grammatical subsystems viewed synchronically within a single language – is also a relevant source of typological anomalies, and can accordingly

Differential object marking and split plurality

99

deepen our understanding of each of the subsystems themselves. This is particularly true for grammatical subsystems that share some organizing principle, such as the Animacy Hierarchy, but in which this organizing principle applies in slightly different ways. In Hup, as we have seen, the intersection of object marking and number marking patterns yields just such a typologically unusual result – the obligatory object marking of plural-marked nouns regardless of their animacy. While this phenomenon appears highly anomolous from the perspective of either subsystem taken on its own, it makes good typological sense when we consider the language as a system of interrelated parts, rather than as a set of independent subsystems. Accordingly, the Hup case also provides a clue to the somewhat mysterious relationship between number marking and other animacy effects noted cross-linguistically. The situation in Hup, and its explanation, suggest that the seeming randomness of this interaction in other languages may in fact be attributable to two competing motivations: the ranking of more individuated entities over less individuated ones, and the use of animacy-based number marking as a default basis for object marking in the interest of efficiency.

Notes 1.

2.

The data in this paper come from original fieldwork on the Rio Tiquié, Amazonas, Brazil, conducted in 2000–2004. I am grateful to the Hupd’ǝh for teaching me their language, and to the Museu Parense Emílio Goeldi and the Instituto Socioambiental for their assistance; support from a Fulbright-Hays Dissertation Research Grant, National Science Foundation Dissertation Improvement Grant no. 0111550, and by the Max Planck Institute for Evolutionary Anthropology (Leipzig) is also gratefully acknowledged. In addition, I would like to thank Georg Bossong, Orin Gensler, and Fernando Zúñiga for their very helpful comments on the material discussed here. Hup (aka Hupda, Jupde) is spoken by approximately 1500 people in northwest Brazil and eastern Colombia. The Hup people live in scatterered villages throughout this region and subsist by hunting, gathering, and small-scale manioc farming. The family name ‘Nadahup’ is preferred to ‘Makú’ because the latter name is used as an ethnic slur in the region and has also been applied to several unrelated language groups in Amazonia. ‘Nadahup’ combines the names of the four established member languages in this family (Hup, Yuhup, Dâw, Nadëb). The name Vaupés-Japura (Uaupés-Japura) has also been used in place of Makú (see Ramirez 2001).

100 3.

4.

5.

6.

7. 8.

Patience Epps Differential object marking and its role of disambiguating subjects and objects applies primarily to nominative-accusative languages; ergative-absolutive systems often make use of DOM’s mirror image, differential subject marking (cf. Silverstein 1976, de Hoop and de Swart 2008). Most of the examples appearing in this paper come from natural discourse (texts, conversation, etc.). The following abbreviations are used: COOP cooperative, DECL declarative, DEP dependent, DIR directional, DIST distributive, DST.CNTR distant contrast, DYNM dynamic, EMPH emphasis 1, EXCL2 secondary exclusive, FLR filler syllable, FUT future, INCH inchoative, INFR inferred evidential, ITG intangible (demonstrative), MSC masculine, NEG negative, NEG:EX negative existance, NEG:R negative reinforcement, NMLZ nominalizer, OBJ object, OBL oblique, PERF perfective, PL plural, POSS possessive, REP reported evidential, SEQ sequential, SG singular, TEL telic, VENT venitive. A further result of the frequent co-occurrence of these two markers is their phonological fusion into the portmanteau form =n’ǎn – a typical outcome of such grammaticalization processes, but one which occurs only rarely elsewhere in Hup. The same explanation probably accounts for another context in which object marking occurs obligatorily in Hup, regardless of the position of the referent along the Animacy Hierarchy: the goal of a ditransitive clause. Given that goals of ditransitives are almost always animate, they are an obvious place in which default object marking might be expected to occur. Both split plurality and DOM exist in many of the other Vaupés languages, but there is no evidence that their interaction results in anything unexpected. Note, however, that some tripartite systems appear to be relatively stable, so these are not all easily explained as representing a transitional stage.

References Aikhenvald, Alexandra Y. 2002 Language Contact in Amazonia. Oxford and New York: Oxford University Press. Aissen, Judith 2003 Differential object marking: Iconicity vs. economy. Natural Language and Linguistic Theory 21 (3): 435–483. Blake, Barry J. 2001 Case. Cambridge: Cambridge University Press. Bossong, Georg 1983 Animacy and markedness in universal grammar. Glossologia 2: 7– 20.

Differential object marking and split plurality 1985a

101

Empirische Universalienforschung: DOM in den neuiranischen Sprachen. Tübingen: Narr. 1985b Markierung von Aktantenfunktionen im Guaraní. Zur Frage der differentiellen Objektmarkierung in nicht-akkusativischen Sprachen. In Relational Typology, Frans Plank (ed.), 1–29. Berlin/New York: Mouton de Gruyter. 1991 Differential object marking in Romance and beyond. In New Analyses in Romance Linguistics, Douglas Kibbee and Dieter Wanner (eds.), 143–170. (Current Issues in Linguistic Theory 69.) Amsterdam/Philadelphia: John Benjamins. 1998 Le marquage differéntiel de l’objet dans les langues d’Europe. In Actance et Valence, J. Feuillet (ed.), 193–258. Berlin/New York: Mouton de Gruyter. Comrie, Bernard 1977 Subjects and direct objects in Uralic languages: a functional explanation of case-marking systems. Études Finno-Ougriennes 12 (1975): 5–17. 1989 Language Universals and Linguistic Typology: Syntax and Morphology, 2nd ed. Chicago: University of Chicago Press. Corbett, Greville G. 2000 Number. Cambridge: Cambridge University Press. Croft, William C. 1990 Typology and Universals. Cambridge: Cambridge University Press. Dixon, Robert M. W. 1972 The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. Drabbe, P. 1955 Spraakkunst von het Marind: Zuidkust Nederlands Nieuw-Guinea. (Studia Instituti Anthropos 11.) Wien-Mödling: Drukkerij van het Missiehuis St. Gabriel. Epps, Patience 2007 Birth of a noun classification system: the case of Hup. In Language Endangerment and Endangered Languages: Linguistic and Anthropological Studies with Special Emphasis on the Languages and Cultures of the Andean-Amazonian Border Area, Leo Wetzels (ed.), 107–128. (Indigenous Languages of Latin America series [ILLA].) Leiden University, The Netherlands: The Research School of Asian, African, and Amerindian Studies (CNWS). 2008 A Grammar of Hup, (Mouton Grammar Library 43). Berlin: Mouton de Gruyter. Filimonova, Elena 2005 The noun phrase hierarchy and relational marking: problems and counterevidence. Linguistic Typology 9.1: 77–113.

102

Patience Epps

Haspelmath, Martin 2005a Universals of differential case marking. Handout for the course ‘Explaining Syntactic Universals’, LSA Institute, July 2005. (Available at http://email.eva.mpg.de/~haspelmt/2.DiffCaseMarking.pdf.) 2005b Occurrence of nominal plurality. In World Atlas of Language Structures, Matthew S. Dryer, Martin Haspelmath, David Gil, and Bernard Comrie (eds.). Oxford: Oxford University Press. de Hoop, Helen, and Peter de Swart (eds.) 2008 Differential Subject Marking. Dordrecht: Springer. Hopper, Paul J., and Sandra A. Thompson 1980 Transitivity in grammar and discourse. Language 56 (2): 251–299. Lazard, Gilbert 2001 Le marquage différentiel de l’objet. In Language Typology and Language Universals: An International Handbook, vol. 2, Martin Haspelmath (ed.), 873–885. Berlin: Mouton de Gruyter. Martins, Silvana 2004 Fonologia e gramática Dâw [Phonology and grammar of Dâw]. Ph.D. dissertation, Vrije Universiteit, Amsterdam. Amsterdam: LOT. Ospina Bozzi, Ana María 2002 Les structures élémentaires du Yuhup Maku, langue de l’Amazonie Colombienne: Morphologie et syntaxe. Ph.D. diss., Université Paris 7 – Denis Diderot. Øvrelid, Lilja 2004 Disambiguation of grammatical functions in Norwegian: Modeling variation in word order interpretations conditioned by animacy and definiteness. In Proceedings of the 20th Scandinavian Conference of Linguistics, F. Karlsson (ed.). University of Helsinki: Department of General Linguistics. Ramirez, Henri 1997 A Fala Tukano dos Ye’pa-Masa, Vol. 1: Gramática [The Tucano Language of the Ye’pa-Masa: Grammar]. Manaus: Inspetoria Salesiana Missionária da Amazônia, CEDEM. 2001 Família Makú ou família Uaupés-Japura? [Makú family or UaupésJapura family?]. Paper presented at the meeting of ANPOLL, Belém, Brazil. Rice, Keren 1989 A Grammar of Slave. (Mouton Grammar Library 5.) Berlin: Mouton de Gruyter. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, R. M. W. Dixon (ed.), 112–171. Canberra: Australian Institute of Aboriginal Studies.

Differential object marking and split plurality

103

Smith-Stark, T. Cedric 1974 The plurality split. In Papers from the Tenth Regional Meeting, Chicago Linguistic Society, Michael La Galy, R. A. Fox and A. Bruck (eds.), 657–671. Chicago: Chicago Linguistics Society. Stenzel, Kristine 2004 A reference grammar of Wanano. Ph.D. diss., University of Colorado. 2008 Kotiria ‘differential object marking’ in cross-linguistic perspective. Manuscript, Museu Nacional / Universidade Federal do Rio de Janeiro. Stilo, Donald 2005 Iranian as buffer zone between the universal typologies of Turkic and Semitic. In Linguistic Convergence and Areal Diffusion. Case Studies from Iranian, Semitic and Turkic, E.A. Csato, B. Isaksson and C. Jahani (eds.), 35–63. London: Routledge Curzon. de Swart, Peter 2007 Cross-linguistic variation in object marking. Ph.D. diss., Radboud Universiteit Nijmegen. Zeevat, Henk, and Gerhard Jäger 2002 A reinterpretation of syntactic alignment. In Proceedings of the 3rd and 4th International Symposium on Language, Logic and Computation, D. de Jongh, H. Zeevat and M. Nilsenova (eds.). Amsterdam: ILLC. At http://www.uni-bielefeld.de/lili/personen/gjaeger/. Zúñiga, Fernando 2007 The discourse-syntax interface in northwestern Amazonia. Differential object marking in Makú and some Tucanoan languages. In Language Endangerment and Endangered Languages: Linguistic and Anthropological Studies with Special Emphasis on the Languages and Cultures of the Andean-Amazonian Border Area, Leo W. Wetzels (ed.), 209–227. Leiden: Publications of the Research School of Asian, African, and Amerindian Studies (CNWS), University of Leiden. 2008 How many hierarchies, really? Evidence from several Algonquian languages. In Scales (Linguistische Arbeitsberichte 86), Marc Richards and Andrej Malchukov (eds.), 99–129. Leipzig: University of Leipzig.

Syncretisms and neutralizations involving morphological case: Challenges for markedness theory Peter M. Arkadiev

1. Introduction1 It is well known that the notion of markedness has great relevance for the phenomenon of neutralization, both in phonology (Trubetzkoy 1939) and morphosyntax (Jakobson 1932, 1936; Greenberg 1966; Boeder 1976; Battistella 1990; Janda 1995). Generally speaking, there is a tendency, quite often claimed to constitute a universal law, for neutralizations (or syncretisms) of the values of one morphosyntactic feature to occur only in the context of the marked value(s) of other features, but not in the context of the unmarked value(s). In this chapter, I bring forth some empirical evidence from the domain of morphological case that explicitly contradicts the above-mentioned generalization, and, moreover, in some cases at least, shows systematic and diachronically stable behaviour. On the basis of these data, I claim that the relation between markedness and neutralization in morphology is much less trivial than some linguists used to believe. The goal of this chapter is primarily empirical, i.e. to bring forward some interesting facts coming from languages both relatively well-known (e.g. Slavic) and less so (e.g. Dardic) which have not been given attention in work on syncretism (even the recent comprehensive typological study of this phenomenon, viz. Baerman, Brown, and Corbett [2005], does not discuss them); I abstain from proposing any far-reaching theoretical interpretations, let alone explanations of these facts, although I speculate on some motivations alternative to those based on the notion of markedness. However, I believe that the data to be discussed are of relevance to the typology and theory of morphological neutralization, and that theoretical principles and typological generalizations in this domain should be altered to take them into account. The structure of the chapter is as follows. In Section 2 I review the basics of the classic markedness theory in morphology and briefly discuss its

106

Peter M. Arkadiev

recent critique by Haspelmath (2006). Sections 3 and 4 are entirely devoted to the presentation of the empirical data. In Section 5 I present a general discussion of the material and draw some preliminary conclusions.

2. Markedness theory and neutralization in morphology The notion of markedness has played one of the principal parts in the linguistic thought of the last three quarters of a century (cf. Eckman, Moravcsik and Wirth 1983; Andrews 1990; Battistella 1996). Markedness has been employed as an important explanatory notion in different fields of the science of language, ranging from phonology to semantics and pragmatics, and it cross-cuts major theoretical and methodological divisions, such as that between ‘formal’ vs. ‘functional’ theories, which both employ this notion. Markedness has been defined in various ways (see Haspelmath 2006 for a critical survey); for the goals of this chapter a simplified version of the definition by Givón seems to be suitable: Markedness involves Structural complexity: The marked structure tends to be more complex (or larger) than the corresponding unmarked one. (ii) Frequency distribution: The marked category tends to be less frequent than the corresponding unmarked category. (iii) Cognitive complexity: The marked category tends to be cognitively more complex – in terms of mental effort, attention demands or processing time – than the unmarked one. (Givón 1995: 28) (i)

Thus, the marked member of a morphological opposition (e.g. Plural) is supposed to be more complex as regards its semantics than the unmarked member (e.g. Singular), to show smaller discourse frequency, and to have a more complex (e.g. longer) formal expression. Also, the following correlates of markedness have been widely cited in the literature (cf. Haspelmath 2006): the presence of a marked value in a language presupposes the existence of the unmarked one; marked values have a distribution restricted in comparison to the unmarked value; marked categories are acquired later by children and are more likely to get lost in language change, etc. The explanatory force of the notion of markedness, as has often been stated (cf. Moravcsik and Wirth 1983: 3), is based on the assumption that different

Syncretisms and neutralizations involving case

107

markedness parameters are intercorrelated, i.e. that structural complexity goes hand in hand with cognitive complexity and frequency distribution, as well as with diachronic stability and skewed acquisition patterns. Recourse to markedness as an explanatory device has been perhaps most popular in typologically oriented linguistics (see e.g. Greenberg 1966; Croft 1990; Givón 1995: Ch. 2), where universal markedness hierarchies have been proposed, such as the person/animacy hierarchy applying to case-marking (Silverstein 1976, Dixon 1979), number-marking (SmithStark 1974), and, more generally, to the choice of grammatical functions in the clause (Kibrik 1997, Aissen 1999). The notion of markedness has been considered important for the phenomenon of neutralization, i.e. partial or full elimination of a certain overt morphosyntactic distinction otherwise present in the grammatical system. A simple example of neutralization comes from Russian, where in adjectives three genders are distinguished in the Singular but collapsed in the Plural (cf. bol’šoj dom ‘a large house (masc.)’, bol’šaja kniga ‘a large book (fem.)’, bol’šoe zadanije ‘a large assignment (neut.)’ vs. bol’šie doma / knigi / zadanija ‘large houses / books / assignments’). In this example gender is the category neutralized, while number is the dominant category (terminology proposed in Hjelmslev 1935–1937). The most general constraint on neutralization or syncretism of morphosyntactic distinctions2 explicitly relates to the notion of markedness, in particular, to markedness hierarchies. This constraint specifies possible contexts for syncretism and is stated as follows: If a certain distinction is suspended in the environment of a given category C [the dominant category, in our terms – P.A.], then ... it is either neutralized only for the marked value of C or for both the marked and the unmarked value, but never for the unmarked value alone. (Bierwisch 1967: 254)

This principle predicts, for instance, that, assuming a relatively uncontroversial markedness hierarchy for number values, i.e. Singular < Plural,3 neutralization may occur in the context of Plural only, or both in the Plural and in the Singular (cf. the above-mentioned examples from Russian); however, it is not possible for a category to be neutralized only in the Singular to the exclusion of the Plural. Formulated in a more succinct manner, the constraint looks like a familiar implicational universal:

108

Peter M. Arkadiev

Markedness constraint on neutralization in morphology: If the values of a certain morphosyntactic feature are neutralized in the context of the unmarked value of the dominant morphosyntactic feature, then they are also neutralized in the context of the marked value(s) of the same feature.

The rationale behind this constraint is admittedly ‘functional’: neutralization, which otherwise only creates ambiguity and parsing difficulty for the hearer, applies ‘in order to’ reduce the total number of marked values in the expression (cf. Boeder 1976); compare the following recent formulation: When a set of forms is relatively unmarked on one featural dimension, it generally allows a broader range of contrasts along other dimensions. The presence of more marked values for one feature triggers neutralization of other features. [...] This suggests that a grammar can limit the overall degree of markedness allowed in a set of feature specifications. [...] (Béjar and Hall 1999: 4)

Thus, if a word-form is specified for a marked member of one category, e.g. [Number:+Pl], it is more ‘economical’ to have it underspecified for the values of some other category, e.g. gender so that the overall markedness of the word-form remains lower than if gender is specified, too. Whether this kind of reasoning is to be taken seriously does not really matter for the purposes of this chapter; however, in Section 3 I hint at how ‘markedness minimization’ could be operative in the material I discuss. Taken together with the commonly assumed markedness hierarchies for various morphosyntactic features, such as that for number stated above (cf. Greenberg 1966: Ch. 3; Battistella 1990: Ch. 3; Givón 1995: Ch. 2), the constraint generates a whole family of testable predictions concerning the set of possible and impossible neutralization patterns in the languages of the world. Interestingly enough, I am not aware of any contribution where a substantial set of these predictions would have been tested against a broad cross-linguistic sample (cf. however, a typological survey of interactions between various grammatical categories in Aikhenvald and Dixon 1998). In a recent critical paper, Haspelmath (2006) suggests (a) that the label ‘markedness’ in fact covers a whole set of not always directly related notions pertaining to different levels of language and different theoretical and methodological approaches, (b) that almost all particular uses of this term may be essentially reduced to more literal and ‘substantial’ notions such as

Syncretisms and neutralizations involving case

109

‘phonetic difficulty’, ‘length of formal expression’, ‘frequency in texts’ etc., and (c) most importantly, that the recourse to markedness as an explanatory device in linguistics should be abolished in favour of these more basic and more directly accessible and measurable notions. In this chapter I bring forward some counterexamples that explicitly falsify the predictions of the markedness constraint on neutralization, coming from the domain of morphological case. The empirical evidence I adduce shows that the purported relation between markedness and neutralization is at best a very non-trivial one, and that Haspelmath’s claim that linguistic phenomena must be explained with recourse to more concrete notions than markedness is supported in this domain, too. In Section 3 I discuss material where case serves as the neutralizing category, while the data in Section 4 show case as the dominant category. The data discussed come mainly from Indo-European languages; however, this circumstance by no means invalidates its relevance: all the phenomena in question have emerged separately in different subgroups of the Indo-European family and neither could have been inherited from the proto-language, nor could have possibly evolved under mutual influence.

3. Case as the neutralizing category Examples where case, in accordance with the markedness constraint, is completely or partially neutralized in the context of the marked member of some other category (most often, number) are numerous. Here I give, for illustrative purposes only, two examples from non-Indo-European languages. In Yaqui, a Uto-Aztecan language spoken in Northern Mexico, the two cases (Direct and Oblique) are distinguished in the Singular but syncretized in the Plural (Lindenfeld 1973), see Table 1. Table 1. Case neutralization in the Plural in Yaqui Direct Oblique

Sg

Pl

misi ‘cat’ misi-ta

misi-m misi-m

In Old Georgian, oblique cases in the Plural shared the same cumulative suffix -ta, in contrast both to the Singular where the five cases were overtly distinguished and to the Nominative and Vocative Plural (Schanidse 1982), see Table 2.

110

Peter M. Arkadiev

Table 2. Case syncretism in the Plural in Old Georgian Nom Erg Dat Gen Ins Adv Voc

Sg k’ac-i ‘man’ k’ac-man k’ac-s k’ac-is k’ac-it k’ac-ad k’ac-o

Pl k’ac-n-i k’ac-ta k’ac-ta k’ac-ta k’ac-ta k’ac-ta k’ac-n-o

Turning to the data that pose more or less severe problems for the markedness constraint, let us first consider material from the Slavic languages, which, though by no means ‘exotic’, has been somehow disregarded by the proponents of the ‘classic’ markedness theory. In particular, we will see that in these languages Singular tends to show fewer overt case distinctions than Plural. In Russian, Polish (De Bray 1980a: 265), Czech (De Bray 1980a: 62), Slovene (De Bray 1980b: 339) and other Slavic languages, nouns belonging to the so called i-declension systematically distinguish more case forms in the Plural than in the Singular, see Table 3 (the total number of distinct word-forms is given in the last line of this and the following tables). Table 3. Slavic i-declension, ‘bone’ Russian Sg Nom kost’ Acc kost’ Gen kosti Loc kosti Dat kosti Ins kost’ju Distinct 3

Pl kosti kosti kostej kostjax kostjam kostjami 5

Polish Sg kość kość kości kości kości kością 3

Pl kości kości kości kościach kościom kośćmi 4

Czech Sg kost kost kosti kosti kosti kostí 3

Pl kosti kosti kostí kostech kostem kostmi 5

Slovene Sg kôst kôst kostî kósti kósti kostjó 4

Pl kostî kostî kostí kostéh kostém kostmí 5

It is clear from Table 3 that the major factor leading to this type of paradigmatic organization is the syncretism of Genitive, Locative, and Dative in the Singular; by contrast, in the Plural these cases have distinct endings. This syncretism goes back to the Common Slavic period (cf. Meillet 1934) and is due to the processes of phonological reduction that have obliterated the distinctions between the formerly non-identical endings. All

Syncretisms and neutralizations involving case

111

three cases were syncretized only in nouns with stable stress; nouns with mobile stress had a means to differentiate these cases (Dybo 1981: 28). The relics of this archaic pattern are still operative in Slovene and SerboCroatian (though the actual realization of the stress-based differentiation of cases is not identical in these languages and differs from that reconstructed for Common Slavic), but have been completely abolished not only in those languages where stress became fixed (Czech and Polish), but also in Russian where stress mobility still flourishes. In this respect data from (literary) Serbo-Croatian (De Bray 1980b: 253–254) are of special interest. In this language, Dative, Locative and Instrumental cases have been collapsed in the Plural of all declensions, thus reducing the maximal number of distinct case forms in the plural to four. In the i-declension with its non-distinct Nominative and Accusative, the number of case forms in the Plural is just three. Since Serbo-Croatian, much like Slovene, shows a prosodic distinction between Locative vs. Genitive and Dative in the Singular, the latter exhibits an extra case form in comparison to the Plural. This situation clearly goes against the panSlavic tendency to have richer case distinctions in the Plural of the i-declension than in the Singular. Remarkably, however, there is a tendency to use in the Instrumental Singular of this inflection class a novel form identical to that of Genitive-Dative Singular, thus ‘compensating’ for the extra distinction and reducing the number of case-forms in the Singular so that it does not exceed that of the Plural, see Table 4. Table 4. Serbo-Croatian i-declension Nom Acc Gen Dat Loc Ins Distinct

Sg stvâr ‘thing’ stvâr stvâri stvâri stvári stvâri (new form), stvârju (old form) 3 (4)

Pl stvâri stvâri stvárī stvárima stvárima stvárima 3

The data shown above suggest that the uneven distribution of case syncretisms between number subparadigms of the i-declension in the Slavic languages has led to the retention of a larger number of separate case forms in the Plural than in the Singular. This important and diachronically stable

112

Peter M. Arkadiev

characteristic of a whole inflection class must reflect something deeper than merely a result of an accidental phonological development. In Czech (De Bray 1980a: 65) and Polish (De Bray 1980a: 279) we may also find instances of complete neutralization of case distinctions in the Singular alongside with their retention in the Plural, see Table 5. Table 5. Complete neutralization of cases in Czech and Polish Czech Sg Nom paní ‘lady’ Acc paní Gen paní Loc paní Dat paní Ins paní

Pl paní paní paní paních paním paními

Polish Sg muzeum ‘museum’ muzeum muzeum muzeum muzeum muzeum

Pl muzea muzea muzeów muzeach muzeom muzeami

It is necessary to note that in Table 5 two substantially different situations are shown: whereas in Czech we are dealing with the result of a regular phonetically driven process which happened to lead to complete homophony of all case endings in the Singular of certain nouns (in other inflection classes and in different phonological environments these endings are distinct from each other), Polish demonstrates a phonologically unmotivated lack of declinability in a formally homogeneous class of borrowed lexemes. These examples show that complete neutralization in the context of the unmarked value of the dominant category is indeed possible, and that such ‘unlawful’ neutralizations may emerge due to quite unrelated processes. Better retention of case distinctions in the Plural in comparison to the Singular was also characteristic of the old Germanic languages. Both in Old Icelandic (Steblin-Kamenskij 1955: 50–75) and Old High German (Jolivet and Mossé 1942: 76–117), the Plural of all declension types distinguished three case forms, viz. Nominative, Genitive (with Accusative coinciding with one of the former) and Dative, whereas in the Singular, depending on the inflection class, the number of distinct forms could range from four to one, see Tables 6 and 7.

Syncretisms and neutralizations involving case

113

Table 6. Case syncretism in Old Icelandic

Nom Acc Gen Dat Distinct

‘arm’ Sg armr arm arms armi 4

Pl armar arma arma ǫrmum 3

‘chief’ Sg hersir hersi hersis hersi 3

Pl hersar hersa hersa hersum 3

‘edge’ Sg skǫr skǫr skarar skǫr 2

Pl skarar skarar skara skǫrum 3

‘heart’ Sg hjarta hjarta hjarta hjarta 1

Pl hjǫrtu hjǫrtu hjartna hjǫrtum

3

Table 7. Case syncretism in Old High German ‘day’ Sg Nom tag Acc tag Gen tages Dat tage Ins tagu Distinct 4

Pl taga taga tago tagum tagum 3

‘knee’ Sg kneo kneo knёwes knёwe knёwe 3

Pl kneo kneo knёwo knёwum knёwum 3

‘heart’ Sg hёrza hёrza hёrzen hёrzen hёrzen 2

Pl hёrzun hёrzun hёrzōno hёrzōm hёrzōm 3

‘mother’ Sg Pl muoter muoter muoter muoter muoter muotero muoter muoterum muoter muoterum 1 3

It is also worth noting that the paradigms of OHG words kneo ‘knee’ and muoter ‘mother’ display yet another kind of peculiar neutralization, viz. lack of a number distinction in the context of the putatively unmarked values of the category of case. Nominative and Accusative, in contrast to the oblique cases, are identical in both numbers. Old English, to which I now turn, exhibited the same pattern, see Plank (1987) and Section 4. In Old English (Smirnickij 1955: 213–244), where the process of disintegration of declension has gone farther than in other Germanic languages of that period, the situation was somewhat more complicated. Both number subparadigms disposed of a considerable number of case syncretism patterns (see Plank 1990), and the only form which was never homonymous with any other case form was the Dative Plural. The norm was for the Singular to exhibit no more distinct case forms than the Plural;4 see Table 8.

114

Peter M. Arkadiev

Table 8. Case syncretism in Old English

Nom

‘ship’ Sg Pl scip scipu

Acc

scip

Gen

scipes scipa

Dat scipe Distinct 3

scipu

scipum 3

‘care’ Sg Pl caru cara, care care cara, care care cara, carena care carum 2 2 (3)

‘friend’ Sg Pl wine wine, winas wine wine, winas wines winia, wina wine winum 2 3

‘sister’ Sg Pl sweostor sweostor sweostor sweostor sweostor sweostra sweostor sweostrum 1 3

If we try to find a common rationale for the above peculiar distribution of case syncretisms across Singular and Plural numbers in the Slavic and Germanic languages, the most obvious one seems to lie in the fact that Plural subparadigms in both groups of languages exhibit a much greater degree of morphological unification than Singular ones. This unification is reflected both in a greater similarity of desinences across different inflectional classes in the Plural than in the Singular, and in the greater mutual affinity of more abstract paradigmatic structures, such as patterns of homonymy between different cells of a paradigm. The tendency to reduce the number of different inflection classes in the Plural has obviously resulted in the reduction of the number of possible syncretisms5 (and thus in the retention of more overt case distinctions) in comparison to the Singular, which, at least in the languages discussed here, has been more or less successfully resisting paradigmatic unification (cf. for instance Russian, where at least three fairly different inflectional classes are distinguished in the Singular, but no such diversity is observed in the Plural). Moreover, it is tempting to speculate that the trend towards greater unification of the Plural subparadigms might be a consequence of the need to ‘minimize’ markedness – in so far as it is less marked for a word form to bear just values of number and case (and probably gender) than to be in addition explicitly specified for a more or less arbitrary declension class. Thus it is possible that here we are dealing with a wellbehaved instance of neutralization; i.e., values of a category (here, inflection class) are partially or wholly neutralized in the context of the marked value of another category (here, number), but not in the context of the unmarked value of the same category. Indeed, none of the languages discussed shows less inflection class distinctions in the Singular than in the

Syncretisms and neutralizations involving case

115

Plural. Somewhat paradoxically, we must conclude that ‘canonical’ neutralization of the values of one morphological feature (inflection class) may lead to a rather ‘non-canonical’ behaviour of another morphological feature (case); such trade-offs are, however, quite widespread in the languages of the world. Moving to Asia, we find quite a number of counterexamples to the markedness constraint on neutralization in the Iranian languages. Thus in two Pamir languages, viz. Wakhi (Paxalina 1975: 41–42) and Sarykoli (Paxalina 1966: 21), overt opposition between the Direct and Oblique cases on nouns is found only in the Plural, while in the Singular the case forms are identical, see Table 9. Table 9. Case neutralization in the Pamir languages Wakhi Sg Direct xūn ‘house’ Oblique xūn

Sarykoli Sg wern ‘ram’ wern

Pl xūn-išt xūn-ǝv

Pl wern-xɛyl wern-ef

An important difference between this case and the ones found in the Slavic and Germanic languages lies in the fact that in Wakhi and Sarykoli the case syncretism in the Singular is not restricted to a particular inflection class but is shared by all nominals except pronouns. The diachronic origins of this situation are not very clear (see Molčanova 1975: 222–223); however, one should bear in mind that, according to the very same markedness theory, Singular endings tend to have less phonological material than the Plural ones, and, as a consequence, Plural forms may be more resistant to phonetic erosion than the Singular ones. An interesting pattern involving gender in addition to number is found in Chali, a Central-Iranian dialect (Yar-Shater 1969: 75–76). In the Plural of both genders, Direct and Oblique cases are expressed by distinct suffixes; however, in the Singular the case opposition manifests itself overtly with Masculine nouns only, the Feminine ones showing case syncretism, see Table 10. Table 10. Case syncretism in Chali Masculine Sg Direct bar ‘door’ Oblique bar-e

Pl bar-e bar-ō

Feminine Sg barra ‘spade’ barra

Pl barr-e barr-ō

116

Peter M. Arkadiev

Two remarks are in order here. First, the situation in Chali seems to constitute an intermediate diachronic step between a more conservative system found in a closely related Eshtehardi dialect (Yar-Shater 1969: 79), showing no case syncretism across all numbers and genders, and a more advanced system found in the Takestani dialect (Yar-Shater 1969: 78) with no case opposition in the Plural at all. Second, one may see that in the Plural of Chali nouns there is no overt distinction between the Masculine and Feminine genders, the respective endings being identical for both types of nouns. Thus, one might hypothesize that the lack of case neutralization in the Plural is in a sense ‘compensated for’ by the neutralization of genders. An explanation involving a ‘trade-off’ between different types of neutralization leading to the overall reduction of markedness is hardly possible for the following data, coming from Burushaski, an isolate language of the Hindukush region (Berger 1974: 20–21; Klimov and Èdel’man 1970: 41– 42). Nouns and pronouns in Burushaski fall into four genders (Masculine, Feminine, and two genders for inanimate nouns) and normally distinguish two cases, viz. Direct and Oblique, except for the Feminine which ‘splits’ the general Oblique case into distinct Ergative and Genitive, see Table 11. Table 11. Case syncretism in Burushaski Masculine Dir Erg Gen

hiles ‘boy’ hiles-e hiles-e

Neuter dan ‘stone’ dan-e dan-e

Feminine gus ‘woman’ gus-e gus-mo

According to the predictions of the markedness constraint on neutralization, we would expect a situation opposite to that actually found in Burushaski, viz. more case distinctions in the unmarked genders and syncretism in the marked one, rather than vice versa (that Feminine in Burushaski is a marked gender is evidenced on purely language internal grounds by the very fact that it is opposed to the other three genders taken as a whole).

4. Case as the dominant category Instances of case being neutralized in the context of certain values of other categories, some of which were discussed in the previous section, are far less numerous than the opposite ones, where a certain value of case triggers neutralization of another category. One of the obvious examples comes from the Indo-European languages (cf. Russian in Section 2), where in the

Syncretisms and neutralizations involving case

117

oblique cases the distinction between Masculine and Neuter genders is absent. Outside the Indo-European family, one finds a neutralization of numbers in the Oblique cases in Chukotko-Kamchatkan languages, see Table 12 for a partial paradigm from Koryak (Žukova 1972: 101). Table 12. Number syncretism in Koryak Absolutive Ergative Locative Dative Allative Translative Ablative Comitative

Sg kaiŋǝ-n ‘bear’

Du kaiŋǝ-t kaiŋ-a kaiŋǝ-k kaiŋǝ-n kaiŋǝ-tǝŋ kaiŋǝ-pǝŋ kaiŋǝ-ŋqo ɣa-kaiŋ-a

Pl kaiŋ-u

Let us now go back to Old High German words for ‘knee’ and ‘mother’; see Table 7. We have seen that the ‘direct’ cases (Nominative and Accusative), which are usually considered to be unmarked with respect to the ‘oblique’ cases (Genitive and Dative) show a pattern of number neutralization that is unexpected given the markedness constraint. Such a situation was also found in Old English (see Plank 1987; this paper also gives an overview of number neutralization in some other Indo-European languages) and marginally in Gothic (Streitberg 1920: 112–113). In Old High German and Gothic (but not in Old English, see below) such patterns are rather sporadic and clearly constitute a result of phonological change, which has obliterated the difference between the once distinct endings. However, if we go back to Central Asia, we find there quite systematic instances of neutralizations in the context of the putatively unmarked case value. In several Indo-Iranian languages belonging to different genetic subgroups and geographical areas, e.g. Nuristanian Kati (Edelman 1983: 60– 61) and Dardic Kalasha (Edelman 1983: 204) spoken in the Hindukush region of Afghanistan and Pakistan, and West-Iranian Kurmanci spoken in Turkey and Iran (Cukerman 1986: 72), we find neutralization of number and sometimes also gender in the context of the Direct case, alongside with a stable retention of number contrast in the Oblique case; see Table 13 for Kati paradigms and Table 14 for Kalasha and Kurmanci6 paradigms.

118

Peter M. Arkadiev

Table 13. Number and gender syncretism in Kati Sg Direct Oblique

Pl

ǰuk ‘girl (fem.)’ ǰuk-a ǰuk-o

Sg

Pl

manči ‘man (masc.)’ manč-e manč-o

Table 14. Number syncretism in Kalasha and Kurmanci

Direct Oblique

Kalasha Sg Pl mōč ‘man’ mōč-es mōč-en

Kurmanci Sg Pl k’itêb ‘book’ k’itêb-ê k’itêb-an

Although it is true that the only systematic examples of neutralization triggered by the unmarked case value are found in the Indo-Iranian languages, they can be explained neither by genetic nor by areal factors. Indeed, since the history of this linguistic group is relatively well documented, it is obvious that the languages in question could not have inherited this kind of syncretism from their common ancestor; on the other hand, any contact influence between Kurmanci, situated in the Western end of the Indo-Iranian area, and the languages of Hindukush is highly improbable, too. Thus the only plausible way to explain the attested patterns is to assume an independent parallel development. The process responsible for the emergence of the peculiar syncretisms discussed here, as well as for the majority of other morphological changes in the Indo-Iranian languages during the last two millennia, is certainly phonological reduction and erosion, which have led to the eventual loss of monosyllabic inflectional morphemes (see Molčanova 1975: 220–222). Since the Direct case is less ‘marked’ than the Oblique case, it is only natural that its endings have less phonological weight, hence being more susceptible to loss through phonetic erosion. An example similar to the just discussed Indo-Iranian one was attested in Old French (Foulet 1928: 45; Pope 1934: 311–312), where the definite article did not distinguish number in the Direct case. Interestingly, no ambiguity usually arose since the article always combined with nouns, and most of these had paradigms which were syncretistic in other ways, so that the article and the noun jointly supplied all the information about the values of number and case of the phrase; see Table 15.

Syncretisms and neutralizations involving case

119

Table 15. Article and noun in Old French Sg Direct li murs ‘the wall’ Oblique le mur

Pl li mur les murs

Moreover, in Old French a paradigmatic type existed, where not only number values were neutralized in the context of the Direct case, but also case values were syncretized in the context of the Singular (Pope 1934: 311), cf. Table 16. Similar paradigms exist also in such Indo-Iranian languages as Dardic Tirahi (Edelman 1983: 193). Paradoxically, they conform to the predictions of markedness theory: that member of the paradigm which is the most marked ‘semantically’ (Oblique + Plural) is the most marked morphologically, being the only one to receive an overt ending. Table 16. Case and number syncretism in Old French Sg Direct pedre ‘father’ Oblique pedre

Pl pedre pedre-s

Finally, if we look more closely at Old English, we find that in this language number neutralizations “indeed are about as frequent as case neutralizations” (Plank 1987: 187), and are attested in all cases. However, the frequency and systematic nature of number neutralization increases in the following progression (Plank 1987: 187–188): Genitive (only one declension type) < Dative (several declension types, including adjectives and demonstrative pronouns) < Accusative (several declension types, including nouns, adjectives and pronouns) < Nominative and Accusative as a group (a whole array of nouns and adjectives of different declension types). Though number syncretism involving just the Nominative by itself is extremely rare in Old English, it is clear that the ‘unmarked’ direct cases in this language are more susceptible to number neutralization than the ‘more marked’ oblique cases.

5. Discussion and conclusion Though undoubtedly infrequent, the patterns of syncretism reported in this article are, in my opinion, very instructive. They clearly show that the explanations for different paradigmatic structures attested in the languages of

120

Peter M. Arkadiev

the world lie not in the rather abstract notions like relative ‘markedness’ of the values of grammatical categories but, among other things, in the nature of forces driving language change. Turning back to the Indo-Iranian case discussed in Section 4, we can draw the following ‘causal chain’ which may have led to the number syncretism in the Direct case; see Figure 1. Direct case endings in both numbers have less phonetic material than Oblique ones

→

Phonetic reduction leads to the loss of the light Direct case endings alongside with the retention of the heavy Oblique case endings

→

Number syncretism in the Direct case, no syncretism in the Oblique case

Figure 1. Emergence of number syncretism in Indo-Iranian

Of course, one may, if so inclined, add a clause like ‘Direct case is less marked than Oblique case’ to the left of Figure 1, but it is not obvious whether such a move is at all meaningful. That in all ancient IndoEuropean languages the case endings of Nominative and Accusative in all numbers have been shorter than those of the oblique cases and that this distribution goes back to the proto-language is a well-known fact (see, e.g., Brugmann 1904: 373–413). But the reasons for this distribution are perhaps to be sought in the rather obscure history of the Proto-Indo-European case system rather than in general theoretical constructs. Indeed, if we look at yet another Indo-Iranian language, Sarykoli, we see that the Direct Plural ending -xɛyl is much longer and heavier than the Oblique Plural ending -ef, which is certainly at odds with the expectation based on the relative ‘markedness’ of the cases in question. The explanation for this situation is, however, quite straightforward (see Molčanova 1975: 222–223): the Direct Plural ending is a relatively young morpheme stemming back to a suffix of nomina collectiva introduced into the paradigm probably in order to get rid of a situation when the only overtly marked value was Oblique Plural; the latter is expressed by the very old and phonetically eroded ending directly reflecting the Proto-Indo-Iranian Dative-Ablative Plural *-Vbhyas. Thus, in order to elucidate the motivations of this paradigm, it is necessary to look at the relative chronology rather than relative ‘markedness’. Similarly, if we try to make sense of the tendency to distinguish more cases in the Plural than in the Singular at least in some of the productive inflectional classes observed in the Slavic and Germanic languages (Section 3), we must take into account the fact that in these languages the Plural subparadigms have undergone paradigmatic unification to a much

Syncretisms and neutralizations involving case

121

greater extent than the Singular ones. The causes of this situation, however, are rather to be sought not in the mysterious ‘markedness’ of the Plural, but in the fact that the Plural paradigms in all old Indo-European languages have been more unified than the Singular ones since prehistoric times. If we consider the putative ways in which paradigmatic unification may proceed, it seems reasonable to assume that native speakers are more prone to confuse those endings that are phonologically similar, rather than those that share little or no phonological material. Thus, in the course of language change, unification more easily affects those parts of the paradigm that are already unified to some extent. The question concerning the ultimate origins of such an uneven distribution of affixes across different parts of the paradigm in Proto-Indo-European – in particular the reasons why in the Plural greater formal similarity was observed between markers with the same value for case but different values for inflection class rather than vice versa – is certainly a fascinating one, but I doubt whether answering it in terms of markedness is any better than pleading ignorance. Moreover, the data discussed here show that even if we take the relative markedness of different morphosyntactic values wholly seriously, it turns out that even those historical developments which conform to predictions of the markedness theory (greater unification of inflectional classes in the Plural than in the Singular, greater phonetic erosion of shorter endings of ‘unmarked’ cases that of the longer ones of the ‘marked’ cases etc.) may quite often lead to structures which violate the markedness constraint on neutralization. All this suggests that looking on particular motivations of particular linguistic structures is often more instructive than trying to explain them by aprioristic global theories. To recapitulate, I would like to stress once again that the goal of this chapter was to bring forward a collection of facts that may at first glance seem to be more or less disparate and weird but which, in my opinion, form a somewhat homogeneous picture. The data I have discussed, especially if further supported by material I am unaware of, indicates that the typological space of possible syncretisms involving morphological case (either as a neutralizing category or as a dominant category) is much broader than has been assumed by the proponents and advocates of the markedness approach to neutralization. I hope that the future typological and theoretical discussions of neutralization will not ignore the facts presented in this paper, and that non-aprioristic and non-simplistic explanations for them will be proposed.

122

Peter M. Arkadiev

Notes 1.

2.

3.

4. 5.

6.

This study was partially supported by the Russian Science Support Foundation and by the Presidium of the Russian Academy of Sciences. I am grateful to Alexandre Arkhipov, Patience Epps, Martin Haspelmath, Matthew Baerman, and an anonymous reviewer for useful comments on the earlier versions of this chapter. All faults and shortcomings are mine. Note that syncretism is a broader term covering instances where the number of overtly distinguished values of the neutralizing category is higher than zero; cf. again Russian adjectives, where in the Nominative all three genders are distinct, while in the oblique cases Masculine and Neuter fall together to the exclusion of the Feminine: bol’šogo doma / zadanija ‘of a large house / assignment (Genitive)’ vs. bol’šoj knigi ‘of a large book (Genitive)’. Ironically, the very assumption that Singular is ‘cognitively’ unmarked with respect to the Plural has been seriously challenged recently in the formal semantic literature, cf. for instance Sauerland, Anderssen and Yatsushiro (2005). Similarly, the status of the Nominative as the ‘unmarked’ member of the category of case is far from being unproblematic (see König 2006). I disregard the Instrumental case, relics of which were found in the declension of adjectives. It is interesting to consider some statistical data from four Slavic languages reported in Hamilton (1974). In Czech, Polish, Russian and Serbo-Croatian, the number of case syncretisms attested in the Singular subparadigm is considerably higher than that found in the Plural. Moreover, neither of these languages has reduced the number of case syncretisms in the Singular in comparison to the Common Slavic, while in Czech and Polish this number has even increased. By contrast, in the Plural subparadigm the number of case syncretisms has been reduced in all of the languages but Czech. It is necessary to note that in some other Kurdish dialects, e.g. in Suleimani and Mukri (MacKenzie 1961: 58), the number syncretism of the kind attested in Kurmanci has been ousted by means of extending the original Oblique Plural ending -ān to cover all Plural contexts regardless of case.

References Aikhenvald, Alexandra Y., and Robert M. W. Dixon 1998 Dependencies between grammatical systems. Language 74 (1): 56– 80. Aissen, Judith 1999 Markedness and subject choice in optimality theory. Natural Language and Linguistic Theory 17 (4): 673–711.

Syncretisms and neutralizations involving case

123

Andrews, Edna 1990 Markedness Theory: The Union of Asymmetry and Semiosis in Language. Durham (NC): Duke University Press. Baerman, Matthew, Dunstan Brown, and Greville G. Corbett 2005 The Syntax-Morphology Interface. A Study of Syncretism. Cambridge: Cambridge University Press. Battistella, Edwin L. 1990 Markedness: The Evaluative Superstructure of Language. New York: SUNY Press. 1996 Logic of Markedness. Oxford: Oxford University Press. Béjar, Susana, and Daniel Currie Hall 1999 Marking markedness: The underlying order of diagonal syncretisms. In Proceedings of the 15th Eastern States Conference on Linguistics, 1–12. Ithaca: Cornell University Press. Berger, Hermann 1974 Das Yasin-Burushaski (Werchikwar). Grammatik. Texte. Wörterbuch. Wiesbaden: Otto Harrassowitz. Bierwisch, Manfred 1967 Syntactic features in morphology: General principles of so-called pronominal inflection in German. In To Honor Roman Jakobson. Essays on the Occasion of His Seventieth Birthday. Vol. I, 239–270. The Hague, Paris: Mouton. Boeder, Winfried 1976 Morphologische Kategorien. In Grammatik. Akten des 10. Linguistischen Kolloquiums, Kurt Braunmüller, Wilfried Kürschner (eds.). Bd. 2, 117–126. Tübingen: Max Niemeyer Verlag Brugmann, Karl 1904 Kurze vergleichende Grammatik der indogermanischen Sprachen. Strassburg: Trübner. Croft, William C. 1990 Typology and Universals. Cambridge: Cambridge University Press. Cukerman, Isaak I. 1986 Xorasanskij kurmandži. [The Kurmanci of Khorasan]. Moscow: Nauka. De Bray, R. G. A. 1980a Guide to the West Slavonic Languages. 3rd ed. Columbus (Ohio): Slavica Publishers. 1980b Guide to the South Slavonic Languages. 3rd ed. Columbus (Ohio): Slavica Publishers. Dixon, Robert M. W. 1979 Ergativity. Language 55 (1): 59–138. Dybo, Vladimir A. 1981 Slavjanskaja akcentologija. [Slavic Accentology]. Moscow: Nauka.

124

Peter M. Arkadiev

Eckman, Fred R., Edith A. Moravcsik, and Jessica R. Wirth (eds.) 1983 Markedness. New York; London: Plenum Press. Edelman, Dzhoi I. 1983 The Dardic and Nuristani Languages. Moscow: Nauka. Foulet, Lucien 1928 Petite syntaxe de l’ancien français. 3-ème éd. revue. Paris: Champion. Givón, Talmy 1995 Functionalism and Grammar. Amsterdam, Philadelphia: John Benjamins. Greenberg, Joseph H. 1966 Language Universals. The Hague etc.: Mouton. Hamilton, W. S. 1974 Deep and surface changes in four Slavic noun systems. Linguistics 127: 27–74. Haspelmath, Martin 2006 Against markedness (and what to replace it with). Journal of Linguistics 42 (1): 25–70. Hjelmslev, Louis 1935–37 La catégorie des cas. Étude de grammaire générale. P. I. København, 1935; P. II. København, 1937. Jakobson, Roman 1932 Zur Struktur der russischen Verbums. In Charisteria Guilelmo Mathesio quinquagenario a discipulis et Circuli Linguistici Pragensis sodalibus oblata, 74–84. Prague. 1936 Beitrag zur allgemeinen Kasuslehre. Gesamtbedeutungen der Russischen Kasus. Travaux du cercle linguistique de Prague 6: 240– 288. Janda, Laura 1995 Unpacking markedness. In Linguistics in the Redwoods: The Expansion of a New Paradigm in Linguistics, Eugene Casad (ed.), 207– 233. Berlin: Mouton de Gruyter. Jolivet, Alfred, and Fernand Mossé 1942 Manuel de l’allemand du Moyen Âge des origines au XVIe siècle. Aubier, Paris: Éditions Montaigne. Kibrik, Aleksandr E. 1997 Beyond subject and object: Toward a comprehensive relational typology. Linguistic Typology 1 (3): 279–346. Klimov, Georgij A., and Dzhoj I. Èdel’man 1970 Jazyk burušaski. [The Burushaski Language]. Moscow: Nauka. König, Christa 2006 Marked nominative in Africa. Studies in Language 30 (4): 705–782.

Syncretisms and neutralizations involving case

125

Lindenfeld, Jacqueline 1973 Yaqui Syntax. Berkeley: University of California Press. MacKenzie, David Neal 1961 Kurdish Dialect Studies. Vol. 1. London: Oxford University Press. Meillet, Antoine 1934 Le slave commun. 2-ème éd. Paris: Champion. Molčanova, E. K. 1975 Kategorija čisla [The category of number]. In Opyt istorikotipologičeskogo issledovanija iranskix jazykov [An HistoricalTypological Study of the Iranian Languages], V. S. Rastorgujeva (ed.), Vol. 2, 200–249. Moscow: Nauka. Moravcsik, Edith A., and Jessica R. Wirth 1983 Markedness – An overview. In Eckman, Moravcsik, and Wirth (eds.), 1–12. Paxalina, Tatiana N. 1966 Sarykol’skij jazyk [The Sarykoli Language]. Moscow: Nauka. 1975 Vakhanskij jazyk [The Wakhi Language]. Moscow: Nauka. Plank, Frans 1987 Number neutralization in Old English: Failure of functionalism? In Explanation and Linguistic Change, Willem Koopman, Frederike van der Leek, Olga Fischer, Roger Eaton (eds.), 177–238. Amsterdam, Philadelphia: John Benjamins. 1990 Paradigm arrangement and inflectional homonymy: Old English case. In Papers from the 5th International Conference on English Historical Linguistics, Sylvia M. Adamson, Vivien A. Law, Nigel Vincent, Susan Wright (eds.), 379–406. Amsterdam, Philadelphia: John Benjamins. Pope, Mildred K. 1934 From Latin to Modern French with especial consideration of AngloNorman. Phonology and Morphology. Manchester: Manchester University Press. Sauerland, Uli, Jan Anderssen, and Kazuko Yatsushiro 2005 The Plural is semantically unmarked. In Linguistic Evidence. S. Kepser, M. Reis (eds.), 413–434. Berlin: Mouton de Gruyter. Schanidse, Akaki 1982 Altgerogisches Elementarbuch. Teil I. Grammatik der Altgeorgischen Sprache. Tbilissi: Tbilisis universit’et’is gamomcemloba. Silverstein, Michael 1976 Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, R. M. W. Dixon (ed.), 112–171. Canberra: Australian Institute of Aboriginal Studies.

126

Peter M. Arkadiev

Smirnickij, Aleksandr I. 1955 Drevneanglijskij jazyk [A Grammar of Old English]. Moscow: Inostrannaja literatura i jazyki. Smith-Stark, T. Cedric 1974 The plurality split. In Papers from the 10th Regional Meeting of the Chicago Linguistics Society, 657–671. Steblin-Kamenskij, Mixail I. 1955 Drevneislandskij jazyk [A Grammar of Old Icelandic]. Moscow: Inostrannaja literatura i jazyki. Streitberg, Wilhelm 1920 Gotisches Elementarbuch. 5. & 6. Aufl. Heidelberg: Carl Winter. Trubetzkoy, Nikolay 1939 Grundzüge der Phonologie. Travaux du Cercle Linguistique de Prague 7, Prague. Yar-Shater, Eshan 1969 A Grammar of Southern Tati Dialects. (Median Dialect Studies, 1.) The Hague, Paris: Mouton. Žukova, Alevtina N. 1972 Grammatika korjakskogo jazyka. [A Grammar of Koryak]. Leningrad: Nauka.

Towards a typology of ‘attachment’ markers: Evidence from East Caucasian languages Dmitry Ganenkov

1. Introduction1 1.1. Attachment marking as a semantic domain It is well known that virtually all East Caucasian (Nakh-Daghestanian)2 languages possess a rich system of spatial marking on nouns. Morphologically, a nominal spatial form typically consists of two separately coded categories, localization and orientation. Localization denotes a spatial domain with respect to a Ground, where a Figure is located. The set of localizations in East Caucasian typically includes meanings such as ‘inside’, ‘on’, ‘behind’, ‘near’, and ‘under’, more rarely ‘in front of’ and varies in number from four to nine. Orientation conveys the notion of movement, indicating direction of the motion of a Figure with respect to a Ground.3 Cf. the following examples4 of nominal forms from Qunqi Dargwa: Table 1. Qunqi Dargwa: two localizations, three orientations qal-li-sa house-OBL-ANTE5 (to) in front of the house

qal-li-sa-b

qal-li-sa-rka

house-OBL-ANTE-ESS in front of the house

house-OBL-ANTE-EL from in front of the house

qal-li-šːa

qal-li-šːa-b

qal-li-šːa-rka

house-OBL-APUD to the house

house-OBL-APUD-ESS near the house

house-OBL-APUD-EL from the house

This chapter discusses attachment marking in the languages of the East Caucasian family. This family of twenty-nine distantly related languages seems to be natural choice in this kind of study, since a characteristic feature of East Caucasian languages is that many languages of the family formally split ON-relation into two localization values, i.e. divide, for example, the scope of use of the Russian preposition na or the English preposition on between two different morphemes. Thus, for example, Agul

128

Dmitry Ganenkov

possesses two localization markers: the ‘Super’ localization with the marker -l and the ‘Cont’ localization with the marker -k. The following examples demonstrate the use of these markers: Agul (East Caucasian, Lezgic) (1) ustːul.i-l kitab ald-e-a. table-SUPER book SUPER-be-PRS ‘There is a book on the table.’ (2)

cil.i-k iškːil k-e-a. wall-CONT picture CONT-be-PRS ‘There is a painting (hanging) on the wall.’

In (1), localization marker -l is used to mark location on the upper surface of the table, while in (2) localization marker -k used to express location of a painting hanging on the wall. Traditionally, markers like the former are called Super, whereas markers similar to the latter are labelled Cont. That languages can differ in how they conceptualize ON-relations was first observed by Bowerman (Bowerman and Pederson 1992; Bowerman 1996; Bowerman and Choi 2001). She noticed that the division of labour between ‘in’ and ‘on’ markers varies across languages, so that there are some spatial configurations expressed by the ‘in’ case marker in Finnish which are conceptualized as ON-relations by speakers of Russian and English (such as ‘a ring on finger’ or ‘a wart on forehead’). Moreover, there are languages that have a distinct marker for this intermediate domain. Well-known examples of the latter type are German (auf vs. an) and Dutch (op vs. aan).6 Recently, Levinson and Meira (2003: 513) pointed out that in the earlier works by Bowerman “the ATTACHMENT area was seen as a category invaded by the encroachment of large IN and ON categories” and proposed to distinguish a separate cross-linguistically relevant category ‘attachment’. As they point out, “[a]dpositions specialized to attachment scenarios turn out to be common enough cross-linguistically” (Levinson and Meira 2003: 496), so that attachment “is an important category that tends to be recognized in language after language” (Levinson and Meira 2003: 514). Although there exist a number of studies investigating attachment markers in individual languages, including the West Germanic languages German and Dutch as well as several languages outside Europe, a typology of attachment marking is still to be built, since almost no comparable data

Towards a typology of ‘attachment’ markers

129

from typologically, genetically and areally diverse languages are available to make valid generalizations about possible variation within this domain.7 Furthermore, I am not aware of any study discussing attachment marking in a cross-linguistic perspective except for preliminary observations found in Bowerman’s papers. Contributions to the recent volume edited by Levinson and Wilkins (2006) also do not discuss cross-linguistic distinctions made within the attachment domain. Only the concluding chapter briefly touches on this issue as a possible line of investigation (Levinson and Wilkins 2006: 520). 1.2. Attachment marking in East Caucasian languages The goal of this chapter is to investigate patterns of variation within the domain of attachment, isolate a number of cross-linguistically relevant parameters of variation (spatial configurations), and reveal semantic conditions determining the choice of a particular spatial marker in a particular spatial configuration. The central issue at question in this chapter is the discussion of the specific parameters that influence the choice of attachment markers in East Caucasian languages. The chapter, therefore, is a first (typologically oriented) description of attachment markers in the East Caucasian family and should be considered as a first step towards the crosslinguistic study of attachment. Language data used in this chapter were collected mostly from native consultants on the basis of a questionnaire that contained Russian sentences describing various spatial ONconfigurations, as well as from reference grammars, dictionaries and other secondary sources. The questionnaire was first compiled using spatial configurations presented in (Bowerman and Pederson 1992; Bowerman 1996; Bowerman and Choi 2001; Levinson and Meira 2003) and then modified and supplemented with additional configurations in the course of the fieldwork. From the point of view of the classification outlined above, East Caucasian languages belong to the Dutch/German type, i.e. they have a distinct attachment marker opposed to a standard ON-marker, cf. examples (1) and (2) from Agul.8 In this respect, they differ both from languages like Finnish which have no distinct attachment marker, but use an ‘in’-marker in some of ON-configurations, on the one hand, and from languages like Tiriyó (Meira 2006) which have several markers specifying different attachment relations, on the other hand.

130

Dmitry Ganenkov

The examination of various ON-configurations in East Caucasian languages that distinguish between two ON-markers shows that they convey a basic semantic contrast between the two following spatial configurations: (i) (ii)

location of a Figure on the supporting (horizontal) surface of a Ground, e.g. ‘book on the table’, location of a Figure attached to a Ground, so that the latter prevents the former from falling down, e.g. ‘picture on the wall’.

This means that in all East Caucasian languages that make a distinction between the two kinds of markers (i) is expressed only by the Super localization, whereas (ii) is described by the Cont localization. This seems to be evidence that these configurations are prototypical for Super and Cont, respectively. All other instances of ON-configurations can formally vary across these languages and are distributed between these two formal means according to certain parameters in each language. Table 2 lists the East Caucasian languages where the distinction between Super and Cont has been found. (The table also shows the genetic affiliation of the relevant languages.) Table 2. East Caucasian languages distinguishing Super and Cont Genetic subgroup Avar Andic Tsezic Dargwa Lak Lezgic Khinalug Nakh

Language — Andi, Bagvalal, Godoberi, Karata, Tindi, Chamalal Bezhta, Hunzib, Tsez, Hinukh, Khwarshi — — Agul, Lezgian, Tabasaran, Rutul, Tsakhur — —

As this table shows, the attachment category is found mostly in Andic, Tsezic and Lezgic languages, whereas Nakh, Avar, Lak, Dargwa and Khinalug apparently lack this distinction. It should be noted here that Avar and Dargwa divide into a number of very divergent dialects, so that some Dargwa dialects (e.g. Kubachi, Icari, Chirag) should be better considered as separate languages. In this study, I investigated only Standard Avar and Standard Dargwa, none of which has the distinction in question. Nonetheless, it may well happen that some

Towards a typology of ‘attachment’ markers

131

other Avar or Dargwa dialects formally distinguish attachment and nonattachment configurations. In what follows, I will take a closer look at various spatial ONconfigurations and describe the properties that can determine the choice of marker across the family. Section 2 discusses semantic properties of typical configurations that influence the choice between attachment and nonattachment markers. Section 3 introduces various ‘non-typical’ ONconfigurations and describes their behaviour with respect to formal marking. In Section 4, I briefly touch on another point which has not been discussed in the previous literature, viz. the interplay between static location and motion and its reflection in the formal spatial marking. Section 5 investigates some non-attachment uses of attachment markers in East Caucasian languages. In Section 6, I conclude the discussion by summarizing spatial configurations that can be expressed by means of attachment markers, and by proposing a semantic map of ON-configurations that constrains and, to some extent, predicts the possible distribution of spatial markers in East Caucasian languages.

2. Typical ON-configurations Typical ON-configurations include various spatial configurations in which a Figure is located on the flat surface of a Ground. In particular, both prototypical configurations (i) ‘a Figure on the supporting (horizontal) surface of a Ground’ and (ii) ‘a Figure attached to a Ground, so that the latter prevents the former from falling down’ belong here. The study of the behaviour of Super and Cont localizations in East Caucasian languages shows that there are two main factors determining the choice between markers, viz. (a) attachment of a Figure to a Ground by some means other than gravity and (b) location of a Figure on a Ground determined by gravity. 2.1. Attachment The first important property of a spatial ON-configuration that conditions the choice between the two spatial markers is attachment of a Figure to the surface of a Ground. A great number of languages in my study use this parameter to define the scope of use of the Cont localization. A good example of languages of this type is Bezhta, where the use of the Cont localization is constrained by this parameter, i.e. Cont is used in all attachment

132

Dmitry Ganenkov

configurations, but cannot express non-attachment relations. Cf. the following examples: Bezhta (East Caucasian, Tsezic) (3) jõso-ł gababek’e boχon-na wall-CONT carpet hang-RES ‘The carpet hangs on the wall.’

gej. (Xalilov 1995: 48) COP

(4)

χiʟ.a-ł biːjk’i micuc’-na. (Xalilov 1995: 57) trousers-CONT agrimony attach-RES ‘An agrimony (pod) stuck to trousers.’

(5)

a. šuša.li-ł ʟibo gej. bottle-CONT paper COP b. šuša.li-ʟ’a ʟibo gej. bottle-SUPER paper COP (a=b) ‘There is a label on the bottle’.

However, no language applies this ‘attachment’ vs. ‘non-attachment’ opposition to choose a marker in a particular configuration in an entirely straightforward manner. Most languages of the family draw finer distinctions within the attachment domain. It is often possible in the languages of the family to express some attachment configurations by means of the Super localization. Even in Bezhta some attachment contexts allow Super, along with Cont (cf. example (5b) above): Most clearly, this further split of the attachment domain into two types of configurations can be seen from Agul data. Here, the demarcation line between use of Super and Cont in attachment contexts is defined by the extent to which the attachment relation between the Figure and the Ground is salient in a given configuration. The Cont localization is used in configurations that strongly imply that the attachment relation between the Figure and the Ground prevents them from separating. In other words, it is salient that the Figure is located on the surface of the Ground due to an attachment relation. Cf. example (6) from Agul: Agul (East Caucasian, Lezgic) (6) we k’eb.ela-k ʡurk’ k-üqː-une. your headscarf-CONT resin CONT-attach-PST ‘Resin stuck to your headscarf.’

Towards a typology of ‘attachment’ markers

133

In contrast, the Super localization is used in configurations where the presence of a sticky substance (or some other means to attach the Figure to the Ground) is not so salient, e.g. ‘lable on the bottle’, ‘stripes on the sleeve’, etc. Most typically, in such configurations the Figure is a thin flat object attached to the flat surface of the Ground, so that there is no special visual evidence of the attachment relation between them. Rather, the configurations are viewed by default as simple location of a Figure on a Ground with no focus on the attachment relation. A particularly important instance includes such configurations where the Figure becomes (or functions as) a part of the Ground. Interestingly, although Super can be replaced by Cont in non-salient attachment contexts, this replacement dramatically changes the interpretation of the spatial configuration denoted by the sentence. A semantic effect usually arises such that the Figure is conceived as being in a very tight attachment relation, so that it is very difficult to separate it from the Ground. In the configuration ‘label on the bottle’, for instance, the use of Cont was interpreted by speakers as indicating that there remain little pieces of the label on the bottle that are very difficult to remove from the surface. In fact, this distinction between two types of attachment configurations (salient vs. non-salient) is itself not so clear-cut. Sometimes, it is up to the speaker to decide how to conceptualize a given configuration, so that both readings (and both localizations) are equally possible. Compare the following example from Agul where the same situation allows both markers, but Super simply indicates that the Figure is located on the Ground, while Cont highlights the presence of the attachment relation between the Figure and the Ground: Agul (East Caucasian, Lezgic) (7) a. markːa al-urqːarq’-e kːanwert.i-l. stamp SUPER-attach-IMP envelope-SUPER b. markːa k-ürqːarq’-e kːanwert.i-k. stamp CONT-attach-IMP envelope-CONT (a=b) ‘Stick a stamp to the envelope.’ Most languages of the family also distinguish between salient and nonsalient attachment configurations, but put it in a slightly different way than Agul does. They do not draw a sharp distinction between the scope of use of Super and Cont. Instead, they typically use Cont in all attachment configurations regardless of salience of attachment relation, but in addition to

134

Dmitry Ganenkov

this, non-salient attachment configurations can be expressed by means of Super as well. However, no additional semantic effects similar to that in Agul arise in other languages. Compare examples (5a) and (5b) from Bezhta, which demonstrate the absence of any clear semantic contrast between two formal means. The only languages that do not use Super in attachment configurations at all are Tsez, Hinukh and Khwarshi (all closely related West Tsezic languages). Here, attachment configurations can be expressed only by Cont irrespective of the degree to which the attachment relation is salient. This is due to the fact that the localization Super in Tsez is strongly associated with location on a surface supporting a Figure from underneath (as is discussed in the next section). 2.2. Gravity The other major parameter that influences the choice between Super and Cont is gravity, which divides all ON-configurations into two types. Super includes such spatial configurations as ‘book on the table’, in which a Ground supports a Figure due to gravity, i.e. the Figure is located on the upper surface of the Ground. Cont includes configurations where a Ground is able to hold a Figure contrary to gravity, i.e. normally the Figure cannot be located in such position, unless it is attached to the Ground or otherwise has the ability to not fall from a non-supporting surface. An example of a language that uses this property to delimit Super and Cont in ON-configurations is Tsez, where the gravity configurations are expressed by Super (example 8), whereas non-gravity configurations are expressed by Cont (examples 9–11): Tsez (East Caucasian, Tsezic) (8) stol.jo-ʟ’ t’ek joł. table-SUPER book is ‘There is a book on the table.’ (9)

karta qido-q jok’a map wall-CONT hang(tr.) ‘to hang the map on the wall’

Towards a typology of ‘attachment’ markers

135

(10) ged.mo-q cet’a-bi raqˁa shirt-CONT button-PL sew ‘to sew buttons on the shirt’ (11) qido-q t’ut’ joł. wall-CONT fly is ‘There is a fly on the wall.’ In most cases, the division made by the gravity parameter corresponds exactly to that made by the attachment parameter. However, there is an important type of ON-configurations, where the two parameters diverge. It is autonomous location of an animate Figure on a non-supporting (vertical) Ground (‘fly on the wall / ceiling’). In such contexts gravity points to the Cont localization, while attachment speaks in favour of the Super localization. In East Caucasian languages, gravity is used as a demarcation line much more rarely than attachment. In fact, only three closely related West Tsezic languages (Tsez, Hinukh, and Khwarshi) use gravity as the only property determining the choice between Super and Cont; i.e., all configurations where a Figure does not fall down in spite of gravity are expressed by Cont, while Super is used only in gravity configurations. Cf. examples (8–11) above from Tsez. Interestingly, both attachment and gravity are employed in Rutul spatial markers, since Cont is used in non-gravity contexts, while Super is used in non-attachment and non-salient attachment contexts. This results in interchangeability of these markers in ‘autonomous location on a nonsupporting surface’ contexts and non-salient attachment configurations: Rutul (East Caucasian, Lezgic) (12) a. mas.alɨ-k dɨd ki wall-CONT fly CONT.be b. mas.al-ɨ: dɨd ʁa wall-SUPER fly SUPER.be (a=b) ‘There is a fly on the wall.’

3. Other ON-configurations This section describes other ON-configurations expressed by Cont and Super localizations in East Caucasian languages. Usually these non-typical configurations lack some of the properties listed earlier in the beginning of

136

Dmitry Ganenkov

Section 2: the Ground has no flat surface (‘wet clothes on the line’, ‘coat on the hook’), the Figure is a part (‘a scratch on the mirror’) or a natural extension (‘leaves / apples on a branch’, ‘wart on the forehead’) of the Ground, or the manner of fixation of the Figure on the Ground is not typical (‘ring on finger’). Most of these ‘non-typical’ configurations can be expressed by both Super and Cont in the languages of the family. Below, I try to reveal the motivations for both options. 3.1. ‘Marks on a surface’ This group includes such non-typical configurations where the Ground has a flat surface and the Figure is a mark / image on the surface or a substance smeared on the surface. These two kinds of Figures slightly differ in how they are coded in the languages. The latter is invariably expressed by the Cont marker in all languages of the family, although some languages allow Super as well. Consider the following example: Bezhta (East Caucasian, Tsezic) miqaː-ʟ’a krem-i m=eχen-nä. (13) öž-di diela miqaː-ł / son-ERG my back-CONT back-SUPER cream-COP CL=put-RES ‘My son put (some) cream on my back.’ This suggests that such configurations are very close to the prototypical Cont configuration ‘location of a Figure attached to a Ground preventing it from falling down’. Note that many languages lack a distinct verb meaning ‘smear’ and the verb ‘put’ is used in such contexts, so that the semantics of smearing is conveyed exclusively by the Cont marker (along with the lexical noun ‘cream’). As for marks and images on a surface, the use of both Cont and Super is attested in my study. Compare example (14) from Tabasaran, where only the Cont localization can be used, and example (15) from Andi, which allows only Super in such contexts: Tabasaran (East Caucasian, Lezgic) (14) sufra.ji-k / cal.i-k q’alušn-ar table_cloth-CONT wall-CONT spot-PL ‘There are spots on the table-cloth / wall.’

k-ad. CONT-be

Towards a typology of ‘attachment’ markers

137

Andi (East Caucasian, Andic) (15) q’enno-lʔa t’ank’il wall-SUPER spot ‘spot on the wall’ From a semantic point of view, these two possibilities are well motivated. The use of Cont is conditioned by the factor that the Figure in such configurations cannot be separated from the surface of the Ground, i.e. the Figure is essentially attached to the Ground. However, the opposite option – the use of Super – is also semantically motivated, since in such configurations there is no attachment of the Figure in the sense of its existing independently of the Ground. Interestingly, three languages in my study, viz. Agul, Tsakhur and Rutul distinguish two kinds of such mark / image configurations. The first one simply describes location of a mark / image on a surface and is expressed by the Super localization in these languages. The second kind of configurations coded by the Cont localization implies that a mark / image on the surface of a Ground is a distinctive feature of the Ground. Compare examples from Agul: Agul (East Caucasian, Lezgic) (16) a. ħüni-n qː˳al.a-l ǯag˳ar t’ant’a cow-GEN side-SUPER white spot b. ħüni-n qː˳al.a-k ǯag˳ar t’ant’a cow-GEN side-CONT white spot ‘The cow has a white spot on its side.’

alde-a. SUPER.be-PRS k-e-a. CONT-be-PRS

3.2. Parts, lumps and natural extensions Another extension of localization Cont that is often found in East Caucasian is to mark the location of a part (Figure) on the whole (Ground). Only a very few languages (namely, Tabasaran, Tsakhur, and Andi) cannot use Cont to mark the location of a part on the whole, and the other languages do so very frequently. Typically, only those parts that ‘hang down’ from the whole or another part, such as ‘fingers on the hand’, ‘ears on the head’, ‘handle on a pan’ allow use of Cont, as in example (17) from Godoberi:

138

Dmitry Ganenkov

Godoberi (East Caucasian, Andic) (17) leli-č’u ĩštuda set’il-da. hand-CONT five fingers-COP ‘There are five fingers on the hand.’ It should be noted here that the location of parts is linked to one more use of Cont in Lezgic languages. Some Lezgic languages, namely, Lezgian, Agul, Tabasaran, and Rutul, conceptualize possession in terms of space and express possessive relations by means of spatial markers. Alienable possession is expressed either by the Post (‘behind’) or the Apud (‘near’) localizations on the possessor, distinguishing between permanent and actual possesion. However, possession of (body) parts is usually expressed by means of a spatial marker that directly reflects actual location of a part with respect to the whole, as in Agul: Agul (East Caucasian, Lezgic) (18) qːizil.di-n silew-ar a-a ruš.a-ʔ. gold-GEN tooth-PL IN.be-PRS girl-IN ‘The girl has golden teeth.’ (lit. ‘There are golden teeth in the girl.’) (19) gi-l č’ar-ar al-dawa. this-SUPER hair-PL SUPER.be-PRS:NEG ‘He has no hair.’ (lit. ‘There is no hair on him.’) This means that the possession of parts ‘hanging down’ from the possessor (fingers, ears, nose, tail, handle) can also expressed by Cont, as in the following Rutul examples: Rutul (East Caucasian, Lezgic) (20) tɨlɨ.je-k ǯɨbɨr ki. dog-CONT tail CONT.be ‘The dog has a tail.’ (21) hemi bedri.je-k t’ul this bucket-CONT handle ‘This bucket has no handle.’

ki-diš. CONT.be-NEG

Another kind of ON-configuration is the location of natural extensions like ‘leaves / apples on a branch’. Although it is semantically very similar to the location of parts more generally, it is nevertheless sometimes ex-

Towards a typology of ‘attachment’ markers

139

pressed by different formal means; for example, a language may express the location of parts by means of Cont, but use Super for natural extensions. Note that the opposite possibility, i.e. location of a part expressed by Super, and natural extension expressed by Cont, is not attested. Finally, one more type of configurations belongs here, namely, location of ‘lumps’ (lump, wart, mole, wound) on the skin or body: Godoberi (East Caucasian, Andic) (22) leli-č’u qirisː ida. hand-CONT wound COP ‘There is a wound on the hand.’ As with marks / images on a surface, both options of formal marking (Super and Cont) for parts, ‘lumps’, and natural extensions are clearly motivated. Here, the motivation for Cont is that the Figure in such configurations looks like it is attached to the Ground. On the other hand, Super is also possible, since there is no real attachment relation between the Figure and the Ground. Although this type of configuration is again quite similar to that involving parts and natural extensions, it behaves differently from these from a cross-linguistic perspective. For example, Tsakhur uses Super for location of parts and natural extensions, but Cont for location of lumps. On the other hand, Bezhta expresses location of parts by means of Cont, while location of lumps is exressed by Super (natural extensions allow both localizations here). The third possibility, which is attested, for instance, in Tabasaran and Godoberi, is to code all three types of configurations by the same localization (Super in Tabasaran and Cont in Godoberi). It should be also noted that, as in the case of marks / images on a surface, Agul can use both localizations to mark location of lumps. The distribution again depends on whether a lump is regarded as a distinctive feature of a Ground, in which case Cont is used, or simply location of a lump on a surface is described, which is marked by Super. 3.3. Manner of attachment The next class of non-typical ON-configurations comprises configurations involving various non-typical manners of location of a Figure on a Ground. Contrary to the configurations discussed above, these do not involve location on a supporting surface or location due to a sticky substance. Two

140

Dmitry Ganenkov

basic types of contexts belong here. The first consists of encirclement with contact, found in contexts like ‘ring on the finger’, ‘glove on the hand’. The second involves a Figure hanging on a hook or a line (‘coat on the hook’, ‘wet clothes on the line’). The motivation for the use of Cont in such configurations is again that the Figure is not located on a supporting surface, but is somehow fixed to the Ground. Moreover, as I have shown above, Cont may be used in configurations where the Figure hangs down from the Ground, which is the case of configurations like ‘coat on the hook’ or ‘clothes on the line’. However, none of these configurations has the component of real attachment, such that the Figure is prevented from separating from the Ground. Only few languages express these ON-configurations by means of Cont. Both types are expressed by Cont in Tsez, Bezhta, and Godoberi, cf. examples (23) and (24). All the other languages use Super with these configurations. Godoberi (East Caucasian, Andic) (23) set’il.a-č’u set’ibešːi-da. finger-CONT ring-COP ‘There is a ring on the finger.’ (24) den hank’u-č’u paltu irχːi. I hook-CONT coat hung ‘I hung my coat on the hook.’ In addition to the above uses of Cont, some East Caucasian languages use the Cont localization in configurations where the Figure is attached to the Ground by means of a string / rope, and is supported from underneath not by the Ground, but by some other object. Most languages do not use either Cont or Super in these configurations. However, such uses of Cont are found in Tsez and Bezhta, as in the following example: Bezhta (East Caucasian, Tsezic) (25) χöχi-ł öže ico-ro. tree-CONT boy tie-Pst ‘They tied the boy to a tree.’

Towards a typology of ‘attachment’ markers

141

4. Detachment configurations So far I have considered various ON-configurations and isolated several parameters of variation (spatial configurations) across the family. However, the previous discussion was confined to static configurations, i.e. contexts that describe the location of a Figure on a Ground. In this section, I describe the behaviour of the localizations in elative (‘motion from’) contexts. It is very typical of East Caucasian languages to use the Cont localization more extensively than the Super in elative configurations. This concerns, first of all, non-salient attachment and location of parts on the whole. In Agul, for instance, in contexts describing separation of a (previously attached) Figure or a part from a Ground the only option is Cont, whereas the Super marker is no longer possible. The same is true for other languages, with the difference that most of them do not exclude Super in such contexts, but use both Cont and Super (even if a language uses only Super in static configurations). In fact, this ‘localization shift’ in elative contexts seems to be very natural. As pointed out in Section 2.1.1, the use of Super in attachment contexts is often possible (or, in some languages, required) when the attachment relation between a Figure and a Ground is not salient. It is evident that in elative contexts when the Figure separates from the Ground the attachment relation is foregrounded, so that the use of Cont becomes more appropriate. The same holds true for parts on the whole, since elative contexts highlight the fact that special efforts are needed to separate a part from the whole. Even more interesting is that the ‘localization shift’ happens not only in elative contexts, but also in contexts that can be called ‘semantically elative’. By this I mean negative sentences describing the absence of a Figure on the Ground and implying that the Figure was separated from the Ground. Mostly, this occurs when the Figure is a part or a natural extension of the Ground. In these cases, Cont is usually required, even if an affirmative sentence allows only Super. Consider the following Tsakhur example: Tsakhur (East Caucasian, Lezgic) (26) mangu-qa χɨlel // χɨlek sa t’ub deš-ob. he-POSS hand-SUPER hand-CONT one finger be.NEG-CL ‘He lacks one finger on his hand.’ (lit. ‘He does not have one finger on hand.’)

142

Dmitry Ganenkov

5. Other topological domains Most languages of the family do not confine the scope of use of the Cont localization to various ON-configurations, but also extend them to express location in other topological domains, viz. near a Ground, inside a Ground, and under a Ground. The use of localization Cont in NEAR-configurations is attested only in West Tsezic, a sub-group comprising closely related languages Tsez, Hinukh, and Khwarshi. Consider the following example from Tsez: Tsez (East Caucasian, Tsezic) (27) uži-j doska-qo darsi eso. boy-ERG blackboard-CONT lesson tells ‘The boy answers the lesson at the blackboard.’ Another use of Cont that seems to be closely related to location near a Ground is the spatial configuration called ‘casual contact’ by Cuyckens (1991), i.e. configurations in which the Figure is supported from underneath (not by the Ground) and is in contact with the Ground, but not attached to its surface. In fact, only languages that use Cont to mark location near a Ground have such uses. Consider the example from Tsez: Tsez (East Caucasian, Tsezic; Xalilov 1999) (28) qido-q-or q’ˁuri riq’ˁik’a wall-CONT-LAT chair place_against ‘to draw up a chair up to the wall (to place a chair against the wall)’ Let us turn now to other topological domains that can be expressed by Cont. In most Lezgic languages, the localization Cont is used to mark location inside a Ground in some contexts. In particular, in Lezgian, Rutul, and Tsakhur, Cont expresses location in a mass and is opposed to the IN localization that expresses ‘location in a hollow Ground’. Consider the ‘inside’ use of Cont in Rutul: Rutul (East Caucasian, Lezgic; Maxmudova 2001: 81) (29) χɨnɨχ xid.i-k ki-r-xu-ri. boy water-CONT CONT-M-fall-PST ‘A boy fell down into the water.’

Towards a typology of ‘attachment’ markers

143

Both Agul and Tabasaran possess a distinct localization marker expressing location in a mass. Instead, the use of Cont is confined to location of a Figure which cannot be separated from the Ground, e.g. ‘salt in soup’ or is considered to be a distinctive feature of a Ground ‘stones in sand’ (stony sand), as in these examples from Tabasaran: Tabasaran (East Caucasian, Lezgic) (30) ǯuk.ra-k kartf-ar k-a-ji ? soup-CONT potato-PL CONT-be.PRS-Q ‘Are there potatoes in the soup?’ (31) aš.di-k q’il gzaf k-a. pilau-CONT salt much CONT-be.PRS ‘There is (too) much salt in the pilau.’ (Xanmagomedov and Šalbuzov 2001: 186) Finally, it is typical of Lezgic languages that the Cont localization is also used to express location under a Ground. Such uses are found in Lezgian, Tsakhur, Rutul, Budukh, Kryz, and a dialect of Agul, as in the following Lezgian example: Lezgian (East Caucasian, Lezgic; Haspelmath 1993: 95) (32) či ajwan.di-n qːaw.u-k čubaruk-r.i-n sa šumud our balcony-GEN roof-SUB swallow-PL-GEN one several χizan jašamiš že-zwa. family living be-IMPF ‘Some families of swallows live under the roof of our balcony.’ Contrary to location inside a Ground, which is a result of the semantic evolution of spatial markers in Lezgic languages, the use of Cont to mark location under a Ground arose due to a historical sound change which led to a merger of non-aspirated stops (the marker -kː of the Sub localization, in particular) and aspirated stops (the marker -k of the Cont localization).

6. Summary and conclusion The different distributions of Cont and Super across the East Caucasian languages allow us to isolate a number of spatial ON-configurations that can be expressed by either of these two localizations. As seen in the previ-

144

Dmitry Ganenkov

ous discussion, spatial configurations expressed by Super and Cont in East Caucasian languages fall into five classes: (i) (ii) (iii) (iv) (v)

the Figure exists independently of the Ground and is located on the surface of the Ground; the Figure is a part of the surface of the Ground; the Figure is a part of the Ground; the Figure is attached to the Ground in a ‘non-typical’ manner; the Figure is located near the Ground.

The distribution of Cont and Super localization markers in East Caucasian languages can be generalized in the form of a semantic map (see Figure 1).9 The map also shows the scope of use of Super and Cont in Agul, as an illustrative case. ‘book on table’

‘fly on wall’

SUPER

‘leaves on branch’

‘clothes on line’

marks / images

‘finger on hand’

‘ring on finger’

‘wart on forehead’

‘coat on hook’

‘put cream on hands’

separation

salient attachment

‘to tie a cow to a tree’

‘to place against’

possession

non-salient attachment

absence ‘painting on wall’ CONT Figure 1. Semantic map of ON-configurations for East Caucasian languages and the distribution of Super and Cont in Agul

One parameter of variation that cannot be displayed on the semantic map is that involving the overlap of the scopes of two markers. In principle, there are three possibilities. The first is that there is no evident semantic contrast between the markers, and both are equally possible in this particular configuration; this is the case of the Cont and Super localizations in non-salient attachment configurations in Bezhta. The second possibility is that there is still no semantic contrast between markers, but that one of

Towards a typology of ‘attachment’ markers

145

them is preferred in the given configuration. An example of this possibility is behaviour of markers in non-salient attachment contexts in Tabasaran. The last option is that each marker has a slightly different interpretation in this configuration. The clearest example of the last possibility is the behaviour of Cont and Super in ‘marks on the surface’ and ‘lumps on the skin’ configurations in Agul and Rutul. It should be recalled that this map is built only on the basis of data from languages of the East Caucasian family. As such, it cannot be considered a fully adequate semantic map of typological generalizations about crosslinguistic variation in the domain of ON-configurations, and it should be compared with and tested against data from genetically and areally distant languages. An interesting problem would be to examine whether the distribution of similar markers in other languages is correctly constrained and predicted by this map, especially in languages such as Dutch, German, Tiriyó, etc., in which the attachment category has been described. A further question is whether this map adequately describes the behaviour of spatial markers in languages such as Finnish that lack a distinct attachment marker but employ the ‘in’-marker in some ON-configurations.

Notes 1.

2.

3.

4. 5.

The research presented here is a part of my dissertation defended in December 2005. I would like to thank once again the supervisor of the thesis Vladimir A. Plungian as well as my opponents Michael A. Daniel and Irina A. Muravyova. I am also very grateful to Patience Epps, Birgit Hellwig and an anonymous reviewer for their comments on the first version of this chapter. East Caucasian is a family of distantly related languages dividing into six branches: Lezgic, Avar-Andic-Tsezic, Nakh, Dargwa, Lak, Khinalug. The languages are mostly spoken in Daghestan and Chechnia (Russia), but also in Azerbaijan and Georgia. Detailed descriptions of East Caucasian case systems can be found in Kibrik (1992); Comrie and Polinsky (1998); Comrie (1999); Kibrik (2003); Daniel and Ganenkov (2008). If there is no source indicated, the data and the examples in this chapter come from my own fieldnotes. Latin based labels for localizations were coined by A. E. Kibrik in the late 1960s and have been being widely used in Dagestanian studies since then. The standard list of localizations includes: APUD ‘near a Ground’, SUPER ‘on a Ground’, IN ‘in a hollow Ground’, INTER ‘between two Grounds / in a mass’,

146

6.

7.

8.

9.

Dmitry Ganenkov CONT ‘attached to a Ground’, SUB ‘under a Ground’, ANTE ‘in front of a Ground’, POST ‘behind a Ground’. For more about semantics of these spatial prepositions, see Cuyckens (1991, 1994); Beliën (2002); Nüse (1999); van Staden, Bowerman, and Verhelst (2006). Recently, several descriptions of attachment markers in typologically diverse languages have been published as contributions to Levinson and Wilkins, eds., (2006), see especially Levinson (2006); Meira (2006); van Staden, Bowerman, and Verhelst (2006). Traditionally, this distinction was described in terms of location on horizontal vs. vertical surfaces. The first who showed that it is rather attachment than horizontality that influences the choice between markers was, to the best of my knowledge, Jakov Testelec (1980). See Haspelmath (2003) on the semantic maps as a method for drawing typological generalizations from cross-linguistic comparison.

References Beliën, Maaike 2002 Force Dynamics in Static Prepositions: Dutch Aan, Op, and Tegen. In Perspectives on Prepositions, H. Cuyckens and G. Radden (eds.), 195–210. Tübingen: Max Niemeyer Verlag. Bowerman, Melissa 1996 Learning how to structure space for language: A cross-linguistic perspective. In Language and Space, P. Bloom, M. Peterson, L. Nadel and M. Garrett (eds.), 385–436. Cambridge, MA: MIT Press. Bowerman, Melissa, and Soonja Choi 2001 Shaping meanings for language: Universal and language-specific in the acquisition of spatial semantic categories. In Language Acquisition and Conceptual Development, M. Bowerman and Stephen C. Levinson (eds.), 475–511. Cambridge: Cambridge University Press. Bowerman, Melissa, and Eric Pederson 1992 Cross-linguistic studies of spatial semantic organization. In Annual Report of the Max Planck Institute for Psycholinguistics 1992, 53– 56. Comrie, Bernard 1999 Spatial cases in Daghestanian languages. In Sprachtypologie und Universalienforschung 52 (2): 108–117. Comrie, Bernard, and Maria Polinsky 1998 The great Daghestanian case hoax. In Case, Typology and Grammar; A. Siewerska, J.J. Song (eds.), 97–113. Amsterdam: John Benjamins.

Towards a typology of ‘attachment’ markers

147

Cuyckens, Hubert 1991 The Semantics of Spatial Prepositions in Dutch. A CognitiveLinguistic Exercise. Ph.D. diss., University of Antwerp. 1994 Family Resemblance in the Dutch Spatial Preposition Op. In Ergebnisse, Probleme, Perspektiven, M. Schwarz (ed.), 179–195. Tübingen: Narr. Daniel, Michael, and Dmitry Ganenkov 2008 Case marking in Daghestanian: Limits of elaboration. In Handbook of Case, Andrey Malchukov and Andrew Spencer (eds.), 668–685. Oxford University Press. Haspelmath, Martin 1993 A Grammar of Lezgian. Berlin/New York: Mouton de Gruyter. 2003 The Geometry of Grammatical Meaning: Semantic Maps and CrossLinguistic Comparison. In The New Psychology of Language. Volume 2, 211–242. Mahwah, NJ: Erlbaum. Kibrik, Aleksandr E. 1992 Principy organizacii imennoj paradigmy v dagestanskix jazykax (sopostavitel’no-tipologičeskie nabljudenija) [Principles of organization of the nominal paradigm in Daghestanian languages (comparative-typological observations]. In Očerki po obščim i prikladnym voprosam jazykoznanija, A.E. Kibrik, 80–101. Moscow: MGU. 2003 Nominal inflection galore: Daghestanian, with side glances at Europe and the world. In Noun Phrase Structure in the Languages of Europe, Frans Plank (ed.), 37–112. Levinson, Stephen C. 2006 The language of space in Yélî Dnye. In Levinson and Wilkins (eds.), 157–205. Levinson, Stephen C., and Sérgio Meira 2003 ‘Natural concepts’ in the spatial topological domain – adpositional meanings in cross-linguistic perspective: an exercise in semantic typology. Language 79 (3): 485–516. Levinson, Stephen C., and David P. Wilkins 2006 Patterns in the data: towards a semantic typology of spatial description. In Levinson and Wilkins (eds.), 512–552. Levinson, Stephen C., and David P. Wilkins (eds.) 2006 Grammars of Space. Cambridge: Cambridge University Press. Maxmudova, Svetlana M. 2001 Morfologija rutul’skogo jazyka [Morphology of the Rutul language]. Moscow: Institut jazykoznanija RAN. Meira, Sérgio 2006 Approaching space in Tiriyó grammar. In Levinson and Wilkins (eds.), 311–358.

148

Dmitry Ganenkov

Nüse, Ralf 1999

General meanings for German an, auf, in and unter towards a (neo)classical semantics of topological prepositions. Ph.D. diss, Humboldt University, Berlin. Testelec, Jakov G. 1980 Nekotorye voprosy tipologii sistem prostranstvennogo sklonenija v dagestanskix jazykax [Some issues in typology of systems of spatial declension in Dagestanian languages]. Unpublished Diploma Thesis. Department of Structural and Applied Linguistics, Moscow State University. van Staden, Miriam, Melissa Bowerman, and Mariet Verhelst 2006 Some properties of spatial description in Dutch. In Levinson and Wilkins (eds.), 475–511. Xalilov, Madžid Š. 1995 Bežtinsko-russkij slovar’ [Bezhta-Russian dictionary]. Maxačkala: IJaLI. 1999 Cezsko-russkij slovar’ [Tsez-Russian dictionary]. Maxačkala: IJaLI. Xanmagomedov, Bejdullax G.-K., and Kurban T. Šalbuzov 2001 Tabasaransko-russkij slovar’ [Tabasaran-Russian dictionary]. Moskva: Nauka.

Part III. Tense, aspect, and desire

Revisiting perfect pathways: Trends in the grammaticalization of periphrastic pasts Chad Howe

1. Introduction1 The development of morphosyntactic structures in the process of diachronic semantic change involves both the grammaticalization of free lexical items to bound morphology (e.g. Latin cantare habemus > Spanish cantaremos ‘we will sing’) as well as the emergence and possible replacement of forms that exhibit degrees of semantic overlap with other forms in the grammar (e.g. the French compound perfect past or passé composé, which has replaced the simple past in spoken varieties). It has been noted that in the latter process increased frequency of the emergent form vis-à-vis another structure (or set of structures) can be taken as a benchmark for advanced stages of semantic change (Hopper and Traugott 1993: 126). Patterns of overall frequency, however, may obscure more specific trends in the evolution of grammatical structures that are subject to a variety of language-internal and external forces across languages and varieties. Thus, it seems that a more nuanced picture of semantic change requires a discussion of characteristics that go beyond simple frequency, taking into account different functional criteria related to the semantic distribution of these forms. The current analysis questions the role of frequency in describing trends in the distribution of simple and compound forms of past reference across different varieties of Spanish. Various authors have observed that the (Present) Perfect (e.g. Diego ha terminado el trabajo ‘Diego has finished the job’) and the simple perfective past or Preterit(e) (e.g. Diego terminó el trabajo ‘Diego finished the job’) in Spanish can both be used to indicate reference to a past action in Spanish, though the Perfect is also said to carry the additional epistemic value of relevance to the moment of speech (see e.g. Brugger 2001). The distribution of these two forms varies across dialects, with some varieties demonstrating a greater preference (as measured by frequency) for one or the other form for indicating certain types of past reference (DeMello 1994, Penny 2000). In this analysis I develop a

152

Chad Howe

typology for distinguishing between trends of semantic change that, despite similar patterns of overall frequency, are not parallel in their functional distributions. Taking as a representative case study samples gathered from interviews and EUROTYP-style questionnaires conducted in three different varieties of Spanish (see Dahl 1985, 2000), I demonstrate that patterns of perfect usage differ qualitatively with respect to different semantic features commonly attested among perfects cross-linguistically – i.e. cooccurrence with definite past adverbials and use in sequenced narratives. Recent developments in the application of typological methodologies in the study of grammaticalization, specifically in the case of temporal and aspectual categories, have shown that trends in semantic change can be tracked via the analysis of specific features associated with various form-function pairs cross-linguistically (see Aldai 2007). The organization of this chapter is as follows. In Section 2, I discuss some fundamental issues related to semantic change and the role of frequency in describing the development morphosyntactic structures. Next, I provide a cross-dialectal overview of the Perfect and the Preterit in Spanish in Section 3 with special attention to variations in overall frequencies and related proposals regarding trends in grammaticalization. In Section 4, I present data from three dialects of Spanish spoken in Spain and Peru (from original fieldwork; see Howe 2006) and analyze usage patterns in conjunction with features attributed cross-linguistically to perfective tense-aspectmood (TAM) constructions. The chapter concludes in Section 5 where I offer some closing commentary.

2. Grammaticalization, frequency, and the perfect > perfective path The current discussion of temporal and aspectual forms in Spanish is couched largely in the approaches outlined by Comrie (1976, 1985), Dahl (1985), and Bybee, Perkins and Pagliuca (1994). These lines of research have developed from comparative studies aimed at discerning salient crosslinguistic categories in the expression of temporal reference. Important to studies of TAM in both the typological and the grammaticalization literatures has been this process of developing sets of functional criteria based on observations across different language families. It is this comparative approach that will serve as the theoretical background for the present analysis.

Revisiting perfect pathways

153

2.1. Trends in semantic change As mentioned above, the shift from ‘less’ to ‘more’ grammatical, as discussed in the grammaticalization literature (see Hopper and Traugott 1993), is not limited only to, for instance, the development of a free lexical item into a fixed expression with specific syntactic constraints. Much of the work on semantic change in diachrony concerns the interplay between forms that display overlap in different functional domains; thus the typical schema presented as A > { A,B } > B depicts the process by which two overlapping forms A and B are used concurrently at some stage in their respective developments. The result of this process is frequently, though not always, the ousting of the older form A by the emergent form B. One commonly cited example is the development of the analytic forms of future express in Spanish (and other Romance Languages), shown in (1). (1)

cantabimusA

>

cantabimusA cantare habemusB

>

cantaremosB vamos a cantarC

A periphrastic expression with the verb habere ‘to have’ as an auxiliary emerged as a alternative to the original synthetic form in Latin, cantabimus ‘we will sing’, with the ‘go’ future – i.e. vamos a cantar ‘we are going to sing’ – arising as another alternative at a later period (see Aaron 2006). What we see in this example is the cyclic nature of semantic change in diachrony with the persistent possibility of variation with other forms in the grammar. The lesson that can be gleaned here is that grammaticalization often involves the negotiation of transfers in semantic spaces between competing structures that can result in the redefining of default forms of, in this case, temporal expression. As competing forms emerge, there is considerable interaction between these and older forms, creating an ongoing tension that can result in the gradual loss of the older structure. The fate of an emergent form is by no means predetermined, though perhaps one of the most robust claims made in the grammaticalization literature is that semantic change, when it does occur, follows typologically common pathways of development and undergoes specific types of syntactic and semantic processes of restructuring. One such pathway, described by Bybee, Perkins and Pagliuca (1994), involves the development of resultative constructions with ‘be’ and ‘have’ auxiliaries into perfectives or evidentials. I will discuss this trend in Section 2.3.

154

Chad Howe

2.2. Frequency and the emergence of grammar One of the hallmarks of grammaticalization is the generalization of the meaning of a particular form to new contexts (e.g. Hopper and Traugott 1993). This process of gradual extension into new semantic spaces, perhaps to the detriment of other forms occupying those spaces, is characterized by a number of formal correlates commonly associated with increased grammaticalization. One important indication is an increase in the overall frequency of the form under question in comparison to other forms expressing similar semantic functions. That is, an increase in frequency signals that a form is expanding into contexts in which it was previously not used. Assuming that the grammar already made use of another form in these contexts, the result of this expansion may be the retraction of the older form in these particular cases. Bybee, Perkins and Pagliuca (1994: 8) note that an increase in frequency is consonant with initial stages of grammaticalization. To return to the example mentioned in (1) above, Aaron (2006: 36) presents quantitative evidence from historical corpora of the distribution of periphrastic and synthetic forms of future reference in Spanish. Her results demonstrate that prior to the 17th century, the periphrastic form (i.e. vamos a cantar) was used in less than 1% of the total number of cases in comparison to the synthetic form (i.e. cantaremos). By the 20th century, this percentage had changed to 25% in the written sample and 59% in the spoken. While Aaron’s analysis treats this variation as the result of the interaction of a variety of different factors, it is clear from observing the changes in overall frequency over time that there has been some adjustment in the distribution of these two forms over time. Another important component in semantic change is the process described by Hopper (1991) as “layering”, whereby two or more forms occupy the same semantic space synchronically, as depicted in (1). For the synthetic and periphrastic futures in Spanish, this space includes future temporal reference. Sankoff and Thibault (1981) argued that morphosyntactic variants need only serve similar discourse functions and that strict semantic equivalence is not required. One consequence of this ‘weak complementarity’ is that competing forms continue to be used in a variety of contexts that are not part of the shared semantic space. In the case of future reference in Spanish, the expression of epistemic modality is attested with the synthetic variant (e.g. Manuela sabrá la respuesta ‘Manuela must know the answer’). Again, Aaron (2006: 91) notes that while only 2% of the cases of synthetic futures were used with modal meanings in the data from

Revisiting perfect pathways

155

the 15th to the 19th century, this number had changed to 24% in the 20th century sample.2 Thus, it would seem that the synthetic future has expanded into semantic domains outside of those shared with the periphrastic form. The step of determining the range and types of contexts in which a particular form overlaps or not with another structure is crucial to the analysis presented in this chapter. That is, we can describe the distribution of a given form in different languages or dialects, at least in part, in terms of the functional spaces that are and are not shared by similar structures. 2.3. Semantic change and the Perfect The evolution of perfective meaning in perfects has been treated in a number of languages and was described by Bybee, Perkins and Pagliuca (1994: 81) as one of the possible pathways of development for forms expressing perfective past reference. Two examples of this development can be seen in German (2) and French (3), both of which have periphrastic forms of perfective past reference that developed from perfects (Bybee, Perkins and Pagliuca 1994: 85).3 German (Indo-European, Germanic) (2) Hans hat gestern den Brief geschrieben. Hans have yesterday the letter write.PST.PTCP ‘Hans wrote the letter yesterday.’ French (Indo-European, Romance) (3) Pierre est arrivé hier. Pierre be arrive.PST.PTCP yesterday ‘Pierre arrived yesterday.’ In addition to other European languages (e.g. Italian and Dutch), this trend is further observed in languages from a number of non-IndoEuropean families, such as some Bantu languages and Mandarin Chinese (Bybee, Perkins and Pagliuca 1994: 81). The commonality across these cases is the fact that the source construction functioned as a perfect (or anterior). In the case of French, the simple perfective past (or passé simple), which in Old French (12th century) was used in variation with the periphrastic form for marking foregrounded events in narratives, is only rarely heard in spoken Modern French and is preserved primarily in written registers (see Squartini and Bertinetto 2000). The simple past in German

156

Chad Howe

(i.e. schrieb ‘wrote’) also has a similar distribution with respect to the periphrastic form, being used largely in non-conversational registers. These two situations, according to Bybee, Perkins and Pagliuca (1994), typify the perfect > perfective pathway and demonstrate that as one form becomes more widely used in a particular context – in this case the compound past in reference to past eventualities – forms sharing that semantic space may be displaced and relegated to distinct registers or dialects. As we shall see in the forthcoming discussion of the perfect in Spanish, this type of marginalization of contextual interpretations does not seem to characterize all situations in which the perfect is reported to have taken on perfective uses. Also germane to the current discussion is the development of perfects with evidential uses, as attested, for example, in Turkish and Bulgarian (Bybee, Perkins and Pagliuca 1994: 95). West (1980: 75) describes this type of construction in Tucano, spoken in parts of Brazil and Colombia, where the speaker uses the verb ‘be’ in conjunction with a non-finite form of another verb to indicate an inference based on results. In the case of the Spanish perfect, it has been noted that the perfect has acquired evidential uses in varieties spoken in contact with Quechua (see Escobar 1997, among others). One result of this extension is the possible co-occurrence with adverbials denoting definite past time reference (e.g. last year), suggesting some parallels in terms of temporal expression with the type of developments attested in the German and French examples. Thus, the surface reflexes of these two developments – i.e. perfectivity in the German and French examples and evidentiality in the cases of Spanish/Quechua contact – may indeed overlap, demonstrating increased frequencies of periphrastic past forms that are a result of expanded contextual compatibility. I will argue, however, that these ‘parallel’ developments obscure language- and dialect-specific mechanisms important in the process of semantic change. 2.4. The perfect in Romance Before turning to the specifics of the Spanish Perfect, let us take a look at the overall situation of the perfect across Romance. According to Harris (1982), the distribution of the perfect across varieties of Romance can be described as a series of stages, differentiated by the types of uses expressed. Harris applies Comrie’s (1976) typology of perfect types (i.e. perfect of result, experiential perfect, perfect of persistent/continuous situation, and perfect of recent past) in his analysis. Thus, the pretérito perfeito composto (compound past) in Portuguese represents an early stage of peri-

Revisiting perfect pathways

157

phrastic past development in Romance in that it can only have a durative or iterative function, which Harris argues are diachronically closer to the functions of the original resultative source construction. Note the incompatibility of the adverbial uma vez ‘once’ in example (4), demonstrating the unavailability of an experiential interpretation. Conversely, the French passé composé, shown in (5), is limited to perfective past situations and is thus incompatible with expressions requiring a durative or continuative meaning. Portuguese (Indo-European, Romance) (4) Eu tenho estado no Recife I have be.PST.PTCP in-the Recife ‘I have been in Rio (*once).’

(*uma vez). one time

French (Indo-European, Romance) (5) J’ ai été à Paris (*depuis hier). I have be.PST.PTCP in Paris since yesterday ‘I was in Paris (*since yesterday).’ Regarding the Spanish Perfect, Harris explains that some dialects follow the more conservative Portuguese pattern (e.g. Mexican Spanish), while others are more similar to the French passé composé (e.g. some varieties of Peninsular Spanish). Though Harris does not provide a discussion of this distribution as a reflection of different stages of grammaticalization, it is apparent that his analysis can be recast in terms of the discussion by Bybee, Perkins and Pagliuca (1994: 81) of the perfect > perfective pathway. In other words, the distinction between the Portuguese pretérito perfeito composto on the one hand and the French passé composé on the other can be understood as representing varying degrees of perfectivity. Squartini and Bertinetto (2000) describe the evolution of simple and periphrastic past tenses in Romance along these lines, proposing that there is a general tendency for Romance perfects to develop into aorists. This trend, referred to by Squartini and Bertinetto as the “aoristic drift” (2000: 404), is attested in languages outside the Romance family as well – e.g. Modern Greek and Danish (see Bybee, Perkins and Pagliuca 1994: 84). One important question that will be addressed in Section 3 is whether or not the relative frequencies of Perfects and Preterits across dialects of Spanish can be considered a reliable indicator of parallel trends in semantic change. That is to say, does increased Perfect usage in comparison to the Preterit constitute sufficient evidence for the claim that a specific peri-

158

Chad Howe

phrastic form is following the perfect > perfective pathway (i.e. the “aoristic drift” in Squartini and Bertinetto’s approach)?4 Moreover, what explanatory role do cross-linguistically attested functional distinctions in perfect usage have in describing possibly divergent trends in semantic change across different dialects of Spanish? In the following section, I defend the claim that overall frequency provides only prima facie evidence for patterns of grammaticalization and, additionally, that an analysis that takes into account broader typological patterns in the distribution of morphosyntactic forms is ideal for discerning divergent trends in semantic change. The latter part of this hypothesis builds on the work of Aldai (2007) who explores the issue of form-function relation in the grammaticalization of past tense constructions in Basque. This author argues that because a typological approach assumes that a given TAM construction may exhibit different varieties of meaning it is better suited than a strictly structuralist analysis to describing synchronic and diachronic tendencies that are often manifested as variation among discrete morphosyntactic structures. The advantage offered by this approach for the current analysis is the ability to tease apart, via the observation of cross-linguistic prototypes associated with perfect constructions, distinct trends in semantic change that may be conflated by relying on a single, quantitative heuristic.

3. A case study of perfect > perfective pathways We begin this section with a brief overview of some representative approaches to analyzing cross-dialectal variation of the Spanish Perfect. Common to these approaches are methodological issues that obscure significant contextual details. With this background, we can then proceed to Section 4 where I defend the claim that comparison of form frequency is an insufficient and potentially misleading measure of functional distribution. 3.1. The Perfect and the Preterit in Spanish Among the ample literature regarding perfects, there are numerous treatments of the Spanish periphrastic past addressing various issues related to its form and meaning.5 In general, the Spanish Perfect is prototypical in that it can express all of the different functions, or ‘types’ (see Section 2.4), ascribed to perfects cross-linguistically by Comrie (1976). This typol-

Revisiting perfect pathways

159

ogy of perfect types has been applied by a number of authors to the description of the Spanish Perfect as a means of evaluating cross-dialectal variation with the Preterit (see e.g. Hernández 2004). The contextual nature of these functions, however, is not such that they can be uniquely associated with the Perfect. Thus, if we take as an example Comrie’s description of the Perfect of Result, described as the situation in which “a present state is being referred to as being the result of some past situation” (1976: 56), we can see that both the Perfect and the Preterit in Spanish can express this meaning. For example (6), the form ha aprendido ‘has learned’ was attested in a context in which the speaker wished to offer evidence of proficiency in Quechua. The Preterit (i.e. aprendió ‘learned’) could, under similar pragmatic conditions, give rise to the same reading. Peruvian Spanish6 (Indo-European, Romance) (6) Él lo ha aprendido aquí en Pukllasunchis. he it.ACC have learn.PST.PTCP here in Pukllasunchis ‘He has learned it [Quechua] (and thus speaks it) here at Pukllasunchis.’ (Howe 2006) In a recent paper, Schwenter and Torres Cacoullos (2008) discuss the application of the typology of perfect types proposed by Comrie (1976) as a measure of the cross-dialectal form-function mapping of the Preterit and Perfect in Spanish and conclude that the prototypical perfect types cannot “be reliably distinguished beyond ideal examples in a large sample of tokens” (2008: 11). Instead, they opt to demonstrate that dialectal distinctions between the distribution of the Perfect and the Preterit are best modelled via the interaction of a variety of co-occurring contextual elements quantifiable in a corpus. This proposal further highlights the difficulty of relying on contextually licensed interpretations as a means of defining functional spaces. While it may indeed be the case that speakers make use of a particular form for a variety of different functional distinctions, our interpretation as analysts is perhaps not a sufficiently precise measure in assigning discrete semantic labels consistently to a large sample of corpus tokens. 3.2. Cross-dialectal frequencies with the Spanish Perfect One potential solution to the challenges posed by applying functional categories consistently across large numbers of tokens is to rely instead on the overall frequency distribution of competing forms as a measure of overlap

160

Chad Howe

in semantic spaces. With this approach, if an increase in the ratio of periphrastic past forms with respect to the simple past is representative of increased perfectivity then two (or more) varieties displaying parallel statistical trends could, consequently, be described as following the same path of grammaticalization. Here, we will make use of Harris’ (1982) description of the evolution of compound and simple past forms in Romance. Recall that in his analysis Harris categorizes dialects of Spanish based on the expression of the prototypical meanings associated with perfects crosslinguistically as proposed by Comrie (1976). Thus, Mexican varieties of Spanish are described as belonging to Stage II, similar to the periphrastic past in Portuguese, while some Peninsular varieties are described as Stage III perfects with some cases approximating the situation of the French passé composé (i.e. Stage IV).7 What is important to keep in mind is that dialects belonging to the latter stages exemplify the trend toward more perfective uses of a perfect (Squartini and Bertinetto 2000: 414). Consequently, we expect to find higher overall frequencies of perfects in these cases than in more ‘conservative’ dialects since the former situation represents an incipient stage of semantic change (Bybee, Perkins and Pagliuca 1994: 8). Concerning the cases of ‘perfective’ Perfects in Spanish, Schwenter (1994) argued that differences in overall frequency of Perfect/Preterit usage among different age groups in a sample of spoken Spanish from Alicante, Spain constituted evidence of a change in progress. Moreover, this author demonstrates increased usage of the Perfect with adverbs denoting reference to the ‘today’ interval (i.e. hodiernal reference). In interview data collected in Madrid and Valencia, Spain in the summer of 2006, I observed a number of tokens in which perfects were used with these types of adverbials, many of which were produced as responses to the question ¿Cuéntame tu día hoy ‘Tell me about your day today’. Consider example (7), uttered by a speaker from Madrid. Peninsular Spanish (Indo-European, Romance) (7) Me he subido con él a las ocho ACC have walk_up.PST.PTCP with him at the eight ‘I walked up with him (= the dog) at eight o’clock.’ (Howe 2006) In example (7), the speaker makes reference to an interval of time that does not overlap with the time at which the utterance was spoken (in the afternoon) but is contained within the same ‘today’ interval. These types of examples have been observed by numerous other studies, some of whom

Revisiting perfect pathways

161

have argued that these Peninsular cases are also compatible with more distant past adverbials (e.g. ayer ‘yesterday, see Brugger 2001). Squartini and Bertinetto (2000) take these observations of the Perfect in some varieties of Peninsular Spanish to be indicative of the same general tendency that resulted in the French passé composé being used as a perfective past. Claims about perfective uses of the periphrastic past have also been attributed to some South American varieties, typically those in contact with Quechua, in which increased frequency of Perfects is argued to be evidence of a similar trend in perfectivization (see DeMello 1994 and Penny 2000). In fact, Alonso and Henríquez Ureña note that in the Andean areas “the use [of the Perfect] coincides with that of Madrid” (1951: 155, translation mine). Such observations are based on examples such as the following, produced by a speaker from Cusco, Peru: Peruvian Spanish (Indo-European, Romance) (8) Bueno, yo he vivido y he nacido good I have live.PST.PTCP and have be_born.PST.PTCP en Lima in Lima ‘Well, I have lived and was born in Lima.’ (Howe 2006) This example demonstrates a typically Andean use of the Perfect in reference to discrete past events (e.g. someone’s birth), which, at first glance, seems parallel to the types of uses attested in Peninsular Spanish. The resulting interpretation of perfective reference in example (8), however, can be accounted for largely by the influence of the simple telic predicate ‘beborn’, which, unlike ‘live’, cannot be interpreted as indicating a durative eventuality. As mentioned above, Escobar (1997) attributes these uses to contact with Quechua and the adoption of epistemic features, namely evidentiality. Interestingly, these purportedly perfective uses of the Perfect have been observed both in bilingual Spanish/Quechua speakers as well as monolingual speakers of Spanish, which would suggest that the effects of the contact situation extend beyond those cases in which both languages are spoken (see Jara Yupanqui 2006). Despite the considerable influence of an indigenous language substratum, there are still a number of authors who hold the view that this increase in perfective uses of the Perfect is analogous to the Peninsular case. In the current analysis, I claim that the mechanisms responsible for perfective uses of perfects in the Peninsular and Andean (i.e. those varieties spoken in contact with Quechua) cases are not identical even though both dialects exhibit some typological trends sugges-

162

Chad Howe

tive of increased grammaticalization (e.g. compatibility with definite past adverbs).8 Of the studies that have explored the purported parallelism between Peninsular and Andean patterns of Perfect usage, DeMello (1996) represents one of only a few that provide corpus evidence demonstrating increased frequencies of Perfect use in different dialects. He argues that the frequencies of Perfect usage are “quantitatively most common” in the samples from cities such as La Paz, Lima, Madrid and Sevilla, claiming further that these tendencies are indicative of parallel functional distributions as well (1996: 619). A closer look at these cases, however, reveals that the types of contexts that demonstrate an increased number of Perfects are not equivalent across these samples, suggesting that these developments are in fact not parallel. In a variationist analysis of corpus data from Lima, Peru, Howe and Schwenter (2008) propose that comparing absolute frequencies in the distribution of Perfects and Preterits in Spanish obscures distinctions that are pertinent to the functional domains of the two forms. In a comparison of the Perfect/Preterit distribution in three corpora of spoken Spanish, Mexico City (Mexico), Lima, Peru, and Madrid, Spain, Howe and Schwenter (2008: 104) note that the frequency of the Perfect follows the expected trends outlined by Harris, namely that there are significantly more Perfects in the Madrid sample (53.6% or 956 of 1783) than in the Mexican one (14.8% or 331 of 2234). The Lima sample displays an intermediate tendency (26.4% or 526 of 1996). Through multivariate analysis, Howe and Schwenter demonstrate that relative frequency is not in itself adequate for determining development along a specific path of grammaticalization, and that cross-dialectal differences in the interaction of language-internal constraints – namely the type of intended past reference (e.g. definite or indefinite) – can account for these divergent distributions.

4. Beyond frequencies: Typological features of perfects In this section I will take a closer look at data gathered from interviews and questionnaires gathered during the summer of 2005 in Madrid and Valencia, Spain and Cusco, Peru with a small sample of speakers. Of the 32 informants who participated in the study, the average ages were 33 (Madrid), 23 (Valencia), and 31 (Cusco). The distribution of participants in the different research sites was as follows: Madrid = 15, Valencia = 8, and Cusco = 9.9 All of the participants were monolingual Spanish speakers (with the exception of one informant from Cusco who had learned Quechua at an

Revisiting perfect pathways

163

early age) and were born at or near the respective sites. In the following discussion, I will focus on two aspects of the data: (i) the overall frequencies of forms of past reference (i.e. Perfect and Preterit) and (ii) the crossdialectal distinctions in compatibility with adverbials and narrative contexts. 4.1. Overview of form preferences In terms of overall form choice patterns, all three samples exhibited the expected tendencies; the speakers from Madrid and Valencia10 were more likely to choose the Perfect where the Cusqueño informants preferred the Preterit. In Table 1, we see that for Madrid and Valencia the rates of Perfect and Preterit choice are virtually the same, whereas the Preterit is clearly the dominant form in the Cusco sample. One reason for the weakened preference for the Perfect in the Peninsular sample from this study as compared to other analyses (see Schwenter and Torres Cacoullos 2008) is the formal nature of the written sentence judgement task, which frequently has a levelling affect favouring more standard linguistic forms. Nevertheless, the frequencies observed with the sentence judgement task are consistent with the Perfect and Preterit distributions observed in the oral interviews conducted in conjunction with these questionnaires (Howe 2006: 121). Table 1. Overall frequencies of Perfects and Preterits Madrid

Valencia

Cusco

Perfect

42.4% (N = 266)

47.5% (N = 161)

20.5% (N = 75)

Preterit

50.8% (N = 319)

45.4% (N = 154)

71% (N = 260)

Both Perfect & Preterit Total

6.8% (N = 43)

7.1% (N = 24)

8.5% (N = 31)

628

339

366

4.2. Perfects and adverbial co-occurrence As discussed above, the frequencies shown in Table 1 provide at best only a tentative picture of the distribution of the Preterit and the Perfect in these dialect samples. To better understand these data, I further observed the

164

Chad Howe

interaction of these forms with two different patterns commonly discussed in the literature concerning perfects cross-linguistically. To begin, Comrie (1976, among many others) points out that perfects are generally incompatible with definite past time denoting adverbials (e.g. ‘yesterday’, ‘in 1996’, etc.). These types of adverbials generally co-occur with forms expressing perfective past reference. Thus, like the English simple past in ‘John arrived yesterday’, the periphrastic past forms in German (example 2) and French (example 3) are also compatible with this class of adverbials. Again, the French passé composé, according to both Harris (1982) and Squartini and Bertinetto (2000), represents the prototypical perfective periphrastic past in Romance. Subsequently, we would expect similar patterns of adverbial co-occurrence in other varieties of Romance in which the periphrastic past is described as following the same perfect > perfective pathway. Analysis of patterns of adverbial co-occurrence in our three dialect samples, however, does not suggest that each variety is subject to the same constraints. First, as illustrated in example (7), the Perfect in many Peninsular varieties is described as making definite past reference in the ‘today’ interval. This observation is further highlighted in example (9) where the speaker switches between the Perfect (me he levantado) and the Preterit (me quedé) when the reference extends beyond the current day. Peninsular Spanish (Indo-European, Romance) (9) eh hoy me he levantado por la mañana um today ACC have get_up.PST.PTCP in the morning y ayer me quedé a dormir en casa and yesterday ACC stay.PST to sleep in house ‘Um, today I got up in the morning and yesterday I stayed at home to sleep.’ (Howe 2006) It is important to point out that the use of the Perfect with the adverbial por la mañana ‘in the morning’ can be viewed as definite past only if it is understood that the speaker considers the moment of speech to be temporally disjoint from the ‘morning’ interval even though both are properly included in ‘today’. Note, for example, that the English Perfect in ‘John has arrived this morning’ is pragmatically odd if uttered in the afternoon (even more so if an adverbial like ‘at six o’clock’ is included). Schwenter (1994) argues that the Perfect in some varieties of Peninsular Spanish has become the default form for making reference to past events that occurred during the same day as the time of utterance. This observation is corrobo-

Revisiting perfect pathways

165

rated by both the interview data in the current study as well as the questionnaire data in which the speakers from both the Madrid and Valencia samples preferred the Perfect over the Preterit in sample sentences containing these types of adverbials (75.6% and 87.5%, respectively). In the Cusco sample, the rate of Perfect selection was much lower with ‘today’ adverbials (25.9%). The interaction of the Perfect and other adverbials was also distinct across the dialect samples. Again, while cases of the Peninsular Perfect cooccurring with pre-‘today’ adverbials have been attested, they are not a salient context for speakers (see Schwenter and Torres Cacoullos 2008). There were very few of these cases in the interview data for the Madrid and Valencia samples, and the questionnaire results suggested a virtually categorical preference for the Preterit. With Cusco informants, however, the questionnaire task displayed a slight increase in the acceptability of the Perfect with pre-‘today’ adverbials (11.1%, N=6). Moreover, there were a number of tokens in the interview data in which speakers used the Perfect with expressions indicating a discrete past interval prior to the current day. Observe example (10). Peruvian Spanish (Indo-European, Romance) (10) pero no lo he hablado durante mi niñez but not it have speak.PST.PTCP during my childhood ‘…but I did not speak it [Quechua] during my childhood.’ (Howe 2006) In (10), the speaker refers to her use of Quechua during her childhood, an interval that is clearly distinct from the moment of speech given that the speaker was in her thirties. To summarize, these co-occurrence patterns (i.e. ‘today’ and pre-‘today’ adverbials) demonstrate that, despite what appears to be perfective reference across the three dialect samples, the Peninsular informants are constrained to a specific type of past reference (i.e. hodiernal) while the informants from Cusco are more flexible with reference to different definite past contexts. I have argued that increased co-occurrence of the Perfect with definite past adverbials in cases of Spanish/Quechua contact is due to epistemic factors, not temporal ones as in the Peninsular cases (see Howe 2006). Additionally, the greater overall frequency of definite past adverbials with Perfects in the Peninsular sample, albeit limited to a specific type of past reference (i.e. ‘today’), further suggests increased grammaticalization along the perfect > pefective pathway.

166

Chad Howe

I will now discuss another context in which the data from these samples suggest divergent trends. 4.3. Perfects in narratives According to Bybee, Perkins and Pagliuca (1994: 54), perfectives are preferred over perfects for the sequencing of foregrounded events in a narrative. In the case of German and French, the periphrastic past, having developed into a fully perfective form of past reference, is the default structure used for indicating foregrounded events in narratives. For most dialects of Spanish, narrative sequence is generally accomplished via the Preterit, while a sequenced interpretation is not possible with the Perfect. During the interview portion of the current study, all participants were asked to provide a description with different temporal parameters – ‘Tell me about your day today/yesterday’. As expected, the informants from the Madrid and Valencia samples consistently produced Perfects in response to the question Cuéntame tu día hoy ‘Tell me about your day today’. One such response from the Madrid sample is provided in (11) (Perfects are in boldface and glosses have been omitted to facilitate reading). Peninsular Spanish (Indo-European, Romance) (11) Me he levantado a las… a las nueve de la mañana. He desayunado en casa. Me (he) hecho la comida. He ido a la casa de mis padres a… para hacer unas burocracias, y luego he venido a la universidad… ‘I got up at nine o’clock in the morning. I ate breakfast at home. I made lunch. I went to my parents house to take care of some business, and then I came to the university.’ (Howe 2006) The speaker provides a sequence of Perfects in the narration of her day’s activities. It should be noted that without the presence of adverbials like luego ‘then’ indicating an explicit sequence of events, this type of example could be interpreted as a listing of events in which no interpretation of temporal sequence is required, a use that is compatible with nonperfective perfects (e.g. the English Perfect). The questionnaire data also indicate that the speakers from the Peninsular samples have an increased acceptance for the Perfect with ‘today’-bound narratives, though the rates are lower than attested with simple ‘today’ adverbial compatibility (38.1% for Madrid and 55.5% for Valencia). The speakers from the Cusco sample

Revisiting perfect pathways

167

were categorical in their selection of the Preterit in the questionnaire task, and there were no examples of Perfects used in cases of sequenced narratives. Finally, the speakers were also asked to indicate form preference for pre-‘today’ narratives. Assuming extended use in narratives as a possible measure of increased perfectivity, we can also observe whether or not speakers extend these forms to cases of more remote past situations. All of the speakers categorically chose the Preterit with pre-‘today’ narratives in the questionnaire task, and there were no attested examples in the interview data. These results are consistent with the observation that it is in this particular context (i.e. pre-‘today’ modification) that the effects of semantic change in the Peninsular case are neutralized and the two varieties converge. That is, advanced stages of perfect > perfective grammaticalization may give rise to increased usage of a periphrastic past in these contexts, as in German and French. The convergence of the three samples in this context can be explained by reiterating the claim regarding divergent pathways of semantic change. I have argued that the Perfect in Peninsular Spanish is developing along a temporal/aspectual axis (i.e. perfective), one result of which is increased compatibility in perfective contexts (e.g. sequencing in narratives). The fact that speakers from the Peninsular samples disprefer the Perfect in pre-‘today’ narratives is indicative of the stage of grammaticalization, not the trajectory. Concerning the Cusco sample, we have argued that the change exhibited by the Perfect is not temporal; it is epistemic. Consequently, we should not expect to find the Perfect in these varieties of Spanish used in typically perfective contexts. These results corroborate our central claim regarding the problems with using comparisons of relative frequencies as a measure of parallel pathways of semantic development. While an increase in frequency of use is expected in cases of grammaticalization, this is not a sufficient condition for determining semantic trajectory as a given surface statistical tendency may be motivated by a number of contextual and language- or dialect-specific factors (e.g. language contact). 4.4. Discussion The objective of this section has been to provide an alternative perspective on the development of the Perfect in these dialect samples. A comparison of relative frequencies of Perfects and Preterits in the Peninsular and Andean samples provides an overly gross depiction of the distributions of

168

Chad Howe

these two structures, leaving out factors critical to understanding the pathways and mechanisms of semantic change. With the Peninsular samples, the observations presented here concerning the compatibility of the Perfect with only a specific type of past adverbial (i.e. hodiernal) and the limited use in narrative contexts not only provides some background for why we would be observing increased percentages of Perfect use – i.e. due to expansion into new semantic spaces (see Hopper 1991) – but also how these forms overlap (or not) in function with the Preterit. Thus, if co-occurrence with past adverbials and use in sequence narratives are taken to be prototypical contexts for perfective forms, then it becomes clear that the evidence from the Peninsular sample clearly suggests a perfect > perfective trajectory. With respect to the Cusco sample, we started this section with an observation about lower frequency of Perfect selection vis-à-vis the two Peninsular samples. Recall, however, that the claims regarding expansion of the Perfect in the Andean cases was based on comparisons with Latin American varieties (e.g. Mexican Spanish) that demonstrate a ‘conservative’ pattern of usage with a concomitantly lower overall frequency. The results from this study suggest, once again, that broad, statistical trends should be tempered with a more fine-grained analysis of the types of contexts that typify the distribution of overlapping forms (see Howe and Schwenter 2008). In looking at patterns of adverbial co-occurrence and narrative compatibility, the current analysis offers this type of detailed contextual analysis and demonstrates that, despite some superficial similarities, these two sets of samples, Peninsular (Madrid and Valencia) and Cusqueño, do not exhibit parallel trends in the functional distribution of the Perfect. With this analysis, we are left to determine how the Cusqueño sample fits into the overall picture of perfect development in Romance as proposed by Harris (1982). At the very least, these two samples represent distinct stages of grammaticalization, with the Peninsular case being closer to the developments attested in French (i.e. perfect > perfective) and expressing several features concomitant with perfective aspect (e.g. narrative uses). On the other hand, the seemingly perfective tokens from the Cusqueño sample, as in examples (8) and (10), do not display the typical hallmarks of perfective forms, suggesting perhaps that this variety follows the Latin American norm of Perfect usage (DeMello 1994). Nevertheless, Harris’ model does not take into account the type of evidential uses attested with perfects cross-linguistically (see Bybee, Perkins and Pagliuca 1994: 95), relying instead on temporal and aspectal criteria in discerning different stages of developent. Therefore, while the Perfect in some Andean varie-

Revisiting perfect pathways

169

ties of Spanish falls outside of the purview of Harris’ stage-wise historical typology, patterns of Perfect usage in these varieties can be explained by taking into account the broader trends in the development of periphrastic pasts attested outside of the Romance languages.

5. Conclusions Beyond the discussion of Spanish, this analysis has exposed some issues related to how to we describe trends in semantic change cross-dialectally and cross-linguistically. The emergence of grammatical structure is influenced by a multiplicity of different factors that can shape surface distributions in language-specific ways. We know that emergent structures tend to follow certain defined pathways of development and that these pathways are cross-linguistically attested (Bybee, Perkins and Pagliuca 1994). We also know that incipient change is often accompanied by increased frequency of use as a particular structure extends into new functional spaces (Bybee, Perkins and Pagliuca 1994: 8). Gauging adherence to a pathway, however, must involve a discussion of the subtleties related to functional distributions and to do so requires some set of evaluative criteria external to the situations being analyzed. Thus, I have argued that the relative frequencies of overlapping structures – in this case the Perfect and the Preterit in different dialects of Spanish – is an overly gross criterion for discerning trends in semantic change and reliance on frequency alone may obscure factors that may suggest divergent mechanisms of grammaticalization. In sum, the current proposal claims that pathways of semantic change must be evaluated via a set of cross-linguistically attested properties associated, in this case, with perfective TAM constructions. It is only through this type of typological analysis that we can begin to account for variable mechanisms of semantic change, which, in a purely frequency-based analysis, can be overlooked. It may indeed be the case that, under a more nuanced description of functional distributions, we would find it useful to revise our current understanding of discrete pathways of change in favour of a model that better describes subtle language- and dialect-specific deviations.

170

Chad Howe

Notes 1.

I am grateful to Pattie Epps, Alexandre Arkhipov, and one anonymous reviewer for their insightful and thorough comments, which were crucial to the shaping of the final draft. This work was made possible by funds provided through the Department of Spanish and Portuguese, the College of Humanities and the Graduate School at The Ohio State University. 2. The availability of oral data in the 20th century corpora used by Aaron (2006) certainly would have its effect on the distribution of the simple and periphrastic futures since this variation is commonly associated with register distinctions. 3. Bybee, Perkins and Pagliuca (1994) also use the term “anterior” as a partial synonym for perfects. An anterior form, according to these authors, “signals that the situation occurs prior to reference time and is relevant to the situation at reference time” (1994: 54). Abbreviations are as follows: ACC accusative, PST past, PTCP participle. 4. Alternatively, one could also ask the question of whether or not a higher frequency of the Preterit vis-à-vis the Perfect (e.g. in many varities of Mexican Spanish) constitutes a pathway distinct from the perfect > perfective change described here (see Schwenter and Torres Cacoullos 2008). 5. For an overview of issues related to the form and meaning of the Spanish Perfect, see Cartagena (1999). 6. The labels “Peruvian” and “Peninsular” are intended to be general cover terms for the data cited from the fieldwork conducted in Cusco, Peru and Madrid and Valencia, Spain, respectively, and should not be taken to refer to all varieties of Spanish spoken in these dialect zones. 7. Harris’ (1982) stage-wise approach has been criticized as not accurately representing those cases, like the varieties of Peninsular Spanish discussed here, where the Perfect has the functions of a perfective (see e.g. Schwenter 1994). 8. Though it is not my intention here to discuss the specifics of the Spanish/Quechua contact situation, it remains to be seen how this case fits with other treatments of contact-induced structural change (see Heine and Kuteva 2005). 9. The sample was not evenly stratified for men and women (27 women and 5 men). Nevertheless, there is no indication in other studies of Perfect/Preterit variation that gender is a relevant factor (see Jara Yupanqui 2006). Moreover, all speakers were either students or faculty at local schools and universities, suggesting comparable socioeconomic backgrounds. 10. In almost all situations, the speakers from both the Madrid and Valencia samples demonstrated the same patterns of form preference.

Revisiting perfect pathways

171

References Aaron, Jessi Elana 2006 Variation and change in Spanish future temporal expression. Ph.D. diss., University of New Mexico. Aldai, Gontzal 2007 Discreteness and non-discreteness in the design of tense-aspectmood. In New Challenges in Typology: Broadening the Horizons and Redefining the Foundations, Matti Miestamo and Bernhard Wälchli (eds.), 272–291. Berlin/New York: Mouton de Gruyter. Alonso, Amado, and Pedro Henríquez Ureña 1951 Gramática castellana. Segundo curso. 10th ed. Buenos Aires: Losada. Brugger, Gerhard 2001 Temporal Modification, the 24-hour rule and the location of reference time. In Current Studies in Spanish Syntax and Semantics, Luis Silva-Villar and Javier Gutiérrez-Rexach (eds.), 243–270. Berlin/New York: Mouton de Gruyter. Bybee, Joan L., Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the languages of the world. Chicago: University of Chicago Press. Cartagena, Nelson 1999 Los tiempos compuestos. In Gramática descriptiva de la lengua española Vol. 2, Ignacio Bosque and Violeta Demonte (eds.), 2935– 2975. Madrid: Real Academia Española/Espasa Calpe. Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. 1985 Tense. Cambridge: Cambridge University Press. Dahl, Östen 1985 Tense and Aspect Systems. Cambridge: Cambridge University Press. Dahl, Östen (ed.) 2000 Tense and Aspect in the Languages of Europe. Berlin/New York: Mouton de Gruyter. DeMello, George 1994 Pretérito compuesto para indicar acción con límite en el pasado: Ayer he visto a Juan. Boletín de la Real Academia Española 4: 611–633. Escobar, Anna María 1997 Contrastive and innovative uses of the present perfect and the preterite in Spanish in contact with Quechua. Hispania 80: 859–870.

172

Chad Howe

Harris, Martin 1982 The “past simple” and “present perfect” in Romance. In Studies in the Romance Verb, Martin Harris and Nigel Vincent (eds.), 42–70. London: Croom Helm. Heine, Bernd, and Tania Kuteva 2005 Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Hernández, José Esteban 2004 Present perfect variation and grammaticization in Salvadoran Spanish. Ph.D. diss., University of New Mexico. Hopper, Paul J. 1991 On some principles of grammaticization. In Approaches to Grammaticalization, E. C. Traugott and B. Heine (eds.), 17–35. Amsterdam: Benjamins. Hopper, Paul J., and Elizabeth C. Traugott. 1993 Grammaticalization. Cambridge: Cambridge University Press. Howe, Chad 2006 Cross-dialectal features of the Spanish present perfect: A typological analysis of form and function. Ph.D. diss., The Ohio State University. Howe, Chad, and Scott Schwenter 2008 Variable constraints on past reference in dialects of Spanish. Proceedings of the Fourth International Workshop on Spanish Sociolinguistics, M. Westmoreland and J.A. Thomas (eds.), 100– 108. Somerville, MA: Cascadilla Proceedings Project. Jara Yupanqui, Ileana Margarita 2006 The use of the preterite and the present perfect in the Spanish of Lima. Ph.D. diss., University of Pittsburgh. Penny, Ralph 2000 Variation and Change in Spanish. Cambridge: Cambridge University Press. Sankoff, David, and Thibault Pierrette 1981 Weak complementarity: tense and aspect in Montreal French. In Syntactic Change, vol 25, B.B. Johns and D.R. Strong (eds.), 205–216. Ann Arbor: University of Michigan. Schwenter, Scott A. 1994 The Grammaticalization of an Anterior in Progress: Evidence from a Peninsular Dialect. Studies in Language 18: 71–111. Schwenter, Scott A., and Rena Torres Cacoullos 2008 Defaults and indeterminacy in temporal grammaticalization: The ‘perfect’ road to perfective. Language Variation and Change 20: 1–39.

Revisiting perfect pathways

173

Squartini, Mario, and Pier Marco Bertinetto 2000 The simple and compound past in Romance languages. In Tense and Aspect in the Languages of Europe, Östen Dahl (ed.), 403–439. Berlin: Mouton de Gruyter. West, Birdie 1980 Gramática popular del Tucano. Bogotá: Ministerio del Gobierno, Summer Institute of Linguistics.

Individual-level meanings in the semantic domain of pluractionality Andrey Shluinsky

1. Introduction1 Pluractionality is a domain of meaning that deals with the plurality of events. Pluractional meanings remain under-investigated, although they are an important and relevant aspectual distinction in many languages. In this chapter a set of so-called Individual-Level pluractional meanings (henceforth IL-meanings) is explored. These meanings are typically expressed by pluractional markers, but sometimes also by more general aspectual grams. They all have one thing in common: they introduce an individual-level predicate, i.e. a predicate that denotes a permanent feature of an individual. The majority of IL-meanings are semantically and/or pragmatically related to iteration, and all of these are typically expressed by the same markers as are plural events. This Individual-Level type of pluractional meanings has never been concisely discussed in the literature on pluractionality. However, it is crucial in understanding how pluractionality is generally organized. IL-meanings are often expressed by pluractional markers, although not all pluractional markers are able to express these individual meanings. The main goal of this chapter is to determine how IL-meanings are related to other meanings in the semantic domain of pluractionality. I present here the results of a cross-linguistic study based both on a questionnaire and on published secondary data. First-hand data from questionnaires is available for a sample of 20 languages (Agul, Adyghe, Basque, Bulgarian, Czech, Danish, Enets, French, Hebrew, KarachayBalkar, Korean, Lithuanian, Mandarin Chinese, Maori, Mari, Nenets, Russian, Susu, Tajik, Udmurt). For a substantially larger sample of languages, the data was drawn from published descriptions. In Section 2, the notion of pluractionality and the corresponding pluractional markers are introduced, and a brief survey of the semantic domain of pluractionality is given. In Section 3, the distinction between individuallevel vs. stage-level predicates is explained, and the interaction between

176

Andrey Shluinsky

individual-level predicates and pluractionality is discussed. The bulk of the chapter is devoted to defining five distinct categories of pluractional ILmeanings, namely INDIVIDUAL-LEVEL STATE, PROPERTY, CAPACITIVE, QUALITATIVE, and GENERIC. In the final section, the relationships among these categories are represented by means of a semantic map. INDIVIDUAL-LEVEL STATE (Section 4) is the label for the meaning of a pluractional marker that, combined with a lexical individual-level predicate, expresses a single permanent state. In this case, the resulting form is automatically an individual-level predicate, and the connection to iteration is the most obscure. PROPERTY (Section 5) is the label for the meaning of a pluractional or aspectual marker that, combined with a lexical stage-level predicate, introduces a permanent characteristic of an individual. This characteristic typically shows up in recurrent events. CAPACITIVE (Section 6) is the label for the meaning of a pluractional or aspectual marker that is used to express the permanent capacity of an individual to perform a certain event. This permanent capacity is pragmatically related to iteration: if one repeatedly takes part in an event, it means that this person (or device) is capable of doing it. QUALITATIVE (Section 7) is the label for the meaning of a pluractional or aspectual marker that, combined with a lexical stage-level predicate, indicates that an individual permanently belongs to a class that is characterized by regular participation in a specific event. GENERIC meaning (Section 8) is the label for the meaning of a sentence (and a pluractional or aspectual marker used in this sentence) that introduces a permanent state of affairs. This state of affairs is a generalization of different repeated events with different individuals. In contrast to other IL-meanings, GENERIC meaning is used to characterize a permanent feature not of a single individual, but of a whole class of individuals.

2. The semantic domain of pluractionality: an overview The semantic domain of pluractionality refers to the relatively large set of meanings corresponding to events that are ‘plural’ in any sense – i.e. that are repeated or have a plural-like internal structure. Normally, pluractional meanings are classified as a subset of a bigger set of aspectual meanings. However, studies on pluractionality are few compared to the extensive work on aspect more generally. The term pluractionality that I use here was originally proposed by Paul Newman (1980, 1990) in his case study of

Individual-level meanings in pluractionality

177

Chadic languages, but now the term has been adopted as a suitable label for the complete variety of different meanings involving any kind of plurality of events (also compare studies from a more formal perspective, e.g. Lasersohn 1995, Yu 2003, van Geenhoven 2005, and a non-formal study by Wood 2007 who use the term in a wider sense than Newman originally did). Grammatical markers that are used mainly for pluractional meanings are labelled accordingly as pluractional markers. I am aware of five crosslinguistically oriented studies of pluractionality (Dressler 1968, Cusic 1981, Xrakovskij 1989b / Xrakovskij 1997, Šluinskij 2005, Wood 2007). A first semantic distinction between different clusters of pluractional meanings was proposed by Dressler (1968), in which distributive and nondistributive pluractionality are differentiated. The difference depends on whether a new participant is involved in each iteration of an event (i.e. distributive), or not (i.e. non-distributive). In the following discussion of pluractional meanings, distributive pluractionality is not considered. A second semantic distinction was formulated independently by Cusic (1981) and by Xrakovskij (1989a), and consists in differentiating eventinternal (in Xrakovskij’s terms, “multiplicative”) vs. event-external (in Xrakovskij’s terms, “iterative”) pluractionality. If a single event consists of a series of repeated sub-events, as in (1), it is classified as event-internal pluractionality. If a whole event is repeated, as in (2), it is called eventexternal pluractionality (for further discussion, see Šluinskij 2006). English (Indo-European, Germanic) (1) John is coughing. (2) John does his morning exercises every day. Both event-internal and event-external pluractionality are clusters of specific meanings. For instance, typical examples of event-external pluractional meanings are the standard habitual and usitative. Habitual describes an event that takes place regularly, once (or a fixed number of times) per period during a certain time (3). Usitative corresponds to an event that takes place under a certain condition (4).2 Lithuanian (Indo-European, Baltic; Ambrazas 1997: 246) (3) kasdien ei-dav-au tavęs pasitik-ti. every_day go-HAB.PST-1SG you.ACC meet-INF ‘Every day I came to see you.’

178

Andrey Shluinsky

Nenets (Uralic, Samoyedic) (4) nʲeka-w tur-pʔ, xada-w elder_brother-1SG come-CONV grandmother-1SG ja-mʔ pʲirʲe-mba-sʲti. soup-ACC.SG cook-DUR-HAB ‘When/if my elder brother came, my grandmother cooked soup.’ Pluractional markers used for event-internal meanings are usually called event-internal pluractionals, or multiplicatives. The term ‘wide pluractionals’ refers to those more rare ones that are used both for event-internal and event-external meanings.

3. Individual-level predicates The distinction between stage-level and individual-level predicates is wellestablished in formal semantics. It was originally introduced by Carlson (1977), who used the terms stage-level vs. individual-level properties. Essentially the same opposition, though outside any formal framework, was proposed independently by Bulygina (1982), who defines stage-level predicates as occurrences and individual-level predicates as qualities. Individual-level predicates express potentially stable characteristics that normally do not change throughout the whole period in which an individual exists (5). Stage-level predicates are used for transitory properties that correspond to separate, relatively short stages of the individual’s existence (6). English (Indo-European, Germanic; Manninen 2001: 1) (5) John loved Mary. (6) John kissed Mary. Individual-level predicates form a subset of stative predicates. All the stable features of an individual are states, but many states are temporary and therefore stage-level. For example, love and be white are stative individual-level predicates, sleep and be standing are stative stage-level predicates, and kiss and run are dynamic stage-level predicates. Various semantic features that distinguish stage-level from individuallevel predicates were identified by Diesing (1988), Kratzer (1995), and Chierchia (1995), based mainly on English data. Among these features is the ‘lifetime effect’ (Musan 1997), a property of individual-level (but not stage-level) predicates in a zero context. For example, John in (5) is likely

Individual-level meanings in pluractionality

179

to be presumed dead, but in (6) he is not. The lifetime effect seems to be a substantial property of individual-level predicates, although it can be eliminated contextually. The lexical meaning3 of individual-level predicates does not allow them to be combined with pluractional meanings. An event that corresponds to an individual-level predicate does not change during the long-term period in which it takes place, so an individual cannot be involved in such an event more than once. Therefore it is to be expected that a verb whose meaning is individual-level will not combine with pluractional markers at all (at least as understood with individual-level lexical meaning). Indeed, in some languages verbal4 individual-level predicates (frequently referred to in grammatical descriptions as verbs of state) have a reduced compatibility with many verbal markers. For instance, in Maninka (Keita 1986: 115) the only grammatical markers that can be combined with the verbs of state are positive and negative particles, cf. the use of a positive particle in (7). In contrast, stage-level predicates, labelled verbs of action, can have a wide set of possible tense-aspect-modality markers, among them the Habitual, an event-external pluractional marker (8). Maninka (Niger-Congo, Western Mande) (7) kě tùn ká kùdù. this man POS little ‘This man was short.’ (Keita 1986: 115) (8)

mùsû yé ná tòbì-là. woman COP sauce cook-HAB ‘The woman cooks the sauce.’ (Keita 1986: 111)

In other languages verbal individual-level predicates follow the standard verbal paradigm, but their pluractional forms are semantically problematic or even impossible. Compare the normal use of the Habitual marker in Nenets on a stage-level verb in (9) with the obligatory reinterpretation of an individual-level predicate into a stage-level one in (10) and the complete ungrammaticality of (11), where such a reinterpretation is semantically impossible. Nenets (Uralic, Samoyedic) to-sʲti. (9) wanʲa Vanya comes-HAB ‘Vanya comes regularly.’

180

Andrey Shluinsky

(10) wasʲa maša-mʔ menʲe-sʲti. Vasya Masha-ACC.SG love-HAB 1. ‘Vasya loves Masha regularly.’ {i.e. regularly falls in love with Masha} 2. *‘Vasya (permanently) loves Masha.’ (11) *wasʲa nʲenčʲa wada-mʔ tʲenʲewa-sʲti. Vasya Nenets language-ACC.SG know-HAB exp. ‘Vasya knows Nenets.’ {i.e. can speak Nenets} However, contrary to the expectation formulated above, Maninka and Nenets are not the most typical cases cross-linguistically. In many languages verbal individual-level predicates may combine with event-external pluractional markers – a phenomenon which I call the INDIVIDUAL-LEVEL STATE. Furthermore, in some cases pluractional markers, when applied to lexical stage-level verbs, cause such verbs to acquire some individual-level features. This is the case with other individual-level pluractional meanings.5

4. INDIVIDUAL-LEVEL STATE as a pluractional function It is well known that event-external pluractional markers can be combined in various languages with verbal individual-level predicates to express the permanent state that is lexically expressed by the verb.6 Consider the following Komi-Zyrian examples: in (12) the -l- suffix serves as a marker for an iterated event, but in (13), with a lexical individual-level predicate, it expresses a permanent state of an individual. Komi-Zyrian (Uralic, Finnic) (12) mʲe mʲešök-jas nov-l-i, šybʲit-l-i... I bag-ACC.PL carry-ITER-PST.1SG throw-ITER-PST.1SG ‘I (repeatedly) carried the bags and threw them.’ (13) mʲe töd-l-i vojna-tö lʲitʲeratura sʲertʲi. I know-ITER-PST.1SG war-ACC.SG literature by ‘I used to know the war from the literature.’ (Serebrennikov 1960: 87)

Individual-level meanings in pluractionality

181

A similar phenomenon can also be observed in languages that have a progressive vs. non-progressive distinction. For example, in English a single individual-level event and a repeated stage-level situation with a present tense reference are expressed by the Simple Present, but for single-event uses of stage-level predicates the Continuous Present is used. Exactly the same distribution can be found in Maasai: typically, the simple form is used for habitual (14) and the progressive form is used for the imperfective viewpoint (15). However, a stage-level predicate requires the simple form, although no repetition is involved in this case (16). Maasai (Nilo-Saharan, Nilotic; Tucker and Mpaayei 1955: 60) (14) e-nya nkishu nkujit. 3SG-eat cattle grass ‘Cows eat grass.’ (15) a-pik-ita nkiri enkima. 1SG-put-PROG meat fire ‘I am putting meat on the fire.’ (16) a-nyor inkiri. 1SG-like meat ‘I like meat.’ The fact that permanent states are semantically close to habituals is well known. Kučera (1981: 181) claims that “iteratives represent states, not activities”; Padučeva (1985: 223) includes directly habitual events in the set of consistent states; de Swart (2000) claims that a habitual interpretation of an imperfective verbal form is a tool for stativizing a dynamic predicate. However, it is necessary to make two remarks about this relation between states and habituals. First, the set of verbs that describe a single event with a habitual marker can be wider than the set of clearly individual-level predicates. Bagvalal, according to Kibrik (2001: 237–239), provides an example of this; see example (17) with a habitual form of the individual-level verb ī- ‘know’ and example (18) with a habitual form of the verb hal- ‘be ill’ that would be expected to be stage-level, but in fact does not have the meaning of a repeated state, but instead that of a single one.

182

Andrey Shluinsky

Bagvalal (East Caucasian, Andic) (17) di-ba he-r-ʕagila sajuz-ił ̄ mis’-abi I-AFF what-N.PL-every Union-GEN language-PL r-ī-r-ō-r. N.PL-know-IPFV-HAB-N.PL ‘I know all the languages of the (Soviet) Union.’ (Kibrik 2001: 233) (18) ʕali hal-ō-w. Ali be_ill- IPFV-HAB-M ‘Ali is ill.’ (Kibrik 2001: 237) Second, the use of event-external pluractional markers for a single INlike in (13) are typical, but not universal; see the Nenets examples (10–11) above and the example from Karachay-Balkar in (19). In Karachay-Balkar, the habitual form of the predicate süj- ‘be in love / fall in love’ cannot have the meaning of a single state ‘be in love’, but can be applied only to the inchoative meaning and involves an iteration of the event ‘fall in love’. DIVIDUAL-LEVEL STATE

Balkar (Altaic, Turkic) (19) kerim lejla-nɨ süj-üücü-dü. Kerim Lejla-ACC love-HAB-3SG ‘Kerim usually falls in love with Lejla.’ / *‘Kerim loves Lejla.’ To summarize, the INDIVIDUAL-LEVEL STATE is a common use of eventexternal pluractional markers, and therefore should be considered within the semantic domain of pluractionality.7

5. PROPERTY as a pluractional function Pluractional marking may function to express a basic characteristic of an individual that becomes apparent in specific regular events, a function of pluractionality I will call PROPERTY. To illustrate, consider the following. The English sentence in (20) is ambiguous: it can either be used for a single occasion of John’s smoking and therefore have a normal single-event perfective interpretation (default), or it can indicate John’s habit of smoking with no reference to any specific event. In Russian, the analogous sentence (21) is normally understood in a zero context as a statement that

Individual-level meanings in pluractionality

183

characterizes the referent with the meaning of PROPERTY (21.1), and the episodic interpretation (21.2) can only be induced by context. English (Indo-European, Germanic) (20) John smoked. Russian (Indo-European, Slavic) (21) vas-ja kuri-l. Vasya-NOM smoke-PST.M.SG 1. ‘Vasya was a smoker.’ 2. ‘Vasya was smoking.’ Example (2) above (John does his morning exercises every day) is used to express just the fact of the regular repetition of an event; (21.1) Vasya was a smoker focuses on internal features of an individual and as a result its relationship to repetition is mediated. Although, logically, every repeated situation can be interpreted as a property of its participants, the semantic difference between standard habitual contexts like (2) and PROPERTY contexts like (21.1) is substantiated by linguistic evidence. PROPERTY contexts are expressed more often like INDIVIDUAL-LEVEL STATE contexts than like habitual contexts. In Lithuanian, the pluractional marker -dav- is restricted by proper event-external uses. Both INDIVIDUAL-LEVEL STATE (22) and PROPERTY (23a) are encoded by the simple form of Past tense; the -dav- marker in (23b) involves a standard repetition that has no focus on the individual’s properties. Lithuanian (Indo-European, Baltic) (22) mano močiut-ė my grandmother-NOM.SG ‘My grandmother knew tales.’

pasak-as. mokėj-o know-PST.3SG tale-ACC.PL

(23) a. mano senel-is rūk-ė. my grandfather-NOM.SG smoke-PST.3SG ‘My grandfather was a smoker.’ b. mano brol-is kasdien rūky-dav-o. my brother-NOM.SG every_day smoke-ITER.PST.3SG ‘(Last year), my brother smoked every day.’ When a pluractional marker is used to express PROPERTY, then it is also regularly used for the INDIVIDUAL-LEVEL STATE. For instance, in Agul the

184

Andrey Shluinsky

so-called Generic Past (see Merdanova 2004 for the details of the Agul tense-aspect system) can be used to express habitual meanings (24), PROPERTY (25) and INDIVIDUAL-LEVEL STATE (26). Agul (East Caucasian, Lezgic) (24) ze ħabaw-a har jaʁ-a šurpa rüx-e-f-ij. my grandmother-ERG every day-TMR soup cook-IPFV-SBST-PST ‘My grandmother cooked soup every day.’ (25) ze ħadad-a p’ap’ruc-ar du-a-f-ij. my grandfather-ERG cigarette-PL pull-IPFV-SBST-PST ‘My grandfather was a smoker [=sucked cigarettes].’ (26) ze ħabaw-as ħakijat-ar my grandmother-DAT tale-PL ‘My grandmother knew tales.’

ħa-f-ij. know-SBST-PST

In summary, the PROPERTY meaning characterizes some specific properties of an individual, but also involves iterative semantics and therefore belongs to the pluractional semantic domain.

6. CAPACITIVE as a pluractional function The meaning of CAPACITIVE has been discussed in some cross-linguistic studies on verbal systems in general (see, e.g., Plungjan 1997), but has not been considered in the context of the typology of pluractionality. The CAPACITIVE meaning relates to the capacity of an individual to participate in an event (ability in terms of Bybee, Perkins and Pagliuca 1994; participantinternal possibility in terms of van der Auwera and Plungian 1998). Logically, CAPACITIVE can be described as a special case of the meaning PROPERTY (viz. a capacity to participate in an event is also an inherent feature that can become explicit in repeated events). However, it can nonetheless be expressed differently from typical PROPERTY contexts and therefore should be described as a distinct meaning. There are two types of CAPACITIVE contexts that tend to have different kind of expressions: inherent and acquired. Consider the Russian examples in (27) and (28).

Individual-level meanings in pluractionality

185

Russian (Indo-European, Slavic) (27) ivan plava-et brass-om. Ivan swim-PRES.3SG breaststroke-INST ‘Ivan can swim using the breaststroke.’ (28) naš-a stiral’n-aja mašin-a our-F.NOM washing-F.NOM machine-NOM otžima-et bel’-e. wring_out-PRES.3SG linen-ACC ‘Our washing machine can wring out the linen.’ Both of these sentences are typical examples of CAPACITIVE contexts, but they differ in that (27) necessarily implies that the individual whose capacity is characterized has participated in the event (acquired capacity) at least once,8 while (28) can refer to a new device that has never been used (inherent capacity). These two semantic types of CAPACITIVES can also be distinguished linguistically: in some languages there are examples of verbal markers that can be used only for inherent CAPACITIVE, but not for acquired CAPACITIVE. For example, in Lithuanian a Simple Past form can be used in the inherent CAPACITIVE context (29a), but the acquired CA9 PACITIVE context requires a modal verb (29b). Lithuanian (Indo-European, Baltic) (29) a. ši mašin-ėl-ė vir-ė sriub-ą. this machine-DIM-NOM.SG cook-PST.3SG soup-ACC.SG (Once upon a time my father bought an amazing machine;) ‘that machine could cook soup itself.’ mokėj-o b. mano ses-el-ė jau my sister-DIM-NOM.SG already can-PST.3SG iš-vir-ti sriub-ą. PREF-cook-INF soup-ACC.SG (When I returned home,) ‘my little sister could already cook soup herself.’ More frequently, however, these two types of CAPACITIVES are not formally distinguished and are expressed by verbal forms with a general meaning or by event-external pluractionals. If an event-external pluractional marker is used for the CAPACITIVE meaning, then it is also used for the INDIVIDUAL-LEVEL STATE; consider examples (30–31) from Susu.

186

Andrey Shluinsky

Susu (Niger-Congo, Western Mande) (30) n xunya ginɛ ma nu bɔrɛ I younger_sibling female to RETR soup ɲin-ma a yɛtɛ ra. cook-HAB (s)he self with (When I returned,) ‘my younger sister could cook soup herself.’ (31) n mama nu gaxo-ma barɛ I grandmother RETR be_afraid-HAB dog ‘My grandmother was afraid of dogs.’

yara. front

In summary, the CAPACITIVE meaning characterizes an inherent property of an individual, but it also involves (potential) iteration. This meaning connects the semantic domain of pluractionality with the semantic domain of modality – see Tatevosov (2005) for further discussion.

7. QUALITATIVE as a pluractional function Sentences with a QUALITATIVE meaning are used to characterize an individual as belonging to a specific ontological class that is defined by participating regularly in an event. A typical case of such use is an expression that a person belongs to a particular profession. In some languages the semantics of general event-external pluractional markers include such QUALITATIVE contexts. For instance, Robins (1958: 82) gives some examples of the Yurok infix -eg- with event-external plueg-·ay ‘pass regularly, use a certain ractional semantics (la·y ‘pass’ > l-egtrack’) and some QUALITATIVE examples (kemol ‘steal’ > k-egeg-emol ‘be a thief’). However, some languages have a special Qualitative marker that is not used to indicate other types of pluractionality. This is the case of the Nivkh Qualitative suffix -xy-: Nivkh (isolate; Panfilov 1965: 75) (32) hy n’ivx q‘otr lyi-xy-d’. this man bear kill-QUALIT-NONFUT ‘This man kills bears.’ As Panfilov (1965: 75) shows, the -xy- suffix in Nivkh can be contrasted with reduplication, which marks other kinds of event-external plu-

Individual-level meanings in pluractionality

187

ractionality. The example in (33a), with a reduplicated form, has the meaning of a repeated event; (33b), with the Qualitative marker, has the meaning of assigning a group of individuals to a specific class. Nivkh (isolate; Panfilov 1965: 75) (33) a. n’yŋ qan-gu vava-d’-ɣy. our dog-PL fightRED-NONFUT-PL ‘Our dogs fought repeatedly.’ b. n’yŋ qan-gu va-xy-d’-ɣy. our dog-PL fight-QUALIT-NONFUT-PL ‘Our dogs are pugnacious.’ Specialized QUALITATIVE markers differ according to whether the regular iteration of the event is obligatory. This is the case for the Nivkh sentences (32) and (33b), which normally imply that the events of killing and fighting occur more or less regularly. In contrast, the specialized marker of the QUALITATIVE in Agul, the so-called Intentional form marked with -je-fis used both for QUALITATIVE contexts that involve a regular iteration (34) and for QUALITATIVE contexts where it is not assumed that the event takes place regularly (35). Agul (East Caucasian, Lezgic) (34) ze ħabaw-a šurpa rüx-e-je-f-ij. my grandmother-ERG soup cook-IPFV-PART-SBST-PST ‘My grandmother cooked soup professionally.’ {i.e. she was a cook} (35) ze ħabaw-ak itːal k-e-je-f-ij. my grandmother-CONT illness CONT-be_situated-PART-SBST-PST ’My grandmother was disposed to illness.’ The set of the possible uses of the specialized QUALITATIVE marker -tyin Selkup (labelled in Kuznecova, Xelimskij and Gruškina 1980 as characterization Aktionsart) is even wider. There are standard QUALITATIVE examples (tə̄ly- ‘steal’ > tel-ty- ‘be a thief’), examples that do not imply a regular iteration (sɔ̄ty- ‘bite’ > sat-ty- ‘be a biter’) and, finally, examples that do not even imply that the relevant situation has ever taken place at all (tal’y- ‘break’ > tal’-ty- ‘be fragile’, Kuznecova, Xelimskij and Gruškina 1980: 232–233). While the QUALITATIVE meaning typically is indicated via a specialized marker, languages that lack such a marker normally express the QUALITA-

188

Andrey Shluinsky

meaning in the same way as they mark standard event-external pluractionality. See, for example, the Udmurt -l- pluractional suffix in a standard habitual context (36a) and in a QUALITATIVE (36b) one: TIVE

Udmurt (Uralic, Finnic) (36) a. baba-je každyj nunal šyd pöžʲt-yl-i-z grandmother-1SG every day soup cook-ITER-PST-3SG ‘My grandmother cooked soup every day.’ stolovyj-yn šyd pöžʲt-yl-i-z b. baba-je grandmother-1SG canteen-LOC soup cook-ITER-PST-3SG ‘My grandmother cooked soup in a canteen.’ In summary, the meaning of QUALITATIVE can be expressed by a special marker, but if such a marker is not available, then such meanings are normally expressed by event-external pluractionals.

8. GENERIC uses of pluractionals Whereas the QUALITATIVE function uses certain features of a class to make statements about its members, the GENERIC meaning involves statements about a class of individuals as a whole. GENERIC refers to a ‘timeless’ event (and therefore it has truth conditions that do not depend on a point in time). For example, (37) is true in present, past and future. English (Indo-European, Germanic) (37) Cows eat grass. There is a vast literature on generics; see Carlson and Pelletier (1995) for a discussion on genericity and further references. Here I will discuss only the connection between genericity and pluractionality. An important feature of GENERIC sentences is the generic reference of the noun phrases involved.10 As Givón (1984: 406–408) shows, the generic reference is semantically close both to definite and non-referential functions. Definite and generic references are similar in that a class of individuals that is introduced by a generic noun phrase can be regarded as a single individual with some specific features. Therefore a GENERIC sentence contains an individual-level predicate with a term referring to a whole class of objects. This explains why GENERIC sentences refer to ‘timeless’ events: the lifetime of the whole class of objects11 is pragmati-

Individual-level meanings in pluractionality

189

cally predicted to be infinite. For example, consider (38), a GENERIC sentence with a past time reference, termed a past generic (Dahl 1975, 1985). English (Indo-European, Germanic) (38) Dinosaurs ate kelp. In (38) we observe a standard lifetime effect of a predicate with an ILmeaning, as defined in this chapter. Truth conditions of (38) are exactly limited by the period during which the necessary class of objects exists. The fact that GENERIC sentences can be defined as sentences with generic NPs is a natural explanation of the fact that special verbal markers for GENERIC meaning are quite rare, although there are some examples, for instance the Usual Aktionsart in Selkup (Kuznecova, Xelimskij and Gruškina 1980: 219) as illustrated in (39), and the Generic Present12 in Agul (Merdanova 2004: 109) as illustrated in (40). Selkup (Uralic, Samoyedic; Kuznecova, Xelimskij and Gruškina 1980: 219) (39) šettyr-qyn šīpa halqal-k-a. spring-LOC duck shed_feathers-GENER-PRES ‘A duck sheds its feathers in spring.’ Agul (East Caucasian, Lezgic; Merdanova 2004: 109) (40) ʡurd-ana ixp̄ -ar uʁ-a-f-e. winter-TMR snow-PL snow/rain-IPFV-SBST-PRES ‘In winter it snows.’ More typically, the GENERIC meaning is expressed either by an eventexternal pluractional marker (that may also have some IL-meanings) or by a simple non-pluractional form (that may contrast with a pluractional form used in other contexts). The first case is exemplified by the Habitual in Swahili; compare a standard habitual sentence (41) and a GENERIC sentence (42). Swahili (Niger-Congo, Central Bantu) (41) mimi hu-soma asubuhi. I HAB-read in_the_morning ‘I usually read in the morning.’ (Gromova and Oxotina 1995: 232)

190

Andrey Shluinsky

(42) paka hu-kamata panya a-ka-wa-la. cat HAB-catch mouse.PL CL-CONS-CL-eat ‘A cat catches mice and eats them.’ (Gromova and Oxotina 1995: 234) The second case is exemplified by Kolyma Yukaghir (Maslova 2003: 140), where a habitual form is used in (43), but the GENERIC context in (44) requires the simple imperfective form. Yukaghir (isolate) (43) tuda: ta:t ed’u-t moda:-nun-d’i:l’i. long_ago so live-SS live-HAB-1PL ‘We used to live that way long ago.’ (Maslova 2003: 140) (44) puge-d-in el-al’a:-čuon qodo:-nu-j. summer-POSS-DAT NEG-melt-CAR lie-IPFV-3SG ‘It lies without melting till summer.’ (Maslova 2003: 131) Similarly, Dahl (1985: 103) assumes that English periphrastic past habitual forms seem to be impossible in past generics;13 compare (38) with (45). English (Indo-European, Germanic; Dahl 1985: 103) (45) ???Dinosaurs used to eat kelp. Thus the GENERIC meaning may be understood as marginally associated with the individual area of the pluractional semantic domain in that it is semantically close both to pluractionality and to the description of a single event.

9. Conclusion: a semantic map for individual-level pluractionals In this chapter, I have introduced five meanings of pluractional markers that are usually left out of descriptions of this semantic domain, although all of them are generally well-known in grammatical typology. These meanings are INDIVIDUAL-LEVEL STATE, PROPERTY, CAPACITIVE, QUALITATIVE and GENERIC. Table 1 presents a list of diagnostic sentences for these meanings.

Individual-level meanings in pluractionality

191

Table 1. Diagnostic sentences for IL-meanings Meaning INDIVIDUAL-LEVEL STATE PROPERTY CAPACITIVE QUALITATIVE GENERIC

Diagnostic sentences My grandmother knows tales. John loves his wife. My grandfather smokes / is a smoker. Our door creaks / is creaky. Our child can speak. This machine can dry the linen. My mother cooks in a restaurant / is a cook. My grandmother is prone to be ill. Cows eat grass. A peasant grows vegetables.

According to my data, pluractional IL-meanings can be organized in a semantic map in the following way (see Figure 1). This map follows the standard methodology of semantic maps: if two meanings are regularly expressed in a language by the same marker, then they are semantically close. This fact is indicated by the lines on the semantic map. This semantic map includes no diachronic perspective, so there are lines but no arrows; these lines are not grammaticalization paths. The figure obeys the principle of continuity: if a marker is used for two meanings that are not directly connected by a line, it should also be used for all the meanings between these two. A semantic map is thus used here as a tool to clarify the observed restrictions on expressing IL-meanings. We can see that the possible set of IL-meanings that a marker can express is limited by the semantic relations between them.

Figure 1. Semantic map for IL-meanings

Due to lack of space, the argumentation for this semantic map is presented in a brief tabular form. In Table 2 information about possible and impossible IL-meanings of some event-external pluractional and imperfec-

192

Andrey Shluinsky

tive markers is summarized. Only the information about the meanings from Figure 1 is given in Table 2. Markers with identical clustering of the meanings in focus are separated by a thinner horizontal line. Table 2. Possible and impossible IL-meanings of specific markers Language Basque

KarachayBalkar Udmurt Mandarin Chinese Danish Tajik Agul English Basque

French

Marker QUAL Imperfective Progres- + sive (Imperfective Gerund in -ten/-tzen + aritzen + Past auxiliary) Habitual (-ucu) +

GEN –

IL-S –

PROP –

CAP –

+

–

–

–

Iterative (-l) zero form

+ +

+ +

– +

– +

– –

Past Imperfect (me-) Habitual Past (-ji) used to construction14 Imperfective Past (Imperfective Gerund in -ten/-tzen + Past auxiliary) Imparfait

+ + – + +

+ + – – +

+ + + + +

+ – + + +

– + + – +

+

+

+

+

+

(The following abbreviations are used: QUAL – QUALITATIVE, GEN – GENERIC, IL-S – INDIVIDUAL-LEVEL STATE, PROP – PROPERTY, CAP – CAPACITIVE.)

To conclude, IL-meanings form an important part of the semantic domain of pluractionality. By characterizing these meanings as a whole and describing each of them in depth, this chapter contributes to an overlooked aspect of this area of grammar.

Notes 1.

The help of many people in collecting data for cross-linguistic research cannot be overestimated. Mukadas Abdullaeva (Tajik), Oumar Camara (Susu), Mads Eskildsen (Danish), Natalja Inokaitėnė (Lithuanian), Solmaz R. Merdanova

Individual-level meanings in pluractionality

2.

3.

4.

5.

193

(Agul), Iker Sancho (Basque) were my consultants. Udmurt and Chinese data were collected by Natalia Serdobolskaya and Anna Leontieva, respectively. Karachay-Balkar and Nenets data were collected during fieldtrips organized by Moscow State University, partly by myself and partly by Anna Pazelskaya. I am strongly indebted to all these people and also to the consultants of the other languages that were not cited in this chapter. I also thank: Sergei G. Tatevosov, who was my supervisor during my work on my dissertation; Yury A. Lander, Timur A. Maisak, Elena V. Paducheva and Vladimir A. Plungian who were the reviewers; others colleagues for their valuable remarks and fruitful discussions; Bernard Comrie and Julia Kuznetsova who read this chapter and made lots of corrections; Pattie Epps, Michael Cysouw and an anonymous reviewer for their multiple notes and corrections; Joshua Wilbur for improving my English. Examples with no reference come from my own work with native speakers. Abbreviations in glossing: 1, 3 – 1st, 3rd person, ACC accusative, ADV adverbial marker, AFF affective, CAR caritive, CL (noun) class marker, CONS consecutive, CONT contact localization, CONV converb, COP copula, DAT dative, DIM diminutive, DUR durative, ERG ergative, F feminine, GEN genitive, GENER generic, HAB habitual, INF infinitive, INST instrumental, IPFV imperfective, ITER iterative, LOC locative, M masculine, N neutral (gender), NEG negative, NOM nominative, NONFUT non-future, PART participle, PL plural, POS positive particle, POSS possessive, PREF (verbal) prefix, PRES present, PROG progressive, PST past, QUALIT qualitative, RED reduplication, RETR ‘retrospective shift’, SG singular, SS same-subject, SBST substantivizer, TMR temporalis form. To be more precise, verbs are usually polysemantic, and therefore the same verb understood in one meaning can be an individual-level predicate, but understood in another meaning a stage-level one. Accordingly, here I am speaking not about a lexical meaning of a verb in general, but about a lexical meaning of a verb taken in a single specific interpretation. The majority of individual-level predicates are nominal, such as in the sentence John is a linguist. For a formal semantic description, the difference between nominal and verbal predicates is not crucial, but in the context of a typology of verbal markers and their uses I speak first of all about the properties of the individual-level predicates that are expressed by verbs. My term ‘individual-level meanings’ is an extension of the standard term ‘individual-level predicates’ that originates from formal semantics. This extension seems to be useful for a cross-linguistic study, although it would not be possible to make this extension within a formal semantics framework. From the point of view of the formal semantics tradition, if an event is repeated, it is classified automatically as a stage-level predicate. However, I claim that if a predicate introduces a repeated event, but is still used for expressing a permanent (or quasi-permanent) characteristic of an individual, this predicate has substantial features in common with prototypical individual-level predicates.

194 6.

7.

8. 9.

10.

11. 12. 13. 14.

Andrey Shluinsky There seems to be no substantial semantic difference between individual-level verbs with and without a pluractional marker. If one changes the iterative form tödli ‘knew’ in (16) with a simple past form tödi ‘knew’, the meaning of the sentence is more or less the same. An alternative analysis is possible that treats pluractionality and the meaning of INDIVIDUAL-LEVEL STATE as related parts of the same larger category. My decision to describe the meaning of INDIVIDUAL-LEVEL STATE – and also other pluractional IL-meanings – as based on the fact that the uses of pluractional markers for INDIVIDUAL-LEVEL STATE are quiet marginal. At least non-zero markers of this type are systematically described for different languages as habitual markers and not as durative/imperfective markers. Typically, a CAPACITIVE context implies that an individual participates in the event more or less regularly, or at least has participated in it many times. Note that as a result (some) CAPACITIVE meaning can be lacking in a particular verbal system. In some languages only modal verbs can be used to express an individual’s capacity. Usually, sentences that refer to some features of unique objects, such as ‘The Earth is round’ are also labelled GENERIC sentences. But for such objects the difference between different kinds of referential status seems to be irrelevant, cf.: *Any Earth is round, *Every Earth is round, etc. Sentences like those mentioned in note 10 refer to unique objects that also have an infinite lifetime. But not the parallel Generic Past, see examples (27–29). Note, however, that speakers’ judgements on this example differ. Data on used to constructions is from Google.com and from previously mentioned sources. See the relevant examples. QUALITATIVE: My father used to sell pizzas. GENERIC: *Dinosaurs used to eat kelp. INDIVIDUAL-LEVEL STATE: I used to love you, Mary Jane. PROPERTY: Churchill used to smoke. CAPACITIVE: John used to swim *‘John could swim’.

References Ambrazas, Vytautas (ed.) 1997 Lithuanian Grammar. Vilnius: Baltos lankos. Bulygina, Tat’jana V. 1982 K postroeniju tipologii predikatov v russkom jazyke [Towards a typology of predicates in Russian]. In Semantičeskie tipy predikatov [Semantic types of predicates], Olga N. Seliverstova (ed.), 7–85. Moscow: Nauka.

Individual-level meanings in pluractionality

195

Bybee, Joan L., Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect and Modality in the Languages of the World. Chicago: University of Chicago Press. Carlson, Gregory N. 1977 Reference to kinds in English. Ph.D. diss., University of Massachusetts, Amherst. Carlson, Gregory N., and Francis J. Pelletier (eds.) 1995 The Generic Book. Chicago – London: University of Chicago Press. Chierchia, Gennaro 1995 Individual-level predicates as inherent generics. In Carlson and Pelletier 1995. Cusic, David D. 1981 Verbal plurality and aspect. Ph.D. diss., Stanford University. Dahl, Östen 1975 On generics. In Formal Semantics of Natural Language; Edward L. Keenan (ed.), 99–111. Cambridge: Cambridge University Press. 1985 Tense and Aspect Systems. Oxford: Blackwell. de Swart, Henriette 2000 Tense, aspect and coercion in a cross-linguistic perspective. In Proceedings of the Berkeley Formal Grammar conference, University of California, Berkeley; Miriam Butt and Tracy H. King (eds.). Stanford: CSLI Publications. Available at http://csli-publications. stanford.edu/LFG/5/bfg00/bfg00deswart.pdf. Diesing, Molly 1988 Bare plural subjects in stage/individual contrast. In Genericity in Natural Languages, Manfred Krifka (ed.), 107–154. Tübingen: University of Tübingen. Dressler, Wolfgang 1968 Studien zur Verbalen Pluralität. Wien: Österreichische Akademie der Wissenschaften. Givón, Talmy 1984 Syntax: A Functional-Typological Introduction. Vol. 1. Amsterdam – Philadelphia: John Benjamins. Gromova, Nelli V., and Natal’ja V. Oxotina. 1995 Teoretičeskaja grammatika jazyka suaxili [A theoretical grammar of Swahili]. Moscow: Nasledie. Keita, Boniface 1986 Eléments de description du Malinke de Kita (Mali). (Publications du centre de dialectologie africaine, №6). Grénoble: Université de Grénoble III. Kibrik, Aleksandr E. (ed.) 2001 Bagvalinskij jazyk. Grammatika. Teksty. Slovari [Bagvalal. Grammar. Texts. Dictionaries]. Moscow: Nasledie.

196

Andrey Shluinsky

Kratzer, Angelica 1995 Stage-level and individual-level predicates. In The Generic Book; Gregory N. Carlson and Francis J. Pelletier (eds.), 125–175. Chicago, London: University of Chicago Press. Kučera, Henry 1981 Aspect, markedness and t0. In Tense and Aspect, (Syntax and Sematics. Vol. 14), Philip J. Tedeschi and Annie Zaenen (eds.), 177–189. New York: Academic press. Kuznecova, Ariadna I., Evgenij A. Xelimskij, and Elena V. Gruškina 1980 Očerki po sel’kupskomu jazyku [Essays on Selkup]. Moscow: Izdatel’stvo Moskovskogo universiteta. Lasersohn, Peter 1995 Plurality, Conjunction and Events. Dordrecht, Boston: Kluwer Academic Publishers. Manninen, Satu 2001 A minimalist analysis of stage level and individual level predicates. The department of English in Lund Working Papers 1. http://www. englund.lu.se/images/stories/pdf-files/workingspapers/vol01/Satu.pdf Maslova, Elena 2003 A Grammar of Kolyma Yukaghir. Berlin: Mouton de Gruyter. McGregor, William 1990 A Functional Grammar of Gooniyandi. Amsterdam, Philadelphia: John Benjamins. Merdanova, Solmaz R. 2004 Morfologija i grammatičeskaja semantika agul’skogo jazyka (na materiale xpjukskogo govora) [Morphology and semantics of grams of Agul (based on Xpjuk patois)]. Moscow: Sovetskij pisatel’. Musan, Renate 1997 Tense, predicates and lifetime effects. Natural Language Semantics 5: 271–301. Newman, Paul 1980 The Classification of Chadic within Afroasiatic. Leiden: Universitare Press. 1990 Nominal and Verbal Plurality in Chadic. Dordrecht: Foris. Paducheva [Padučeva], Elena V. 1985 Vyskazyvanie i ego sootnesennost’ s dejstvitelnost’ju [Sentence and its reference to reality]. Moscow: Nauka. Panfilov, Vladimir Z. 1965. Grammatika nivxskogo jazyka [A grammar of Nivkh]. Vol. 2. Moscow, Leningrad: Nauka. Plungian [Plungjan], Vladimir A. 1997 Vid i tipologija glagol’nyx sistem [Aspect and the typology of verbal systems]. In Trudy aspectologičeskogo seminara filologičeskogo

Individual-level meanings in pluractionality

197

faku’lteta MGU [Papers of the seminar on aspectology at the philological faculty, Moscow State University]. Vol. 1, Čertkova, Marina Ju. (ed.), 173–190. Moscow: Izdatel’stvo Moskovskogo universiteta. Robins, Robert H. 1958 The Yurok Language: Grammar, Texts, Lexicon. Berkeley, Los Angeles: University of California press. Serebrennikov, Boris A. 1960. Kategorii vremeni i vida v finno-ugorskix jazykax permskoj i volžskoj grupp [Tense and aspect in Permic and Volgic Finno-Ugric languages]. Moscow: Nauka. Shluinsky [Šluinskij], Andrej B. 2005 Tipologija predikatnoj množestvennosti: količestvennye aspektual’nye značenija [A typology of pluractionality: quantitative aspectual meanings]. Ph.D. diss., Moscow State University. 2006 K tipologii predikatnoj množestvennosti: organizacija semantičeskoj zony [A tentative typology of pluractionality: organization of the semantic domain]. Voprosy jazykoznanija 1: 46–75. Tatevosov, Sergei 2005 From habituals to futures: discerning the path of diachronic development. In Perspectives on Aspect, Henk J. Verkuyl, Henriette de Swart, and Angeliek van Hout (eds.), 181–197. Dordrecht: Springer. Tucker, Archibald N., and J. Tompo Ole Mpaayei 1955 A Maasai Grammar. London etc.: Longmans, Green and Co. van der Auwera, Johan, and Vladimir A. Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. van Geenhoven, Veerle 2005 Atelicity, pluractionality, and adverbial quantification. In Perspectives on Aspect, Henk J. Verkuyl, Henriette de Swart, and Angeliek van Hout (eds.), 107–124. Dordrecht: Springer. Wood, Esther J. 2007 The semantic typology of pluractionality. Ph.D. diss., University of California, Berkeley. Xrakovskij, Viktor S. 1989a Semantičeskie tipy množestva situacij i ix estestvennaja klassifikacija [Semantic types of event plurality and their natural classification]. In Xrakovskij (ed.), 5–53. Xrakovskij, Viktor S. (ed.) 1989b Tipologia iterativnyx konstrukcij [A typology of iterative constructions]. Leningrad: Nauka. 1997 Typology of Iterative Constructions. München: Lincom. Yu, Alan C. L. 2003 Pluractionality in Chechen. Natural Language Semantics 11 (3): 289–321.

The symbiosis of descriptive linguistics and typology: A case study of desideratives Olesya Khanina

1. Introduction1 This chapter illustrates the symbiotic relationship between descriptive linguistics and typology. Typology supplies parameters for description of individual languages, and descriptive grammars provide the data that make the formulation of these parameters possible. If a grammar writer does not have at hand parameters of cross-linguistic variation for some aspect of the language, this aspect may be underdescribed in his/her grammar. Accordingly, if this aspect is underdescribed for a number of languages, it does not receive a typological account. In brief, cross-linguistic generalizations feed better descriptions, and better descriptions allow for more solid crosslinguistic generalizations. This insuperable circular relationship is illustrated in this chapter with a case-study of desiderative constructions. I focus on the ways arguments are encoded in these constructions, as an example of how some data can be recurrently missing in grammars due to the lack of a cross-linguistic awareness of the topic, and how the inclusion of this data could in turn improve the typology of desideratives. The layout of the presentation is as follows. First, I introduce my typological research on desiderative constructions and the parameters of crosslinguistic variation connected to their argument structure (Section 2). I then discuss the parameters that are usually overlooked in grammatical descriptions (Section 3), and comment on the types of cross-linguistic generalizations that are currently missing in typology due to the lack of consistent data on these parameters (Section 4). I conclude with some general remarks on this symbiosis.

200

Olesya Khanina

2. Typological research on desiderative constructions My dissertation (Xanina 2005) and postdoctoral research (Khanina 2008, in prep.) have explored desiderative constructions cross-linguistically and have proposed some parameters for cross-linguistic variation, as listed below. I have used a balanced sample of 73 languages, built on the principles presented in Rijkhoff and Bakker (1998).2 I understand desideratives as the means languages use to express the concept of wanting, cf. English Peter wants to leave, Peter wants Mary to leave. Their cross-linguistic variability is illustrated by the following sentences: (1)

Bukiyip (Torricelli, Kombio-Arapesh): desiderative verb n-a-dúk-anú tamiok lowénam. n-a-kli 3SG.M_A-R-want [3SG.M_A-R-kill-3SG.M_P axe Lowenam]3 ‘He wanted to kill Lowenam with an axe.’ (Conrad and Wogiga 1991: 67)

(2)

Ingush (East Caucasian, Nakh): desiderative noun б-а.4 са Iаж б-аа безам sy wazh b-a’a biezam b-y. I.GEN [apple.NOM CL-eat] desire.NOM CL-be ‘I want to eat an apple.’ (Sultan Mereshkov, p.c.)

(3)

Kham (Sino-Tibetan, Bodic): desiderative affix yahm ta-ruhp-yo, ge-bahl-si-rih-zya. door PROH-close-IMP 1PL-watch-MM-DESID-CONT ‘Don’t close the door, we want to watch.’ (Watters 2002: 108)

(4)

Mapudungun (Araucanian): desiderative particle küpá amu-fu-y-iñ welu DESID go-IPD-IND-1NON_SG-PL BUT elu-ñma-nge-la-y-iñ. give-IO-PASS-NEG-IND-1NON_SG-PL ‘We(pl) wanted to go but were not allowed to.’ (Smeets 1989: 301)

The symbiosis of descriptive linguistics and typology

(5)

201

Telugu (Dravidian, South-Central): desiderative construction naaku peekaaTa dwaaraa Dabbu I.DAT [card_game through money sampadinc-aal an(i) undi. earn-OBG COMP] be.PRS.3SG ‘I want to win money at cards.’ (Krishnamurti and Gwynn 1985: 369)

These examples show that languages may use constructions of various morphosyntactic types to express wanting: verbs (1), nouns (2), affixes (3), particles (4) or other conventionalized constructions5 (5). In all cases, the real-world states of affairs denoted by these constructions are of the same nature, and it is this expressive function that defines desiderative constructions as a class.6 Note that here I understand ‘construction’ as a very general neutralizing term encompassing both lexemes (verbs and nouns) with their argument slots, and affixes/particles with argument or distributional limitations they may bring to their verbal hosts (consistent with the idea of constructions as understood in the Construction Grammar approach; see Goldberg 1995, 2006, and others). My study has revealed a number of parameters of cross-linguistic variation applicable to desiderative constructions. These are concerned, on the one hand, with the semantic content of these constructions (Khanina 2008), and, on the other hand, with the encoding of arguments involved in the event of someone’s wishing for something. The discussion in this chapter is restricted to the argument problematic only. Even though desiderative constructions vary considerably in their morphosyntax, each type can have all participants of the want-event expressed. These universal conceptual participants are the Wanter, the one who wants, and the Wanted, the object of Wanter’s wishes. The Wanted can be one of two subtypes, the Wanted-object and the Wanted-event; and Wantedevents can also be one of two subtypes, with subjects coreferential to the Wanter (i.e. same-subject wanting) and with subjects non-coreferential to the Wanter (i.e. different-subject wanting). I illustrate this with some examples: English (Indo-European, Germanic) (6) Peter wants an apple. (7) Peter wants to leave. (8) Peter wants Mary to leave.

202

Olesya Khanina

In (6), Peter is the Wanter and an apple is the Wanted, specifically the Wanted-object. A Wanted-event is exemplified in (7–8): in (7) the Wantedevent ‘Peter leaves’ has a subject coreferential with the Wanter, and in (8) the Wanted-event ‘Mary leaves’ has a subject non-coreferential with the Wanter. My study has shown that all these semantic participants, the Wanter, the Wanted-object, the Wanted-event with a coreferential subject, and the Wanted-event with a non-coreferential subject, can in principle be expressed within desiderative constructions of any morphosyntactic type. In other words, desiderative lexemes as well as desiderative grams can allow the expression of all of these arguments. This is illustrated in (9–15): all types of semantic participants are expressed with the desiderative verbs in (9–11) and with the desiderative affixes in (12–15). Yurok (Algic) (9) nek skewoksim-e-k’ pa’ah I want-STEM-1SG water7 ‘I want some water.’ (Robins 1958: 19) ḱe-kemey-e-ḱ. (10) skewoksim-e-ˀm hes ki want-STEM-2SG Q [FUT PRON_PREF_2-go_home-STEM-SG] ‘Do you want to go back home?’ (Robins 1958: 168)

(11) ’imi

skewoksi’m ki kolish want.3SG8 [FUT somebody ’u-komchum-e-k’ ku PRON_PREF_3-know-STEM-SG PRTCL ’u-meyoomoy-k’. PRON_PREF_3-be_pregnant-SG] ‘She did not want anyone to know she was pregnant.’ (Kroeber 1911: 425–426; Kroeber 1976: 313–314) NEG

Examples (9–11) come from Yurok, an Algic language of California. All of them show the Wanter expressed either by an overt noun phrase, as in (9), or by a personal cross-reference marker on the desiderative verb. The capability of this verb to be used with Wanted-objects is exemplified in (9), where pa’ah ‘water’ denotes this semantic argument. Wishing for an event is illustrated in (10) and (11): in (10) the subject of the Wanted-event is coreferential with the Wanter ‘you’, while in (11) the subject of the Wanted-event kolish ‘somebody’ is not coreferential with the Wanter ‘she’.

The symbiosis of descriptive linguistics and typology

203

The following examples show that the same possibilities are available when desiderative affixes are used. Imbabura Quechua (Quechuan) (12) ñuka-ta yaku-naya-n. I-ACC water-DESID-3 ‘I want water; I am thirsty.’ (Cole 1982: 180) (13) ñuka-ta miku-naya-n. I-ACC eat-DESID-3 ‘I want to eat.’ (Cole 1982: 181) Yagua (Peba-Yaguan) (14) siimiryú̜ú̜y. sa-jmiy-rú̜ú̜y 3SG-eat-DESID ‘He wants to eat.’ (Payne 1985: 289) (15) siimimyuuryú̜ú̜tyánníí. sa-jmiy-muuy-rú̜ú̜y-tániy-níí 3SG-eat-COMPL-DESID-CAUS-3SG ‘Hei wants himj,*i to finish eating.’ (Payne 1985: 290) The example from Imbabura Quechua in (12) illustrates desiderative affixes with the Wanter, ‘I’, and the Wanted-object, ‘water’, expressed, and example (13) does the same for the Wanted-event with a subject coreferential to the Wanter. In the Yagua examples in (14) and (15), a Wanted-event is encoded by a construction with a desiderative affix: in (14) the subject of the Wanted-event is coreferential to the Wanter, while in (15) it is not. The examples above demonstrate that all conceptual participants of a want-event can be expressed, regardless of the morphosyntactic nature of a desiderative – whether it is lexical, as in (9–11), or grammatical, as in (12– 15). This discussion, however, has so far been about possibilities, not probabilities, and as one may guess, some types of wishing are more often expressed by some and not other morphosyntactic types of desideratives. Indeed, grammatical desideratives, affixes and particles, are used much more frequently for denoting desire relating to same-subject events, while lexical desideratives, verbs and nouns, tend to be used more frequently for denoting desire relating to objects and to different-subject events.9 In other words, even if desiderative grams can be attached to nouns, as in (12), and

204

Olesya Khanina

can be used for different-subject wanting, as in (15), this seems to be rather rare cross-linguistically, and also in natural discourse.10 Thus, in my sample of 73 languages, only three languages have been documented as having desiderative affixes that can occur with nominal stems: Imbabura Quechua, Moseten, and Labrador Inuttut (this is also known to be possible in Tuvan (Turkic), though this language was not part of my original sample).11 In languages that have desiderative grams, desiderative lexemes are usually used with reference to the wanting of objects and to different-subject wanting, and same-subject wanting is reserved for the bound morphemes. Thus we come to the main point of this chapter: the symbiosis between linguistic description and typology. I have just exemplified linguistic structures that are possible, but infrequent, and so often omitted in grammatical descriptions and disregarded in typology. In the rest of this discussion, I explore in some detail how linguistic structures of this kind are treated in linguistic descriptions (Section 3), and why they matter for typology (Section 4).

3. Missed linguistic structures: arguments of desiderative constructions If one considers the arguments of desiderative constructions, it becomes possible to compile the following list of data types that are recurrently missing in descriptive grammars: i) information on the ability of a desiderative construction to have each of the possible Wanted arguments: the Wanted-object, the Wanted-event with coreferential subject, and the Wanted-event with non-coreferential subject (in other words, the full information regarding the restrictions on possible arguments of a given desiderative construction); ii) information on alternative ways of expressing the desiderative meaning that cannot be expressed with a desiderative construction: in particular, if a construction cannot take some type of Wanted arguments (the Wanted-object, the Wanted-event with coreferential subject, and the Wanted-event with non-coreferential subject), how wanting of these arguments can be expressed in this language; iii) information on case frame alterations that can be imposed on a verb by a given grammatical desiderative.

In this section, each of these types of data is commented on and illustrated with examples from languages in my sample.

The symbiosis of descriptive linguistics and typology

205

First, as was briefly mentioned above, the mere fact of whether a desiderative construction can or cannot be used with each of the conceptual participants of the want-event may come without any remarks in a grammar. This means that a desiderative construction may be discussed or at least exemplified in the grammar, but only with a limited set of arguments, and no data may be provided on the possibility of its use with other types of arguments. Thus, for twenty of the thirty-four desiderative grams (affixes and particles) attested in the languages in my sample, their descriptions contained no hints as to whether they can or cannot be used with nouns, i.e. whether they can express the wanting of an object. For ten desiderative grams, their descriptions contained no hints as to whether they can or cannot express different-subject wanting. All thirty-four desiderative grams, however, were described as expressing same-subject wanting; that is, their capacity to express arguments like the Wanted-event with a coreferential subject was explicitly stated in all sources. This seems to be a good example of how the most frequent uses are reflected in grammars, while others may not be. Though there are no universal restrictions to prevent any morphosyntactic type of desiderative construction from taking each of the conceptual participants of the want-event, individual desiderative constructions can be confined to some arguments only. In other words, there are desiderative constructions, both lexical and grammatical, that can be used only with some arguments, but not others; cf. Table 1. This table features only desideratives for which I managed to find full information on their possible arguments in the corresponding sources. Note that the number of ‘personal communication’ references indicates the paucity of this information in grammatical descriptions.12 When there are restrictions on possible arguments of desiderative constructions, one also wants to find information on alternative ways to express the meaning in question. These alternative constructions may be regular, i.e. always used to render this meaning, or non-regular or periphrastic, i.e. used occasionally to render this meaning along with other constructions.

206

Olesya Khanina

Table 1. Desideratives (from the balanced 73-language sample) that have restrictions on their arguments13 Restriction on possi- Lexical desiderative ble arguments constructions Only Wanted-object Not attested/recorded14 possible, e.g. Peter wants an apple (not Peter wants to leave or Peter wants Mary to leave)

Grammatical desiderative constructions – Labrador Inuttut affix -guk-/-ɣuɣ- (Fortescue, Jacobson, and Kaplan 1995, Smith 1978) – Moseten affix -wei(Sakel 2004) – Hua affix -su-/-si-/ -sa(Haiman 1980) – special person-number paradigm of Korowai (Enk and de Vries 1997) – Yagua affix -rú̜ú̜y(Payne 1985)

– Ingush verb low (Sultan Mereshkov, p.c.); e.g. Peter wants to – Tukang Besi verb hada leave and Peter wants (Donohue 1999); Mary to leave (not – Fongbe verb bà (Claire Peter wants an apple) Lefebvre, p.c.) – Kham verb (həi) li (David Watters, p.c.); – Thai verb yàak may be used with only certain nouns, e.g. ‘water’ (Iwasaki and Preeya 2005, Smyth 2002) – Labrador Inuttut affix Only Wanted-event – Ingush noun biezam, cf. example (2) -guma- (Bok-Bennema with coreferential 1991, Johns 1999, subject possible, – Ingush verb dog doagha e.g. Peter wants to (Sultan Mereshkov, p.c.) Smith 1978) leave (not Peter wants – Ket verb tabaq (Edward – Kham affix -rih-, cf. an apple or Peter Vajda, p.c.) example (3) (Watters 2002) wants Mary to leave) – Basque nahian ibili may belong here too (Iker San– Japanese affix -(i)ta cho, p.c.) (Onishi 1997) – Tukang Besi affix -um(Donohue 1999) Only Wanted-event possible,

The symbiosis of descriptive linguistics and typology Table 1 continued.

207

Desideratives (from the balanced 73-language sample) that have restrictions on their arguments

Restriction on possible arguments Only Wanted-object and Wanted-event with coreferential subject possible,

Lexical desiderative constructions – Kham verb pəĩ (David Watters, p.c.)

Grammatical desiderative constructions – Semelai particle lən (Kruspe 2004)

Not attested

– Labrador Inuttut affix -tqu-/-kqu- (BokBennema 1991, Johns 1999, Smith 1978)

– Japanese verb hos(h)ii (Onishi 1997)15

Not attested

e.g. Peter wants an apple and Peter wants to leave (not Peter wants Mary to leave)

Only Wanted-event with noncoreferential subject possible, e.g. Peter wants Mary to leave (not Peter wants an apple or Peter wants to leave)

Only Wanted-object and Wanted-event with noncoreferential subject possible, e.g. Peter wants an apple and Peter wants Mary to leave (not Peter wants to leave )

Regular alternatives are documented for Japanese and Labrador Inuttut. As can be seen from Table 1, Japanese uses i) the lexeme hos(h)ii to express wishing for an object (16) and with respect to an event with a non-coreferential subject (17), and ii) the affix -(i)ta to express wishing relating to an event with a coreferential subject (18).

Japanese (16) ore wa kore o hosii I TOP this ACC want ‘I want (to have) this (thing).’ (Onishi 1997: 225) (17) ore wa taroo ni it-te hosii I TOP Taroo DAT go-CONJ want ‘I want Taroo to go.’ (Onishi 1997: 225)

208

Olesya Khanina

(18) ore wa soko ni iki-tai I TOP there LOC go-DESID ‘I want to go there.’ (Onishi 1997: 224) The same table shows that Labrador Inuttut has three different desiderative affixes for each of the subtypes of Wanted: -guk-/-ɣuɣ- is used with Wanted-objects, -guma- with same-subject Wanted-events, and -tqu-/-kquwith different-subject Wanted-events (this is also true for the other Inuit languages). Itelmen, a moribund language of Siberia, provides an example of a nonregular alternative for the desiderative affix -aɬ-/-a-. This affix can have only one type of argument, Wanted-events with coreferential subjects, cf. (19).16 When necessary, the meaning of a wishing event where the Wanter and the doer are not identical is expressed periphrastically, as in (20). Itelmen (Chukotko-Kamchatkan, Southern) (19) тхтум т’ињл̭ -а-∅-с-кипнэн. txtum t’inʲɬ-a-∅-s-kʲipnen.17 Itelmen_boat buy-DESID-PF-PRS-1SG_A.3PL_OBL ‘I want to buy an Itelmen boat for them.’ (Alexander Volodin, p.c.) (20) пɬҳа-ˀн-кэ қ-ч’элы-қзу-с-сх нвонкэ, манк pɬχa-ˀn-ke q-čʼeli-qzu-s-sx nwonke, mank other-PL-ALL 2.IMP-do-IPF-PRS-2PL so as тэзва-н-кэ пɬҳа-ˀн əн-к’-ч’элы-қзу-с-кисхэн, tezwa-n-ke pɬχa-ˀn ən-k’-č’eli-qzu-s-kʲisxen you-OBL-ALL other-PL 3PL_S-SUBJ-do-IPF-PRS-3PL_A.2PL_P тэзва-н-ɬханŏң. tezwa-n-ɬxanŏŋ. you-OBL-manner Luke 6:31: ‘Treat the others in the way you would like the others to treat you. (lit., Treat the others in the way that the others would treat you, in your manner.)’ (Volodin 2002: 54) Sentence (20) shows that even in a prototypical context where no doubt is cast on the presence of a different-subject wish, Itelmen does not use a desiderative lexeme or a gram, but a non-grammaticalized conventionalized construction, as a regular way of rendering this meaning (Alexander Volodin, p.c.). This fact may be due to the level of endangerment of the language, but this is difficult to determine because there is insufficient

The symbiosis of descriptive linguistics and typology

209

information about alternative ways of expressing different-subject wishing in other languages. Grammatical descriptions that report the existence of desiderative constructions relating only to same-subject Wanted-events usually do not provide consistent information on how other types of wanting can be expressed. Though information on alternatives for same-subject constructions in Itelmen was given to me by the Itelmen specialist Alexander Volodin, for Japanese (Onishi 1997) and Tukang Besi (Donohue 1999), commentary on alternatives is included within the descriptive literature. To illustrate, I present the Tukang Besi case in more details. Tukang Besi uses the verb hada in one of its main desiderative constructions: Tukang Besi (Austronesian, Sulawesi) (21) ku-hada [te wila-’a i ’one]. 1SG-want CORE go-NMN OBL beach ‘I want to go to the beach.’ (Donohue 1999: 389) After introducing this verb in the ‘Complementation’ section of his grammar, Donohue writes the following: A normal nominal may not appear as the object of hada, as seen by the ungrammaticality of (22) (22) *ku-hada te po’o. 1SG-want CORE [mango] ‘I want a mango.’ In order to make (22) grammatical using hada [i.e. to express wanting of an object with the verb hada – O.Kh.], a serial verb construction must be used, as in (23), or a complement construction, as in (24). (23) ku-hada-’awa te 1SG-want-get CORE ‘I want to get a mango.’

po’o.

mango

(24) ku-hada [ku-’awa te po’o]. 1SG-want 1SG-get CORE mango ‘I want to get a mango.’ (Donohue 1999: 389)18

210

Olesya Khanina

The last type of data that grammars sometimes fail to provide is relevant to grammatical desideratives only. When a verb is used with a desiderative affix or particle, the case encoding of its arguments may change, as the Shipibo-Konibo examples in (25–26) show. Shipibo-Konibo (Panoan; Valenzuela 2003: 576) (25) e-n-ra wame rete-ai. I-ERG-DIR_EV paiche:ABS kill-INCOMPL ‘I kill paiche.’ wame (26) e-a-ra I-ABS-DIR_EV paiche:ABS ‘I want to kill paiche.’

rete-kas-ai. kill-DESID-INCOMPL

Sentence (25) illustrates the usual case frame of the Shipibo-Konibo verb rete ‘to kill’ (ergative, absolutive), as opposed to (26) where the attachment of the desiderative suffix -kas- changes it to (absolutive, absolutive). Such alterations of the initial case frame is far from universal, and many desideratives grams do not induce it, cf. the Chuvash desiderative suffix -sšăn in (27–28). Chuvash (Altaic, Turkic) (27) epĕ xam-ăn jultaš-a kura-t-ăp. I.NOM self.1SG-GEN friend-DAT/ACC see-PRS-1 ‘I see my friend.’ (personal fieldnotes) xam-ăn jultaš-a kura-sšăn. (28) epĕ I.NOM self.1SG-GEN friend-DAT/ACC see-DESID ‘I want to see my friend.’ (Xanina 2003: 101)

To conclude this section, it is worth noting that gaps in grammatical descriptions, like those discussed here, come to the attention of typologists only because some descriptions provide information that is lacking in others. So typology can by no means blame linguistic description for these gaps; we should instead be grateful that the gaps have become noticeable and have allowed for another round of typological questions and studies to come.

The symbiosis of descriptive linguistics and typology

211

4. Missed typological generalizations: arguments of desiderative constructions There are surely instances in which typological generalizations are overlooked even if the necessary information is present in grammars. However, here I am interested in those cases in which there is not enough data in the descriptions to make generalization possible. For desideratives, such missed typological generalizations can be classified by the three types of information that grammatical descriptions fail to provide. Table 2 reproduces the typology of the missing information in a more concise manner and gives the types of cross-linguistic generalizations that cannot be made due to this lack of information. I will comment on each generalization (1– 4) in turn. Table 2. Missed typological generalizations concerning desideratives, by types of missing data in descriptive grammars Missing data in descriptive grammars Missing typological generalizations I. Information on the ability of a 1. Generalizations about what is indeed desiderative construction to have cross-linguistically rare in the argument each of the possible Wanted ardomain of desideratives guments a. The lack of such generalizations also prevents possible insights that may be gained from the rare cases: why do they exist at all? 2. Generalizations on cross-linguistically recurrent patterns in the expression of each argument in the case of differentsubject wanting encoded by desideratives affixes; 3. Differentiation between concepts that II. Information on alternative ways universally require a conventionalized to express a desiderative meaning structure for their expression and those that do not; III. Information on case frame altera- 4. Generalizations on cross-linguistically tions that can be imposed on a recurrent patterns of expression of the verb by a given grammatical deWanter by grammatical desideratives. siderative

The absence of some information in a descriptive grammar may or may not be meaningful, depending on the nature of the linguistic phenomenon

212

Olesya Khanina

in question.19 For example, if the category of number is described with reference to singular and dual numbers only, one can in most cases deduce from such a description that trial or paucal numbers do not exist in the language. The same applies to distant past or distant future tenses; even when there is no direct statement of their absence, this can be assumed based on the description of the tense system in the language. The absence of phonemic tone can be similarly calculated from the description of phonemic distinctions in the language, even without any direct reference to it. Accordingly, the absence of those features which would be in a paradigmatic relationship with the features overtly mentioned in a grammar does not present a problem to a reader. However, if the category in question is not discussed in the grammar at all, i.e. the category of number, the category of tense, or the phonemic inventory do not get any treatment in a grammar, then no information of this kind can be retrieved from it. This is exactly the case for the argument structures of desiderative constructions. If a grammatical description devotes some space to the discussion of possible arguments of a desiderative, one can assume that those arguments that were not discussed cannot be used with this desiderative. But, as often happens, when there is no such discussion, no data can be retrieved.20 Returning to the case of desideratives, descriptive passages are rarely devoted exclusively to them and their syntax; thus, if descriptive grammars provide information only about the possibility of some of their arguments, but not about the impossibility of others, it is hardly possible to understand whether the latter are rarely mentioned in the literature or indeed rarely attested in the languages of the world. For example, it is rather typical for grammatical descriptions to provide examples in which a desiderative construction is used to render a wishing event with the Wanter identical to the doer, but not any other type of wishing event. Does this imply that it is relatively frequent for desiderative constructions across the world to be restricted to the expression of this type of wishing event only? Not necessarily – and thus we still do not know which restrictions in Table 1 are more widespread in the languages of the world. For example, the Japanese system of desiderative constructions discussed in the previous section is unique in my study in that it has two desideratives, one for wishing for objects and non-coreferential events, and the other for wishing for coreferential events. At first glance there is nothing too strange about such a configuration, and the fact that it is actually attested in one language proves it. If it turns out, however, that this configuration is indeed very rare, we will need to consider it more thoroughly

The symbiosis of descriptive linguistics and typology

213

and think of the implications it may have for our understanding of the semantics of the concept ‘want’. Of the twenty-seven desiderative affixes attested in the languages in my sample, only four are described as being able to take a Wanted argument with a non-coreferential subject. Of the remaining twenty-three affixes, fourteen are documented as prohibiting such arguments, and for nine affixes there is no information available about such arguments. Moreover, the four desiderative affixes that can take a Wanted argument with a noncoreferential subject are not uniform. Of these, Hua -su-/-si/-sa- can take such an argument only in questions with 1st or 3rd person subjects (for more information see Haiman 1980: 441–443), while Labrador Inuttut -tqu/-kqu- can be used with this type of argument but with no others (BokBennema 1991, Johns 1999, Smith 1978). Ultimately, only two of the affixes in question, Chuvash -sšăn (Xanina 2003) and Yagua -rú̜ú̜y (Payne 1985), can be used with any type of Wanted-event, both with coreferential subjects and with non-coreferential ones. Interestingly, both these affixes (and one desiderative particle from a language not in my sample, Lahu gâ; Matisoff 1973: 574) show morphosyntactic uniformity in the encoding of different-subject wishing.21 All of them employ a causative morpheme to augment the argument structure of the lexical verb so that it can take an additional argument, the doer of the Wanted-event; cf. the following examples (29–30) from Chuvash and (31–32) from Yagua. Chuvash (Altaic, Turkic) (29) văl š’ak kĕnekĕ-ne vula-sšăn. (s)he.NOM this book-DAT/ACC read-DESID ‘(S)he wants to read this book.’ (fieldnotes) š’ak kĕnekĕ-ne (30) văl sana (s)he.NOM you.DAT/ACC this book-DAT/ACC vula-ttar-a-sšăn. read-CAUS-STEM-DESID ‘(S)he wants you to read this book’. (Xanina 2003: 101)

Yagua (Peba-Yaguan) (31) siimiryú̜ú̜y. sa-jmiy-rrú̜ú̜y 3SG-eat-DESID ‘He wants to eat.’ (Payne 1985: 289)

214

Olesya Khanina

(32) siimimyuuryú̜ú̜tyánníí. sa-jmiy-muuy-rrú̜ú̜y-ttániy-n niy níí 3SG-eat-COMPL-DESID-CAUS-3SG ‘Hei wants himj,*i to finish eating.’ (Payne 1985: 290) These three unrelated languages, from different parts of the world,22 show the same pattern, but there is no further data by which to determine whether this pattern is (almost) universal, or is essentially non-existent (as its presence in these three languages could be pure coincidence). If descriptions of languages with desiderative affixes paid more attention to the arguments these can take, there would be more data on how these affixes express different-subject wishing, but so far no generalizations are possible regarding cross-linguistically recurrent patterns in the expression of Wanted-events by means of desiderative affixes. From the perspective of major European languages, I consider all possible arguments of desiderative constructions as having more or less equal semantic/pragmatic status in all human languages. All of them would presumably tend to be expressed by some conventionalized lexical or grammatical means, as opposed, say, to local adjuncts that can be expressed by an array of non-conventionalized means: adverbs, noun phrases in various oblique cases, and various adverbial clauses. If patterns similar to Itelmen are, however, yet to be discovered in a significant number of languages, my opinion will have to be reconsidered. As was described in the previous section, this disappearing language does not have any regular way of expressing the wishing of an event in which the doer is not identical to the Wanter. Periphrasis is used instead, and this does not seem to be grammaticalized periphrasis, but occasional periphrasis. For Itelmen the reason for this may lie in language attrition; Itelmen may once have had a grammaticalized way of expressing this type of wishing that has since been lost. Summing up, the lack of information on alternative ways of expressing meanings that cannot be expressed by a given desiderative leaves us in the dark about the universality or non-universality of some types of wishing expressions. The final issue I will discuss here relates to the fact that cross-linguistic generalizations about grammatical desideratives, if they existed, would allow us to evaluate a number of theoretical claims. How should we analyze the Wanter noun phrase when a desiderative affix or particle expresses the wishing for an event whose doer is identical to the Wanter (as illustrated in (29) and (31) above, among other examples)? Should the Wanter noun phrase, e.g. văl in (29), be analyzed as also expressing the doer of the

The symbiosis of descriptive linguistics and typology

215

Wanted-event? Or does it express the Wanter only, and the doer of the Wanted-event should be considered omitted? One type of factual data that would contribute to a cross-linguistic answer to this question can be found in case frame alterations like that illustrated in the Shipibo-Konibo sentences (25–26) above. Such cases are rarely reported in the literature, although one other desiderative affix (Imbabura Quechua -naya-, cf. examples (12–13) with the Wanter in the Accusative case) and one particle (Pilaga setake, cf. Vidal 2001) in the languages in my sample were documented as producing a very similar kind of syntactic demotion of the original main participant of the verb upon the attachment of a desiderative gram.23 Because these are cases in which something overt happens to the noun phrase in question, they provide important data that can be used for the syntactic analysis of that noun phrase. But again, the current pool of grammatical descriptions does not allow for a proper understanding of how frequent this phenomenon is and what syntactic changes to this noun phrase are possible.

5. Conclusion In this chapter, I have presented data relating to the arguments of desiderative constructions that illustrate the symbiotic relationship between linguistic description and typology. When there are no, or few, typological generalizations available for a particular subject area, it is described inconsistently in grammars: some grammars may provide a more detailed description of the topic, while others may give it no comment at all. The data from those grammars that address topics not yet investigated by typologists call attention to new areas of potential typological research. Starting from this incomplete data, typology formulates new questions, which then require a new kind of information from future grammatical descriptions. Such relationships, in which the outcomes of one side feed the necessities of the other, and vice versa, have a perfect right to be regarded as symbiotic. It is ultimately the work done by grammar writers that allows typologists to ask more from other grammar writers. One of the functions of typology is thus to be a kind of ‘router’ for the grammar-writing community, and so to serve grammar writers by providing guidelines for better grammar writing (see also Noonan 2005 on this issue). This is of course not the only function of typology, which also provides linguistic generalizations that inform theories of language and even of the human mind. Yet

216

Olesya Khanina

even while typology uses language descriptions for its own purposes, it has the considerable ‘side-effect’ of helping them to improve. It is hoped that these epistemological reflections may help to better understand both disciplines, linguistic description and typology, and the way they work hand in hand to reveal basic principles of human language.

Notes 1.

2. 3.

4. 5.

6.

I would like to express my most sincere gratitude to the editors of the volume, in particular Pattie Epps, for their patience in dealing with my contribution and their insightful advice that made it all possible. The reviewers’ comments were also very helpful. Needless to say, all the shortcomings of the paper are mine alone. For particulars of my sample see (Khanina 2008) or (Khanina, in prep.). Abbreviations used in this paper are as follows: 1 1st person, 2 2nd person, 3 3rd person, A agent, ABS absolutive, ACC accusative, ACT actual, ALL allative, CAUS causative, CL substantive class, COMP complementizer, COMPL completive, CONJ conjunctive used for clause-chaining, CONT continuous, CORE core case, DAT/ACC dative-accusative case, DAT dative, DESID desiderative, DIR_EV direct evidential, ERG ergative, FUT future, GEN genitive, IMP imperative, INCOMPL incompletive, IND indicative, IO indirect object, IPD impeditive, IPF imperfect, LOC locative, M masculine, MM middle marking, NEG negative, NMN nominalization, NOM nominative, NON_SG non singular, OBG obligative, OBL oblique, P patient, PASS passive, PF perfect, PL plural, PROH prohibitive, PRON_PREF pronominal prefix, PRS present, PRTCL particle, Q question, R real, SG singular, STEM stem marker, SUBJ subjunctive, TOP topic. The first line is in the traditional Ingush orthography; the second line is in the phonemic orthography adopted in Nichols (2004) and Nichols (in prep.). This type of desiderative construction, here referred to as ‘conventionalized’, does not involve any material that would introduce the desiderative meaning, whereas all other desiderative constructions do have such material, either lexical (such as verbs and nouns) or grammatical (such as affixes and particles). All material making up a conventionalized construction of this sort is used somewhere else in the language without any hint of a ‘want’ meaning. It is thus by abstraction and conventionalization that this structure conveys someone’s wishing for something. Note that desideratives are to be distinguished from optatives, e.g. English If only he were here now! Roughly speaking, optatives convey speaker’s wish, while desideratives convey anyone’s wish, see examples (1–5).

The symbiosis of descriptive linguistics and typology 7.

8. 9.

10. 11. 12. 13. 14.

15.

16. 17.

18. 19. 20.

217

My thanks go to Juliette Blevins for her generous help with glossing the Yurok examples and making (11) available to me. The original orthography of the sources is kept. The 3SG form is marked in Yurok by glottalization of the last consonant of the stem. ‘Conventionalized’ desiderative constructions, as exemplified in (5), show properties typical for both lexical and grammatical desideratives, and so they represent a more complicated case that is not discussed in this paper. Some discussion relevant to this may be found in Khanina (2008). In fact, the Imbabura Quechua desiderative suffix -naya- (see example 12) can occur only with a closed class of nouns. See Cole (1982), Sakel (2004), Fortescue, Jacobson, and Kaplan (1994), Anderson and Harrison (1999). In the cases of Ingush and Ket, I have had to rely on personal communication due to the absence of comprehensive reference grammars. Here again, I do not discuss ‘conventionalized’ desiderative constructions (see note 9). Note, however, that because my research focused on desideratives relating to the wanting of an event, I may have overlooked desiderative constructions that could be used with Wanted-objects only. The Japanese desiderative lexeme hos(h)ii represents an interesting case because its inflectional paradigm is that of inflected adjectives and not of verbs of the language (Backhouse 2004, Iwasaki 2002, Kaiser et al. 2001). Unlike most other adjectives, however, hos(h)ii can take an Accusative argument (denoting the Wanted-object, cf. example 16). The Japanese desiderative thus resembles the transitive verbs of the language, while morphologically hos(h)ii is an adjective. Given that no other desiderative adjective was attested in the languages in my sample and that morphosyntactically hos(h)ii indeed behaves as a verb, I have chosen to classify this lexeme as a desiderative verb. This was done only for the purposes of building a meaningful typology and inevitably compromises the actual data. There is no information available on whether the Itelmen desiderative affix can be used with Wanted-objects. The transliteration of the examples to IPA is mine, following the phonological principles of Itelmen as presented in Volodin (2005: 2); “č” stands for the IPA “tʃ”. Donohue (1999: 389) gives different example numbers, but the rest of the citation is reproduced intact. I would like to thank one of the reviewers who directed my attention to this issue. Another complication, however, lies in the fact that paradigmatic relations are not always perceived in the same strict way for syntactic categories as they are

218

Olesya Khanina

for phonological or morphological categories, and the issue of arguments of desiderative constructions is a syntactic question. 21. The Lahu form is a particle, not an affix, but this is a highly isolating language, and the functions of its particles are very similar to those of affixes in such agglutinative languages as Chuvash or Yagua. 22. Chuvash is spoken in western Russia, Yagua is spoken in Peru, and Lahu is spoken in China. 23. These cases are analyzed in detail in Khanina (in prep.).

References Anderson, Gregory D. S., and David K. Harrison 1999 Tyvan. (Languages of the world: Materials 257.) München: Lincom. Backhouse, Anthony E. 2004 Inflected and uninflected adjectives in Japanese. In Adjective Classes, R. M. W. Dixon and Alexandra Y. Aikhenvald (eds.), 50– 73. Oxford: Oxford University Press. Bok-Bennema, Reineke 1991 Case and Agreement in Inuit. Berlin: Foris Publications. Cole, Peter 1982 Imbabura Quechua. Amsterdam: North-Holland. Conrad, Robert J., and Kepas Wogiga 1991 An Outline of Bukiyip Grammar. (Pacific linguistics: Series C, Books; 113.) Canberra: Australian National University. Donohue, Mark 1999 A Grammar of Tukang Besi. Berlin: Mouton de Gruyter. Enk, Gerrit J. van, and Lourens de Vries 1997 The Korowai of Irian Jaya: Their Language in Its Cultural Context. (Oxford Studies in Anthropological Linguistics 9.) New York: Oxford University Press. Fortescue, Michael, Steven Jacobson, and Lawrence Kaplan 1994 Comparative Eskimo Dictionary. (Research Paper Number 9.) University of Alaska Fairbanks: Alaska Native Language Center. Goldberg, Adele 1995 Constructions. A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. 2006 Constructions at Work: The Nature of Generalization in Language. Oxford: Oxford University Press. Haiman, John 1980 Hua: A Papuan Language of the Eastern Highlands of New Guinea. (Studies in Language Companion Series 5.) Amsterdam: Benjamins.

The symbiosis of descriptive linguistics and typology

219

Iwasaki, Shoichi 2002 Japanese. Amsterdam: John Benjamins. Iwasaki, Shoichi, and Ingkaphirom Preeya, 2005 A Reference Grammar of Thai. New York: Cambridge University Press. Johns, Alana 1999 On the lexical semantics of affixal ‘want’ in Inuktitut. International Journal of American Linguistics 65 (2): 176–200. Kaiser, Stefan, Yasuko Ichikawa, Noriko Kobayashi, and Hilofumi Yamamoto 2001 Japanese: A Comprehensive Grammar. London: Routledge. Khanina [Xanina], Olesya 2003 Opyt opisanija semantičeskoj zony želanija i sredstv ee vyraženija (na materiale čuvašskogo jazyka) [Essay of desiderative semantic zone and its expression (based on Chuvash)]. Vestnik MGU, Filologija 5 [Working Papers of Moscow State University, Philology 5]: 101–116. 2005 Jazykovoe oformlenie situacii želanija: opyt tipologičeskogo issledovanija [Linguistic encoding of ‘wanting’: a typological study]. Ph.D. Diss., Moscow State University. 2008 How universal is wanting? Studies in Language 32 (4): 818–865. in prep. Typology of desiderative constructions. Krishnamurti, Bhadriraju, and John Peter Lucius Gwynn 1985 A Grammar of Modern Telugu. Delhi: Oxford University Press. Kroeber, Alfred Louis 1911 The languages of California north of San Francisco. University of California Publications in American Archaeology and Ethnology 9: 414–26. 1976 Yurok Myths. Berkeley: University of California Press. Kruspe, Nicole 2004 A Grammar of Semelai. Cambridge: Cambridge University Press. Matisoff, James A. 1973 The Grammar of Lahu. (University of California Publications: Linguistics 75.) Berkeley: University of California Press. Nichols, Johanna 2004 Ingush-English and English-Ingush Dictionary. London: Routledge Curzon. in prep. Ingush grammar. Noonan, Michael 2005 Grammar writing for a grammar-reading audience. Linguistic Typology 30 (2): 351–365. Onishi, Masayuki 1997 The grammar of mental predicates in Japanese. Language Sciences 19 (3): 219 – 233.

220

Olesya Khanina

Payne, Doris L. 1985 Aspects of the grammar of Yagua: a typological approach. Ph.D. Diss., University of California at Los Angeles. Rijkhoff, Jan, and Dik Bakker 1998 Language sampling. Linguistic Typology 2: 263–314. Robins, Robert H. 1958 The Yurok Language. Grammar, Texts, Lexicon. (University of California Publications in Linguistics 15). Berkeley: University of California Press. Sakel, Janet 2004 A Grammar of Mosetén. (Mouton Grammar Library 33.) Berlin: Mouton de Gruyter. Smeets, Catharina J. 1989 A Mapuche grammar. Ph.D. Diss., Leiden University. Smith, Lawrence R. 1978 A Survey of the Derivational Postbases of Labrador Inuttut (Eskimo). (National Museum of Man Mercury Series. Canadian Ethnology Service Paper No. 45) Ottawa: National Museum of Canada. Smyth, David 2002 Thai: An Essential Grammar. London: Routledge. Valenzuela, Pilar M. 2003 Transitivity in Shipibo-Konibo grammar. Ph.D. Diss., University of Oregon. Vidal, Alejandra 2001 Pilagá grammar (Guaykuruan family, Argentina). Ph.D. Diss., University of Oregon. Volodin, Alexander P. 2002 Mel Lanom Lukank Nkelycen (Blagoe Povestvovanie, Lukoj Napisannoe) [Good narration written by Luka]. Moscow: Bible Translation Institute. 2005 An introduction to the Itelmen language. Talk at Max Planck Institute for Evolutionary Anthropology in Leipzig, 2 March 2005. Watters, David E. 2002 A Grammar of Kham. Cambridge: Cambridge University Press.

Part IV. Clause structure and verbal derivation

Comitative as a cross-linguistically valid category Alexandre Arkhipov

1. Introduction1 The category of comitative has enjoyed a considerable increase of interest in the last decades, both in language-internal and in typological studies. Among the latter, Lehmann and Shin (2005) and the fundamental study by Thomas Stolz, Cornelia Stroh and Aina Urdze (2006) devoted to comitatives and related categories should be mentioned as the most significant recent works. However, there still remain some conceptual aspects of comitative that appear to be not quite established. One of them, ironically, is the very notion of comitative: how to delimit its scope and what kind of variation can be taken as internal or external for this category. The purpose of this chapter is to give a definition of comitative which would be easily applicable cross-linguistically and at the same time yield the class of constructions which is most distinct from other established categories, such as coordinating, sociative, and depictive constructions. In a nutshell, comitative is defined as a particular construction type used to ‘pluralize’ a participant – that is, to predicate the same state of affairs of two individual participants, such that the main predicate itself is not repeated and the two participants are not equal in their syntactic status. Section 2 will present a slightly more formal definition of the comitative construction. Section 3 will consider a number of categories which are to be distinguished from the comitative proper.

2. Comitative as a pluralizing construction The definition proposed below is generally inspired by Stassen’s (2000) definitions of comitative and coordinate strategies of NP-conjunction. It aims to provide a solid basis for identifying comitative constructions crosslinguistically, relying more on the formal properties with as little reference to semantic details as possible. This position is taken not to the detriment of the semantic component in general, but in order to make the definition most helpful given possible limitations on the data at hand. At the same

224

Alexandre Arkhipov

time, it is not accidental. I argue that formal properties are essential for the comitative to be considered an autonomous category, consistently differentiable from its relatives across languages. 2.1. Preliminary notions Following E. Maslova (1999; 2007), I use the label of participant set to refer to cases when “two or more separate individuals are ascribed the same type of participation in the event” (Maslova 2007: 337). A participant set can be referred to by different means, e.g. by a non-singular NP, a pair of coordinate NPs, etc., as in The children/They/Jim and Sarah were running in the garden. A comitative construction, as in Jim was running in the garden with Sarah, is another such strategy.We will only consider cases with two individual participants here. An opposition in terms of structural rank (along the lines of Stassen 2000) is important in characterizing the asymmetry of the comitative construction. This is a generic morphosyntactic characteristic of a noun phrase corresponding principally to the relational hierarchy: subject > direct object > indirect object > oblique (cf. Johnson 1977). However, in languages where grammatical relations are claimed to be irrelevant, e.g. in Archi (Kibrik 1977) or Acehnese (Durie 1987), the scale of structural ranks will only include two positions: core > periphery. 2.2. Comitative construction A (genuine) comitative construction (=ComC) is a morphosyntactic construction used to express a non-obligatory participant set in a given situation S, such that: (i) the predicate denoting S is not repeated more than once; (ii) the individual participants making up the participant set are expressed separately; (iii) the expressions denoting these participants differ in structural rank.

The constraint (i) excludes complex constructions of the sort John came and Mary came, too. On the other hand, it does not exclude cases where comitative is construed with a supplementary predication headed by a special ‘comitative’ predicate (e.g. in serializing languages). Because of these

Comitative as a cross-linguistically valid category

225

constructions, (i) cannot be reformulated in terms of a ‘one predicate’ limitation. Next, (ii) opposes comitative and coordinating constructions, both falling under Stassen’s (2000) NP-conjunction, to other pluralization strategies. Finally, (iii) is what distinguishes comitative from coordination. Comitative markers are often highly polysemous, and the patterns of this polysemy are a fertile ground for study. Given a marker M which can be used in a genuine ComC in a given language, all the constructions using M will be referred to as quasi-comitative constructions (quasi-ComCs).2 A note on the prerequisites to the above definition is in order. It is mentioned that the participant set expressed with the help of the construction in question should not be obligatory for the given situation. This is meant to exclude the (quasi-)ComCs used to introduce the participants of polyadic predicates, in particular of reciprocals. This rather controversial issue is discussed in 3.3. In general, the proposed definition is rather restrictive, which nevertheless does not exclude the full army of quasi-ComCs from further study. On the contrary, it permits us to highlight the core of the comitative-related phenomena and thus shed more light upon the intricate relationships between comitative and the neighbouring domains. In a comitative construction, we will refer to the noun phrase with the higher structural rank as core NP, and to the one with the lower structural rank as comitative NP. The participants denoted by these noun phrases will be referred to as core participant and comitative participant, respectively. We will also recur to the less specific term peripheral NP (instead of comitative NP) in the discussion of quasi-comitative constructions in 3.3. 2.3. Tests for comitative status Given the definition proposed above, a number of tests can be suggested which could help to qualify a given construction as a genuine ComC. Let us just mention two simple tests that will serve us later in section 3. First, a ComC should be replaceable with another pluralizing device – a plain or coordinate plural noun phrase – without change in the propositional meaning;3 see (i) below. Second, as the same type of participation is ascribed to each member of the participant set, we should be able to rephrase the ComC via a conjunction of propositions (ii); see the discussion of a similar test, along with others, in Schlesinger (1979). (i) ‘Ann came with Harry.’ = ‘TheyAnn, Harry came.’ = ‘Ann and Harry came.’ (ii) ‘Ann came with Harry.’ = ‘Ann came.’ & ‘Harry came.’

226

Alexandre Arkhipov

3. The neighbourhood In this section, we will discuss seven selected categories of structures most closely related to the comitative. The evidence for this closeness comes, among other things, from a fair number of syncretistic markers in various languages. Each of the categories in question may share its markers with the comitative (i.e. be expressed by a quasi-comitative construction). Sometimes a single marker covers a vast zone including most of these functions as well as a number of others, not mentioned here; cf. the analysis of Riau Indonesian sama by Gil (2004) who notably argues that it should be assigned a single general function rather than a manifold of specific ones. Importantly, these functions belong to quite different domains involving nominal and verbal number, participant relations, secondary predication, coordination and subordination, etc. Due to this diversity, a good deal of confusion has accumulated in studies of these categories, both in terminology and in distinguishing between or unifying particular categories. 3.1. Comitative vs. accompaniment One of the most common assumptions about comitative, which is restated now and then, orally or in writing, is that “comitative is the case expressing accompaniment”.4 This definition will be shown to be inadequate for several reasons. One of them is the possibly misleading use of the term accompaniment; see below for Stolz’s use of this term. Unless explicitly defined, the term ‘accompaniment’ is naturally interpreted in its common language sense, which suggests a certain asymmetry: a person or entity which accompanies someone or something is usually understood as being clearly secondary with respect to the situation in which both are involved. Consider such dictionary examples as Disease is often an accompaniment of famine (Hornby and Cowie 1982: 6) (from which it does not follow that Famine is often an accompaniment of disease). However, this is not always applicable to comitative: one and the same situation can often be described using comitative marking with either of the two participants.5 For instance, the choice between (1a) and (1b) can be determined by the focus of the preceding question, not by the difference in, say, the degree of conrol or involvement in the situation or other participant properties. In (2), the choice of the subject of the relative clause (he)

Comitative as a cross-linguistically valid category

227

and the comitative marking on his mother is conditioned by Jin Han’s being the topic of the preceding text. English (Indo-European, Germanic) (1) a. [— Where’s Jim?] — He’s taking a walk with Sarah. b. [— Where’s Sarah?] — She’s taking a walk with Jim. (2)

The hat in the title refers to Jin-Han’s first memory, in which he was taking a walk with his mother and his hat blew off. (Internet)

Thus the inherent asymmetry of the comitative construction does not reflect specifically an asymmetry of participant relations, being available to other kinds of motivation. Besides the motivation of information structure, as in (1–2), complex syntactic constructions, e.g. the ones with samesubject restrictions, also favor the use of comitative. All this makes definitions based on the difference in degrees of involvement problematic; cf. Anderson (1994: 453): “The ‘comitative’ usually marks the animate (typically human) which is conceived of as accompanying the participation of some more centrally involved participant in a predication.” Another observation concerns the lexemes most closely associated with the term ‘accompaniment’. Compare the following English sentences: English (Indo-European, Germanic; Walmsley 1971: 505) (3) a. Jim accompanied Harry to the station. b. Harry went to the station with Jim. Taken literally, the label of ‘accompaniment’ (and, consequently, that of ‘comitative’) would be applied to both cases. But, curiously enough, the English comitative does not occur with the verb accompany: (3a) does not contain an instance of comitative, while (3b) does. Perhaps the most important objection behind these two points is that comitative does not “express” accompaniment (or whatever else) in the same sense as the cases considered to be semantic ones, e.g dative or instrumental, express a participant relation – ComCs can be applied to pluralize different participant roles. So for Stolz, Stroh and Urdze (2006), the relation of Accompaniment that they posit as the basis of the comitative is actually an interparticipant relation, i.e. it holds between two participants, not between the predicate and the participant. However, for them “syntax is not decisive” for the definition (Stolz, Stroh and Urdze 2005: 35), and their

228

Alexandre Arkhipov

thematic role of Comitative also covers depictive and attributive (adnominal) quasi-ComCs (see 3.7). 3.2. Comitative vs. sociative (togetherness) In Russian-speaking linguistic tradition comitative is associated with the notion of togetherness (sovmestnost’).6 Consider the succint definition given by Igor Mel’čuk in his Cours de morphologie générale (1998: 338): “‘comitative’: ‘together with’.” However, togetherness per se can also be shown to be orthogonal to the comitative. Although there is a certain affinity between these categories – for instance, there is evidence in a number of languages that comitative markers can take their source in expressions meaning ‘together (with)’ (e. g. Russian vmeste s, common Turkic *birlen) – it is by no means absolute. Following Zaliznjak and Šmeljov (1999), I consider togetherness to be a gradual characteristic of a situation which possesses a participant set, i. e. two or more individual participants with the same semantic role (e. g. The children/They/Jim and Sarah were running in the garden). Plurality (the existence of a participant set) is not sufficient for the situation to qualify as sociative (joint): it must also be perceived as a whole with respect to those individual participants – or, taking up the wording of E. König (1995: 66), exhibit a certain perceptual unity. “A joint action differs from identical actions of two people (or more) – even if they happen simultaneously and at the same place – in that the individual actions are conceived of as parts of one common action” (Zaliznjak and Šmeljov 1999: 450): e.g. when several workers are building a house, each of them is also building a house, no matter how different their actions are. Also, in case of human participants, participation in a [+sociative] situation results in some shared experience. The opposite of sociative predicates are distributives, which also require one of the participant slots to be filled by a participant set. Distributives emphasize the multiplicity of individual manifestations of a situation, and hence usually feature many individual participants rather than just two; cf. The kids scattered across the garden. – ??Jim and Sarah scattered across the garden. There are languages in which two comitative markers differ by the [±sociative] feature. Cole (1985: 80) reports the existence of two comitative markers in Imbabura Quechua, -wan meaning “together with but not forming a single entity” (4a) and -ndi “together with and forming a single entity” (4b). This reflects quite precisely the notion of sociative as under-

Comitative as a cross-linguistically valid category

229

stood here. Note that in the same language the sociative (“joint action”) can be marked on the verb (5). Compare also the reinforcement of existing comitative markers by adverbs meaning ‘together’, e.g. in English together with, German zusammen mit, Russian vmeste s, acting as a secondary comitative construction. Imbabura Quechua (Quechuan) (4) a. ñuka-ka wambra-wan puri-ni 1SG-TOP child-COM go-1SG7 ‘I walk with the child.’ {We are walking separately and only happened to be together} (Cole 1985: 80) [–sociative] b. ñuka-ka wambra-ndi puri-ni 1SG-TOP child-COM go-1SG ‘I walk with the child.’ {We are engaged together in going to a joint destination} (Cole 1985: 81) [+sociative] (5)

a. shamu-rka-nchi come-PST-1PL ‘We came (perhaps separately).’ (Cole 1985: 185) [–sociative] b. shamu-naju-rka-nchi come-SOC-PST-1PL ‘We came together.’ (Cole 1985: 185) [+sociative]

On the other hand, it has been noted that comitative constructions typically impose a unity of time and place on the subevents ascribed to the two participants, which is not the case with coordination.8 Thus we can suggest a three-level scale of subevents integration, in which all the three levels of interpretation can be accessible for a coordinate construction, but only two options are available to ComCs: Table 1. Interpretations available for comitative and NP coordination Subevents independent in space and time [– sociative] coordination only

Subevents in the same space and time frame [– sociative] coordination and comitative

Subevents in the same space and time frame [+ sociative] coordination and comitative

230

Alexandre Arkhipov

3.3. Comitative vs. polyadic participants9 Traditionally, one of the most common contexts exemplifying comitative has been the marking of polyadic participant roles – that is, roles necessarily requiring a participant set (see 2.1), i.e. shared by minimally two individual participants. Predicates which assign polyadic participant roles are in turn called polyadic predicates; this includes reciprocals, sociatives, distributives, assistives and several other classes of predicates. The participant set licensed by polyadic predicates can generally be expressed, as elsewhere, in several ways. Notably, simple constructions where it is denoted by a plain or coordinate plural-reference noun phrase (They met) are contrasted to discontinuous constructions where a comitative marking is used on one of the noun phrases (He met with Julia). I will try to show why discontinuous polyadic constructions should not be treated as genuine comitative constructions and especially why they cannot be used as a diagnostic context for ComCs. The fact that they do not satisfy the proposition conjunction test (‘He met with Julia.’ ≠ ‘*He met.’ & ‘*Julia met.’) amounts to the condition of non-obligatoriness of the participant set that we somewhat arbitrarily imposed in our definition in 2.2; now let us explain the underlying motivation. To begin with, being a polyadic predicate does not necessarily mean being always polyadic (i.e. in all subsenses/uses). Given that some uses of an otherwise polyadic predicate may not require an obligatory participant set, they can also have other options for marking their participants. The choice of the marker will often depend on the semantics of the predicate and on the interparticipant relations in the situation described. The problem is then to identify the comitative marker among all those markers which can possibly occur with the peripheral NP in a given language; cf. for instance English Sigurd fought with Harald vs. Sigurd fought against Harald; speak to someone vs. speak with someone. Consider also French divorcer d’avec quelqu’un ‘divorce from someone’, in which case the general comitative marker avec combines with the preposition de ‘from’ indicating separation. Conversely, as different polyadic predicates in a given language allow for a variety of options in marking their participants, comitative is not always possible with all of them. Thus in standard Russian the verb poženit’sja ‘marry’ only occurs with a plural subject, while poznakomit’sja ‘get acquainted (with smb.)’ allows both a simple and a discontinuous construction:

Comitative as a cross-linguistically valid category

231

Russian (Indo-European, Slavic) (6) a. Oni poženi-li-s’/ poznakomi-li-s’ nedavno. they marry-PST:PL-REC get_acquainted-PST:PL-REC recently ‘They got married/acquainted recently.’ + b. On *poženi-l-sja/ poznakomi-l-sja he marry-PST.SG.M-REC get_acquainted-PST.SG.M-REC s nej nedavno with she.INS recently ‘He *married/+met her recently.’ This can be true not only for particular predicates in a language, but also for a language as a whole. According to Maslova and Nedjalkov (2008), 16 of 175 languages in their sample do not allow any non-iconic (i.e. without the repetition of the predicate) expressions of reciprocal, which excludes comitative altogether.10 In addition, in at least one language, Pama-Nyungan Ngiyambaa (Nedjalkov 2007: 27, 105), the reciprocal can be expressed non-iconically only if the participant set is referred to by a single noun phrase with plural reference. Furthermore, the recent studies of reciprocals agree in that “a language has the discontinuous reciprocal construction only if it has also the simple one, but not vice versa. …[T]he comitative strategy could only give rise to a discontinuous reciprocal construction, which seems to be impossible in absence of a simple construction” (Maslova 2007: 337). To sum up, in a considerable number of languages the polyadic participant cannot be expressed by a ComC. Furthermore, in many instances it is difficult to tell for sure if a given use of a predicate is truly polyadic or not. Discontinuous (i.e. comitativelike) constructions are often ambiguous between a polyadic (symmetrical) and a non-symmetrical reading (recall the above examples speak with/to smb.), even if the predicate bears formal polyadic, e.g. reciprocal, marking: Russian (Indo-European, Slavic) (7) a. Ja s nim podra-l-sja (*a on so mnoj net). I with him fight-PST.M-REC but he with me no ‘I fought with him (*but he did not [fight] with me).’ b. Ja s nim pozdorova-l-sja, a on so mnoj net. I with him say_hello-PST.M-REC but he with me no ‘I greeted him but he did not (greet) me.’ (Lit.: ‘I greeted-eachother with him but he did not [greet-each-other] with me.’)

232

Alexandre Arkhipov

Finally, marked polyadic predicates tend to develop such uses (called “non-reversible discontinuous constructions” by Ogloblin and Nedjalkov 2007: 1470) where, due to a certain expansion of the predicate’s meaning, a mismatch of semantic classes of the two participants is allowed, making both the reversal of the marking and the simple construction impossible. Fulfulde Maasina (Niger-Congo, Northern Atlantic; Koval’ and Gnalibouli 1997: 165, glossing mine) e deƴƴere mum woownde (8) Agay jooɗ-d-ii Agay sit-SOC-PF with silence his usual ‘Agay was sitting in (lit.: together with) his usual silence.’ Indonesian (Austronesian, Sundic; Ogloblin and Nedjalkov 2007: 1470) (9) a. Kami ber-nadap-an dengan kesulitan we REC-face-REC with difficulties ‘We are facing difficulties.’ b. * Kami dan kesulitan ber-nadap-an we and difficulties REC-face-REC ‘We and difficulties are facing each other.’ It is then not so surprising to encounter such evolutions of comitative markers as reported by Chappell (2006) for several groups of Sinitic dialects of Southeastern China, where various comitative markers gave rise to object markers. Thus, the object marker in Min dialects, kā, amounts to a comitative marker evolved from a (polyadic) verb ‘to gather, to share’ (Chappell 2006: 453). Marking of polyadic participant roles appears to be a plausible link in the chain: a polyadic verb becomes involved in a serial construction where it marks a polyadic participant, which can in turn lead to comitative uses and then, perhaps independently, to object marking.11 In sum, one cannot be sure that any arbitrary discontinuous reciprocal or sociative construction is indeed an instance of a ComC. In all of the constructions discussed in (6–9) mismatches are observed between polyadic semantics and/or polyadic marking and comitative marking. Evidently, mere translations of such expressions into English or any other intermediary language are not indicative of the nature of the particular construction used.

Comitative as a cross-linguistically valid category

233

3.4. Comitative vs. comitative verbs As pointed out by Ekkehard König (p.c.), the decision to exclude discontinuous reciprocal constructions from the class of genuine ComCs also implies the exclusion of the (pure) ‘comitative verbs’, which in his opinion is objectionable. I propose, however, that this is a desirable restriction. The label ‘(pure) comitative verb’ arguably conceals a considerable crosslinguistic variation in the origin, semantics and syntax of such predicates, which in their independent uses often have little in common with the comitative. Let us briefly illustrate this point. Comitative verbs may be understood either (i) as the verbs acting as comitative markers in serial constructions, or (ii) as the verbs rendered in English as ‘be with’ (the sense intended by E. König). Close to the latter are (iii) the occurences of comitative adpositions/case forms with a zero verb or copula – that is, complex expressions whose parts are rendered as ‘be’ plus ‘with’. For a comitative verb (in its use as a main predicate) to count as an instance of ComC we should expect that it assigns the same semantic role to both of its arguments, and that at least the plural NP test is satisfied. Now, comitative markers in serial ComCs have their origin in (usually transitive) verbs with different semantics, including ‘follow’, ‘accompany’, ‘take’, ‘have’, ‘hold’ (Stassen 2000: 20), ‘be included among’, ‘be with’, ‘be together with’, ‘be in the company of’ (Lord 1993: 57, 62). For most of these, from ‘folow’ to ‘be included among’, the ‘same role’ condition is not met. The ‘same role’ status could be argued for the arguments of ‘be with’ and variations thereof, but even with these verbs the plural NP test does not seem to work well since the verb is bivalent: ‘I am with you.’ ≠ ‘*We are.’ An acceptable paraphrase like ‘We are together’ would in fact clarify the semantics hidden behind the apparent reduction of the form. A cursory glance reveals some more popular semantic flavours of the ‘be with’ verbs (see Arxipov 2009 for more details). With animate referents, such verbs can be interpreted as indicating togetherness, like in (10a) from Tetun. The same verb in Tetun can also indicate possession of various kinds, including possession of animate nouns (10b). It has also a frozen intransitive third person existential form (van Klinken 2000: 361–362). At least the first two meanings appear to be frequent cross-linguistically. However, many other nuances are also possible;12 thus, in Archi (East Caucasian) ‘be’ plus an artifact noun in comitative case is most naturally interpreted as being involved in an activity of repairing a device (e.g. a gun or a car) (Michael Daniel, p.c.).

234

Alexandre Arkhipov

Tetun (Austronesian, Central Malayo-Polynesian) (10) a. ha’u k-oo emi 1SG 1SG-be_with 2PL ‘I am with you. (So don’t be afraid.)’ (van Klinken 2000: 358) b. ha’u k-oo ina kaa lale? 1SG 1SG-have mother or not ‘Do I have a mother or not?’ (van Klinken 2000: 361) 3.5. Comitative vs. NP coordination The relations between comitative and coordination have been discussed at length in such typological work as (Stassen 2000; Haspelmath 2004b); see (Haspelmath 2004a: 14–24) for a brief overview. Let us just summarize the most important findings to date. Comitative and NP coordination are seen as two major options of what has been called NP-conjunction by Stassen (2000). Both are used to represent two participants of a situation as two distinct noun phrases, but coordinating constructions are largely symmetric while comitative constructions are clearly non-symmetric.13 Comitative constructions are thought to be universal, while coordination has been shown for many languages to be a relatively recent innovation; in some languages (called WITH-languages by Stassen) there is no other formal means of NP-conjunction than the ComC. However, WITH-languages tend to borrow or develop their own coordinating constructions and thus become AND-languages (this phenomenon is called AND-drift). Many of them recur to an intermediate solution (mixed WITH-type): the normal comitative marker is also used in constructions exhibiting some (or all) features of coordination proper (e.g. Russian Saša s Mašej ‘Sasha and Masha’). These are called here coordinating comitative constructions.14 Several types of constructions may coexist in a language; sometimes more than one coordinating ComC can be found, with differing degrees of similarity to coordination proper. Thus, the languages differ according to how many types of constructions they have; this has a direct impact on the translatability of constructions between languages. Arxipov (2005) therefore proposes a quantitative cardinality measure to complement Stassen’s binary opposition of AND- and WITH-types. The term should be understood in its set-theoretical sense as the number of distinct constructions in the inventory of a language. For instance, a pure WITH-language in Stassen’s terms (e.g. Jacaltec) will be assigned cardinality 1, disposing of only one

Comitative as a cross-linguistically valid category

235

construction to express NP-conjunction. AND-languages, such as English, distinguish between comitative and coordination (cardinality 2). A mixed WITH-language, such as Hausa (Abdoulaye 2004), also has cardinality 2; from the translatability perspective, this means that each of the Hausa constructions has a counterpart in English, although the markers used are different. On the other hand, some AND-languages, e.g. Russian and Meadow Mari (Uralic), possess a third construction – coordinating ComC – which has no direct translational equivalent in English, so they should be assigned cardinality 3. Finally, in the Papuan language Kobon (Davies 1981: 73–75) two coordinating ComCs with different sets of coordination-like features exist in addition to a genuine ComC and a coordinating construction; Kobon thus has cardinality 4. 3.6. Comitative vs. inclusory constructions Inclusory constructions are another means of plural reference. An inclusory construction is formed by two noun phrases, one of which itself has a plural reference (most often, it is a non-singular personal pronoun). The referent of the second noun phrase is included into the set of referents of the first, hence the name inclusory. A typical example is found in Russian: Russian (Indo-European, Slavic) rybu u pričala. (11) My s bratom lovi-l-i [we with brother:INS] catch-PST-PL fish:ACC at jetty:GEN ‘My brother and I (lit.: We with [my] brother) were fishing at the jetty.’ Here the two noun phrases form a single constituent triggering plural agreement on the verb. This type can be called phrasal inclusory construction, as opposed to split inclusory constructions in which the total set of referents is only signaled by a pronominal clitic or affix on the verb, thus not forming a constituent with the ‘included’ noun phrase: Spanish (Indo-European, Romance; Camacho 2000: 366) (12) Con Juan vamos al cine. with Juan go:1PL to:the movies ‘Juan and I are going to the movies.’

236

Alexandre Arkhipov

Bunuba (Australian; Singer 2001: 56) (13) ngay wiyi-guda wardburrayntha yes woman-COM go:PST:3:DU ‘(Yes,) he and the woman went.’ Inclusory constructions are not structurally uniform across languages; they may be realized as dedicated construction types or parallel some other constructions. For example, the marker that joins the two noun phrases in (11) is the standard Russian comitative marker; the fact that the two cannot be detached from each other indicates that the underlying structure is that of a coordinating comitative (see 3.5). In Hausa, inclusory constructions can be based either on a coordinating comitative, as in Russian, or on a genuine comitative: Hausa (Afro-Asiatic, West Chadic; Abdoulaye 2004: 186) (14) mun jee kàasuwaa dà abdù 1PL.PFV go market with Abdu ‘We went to the market, Abdu and me.’ Juxtaposition (Mparntwe Arrernte) and coordinate marking (Basque) are also attested in inclusory constructions; see also (Singer 2001) for examples of dedicated inclusory marking, and (Haspelmath 2007: 33–35) for a discussion. Mparntwe Arrernte (Australian, Pama-Nyungan; Ohori 2004: 48) (15) kake ilerne elder_brother 1DU ‘elder brother and I (=we-two)’ Basque (isolate, Europe; examples adapted from Internet) (16) a. Ama eta bi-ak bizi ziren bertan. mother and two-PL live AUX.PST.3PL there ‘She lived there with (her) mother.’ (lit. ‘they two and mother’) b. Guraso-ak eta laur-ok etxe-tik atera ginen. parents-PL and four-PROX:PL house-ABL go_out AUX.PST:1PL ‘(My) parents and us (four people in total; lit. ‘we four and parents’) came out of the house.’ Incidentally, both Mparntwe Arrernte and Basque exhibit a word order which is clearly a minority option, seemingly restricted to languages with

Comitative as a cross-linguistically valid category

237

head-final properties: the ‘included’ noun phrase is followed by the conjunction followed by the inclusory noun phrase (head). (The Basque construction is also remarkable because the inclusory noun phrase is not a pronoun but a substantivized numeral corresponding to the overall number of individual referents.)15 3.7. Comitative vs. depictives Depictives, as described by Himmelmann and Schultze-Berndt (2005) and exemplified in (17), are a kind of adjuncts which convey a predication with respect to a participant of the main predication (hence the full denomination “depictive secondary predicates”). They are thus participant-oriented adjuncts, as opposed to adverbials, which are considered event-oriented adjuncts. A key feature that distinguishes depictives from other secondary predicates, e.g. resultatives, is that the time frame of the depictive is included into (or overlaps with) the time frame of the main predicate. Most widely discussed depictives express a physical or emotional state of the controller participant. So in (17a), angry denotes the emotional state in which John was while leaving the party, and fresh in (17b) expresses the physical state of carrots at the moment when they were bought by George. English (Indo-European, Germanic) (17) a. John left the party angry. (HSB 2005: 8)16 b. George bought the carrots fresh. (HSB 2005: 1) Depictives are better known to be expressed by adjectives. However, as noted in HSB (2005: 67), there is a considerable number of other types of expressions, including noun phrases (especially inflected ones), adpositional phrases, ideophones and subordinate clauses, which can manifest all the same features as do the more traditional depictives. Variation in the semantic nature of relations expressed by depictives is also considerable. The semantic range of participant-oriented adjuncts surveyed in HSB (2005) includes expressions of mental or emotional condition, posture and configuration, manner, comparison, function, role and life stage, time and location, among others. The functions of interest here will be those of “concomitance and association” (HSB 2005: 36–38). HSB remark that, although most often regarded as adverbials, expressions of concomitance like with his accomplice in (18) in fact exhibit participant orientation. This orientation is not explicitly encoded in English,

238

Alexandre Arkhipov

but is marked overtly in, for example, a number of Australian languages where a comitative NP bears the same case marking as the core NP, as in (19a–b). English (Indo-European, Germanic; HSB 2005: 36–37) (18) They brought in the prisoner with his accomplice. Djabugay (Australian, Pama-Nyungan) (19) a. ngawu mulam-djada buga-na minya I.ERG juice-COM eat-FUT fish ‘I will eat fish with gravy.’ (Patz 1991: 271) b. ngawu magumagu-djada-nggu buga-na minya I.ERG sister-COM-ERG eat-FUT fish ‘I and my sister will eat fish / I will eat fish together with my sister.’ (Patz 1991: 293) Agreeing that comitative NPs in genuine comitative constructions can be treated as participant-oriented adjuncts, I argue that they are distinct from depictives proper. The difference between them can be made apparent with a paraphrase test applicable for depictives (HSB 2005: 11). Example (17b) can be paraphrased as (20), where the depictive is promoted to primary predicate: English (Indo-European, Germanic) (20) The carrots were fresh (when George bought them). (HSB 2005: 11) But the meaning of a comitative construction is stronger than such a paraphrase; e.g. (19b) does not only entail the mere presence of my sister when I eat fish (I will be with my sister), but also that my sister will eat fish, too.17 A genuine ComC is thus different from depictives in that it does not by itself denote a situation that holds within the time frame of the main predicate; rather, it shows that the same situation as described by the main predicate for the core participant also holds for the one introduced by the comitative NP. This explains the fact that “expressions where the concomitant is more closely associated with, and less independent of, the controller” (in particular, it can be “any inanimate entity carried in close physical contact”) are in a number of languages encoded distinctly from ComCs (HSB 2005: 37), although both types of expressions receive the same marking in many other languages; cf. English He came with a friend vs. He came with a bag. The

Comitative as a cross-linguistically valid category

239

following Djabugay examples show a marker -:rr/-nydji/-i,18 in contrast with -djada in (19). As Patz (1991: 294) explicitly warns, (21c) “only allows the interpretation of the dog being carried. If the dog walks along as company, comitative -djada has to be used”: Djabugay (Australian, Pama-Nyungan) (21) a. warrdjan gulu walngga-l bama-:rr, bibunba(y)-nydji raft this float-RES people-PROPR children-PROPR ‘This raft floats with adults and children (aboard).’ (Patz 1991: 270) b. gurrungga burra-ng djama-:rr, djamu-: wanirri-l kookaburra fly-PRS snake-PROPR beak-LOC carry-PRS ‘The kookaburra flies with a snake, (it) carries (it) in (its) beak.’ (Patz 1991: 293) c. ngawu gali-ng gurra:-rr 1SG go-PRS dog-PROPR ‘I go carrying a dog.’ (Patz 1991: 294) The proprietive-marked participant does not have the same semantic role as the core participant – e.g. snake in (21b) is not an agent like kookaburra – and hence no participant set is formed; thus the construction does not qualify for genuine ComC. At the same time, such expressions seem to be indeed depictive, as they convey a spatial configuration and/or a situation of causative locomotion limited by the time frame of the main predicate. The proprietive marker “also acts as a highly productive suffix in the formation of adjective and noun stems” (Patz 1991: 270); proprietivemarked nouns thus can also function as adjectives (22) and as main predicates (23). Djabugay (Australian, Pama-Nyungan) (22) a. gambarr-i b. gidjarr-i c. djabuga-nydji cloud-PROPR paint-PROPR Djabugay-PROPR ‘cloudy’ ‘painted’ ‘Djabugay-speaking people’ (Patz 1991: 270) (23) gulu murra-nydji 3SG fever-PROPR ‘He is sick (has a fever).’ (Patz 1991: 294)

240

Alexandre Arkhipov

Let us add that in languages which use identical marking for (genuine) comitative and comitative-like depictives, the same marker can also cover the adnominal and predicative functions: Khalkha Mongolian (Altaic, Mongolic) (24) a. cas-taj övöl snow-COM winter ‘snowy winter’ (Cèdèndamba 1978: 67) b. ene xüüxed xar nüd-tej this child black eye-COM ‘This child has black eyes’ (Cèdèndamba 1978: 18)

4. Conclusion Comitative has been analyzed as a type of pluralizing construction, in contradistinction with its received inclusion into semantic cases, i.e. cases expressing a specific participant relation (semantic role). Comitative appears when two participants share the same role. However, there is more to the comitative than the “same role” relation. Comitative is a specific type of grammatical construction: it is an asymmetrical construction employed to introduce a non-obligatory participant with the same role as one of the core participants. The asymmetry of the construction does not necessarily reflect an underlying asymmetry of participation. The asymmetry is motivated rather by the pragmatic rankings of participants, information structure and so on. One and the same situation can often be described with either of the two participants bearing the comitative marking. Comitative being the focal point of the intertwinement of several domains – plurality, coordination, participant relations, secondary predication, possession, togetherness, and more – it appears vital to draw distinctions between comitative and neighbouring categories. Hopefully, the proposed definition of the comitative construction will serve this purpose. Given the limited amount of space, a considerable number of issues were necessarily left out of this chapter’s scope. These include the relations of comitative to various types of possession, instruments and agents, sociative causation; the tendency of comitative to mark an underspecified participant; as well as most of the questions concerning the formal marking strategies in comitative constructions.

Comitative as a cross-linguistically valid category

241

Notes 1.

2.

3.

4.

5.

6.

7.

This chapter is based principally on the author’s dissertation ‘Typology of comitative constructions’ (Arxipov 2005), now published as Arxipov (2009). The main sample of the study included over 80 languages of diverse families: Indo-European, Uralic, Altaic, Dravidian, Yenissean, Sino-Tibetan, East Caucasian, Chukotko-Kamchatkan, Na-Dene, Tai-Kadai, Austronesian, AustroAsiatic, Afro-Asiatic, Nilo-Saharan, Niger-Congo, Trans-New Guinea, SepikRamu, Pama-Nyunga, Tiwian, Mayan, Uto-Aztecan, Quechuan, Arawakan, Tupian, Carib, isolates Basque and Yukaghir and several creoles. I would like to thank the supervisor of my thesis, Aleksandr E. Kibrik, for his everlasting support, both scientific and personal, and the reviewers of the thesis Vladimir A. Plungian, Anna A. Zalizniak, Nina R. Sumbatova, Anna S. Panina and Yury A. Lander for their insightful comments. I am also grateful to the many colleagues for fruitful discussions, in particular to Peter Arkadiev, Hilary Chappell, Michael Daniel, Nick Evans, Ekkehard König, Sergej Krylov, Christian Lehmann, Timur Maisak, Susanne Michaelis, Claire MoyseFaurie, Heiko Narrog, Barbara H. Partee, Françoise Rose, Andrey Shluinsky, Ruth Singer. And, of course, a huge thanks to Pattie Epps who took the lion’s share of the work on this volume, making it all possible. This text has significantly improved thanks to comments from Andreas Jäger, Michael Daniel and Pattie Epps. All errors and omissions are mine. The term ‘quasi-comitative’ should not be taken as implying the primacy of the (genuine) comitative use for any given construction. It is only a convenient label for a study focussed on comitatives. In contrast, the information structure is likely to change, since comitative NPs show a tendency to be in focus; cf. a similar observation for depictives in (Himmelmann and Schultze-Berndt 2005: 15–19). See for example Anderson (1994: 453), Blake (2001: 154) and the SIL Glossary of linguistic terms by E. Loos et al. (http://www.sil.org/linguistics/ GlossaryOfLinguisticTerms/). But see Stolz, Stroh and Urdze (2006: 27–28; 59; 314–315; 349–353) for discussion of languages which explicitly distinguish between ‘accompaneeorientation’ and ‘companion-orientation’. The closest equivalent of ‘togetherness’ seems to be ‘sociative’, although this term has been employed for various categories including comitative as well; I will treat them here as synonyms. The term ‘joint action’ is also found, e.g. in descriptions of verbal sociative markers. The following abbreviations are used in glosses: 1/2/3 first/second/third person, ACC accusative, AUX auxiliary, COM comitative, DU dual, ERG ergative, FUT future, GEN genitive, INS instrumental, LOC locative, M masculine, PF perfect, PFV perfective, PL plural, PROPR proprietive, PROX proximal, PRS present, PST past, REC reciprocal, RES resultative, SG singular, SOC sociative, TOP topic.

242 8.

9. 10.

11.

12.

13.

14. 15.

16. 17.

Alexandre Arkhipov See for instance Abdoulaye (2004: 182–184) for evidence from Hausa; see also Shibatani and Pardeshi (2002) for the importance of the spatiotemporal overlap in the domain of causative, notably in the sociative causation. Much of the terminology in this section is borrowed from Maslova (1999; 2007) and Nedjalkov (2007). Although Maslova and Nedjalkov warn that this number may be overestimated due to lack of detailed data on some languages, the number of the languages lacking non-iconic reciprocal constructions in the ‘core’ WALS sample of 100 languages is even higher (ca. 13%) since the authors were deliberately looking for most varied patterns related to non-iconic reciprocals. As suggested by examples from well-known European languages, ‘being with’ can have many facets, for instance being together in the same place and time, or being of the same opinion or party, or even being intimate with someone. The object marking function could arise directly from participant marking in in non-symmetric uses of polyadic predicates (cf. speak to smb.). For Hakka object marker lau, originating from a verb ‘to mix together’ (Chappell 2006: 458), a ‘comitative’ function is suggested by Chappell (2006) and Lai (2003). However, as far as one can judge from their examples, it seems to be restricted to polyadic predicates, and thus is not a genuine comitative. Another frequent function of lau, supporting the proposed development, is marking the adressee with verbs of communication such as ‘talk to’, ‘introduce’, ‘explain’ (Lai 2003: 361–362). Opinion is divided, however, as to whether the semantic or the formal symmetry or both should be taken as the defining feature of coordination in general as opposed to subordination, since there exist a good number of rather intricate intermediate cases; see Haspelmath (2004a: 33–37) for a discussion. Haspelmath (2007: 29–33) uses the term comitative conjunction. The first person interpretation in (16b) is triggered by the proximal article -ok; cf. third person interpretation with the neutral plural article -ak (16a). This correlation is not absolute: the neutral form can also be used to denote a participant set including the speaker, and the proximal form can be used with reference not to the speaker him/herself but to someone closely related to him/her in some way. When used with an inanimate noun phrase, the proximal plural article rather indicates spatial proximity: liburu-ak ‘the books’ vs. liburu-ok ‘these books’. Haspelmath (2007: 34) reports to have found “only a single language in which the inclusory word is not a non-singular pronoun, but a nonsingular full noun”, viz. Margi, a Chadic language of Nigeria; apparently, Basque is another such language. Throughout this subsection, HSB (2005) will be used as shorthand for Himmelmann and Schultze-Berndt (2005). Quasi-comitative constructions which only entail the presence of the second participant are reported in Bagvalal (Kazenin 2001: fn. 1, 592).

Comitative as a cross-linguistically valid category

243

18. I gloss Djabugay -:rr/-nydji/-i as PROPR[ietive], although Patz gives it a general COM label (the same as -djada (19) and one other marker).

References Abdoulaye, Mahamane L. 2004 Comitative, coordinating, and inclusory constructions in Hausa. In Coordinating Constructions, Martin Haspelmath (ed.), 165–193. (Typological Studies in Language 58.) Amsterdam: Benjamins. Anderson, John M. 1994 Case. In The Encyclopedia of Language and Linguistics, R. E. Asher and J. M. Y. Simpson (eds.), Vol. 2, 447–453. Oxford: Pergamon Press. Arkhipov [Arxipov], Alexandre V. 2005 Tipologija komitativnyx konstrukcij [Typology of comitative constructions]. Ph.D. diss., Moscow State University. 2009 Tipologija komitativnyx konstrukcij. Moskva: Znak. Blake, Barry J. 2001 Case. 2nd ed. Cambridge: Cambridge University Press. Camacho, José 2000 Structural restrictions on comitative coordination. Linguistic Inquiry 31 (2): 366–375. Cèdèndamba, C. 1978 Sopostavitel’nyj sintaksis russkogo i mongol’skogo jazykov [A comparative syntax of Russian and Mongolian]. Ulan-Bator: Izd-vo Ministerstva narodnogo obrazovnija MNR. Chappell, Hilary 2006 From Eurocentrism to Sinocentrism: The case of disposal constructions in Sinitic languages. In Catching Language: The Standing Challenge of Grammar Writing, Felix K. Ameka, Alan Dench, and Nicholas Evans (eds.), 441–486. Berlin/New York: Mouton de Gruyter. Cole, Peter 1985 Imbabura Quechua. (Croom Helm Descriptive Grammars.) London: Croom Helm. Davies, John 1981 Kobon. (Lingua Descriptive Studies 3.) Amsterdam: North Holland. Durie, Mark 1987 Grammatical Relations in Acehnese. Studies in Language 11 (2): 365–400.

244

Alexandre Arkhipov

Gil, David 2004

Riau Indonesian sama: Explorations in macrofunctionality. In Coordinating Constructions, Martin Haspelmath (ed.), 371–424. (Typological Studies in Language 58.) Amsterdam: Benjamins. Haspelmath, Martin 2004a Coordinating constructions: An overview. In Coordinating Constructions, Martin Haspelmath (ed.), 3–39. (Typological Studies in Language 58.) Amsterdam: Benjamins. 2007 Coordination. In Language Typology and Syntactic Description. Second edition. Vol. 2: Complex Constructions, Timothy Shopen (ed.), 1–51. Cambridge: Cambridge University Press. Haspelmath, Martin (ed.) 2004b Coordinating Constructions. (Typological Studies in Language 58.) Amsterdam: Benjamins. Himmelmann, Nikolaus P., and Eva F. Schultze-Berndt 2005 Issues in the syntax and semantics of participant-oriented adjuncts: An introduction. In Secondary Predication and Adverbial Modification, Nikolaus P. Himmelmann and Eva F. Schultze-Berndt (eds.), 1– 67. Oxford: Oxford University Press. Hornby, A. S., with A. P. Cowie (eds.) 1982 Oxford Advanced Learner’s Dictionary of Current English. Special edition for the USSR. Vol. 1. A–L. Oxford: Oxford University Press; Moscow: Russian Language Publishers. Johnson, David E. 1977 On relational constraints on grammars. In Syntax and Semantics. Vol. 8. Grammatical Relations, Peter Cole and Jerrold Sadock (eds.), 151–178. New York: Academic Press. Kazenin, Konstantin I. 2001 Deepričastnye konstrukcii [Converb constructions]. In Bagvalinskij jazyk: Grammatika. Teksty. Slovari [Bagvalal: Grammar. Texts. Dictionaries], Aleksandr E. Kibrik (ed.), 554–594. Moskva: Nasledie. Kibrik, Aleksandr E. 1977 Opyt strukturnogo opisanija arčinskogo jazyka. T. 2. Taksonomičeskaja grammatika [An experimental structural description of Archi. Vol. 2. The taxonomic grammar]. Moskva: Izd-vo Moskovskogo universiteta. Koval’, Antonina I., and Boureima Amadu Gnalibouli 1997 Glagol fula v tipologičeskom osveščenii [The verb of Fula in typological perspective]. Moskva: Russkie slovari. König, Ekkehard 1995 The meaning of converb constructions. In Converbs in Crosslinguistic Perspective, Martin Haspelmath and Ekkehard König (eds.), 57–95. Berlin: Mouton de Gruyter.

Comitative as a cross-linguistically valid category

245

Lai, Huei-ling 2003 Hakka LAU constructions: A constructional approach. Language and Linguistics 4 (2): 353–378. Lehmann, Christian, and Yong-Min Shin 2005 The functional domain of concomitance: A typological study of instrumental and comitative relations. In Typological Studies in Participation, Christian Lehmann (ed.), 9–104. (Studia Typologica, 7). Berlin: Akademie. Lord, Carol 1993 Historical Change in Serial Verb Constructions. Amsterdam, Philadelphia: John Benjamins. Maslova, Elena 1999 Reciprocals and set construal. In Reciprocals: Forms and Functions, Zygmunt Frajzyngier and Traci S. Curl (eds.), 161–178. (Typological Studies in Language 41.) Amsterdam, Philadelphia: John Benjamins. 2007 Reciprocal and polyadic: Remarkable reciprocals in Bantu. In Reciprocal Constructions, Vladimir P. Nedjalkov (ed.), 335–352. (Typological Studies in Language 71.) Amsterdam, Philadelphia: John Benjamins. Maslova, Elena, and Nedjalkov, Vladimir P. 2008 Reciprocal constructions. In The World Atlas of Language Structures Online, Martin Haspelmath, Matthew S. Dryer, David Gil, and Bernard Comrie (eds.), Chapter 106. Munich: Max Planck Digital Library. Available at http://wals.info/feature/106. Mel’čuk, Igor A. 1998 Kurs obščej morfologii. T. 2 [Course in general morphology. Vol. 2]. Moskva: Jazyki russkoj kul’tury; Vena: Wiener Slawistischer Almanach. Nedjalkov, Vladimir P. 2007 Overview of the research: Definitions of terms, framework, and related issues. In Reciprocal Constructions, Vladimir P. Nedjalkov (ed.), 3–114. (Typological Studies in Language 71.) Amsterdam, Philadelphia: John Benjamins. Ogloblin, Alexandr K., and Vladimir P. Nedjalkov 2007 Reciprocal constructions in Indonesian. In Reciprocal Constructions, Vladimir P. Nedjalkov (ed.), 1437–1476. (Typological Studies in Language 71.) Amsterdam, Philadelphia: John Benjamins. Ohori, Toshio 2004 Coordination in Mentalese. In Coordinating Constructions, Martin Haspelmath (ed.), 41–66. (Typological Studies in Language 58.) Amsterdam, Philadelphia: John Benjamins. Patz, Elizabeth 1991 Djabugay. In The Handbook of Australian Languages, Vol. 4, Barry J. Blake (ed.), 244–347. Oxford: Oxford University Press.

246

Alexandre Arkhipov

Schlesinger, Izchak M. 1979 Cognitive structures and semantic deep structures: the case of the instrumental. Journal of Linguistics 15 (2): 330–395. Shibatani, Masayoshi, and Prashant Pardeshi 2002 The causative continuum. In The Grammar of Causation and Interpersonal Manipulation, Masayoshi Shibatani (ed.), 85–126. (Typological Studies in Language 48.) Amsterdam, Philadelphia: John Benjamins. Singer, Ruth 2001 The inclusory construction in Australian languages. Honours thesis. University of Melbourne. Stassen, Leon 2000 AND-languages and WITH-languages. Linguistic Typology 4 (1): 1–54. Stolz, Thomas, Cornelia Stroh, and Aina Urdze 2006 On Comitatives and Related Categories: A Typological Study with Special Focus on the Languages of Europe. (Empirical Approaches to Language Typology 33.) Berlin/New York: Mouton de Gruyter. van Klinken, Catharina 2000 From Verb to Coordinator in Tetun. Oceanic Linguistics 39 (2): 350–363. Walmsley, John B. 1971 The English comitative case and the concept of deep structure. Foundations of Language 7: 493–507. Zaliznjak, Anna A., and Aleksej D. Šmeljov 1999 O tom, čego nel’zja sdelat’ vmeste [On what cannot be done together]. In Tipologija i teorija jazyka: Ot opisanija k ob”jasneniju. K 60-letiju A. E. Kibrika [Typology and the theory of language: From description to explanation. For the 60th anniversary of A. E. Kibrik], Ekaterina V. Raxilina, Jakov G. Testelec (eds.), 450– 457. Moskva: Jazyki russkoj kul’tury.

Towards a typology of labile verbs: Lability vs. derivation Alexander Letuchiy

1. Introduction1 In this chapter, I analyze some cross-linguistic properties of labile verbs, focusing in particular on the relationship between labile verbs and derivational markers. In this discussion, I consider the traditional definition of a labile verb, adopted in Chikobava (1942), Kibrik et al. (1977) and many other typological and descriptive works: a labile verb is a verb which can be used transitively or intransitively without any formal change. In many other works, such as Haspelmath (1993a) and Ljutikova (2002), the definition is more restrictive: “in labile alternations, the same verb is used both in the inchoative and in the causative sense” (Haspelmath 1993a: 92); thus, a labile verb must have one monovalent (inchoative) and one bivalent (causative) use. However, I show here that neither definition is fully accurate: the first does not take into account the difference between so-called ‘agent-preserving’ and ‘patient-preserving’ lability, and the second definition does not take into account the ‘converse’ labile verbs, in which transitive and intransitive realizations require different participant roles (e.g. Bulgarian xaresvam ‘like / please’). I propose a new definition involving two parameters: transitivity change and change of semantic roles. Previous work on labile verbs may be encountered in descriptive research (e.g. Bokarev 1949, Gišev 1968, Rothemberg 1974), and in typological work; for example, Nedjalkov (1969), Haspelmath (1993a), Ljutikova (2002), and Dixon (1988) propose a definition of lability and analyze the typology of inchoative / causative oppositions. This discussion is based on Letučij (2006a), a full typological study of labile verbs, based on a sample of over 100 languages.2 The primary focus of this chapter is the relationship between labile verbs and derivational markers. A typological classification of labile verbs, as proposed in Section 2, indicates that labile verbs can be classified according to anticausative, reflexive, and other subtypes, which are generally isomorphic to the classification of valency derivations (anticausative, re-

248

Alexander Letuchiy

flexive, etc.); yet previous research offers no precise typological hypotheses about the relations among these mechanisms. Section 3 examines in detail one of these relationships, that between the anticausative subtype of lability and (anti)causative derivation, and shows that in some cases lability and (anti)causative markers are mutually exclusive, but that often labile verbs can take derivational markers in both uses. I propose a threefold classification of the semantic opposition between the uses of labile verbs and the marked (anti)causative, and conclude that the relations between lability and grammatical markers are distinct for different language types. This result is an illustration of how grammatical and lexical phenomena may interact in complicated ways within the language system. Finally, in Section 4, I propose a hypothesis concerning the functional motivation of lability in the language system. I address the following question: ‘Is lability a sort of unmarked valency derivation?’, and determine that it is not, which has implications for our understanding of the function of lability.

2. Labile verbs: a typological overview 2.1. Semantic correlations in labile pairs I begin by classifying labile verbs according to the type of semantic correlation that holds between their two uses, which I denote as a ‘labile pair’. In previous research, there seems to be no such classification, except perhaps in Ljutikova (2002). Anticausative lability:3 Xârâcùù (Austronesian, Oceanic; Moyse-Faurie 1995) bёchâ ‘untie (smth.)’, Lezgian (East Caucasian, Lezgic) q’in ‘kill / die’ etc.; Reflexive lability: Khwarshi (East Caucasian, Tsezic) esanho ‘wash (smth.)’, Alutor (Chukotko-Kamchatkan, Northern; Kibrik, Kodzasov and Murav’jova 2000) psesәwa ‘take one’s shoes off / take smb.’s shoes off’; Reciprocal lability: Estonian (Uralic, Finnic) suudlema ‘kiss (smb.)’, Arabic (Afro-Asiatic, Semitic) iltaqa: ‘meet (smb.)’; Passive lability: Bambara (Niger-Congo, Western Mande) sègin ‘return / be returned’, Kabyle (Afro-Asiatic, Berber) mDl ‘bury / be buried’; Converse lability: Bulgarian (Slavic) xaresvam ‘please / like’, Romanian (Romance; Căluianu 2000) place ‘please / like’, French (Romance) sentir ‘smell (transitive) / smell (intransitive)’.4 The types proposed here, based on Ljutikova’s (2002) and my own analyses, are very similar to the existing classifications of valency deriva-

Towards a typology of labile verbs

249

tions. For instance, the reflexive type of lability is parallel to the reflexive valency derivation: one of the uses of the labile verb is semantically a reflexive of the other. In the reciprocal type of lability, one of the uses denotes a non-reciprocal situation, while the other denotes a mutual action of two participants. The passive subtype includes labile pairs in which one of the uses is transitive (e.g. ‘bury’), and the other is intransitive with a patientive subject (e. g. ‘be buried’), just as is characteristic of passive derivatives in some languages of the world. The converse type is the sole type that is not analogous to any valency derivation; it is instead similar to pairs of verbs like ‘take’ / ‘give’, where both verbs denote the same situation with the same number of arguments, but the participants have different syntactic/semantic status (for instance, the recipient is the subject in the case of ‘take’ and the indirect object in the case of ‘give’). The anticausative subtype, the focus of this chapter, presupposes that the intransitive use of the verb differs from the transitive one by the absence of an agent (e.g., the cup broke (intransitive), cf. I broke a cup (transitive), where I is an agent): Olutec (Mixe-Zoque, Mixe; Zavala 2000: 78)5 seme tuk. (1) a. ∅=jik-pa B3(ABS)=become_dirty-INCOMPL very one ‘One gets very dirty.’ b. ʔi=jik-pe kay+an. A3(ERG)=make_dirty-INCOMPL food ‘He is making food dirty.’ The anticausative subtype is much more frequent in my sample than are all the other subtypes. In most languages that show large classes of labile verbs, independently of their genetic affiliation, the anticausative subtype outnumbers all the others, e.g. in English, French, Avar (East Caucasian), Warekena (Arawakan),6 etc. In contrast, passive lability is extremely rare throughout the world, except in Africa (Kabyle, Bambara; Chaker 1983, Vydrine and Coulibali 1994). 2.2. Syntactic classification of labile verbs This section considers the syntactic classification of labile verbs, corresponding to the full list of semantic subtypes considered above (anticausative, passive, converse, etc.). The syntactic classification I propose relies

250

Alexander Letuchiy

on two factors: (i) valency and (ii) syntactic transitivity. Here I use the term ‘valency’ to denote the number of syntactic arguments of the verb (for instance, in I gave John a pen the verb give is trivalent, because it has three syntactic arguments: the agent I, the recipient John and the patient a pen). The term ‘syntactic transitivity’ is used for the twofold opposition of intransitive vs. transitive verbs. This definition of valency and transitivity allows for intransitive verbs that are not monovalent, if neither of their arguments is a direct object. Moreover, both of the preceding definitions require that the two uses of any labile verb differ in transitivity. Nevertheless, it is easy to show that both uses can have the same valency, i.e. the same number of arguments. Some labile verbs can have two bivalent uses with the participant roles reversed: Bulgarian (Indo-European, Slavic) (2) a. Toj xaresva kniga-ta. he.NOM like/please.3SG.PRS book-DEF ‘He likes the book.’ (transitive) b. Kniga-ta mu xaresva. book-DEF he.DAT like/please.3SG.PRS ‘He likes the book’ (lit. ‘The book pleases him.’) (intransitive) I suggest that the following parameters are necessary to formulate the general definition of labile verbs for any semantic subtype: i) labile verbs have at least two distinct uses, one transitive and one intransitive; ii) the syntactically privileged argument (subject) in these different uses has different semantic roles. (Consider, for instance, 2a where the subject is an experiencer and 2b where the subject is a stimulus.)

Our definition of lability differs from that adopted in Haspelmath (1993a): it includes labile verbs that are necessarily bivalent in both uses. However, our definition is also different from the simplest one, adopted in Chikobava (1942) and Dixon (1994). These authors suppose that lability can be defined on the basis of syntactic transitivity change only. Thus, they regard cases like John drinks / John drinks tea as labile, in the same way as are The stick broke / I broke a stick. In my approach, in the second case the verb (break) is treated as labile proper, since the transitivity change is accompanied by the change of meaning and of the semantic role of the sub-

Towards a typology of labile verbs

251

ject. The first verb (drink) is regarded as ‘quasi-labile’ since the intransitive use differs from the transitive use only by its transitivity. As the patient in John drinks is semantically obligatory, both uses have two semantic arguments. This definition has one important advantage. It lets us consider certain intermediate cases that are labile by only one of the criteria. Verbs that satisfy only the first criterion are examples of agent-preserving lability7 of the type John eats (meat); we will not address these cases here. On the other hand, Russian učit’ ‘teach / learn’ satisfies only the second criterion (the Bulgarian verb uča ‘teach / learn’ behaves similarly): Russian (Indo-European, Slavic) (3) a. Ja uč-u anglijsk-ij. I learn-1SG.PRS English-ACC.SG ‘I learn English.’ b. Pap-a uč-it men’a daddy-NOM teach-3SG.PRS I.ACC ‘(My) father teaches me English.’

anglijsk-omu. English-DAT.SG

The subject of (3a) (learner) becomes the object of (3b). The semantic opposition between the uses is almost identical to that in inchoative / causative pairs like ‘die / kill’ (‘teach’ roughly means ‘make learn (sth.)’). However, in both uses the verb remains syntactically transitive. I consider this verb as ‘quasi-labile’ since it satisfies only one criterion of lability. Table 1. Semantic and syntactic classification of labile pairs Type

Ex.

Intransitive use

Transitive use

Anticausative

Xârâcùù bëchâ Khwarshi esanho Estonian suudlema Bulgarian xaresvam Bambara sègin

‘untie’ (monovalent) ‘wash’ (monovalent) ‘kiss’ (monovalent) ‘please’ (bivalent intransitive) ‘be returned’ (bivalent intransitive)

‘untie sth.’ (bivalent) ‘wash’ (bivalent) ‘kiss’ (bivalent) ‘like’ (bivalent transitive) ‘return’ (bivalent transitive)

Reflexive Reciprocal Converse Passive

252

Alexander Letuchiy

The ability of a particular language to express the non-causative and the causative situation with the same verb, as in (3), is an important typological parameter, and should be considered in combination with other parameters like the overall number of labile verbs and their subtypes (see 2.1). Note that such pairs, where both members are syntactically (in)transitive, can exist for other correlation types (reflexive, reciprocal and others), although I do not discuss these here. The semantic and syntactic classification of labile verbs is briefly summarized in Table 1. 2.3. Lexical distribution of labile verbs For most subtypes of labile verbs, their lexical distribution is crosslinguistically relatively straightforward. For instance, verbs belonging to the reciprocal subtype are principally the so-called ‘inherent reciprocals’ (in Kemmer’s 1993 terms), such as English kiss, Estonian suudlema ‘kiss’, and Arabic ’iltaqa: ‘meet’. Reflexive lability is characteristic for verbs relating to body care, e.g. Khwarshi (East Caucasian) esanho ‘wash’, Piro (Arawakan) šišaptota ‘wash’ (Mattheson 1965). The situation with converse verbs is a bit more complicated, but in general, this subtype of lability is restricted to verbs of emotions and perception (Bulgarian xaresvam ‘like’, French sentir ‘smell’ and so on). The distribution of the passive type is unclear because of its low frequency. In contrast, the distribution of the anticausative subtype varies significantly across languages. While I do not analyze it in detail here, I note that this distribution is not fully accounted for by the parameter of spontaneity (the probability of spontaneous occurrence of an event, i.e. its occurrence without any external agent) proposed by Nedjalkov (1969) and Haspelmath (1993a); see Letuchiy (2008) for detailed argumentation. Spontaneity is indeed relevant for the distribution of anticausative labile verbs in some languages, such as Adyghe (West Caucasian) and French, but not for others. In some cases, the crucial parameter is rather the degree of affectedness of the patient. For instance, in Lezgian (Haspelmath 1993b), lability characterizes primarily destruction verbs, such as xun ‘break’, kun ‘burn’, q:azunun ‘tear’, čurun ‘bake’, q’in ‘kill / die’. In contrast, lability in verbs of destruction is not characteristic for Arabic, Ancient Greek, Awa Pit (Barbacoan; Curnow 1997) and Russian. In all of these languages, verbs with weak semantic transitivity (motion verbs in

Towards a typology of labile verbs

253

Ancient Greek; verbs like ishkwin- ‘be startled / stare’, kaa- ‘be born / bear’, alizh kul- ‘be annoyed / annoy’ in Awa Pit) are labile. Finally, it is also noteworthy that the distribution of anticausative labile verbs within the lexicon does not appear to be formulable in terms of implicational universals. For instance, in Lezgian and Ancient Greek the sets of anticausative labile verbs do not intersect at all.

3. Anticausative lability and derivational markers I turn now to the relationship between anticausative lability and (anti)causative derivational markers. In previous research, lability and derivational markers of causative and anticausative have generally been regarded as mutually exclusive. For instance, in Haspelmath (1993a), Ljutikova (2002), and Nichols, Peterson and Barnes (2004), pairs that include a transitive and an intransitive verb, such as ‘die’ / ‘kill’ or ‘dry (transitive / intransitive)’, are analyzed; for a given language, each pair is considered to belong to one of five types: ‘marked causative’ (the transitive verb is a derived from the intransitive one), ‘marked anticausative’ (the intransitive verb is derived from the transitive one), ‘equipollent’ (both verbs contain derivational markers), ‘labile’ (the same verb designates the transitive and the intransitive variant of the situation), and ‘suppletive’ (the two verbs have different roots / stems). In fact, this view simplifies the real situation, as illustrated by the German and French labile verbs in (4–5): brechen ‘break’ and casser ‘break, tear’. The German verb cannot form a marked anticausative (4b is ungrammatical), while the French one can (5b is grammatical): German (Indo-European, Germanic) (4) a. Die Tasse ist gebrochen. ‘The cup broke.’ b. *Die Tasse hat sich gebrochen. ‘The cup broke.’ c. Ich habe die Tasse gebrochen. ‘I broke a cup.’ French (Indo-European, Romance) (5) a. Le fil a cassé. ‘The thread broke.’ b. Le fil s’est cassé. ‘The thread broke.’ c. J’ai cassé la tasse. ‘I broke a cup.’ The following sections explore cross-linguistic variation in the relations between labile verbs and (anti)causative markers.

254

Alexander Letuchiy

3.1. Labile verbs that do not take a derivational marker In some languages, the clear-cut division of the labile type and the type ‘marked (anti)causative’, as assumed in Haspelmath (1993a), does indeed exist. In these languages, such as German, Adyghe (West Caucasian), Boumaa Fijian (Austronesian, Oceanic), Godoberi (East Caucasian) some or all labile verbs can never take the causative / anticausative marker. These languages show how two very different phenomena – a grammatical derivational marker and a lexical phenomenon of lability – can be in complementary distribution. This fact is typologically relevant in that it indicates that different levels of language are not independent of each other. In some cases, the incompatibility of labile verbs and (anti)causative markers is motivated by very general restrictions. For example, the causative derivation in Fijian (see Dixon 1988) is incompatible with all transitive verbs, including transitive uses of labile verbs. However, in many languages the incompatibility of labile verbs with derivational markers cannot be explained by general restrictions. For instance, in Adyghe and Godoberi, some labile verbs do not take the causative marker in their intransitive uses, while all other intransitive verbs can be causativized. For example, the verbs qWǝten ‘break’ in Adyghe and hiš:i ‘close’ in Godoberi can only be causativized in their transitive uses (see 6d and 7d); causatives of these verbs mean, respectively, ‘make smb. break smth.’ and ‘make smb. close smth.’ Their intransitive uses cannot be causativized (see 6c and 7c); the meanings ‘break smth.’ and ‘close smth.’ can only be expressed by the transitive uses of the labile verbs, not by morphological causatives: Adyghe (West Caucasian, Circassian) (6) a. Čaške-r me-qWǝte. cup-ABS DYN-break ‘The cup broke.’ b. Se čaške-r s-qWǝta-ʁe. I cup-ABS 1SG.A-break-PST ‘I broke the cup.’ c. *Se čaške-r z-ʁe-qWǝta-ʁ. I cup-ABS 1SG.A-CAUS-break-PST Intended: ‘I broke the cup.’ d. Se č̣’ale-m čaške -r je-z-ʁe-qWǝta-ʁ. I boy-ERG cup-ABS 3SG.IO-1SG.A-CAUS-break-PST ‘I made the boy break the cup.’

Towards a typology of labile verbs

255

Godoberi (East Caucasian, Andic; Kibrik 1996: 123) (7) a. Hincu ži-b-da hiš:i. door REFL-3-PCT close.PST ‘The door closed by itself.’ b. Ilu-di hincu hiš:i. mother-ERG door close.PST ‘Mother closed the door.’ c. *Ilu-di hincu hiš:-ali. mother-ERG door close-CAUS.PST Intended: ‘Mother closed the door.’ d. Imu-di ilu-č’u hincu hiš:-ali. father-ERG mother-CONT door close-CAUS.PST ‘Father made mother close the door.’ Andrej Kibrik (1996: 122–123) explains this phenomenon in diachronic terms: if one of the uses of a labile verb is incompatible with the causative marker, this use is considered to be historically secondary. However, this decision is problematic both for Godoberi and typologically. For Godoberi, Kibrik’s claim conflicts with the fact that morphological causatives can be further causativized (e.g. bic’-al-ali ‘make melt’, a causative of bic’ali ‘melt (transitive)’, which is itself a morphological causative). If the ungrammaticality of (7c) is really explained by the fact that the transitive use is secondary for the verb ‘close’, why can the transitive verb ‘melt’, which is morphologically derived, be further causativized? This fact presupposes a fundamental difference between morphologically derived verbs, such as ‘melt’, and diachronically derived uses (7c), which Kibrik does not account for. In principle, one might argue that the secondary uses of Godoberi labile verbs emerged after the morphological causative became productive, and that this is why they cannot be an input for the causative derivation. However, this decision requires us to suppose that intransitive uses of verbs like hiš:i ‘close’ are newer than any other verb of the language: according to Kibrik, all lexemes except some labile verbs and several individual lexemes can be causativized. The Godoberi data is insufficient to conclude that intransitive uses of some labile verbs are the most recent lexemes in this language, though it is theoretically possible. Moreover, Kibrik’s account, originally used for Godoberi, it is not helpful for cases like French and German. The difference between (4b) and (5b) above would mean that the direction of derivation is different for German brechen and French casser. However, according to Greimas (2001)

256

Alexander Letuchiy

and Kluge (1975) both brechen and casser used to be only transitive in Old French and in the preceding stages of German, respectively; their intransitive uses are historically secondary. We must therefore find a synchronic feature of labile verbs or of derivational markers that prevents some uses from taking derivational markers. To solve this problem, we could in principle postulate two different mechanisms of derivation for verbs like Adyghe zebɣǝrǝteqWǝn ‘spill’ and those like qWǝten ‘break’. Within such an account, the intransitive uses of some labile verbs in Godoberi and Adyghe, such as qWǝten, are syntactically similar to derived anticausatives; i.e. a zero derivational marker is postulated. Note that very often derived anticausatives do not take causative markers (see Ljutikova et al. 2006 for Balkar). On the other hand, if a verb can be causativized in both uses (e.g. Adyghe zebɣǝrǝteqWǝn ‘spill’), this would mean that both of them behave synchronically as non-derived. We can then hypothesize that they underwent a purely semantic change similar to a change between two meanings of the same word such as couple ‘a small number’ and couple ‘pair’. The disadvantage of this decision is that no additional tests prove that the mechanism of derivation is different for the two verb classes. I therefore adopt a different hypothesis. This is argued specifically for Adyghe, and may also be true for other languages. I propose that properties of the derivational marker itself prevent it from combining with some labile verbs. This seems to be the case in Adyghe, in which the causative marker ʁe- always adds a new subevent to the event structure (understood in terms of Ljutikova et al. 2006). In Arkad’jev and Letučij (2007), we show that syntactic tests, e.g., the interpretation of negation, prove the existence of complex event structure for verbs with the causative marker. Most Adyghe labile verbs that are incompatible with the causative marker denote momentary events, for which the supposed meaning of the derived causative verb is usually unnatural: events like ‘break (smth.)’ are not usually conceptualized as including two subevents. Thus, the causative marker in Adyghe is semantically incompatible with events like ‘break’ or ‘tear’. Another line of argument is plausible for German and French labile verbs. Rothemberg (1974) shows that in French, lability characterizes very few transitive verbs with strong semantic transitivity – the class with which anticausative markers are most likely to occur. On the contrary, many verbs of this semantic class are labile in German. As a result, in German lability in a sense ‘outcompetes’ the derivational marker, making it incompatible with the destruction class. This is not the case in French.8

Towards a typology of labile verbs

257

3.2. Labile verbs that do take a derivational marker In contrast to the cases above, anticausative labile verbs in some languages do take valency-changing markers. This often leads to apparent synonymy of forms (cf. the French verbs casser (intransitive use) and se casser ‘break, tear’). Cases of this type may seem to violate the economy principle: the language uses two synonymous mechanisms with the same lexeme. But in fact transitive uses of labile verbs are not fully synonymous with morphological causatives. Three types of semantic opposition between the derived (anti)causative and the corresponding member of the labile pair (transitive for causatives, intransitive otherwise) may be distinguished: (i) oppositions of semantic types; (ii) participant properties variation; (iii) situational oppositions.

We can speak of oppositions of semantic types when the correlation type in the labile pair (here the anticausative) does not match the semantics of the morphologically derived verb (e.g. autocausative).9 For example, the intransitive use of the German labile verb hinunterstürzen ‘turn over’ in (8b) expresses the anticausative meaning. The subject is patientive: it does not control the situation. In (8c) the reflexive verb, which may first seem to be synonymous to the labile verb, expresses the autocausative meaning. The subject participant controls the initial phase of the situation. German (Indo-European, Germanic) (8) a. Er stürzte sein Glas hinunter. ‘He turned his glass over.’ b. Er stürzte hinunter. ‘He fell.’ c. Er kletterte auf einen sehr hohen Berg und stürzte sich hinunter. ‘He climbed up a very high mountain and jumped down.’ The second type – participant properties variation – supposes that the labile verb and the derived form differ according to the semantic properties of the participants, mostly those of the subject. Very often this difference is related to control or volitionality: for instance, the unmarked use may have a subject which can be agentive or non-agentive, whereas the subject of the marked causative must be agentive. Compare the Russian labile verb lit’ ‘pour’ and the marked anticausative lit’sja ‘pour’. The latter can designate any type of spontaneous motion of liquid (e.g. 9b). The former is restricted

258

Alexander Letuchiy

to contexts where the patient is autonomous (motion of liquid which is unexpected and not controlled by an agent (9c), but not, for instance, motion of liquid from the tap, which is expected by the agent (9b). Thus, we can speak of a semantic role variation: the (autonomous) subject of the intransitive use in (9c) is semantically different from the (non-autonomous) object of the transitive use in (9a). The ‘classical’ labile verbs, such as break in English, can also have this variation: the subject of the intransitive use can in some contexts be more autonomous than the object of the transitive use. However, this variation in English is not obligatory for all contexts, whereas it is obligatory for all contexts of the intransitive use of Russian lit’: Russian (Indo-European, Slavic) (9) a. My l’j-om masl-o na skovorod-u. we pour-1PL.PRS oil-ACC on pan-ACC ‘We pour (some) oil into the frying pan.’ ?? b. Iz kran-a l’j-ot-sja / l’j-ot vod-a. from tap-GEN pour-3SG.PRS-REFL pour-3SG.PRS water-NOM.SG ‘The water is pouring from the tap.’ c. Trub-a lopnu-l-a, iz nej-o pipe-NOM.SG crack-PST-F from she-GEN l’j-ot-sja / l’j-ot vod-a. pour-3SG.PRS-REFL pour-3SG.PRS water-NOM.SG ‘The pipe has cracked, the water is pouring from it.’10 In Adyghe, participant properties are also relevant for the choice between a causative and a labile verb: the verb qjǝnen ‘leave; stay’ in its transitive use requires both the actor and the undergoer (patient) to be nonvolitional. Volitionality of the undergoer (10c) or of the actor (10d) require the morphological causative to be used: Adyghe (West Caucasian, Circassian) (10) a. Se wǝne-m sǝ-qe-na-ʁ. I home-ERG 1SG.S-DIR-stay-PST ‘I stayed at home.’ (intransitive use) b. Sabǝj-ǝm ǯegWaʁe-xe-r wǝne-m qǝ-r-jǝ-na-ʁ. child-ERG toy-PL-ABS home-ERG DIR-LOC-3SG.A-leave-PST ‘The child left his toys at home.’ (non-volitionally)

Towards a typology of labile verbs

259

c. Šǝbλe-m wǝne-m tǝ-qǝ-r-jǝ-ʁe-na-ʁ / storm-ERG house-ERG 1PL.S-DIR-LOC-3SG.A-CAUS-stay-PST / *tǝ-qǝ-r-jǝ-na-ʁ / 1PL.S-DIR-LOC-3SG.A-stay-PST ‘The storm made us stay at home.’ (volitional object) d. Sabǝj-ǝm ǯegWaʁe-xe-r wǝne-m child-ERG toy-PL-ABS home-ERG ?? qǝ-r-jǝ-na-ʁ. qǝ-r-jǝ-ʁe-na-ʁ / DIR-LOC-3SG.A-CAUS-leave-PST / DIR-LOC-3SG.A-leave-PST ‘The child left his toys at home.’ (volitionally) It may seem that in Russian (9) and Adyghe (10) we observe very different situations: in (9) the unmarked form presupposes an autonomous subject; in (10), in contrast, it designates that none of the arguments is agentive or volitional. However, in fact the opposition of participant properties remains the same in (9) and (10). The difference in the interpretation of the labile verb is due to the fact that the Russian labile verb in (9) ‘competes’ with the anticausative form, whereas the Adyghe lexeme ‘competes’ with the marked causative. The main systemic requirement is that labile verbs must differ semantically from marked (anti)causatives. When lability ‘competes’ with the causative, transitive uses like (10b) choose the semantic feature which is not characteristic for marked causatives: non-agentivity of both arguments. According to Shibatani (2002) and others, causatives in the languages of the world often have an agentive causer and an agentive causee. Thus, labile verbs can have a non-agentive causer and causee. By contrast, in languages like Russian, where lability competes with anticausatives, intransitive uses of labile verbs have semantic features which are not characteristic for anticausative verbs. Anticausatives tend to have patientive subjects (see for instance Haspelmath 1993a), while intransitive uses of labile verbs can have agentive or autonomous subjects (this is also true for some other Russian labile verbs, such as colloquial gnat’ ‘run / urge’ or dvinut’ ‘move smth. / go’). Finally, we can speak of situational oppositions when the derived verb differs from the corresponding use of the labile verb by the semantic properties of the whole denoted situation. For instance, the two may have different aspectual characteristics; in Heerlen Dutch, for example, intransitive uses of labile verbs differ from marked anticausatives of the same lexemes (Cornips and Hulk 1996).

260

Alexander Letuchiy

Heerlen Dutch (Indo-European, Germanic; Cornips and Hulk 1996) (11) a. *Het ei zich 3 minuten lang gekookt heeft. DEF egg REFL 3 minute.PL for cook.PART have.PRS.3SG Intended: ‘…that the egg has boiled for three minutes.’ b. Het ei 3 minuten lang gekookt heeft. DEF egg 3 minute.PL for cook.PART have.PRS.3SG ‘…that the egg has boiled for three minutes.’ Cornips and Hulk (1996) show that the incompatibility of marked anticausatives with for-adverbials of temporal duration in Heerlen Dutch, as in (11a), results from the fact that the marked anticausatives emphasize only the final phase of the situation. In contrast, intransitive uses of labile verbs do not emphasize any particular phase and thus allow for expressions of duration.11 According to Schäfer (2003), similar aspectual differences are observed in Italian, Greek, French and German. However, example (12) shows that in Adyghe we have a different picture: Adyghe (West Caucasian, Circassian) (12) a. Č̣’ale-m depqǝ-r zebɣǝr-jǝ-ʁe-teqWǝ-ʁ. boy-ERG wall-ABS LOC-3SG.A-CAUS-spill-PST ‘The boy destroyed the wall.’ b. Se ṣ̂Weš’ǝʁWǝ-r zebɣǝrǝ-s-teqWǝ-ʁ. I sugar-ABS LOC-1SG.A-spill-PAST ‘I spilled sugar.’ Native speakers distinguish the transitive use of the labile verb zebɣǝrǝteqWǝn ‘spill’ (12b) from the causative form ‘destroy’ (12a). The labile verb does not emphasize the causation: no special efforts of the agent are necessary to spill sugar. On the other hand, the causative form in (12a) is employed to denote the situation of destruction, which presupposes an effort of the causer.12 The difference between (11) and (12) again shows that the semantic properties of labile verbs depend on the type of derivational marker with which they compete (an anticausative marker in 11 and a causative marker in 12) – labile verbs themselves do not have any aspectual restrictions a priori. In (11), the intransitive use can take an adverbial of duration, whereas the marked anticausative is incompatible with such an adverbial. In (12), the competition between lability and the causative leads to another

Towards a typology of labile verbs

261

kind of opposition: since the function of the causative is to introduce the causing situation, lability, in its turn, does not emphasize the causation. 3.3. Summary and discussion The relations between anticausative lability and (anti)causative markers that we examined in this section are far from accidental. Labile verbs may be incompatible with a particular derivational mechanism for the various reasons addressed above. When a labile verb is compatible with a marker of valency derivation, the derived verb has semantic features typical of marked (anti)causatives. Lability then plays the role of a ‘default’ mechanism. The aspectual characteristics and agentivity / patientivity features of the labile verbs and their arguments depend on which derivational marker competes with lability in a given system. Lability, as opposed to valency derivations, follows a simple rule: if A is a labile verb with two uses, A1 and A2, and if the language has also an (anti)causative B which is similar in meaning to A1, then A2 is semantically closer to A1 than to B. Thus, the two uses of any labile verb have some common semantic and syntactic features. For example, when lability competes with a causative marker, as in (10) and (12) above, the marked causative denotes a situation in which the causer is usually volitional and the causation is conceptualized as a separate subevent. At the same time, the transitive use of a labile verb does not introduce a new autonomous situation. Although the semantics of (10b) and (12b) include the meaning of causation, the causer and the causing situation are not emphasized. Thus, the members of a labile pair differ from each other less significantly than do the unmarked verb and the marked causative. The same is true when we observe a competition between the intransitive use of the labile verb and an anticausative marker, as in (9) and (11). The marked anticausative emphasizes the change of properties of the patient and/or the final state of the patient. The intransitive use of the labile verb, on the contrary, does not emphasize the effect on the patient – it can denote the whole situation, not only the final state, as in (11b). Thus, there are more features in common between the two uses of a labile verb than between an unmarked transitive verb and a marked anticausative. The opposition between lability and derivational markers are linked to the general motivation for lability in the language system, the subject of Section 4 (see also Letučij 2006a, Letuchiy 2006b).

262

Alexander Letuchiy

4. Motivations for lability The functional approach to language presupposes that linguistic phenomena are functionally motivated within the language system. The function of lability has not been discussed in previous research. I address this problem here in light of the main focus of this chapter, the relations between anticausative lability and derivational markers, as investigated in Section 3 above. In considering the function of lability, the most important question is whether lability can be regarded as a sort of unmarked valency derivation. Most authors, such as Haspelmath (1993a) and Ljutikova (2002), do not address this, although some research in the generative paradigm (Hale and Keyser 2002) explicitly considers lability to be a derivation. The data considered in this chapter suggest that lability cannot be considered an unmarked derivation. Recall that the principal characteristic of derivation, as explicitly or implicitly assumed in the majority of works on the typology of valency derivations (e.g., Nedjalkov and Sil’nickij 1969, Geniušienė 1980, Haspelmath 1993a, Shibatani ed. 2002), is roughly the following: Derivation tends to derive the situation A′ from situation A, which is different from A′ (although semantically close to it). Lability is more likely a sort of polysemy, similar to other types of polysemy existing in natural languages. We can say that polysemy tends to unify two semantically close situations or entities in one lexeme, while the difference between the two situations is irrelevant (though relevant for derivation). Similarly, the crucial factor that makes lability possible is the proximity between the semantic and syntactic features of the two situations (other than the difference in valency and transitivity, cf. 2.2). Let us review the conclusions drawn in Sections 2–3. First of all, some labile verbs have one bivalent and one monovalent use, while others have two bivalent uses. If lability were an unmarked causative / anticausative derivation, the only syntactic type of lability would be the type with one monovalent and one bivalent use, since causativization, anticausativization and often passivization change the number of syntactic arguments of the verb (the agent of the passive construction often cannot be expressed in the sentence). However, labile verbs of the converse and passive subtypes (not of the anticausative subtype) often have two bivalent uses. By contrast, if lability is a kind of polysemy, this fact is not at all strange. In the languages of the world, there are numerous semantic types of polysemy which do not correspond to any marked derivations. More-

Towards a typology of labile verbs

263

over, if a verb is polysemous it often keeps the same number of arguments in different uses (cf. the English verb hug which can denote an action, as in I hugged my wife, or a spatial configuration, as in The road hugs the river ‘The road goes along the river’). Thus, the existence of the converse type can be explained only if we suppose that lability is a type of polysemy and not a valency derivation. Second, very often the group of anticausative labile verbs includes verbs with weak semantic transitivity (see 2.3). If this type of lability were an unmarked derivation, this distribution would be very peculiar. Typologically, anticausativization tends to be compatible primarily with destruction verbs, such as Russian razbit’ ‘break (transitive)’ / razbit’-sja ‘break’ (intransitive), porvat’ ‘tear (transitive)’ / porvat’-sja ‘tear (intransitive)’, and causativization has often a very wide distribution. We do not find any languages where causativization or anticausativization would be restricted to verbs with a weak semantic transitivity. If lability is not a derivation, however, the fact that labile verbs are often restricted to those having weak semantic transitivity can be explained. Two uses of a polysemous lexeme tend to have common semantic features; thus the members of pairs like ‘begin (transitive)’ / ‘begin (intransitive)’ are closer to each other than the members of pairs like ‘break (transitive)’ / ‘break (intransitive)’. In the former pair, neither of the uses has a strongly patientive or a strongly agentive subject, whereas in the latter pair, one of the uses has a strongly patientive subject, and the other use has a strongly agentive subject. However, this argument is weakened by the fact that there are languages in which all or most labile verbs have strong semantic transitivity. Third, the semantic and syntactic features of a derivational marker are to a certain degree independent of other derivational mechanisms in the same language. For instance, the properties and the lexical distribution of anticausative markers seem to be roughly consistent across languages; I am not aware of any cases in which the features of an anticausative construction depend on the features of other valence-related derivational forms in the same language. In contrast, my analysis of the relations between labile verbs and derivational markers has shown that lability does not offer such an invariable semantic feature. The interpretation of labile verbs depends on the marker (causative or anticausative) it competes with. Again, this situation is explicable if lability is not considered a derivation, but rather a kind of polysemy; in this case lability and valency derivation are simply different mechanisms, which are not obligatorily in complementary distribution in a given language. The variance of interpretation of the labile verbs in (9–12) is also explicable in this approach; given that

264

Alexander Letuchiy

polysemy is not a grammatical phenomenon, the exact features of each meaning of a given lexeme can vary language-internally and languageexternally. What motivates this polysemy in all languages is the existence of semantic features that are common across the two uses of a labile verb, and this is why these two uses are semantically closer to each other than are the base verb and the derived (anti)causative. On the whole, the approach proposed here is problematic only for languages like English, where lability is very productive. In these languages, labile verbs are often incompatible with derivational markers (or the language lacks derivational markers altogether); in such cases the argument that lability is a sort of polysemy which interacts in a complicated way with derivational markers is not so persuasive. In languages of this type, lability may rather take the functional niche of an unmarked valency derivation.

5. Conclusions As I have argued here, labile verbs can be classified according to a number of subtypes – anticausative, reciprocal, reflexive, converse, and passive – most of which correspond to morphologically marked valency derivations. This classification leads to a revised definition of lability: a labile verb is a verb with two uses, which differ in syntactic transitivity and in the semantic role of the subject. This chapter focused on the relations between lability and markers of valency derivation, a particularly relevant question given their functional similarity. Though lability is not a purely grammatical phenomenon, it is not independent of purely grammatical markers. As I have shown for the anticausative subtype of labile verbs, there are cases in which lability outcompetes a derivational marker, preventing it from being compatible with a certain class of verbs. However, very often labile verbs take the derivational marker in all their uses; in such cases, the opposition between a labile verb and a marked (anti)causative derivative obeys a general rule: the uses of a labile verb tend to have more common semantic and syntactic features than do an unmarked verb and its morphological (anti)causative. This point was developed in the final section, in which I argue that the function of lability in the language system is different from that of derivational markers. Lability does not serve to derive one variant of the situation from another one. It is rather a kind of polysemy, which brings together, in one lexeme, two situations that share significant semantic and syntactic properties.

Towards a typology of labile verbs

265

Notes 1.

I would like to thank my Adyghe, Khakas, Russian, German and Bulgarian informants. I am also grateful to Vladimir Plungian, Peter Arkadiev, Yakov Testelec, Nina Sumbatova, Gilles Authier, Francisco Queixalós and others for fruitful discussions. Last but not least, I want to thank Alexandre Arkhipov and Patience Epps for their invitation to participate in the present volume. 2. I concentrate mostly on the data of Indo-European and Caucasian languages. Other families and parts of the world are also represented, e.g. Oceania: Xârâcùù (Oceanic); Africa: Bambara (Western Mande), Songhay (NiloSaharan), Kabyle (Berber); Asia: Mandarin Chinese (Sinitic), Limbu (TibetoBurman), Japanese; and the Americas: Piro (Arawakan), Trumai (isolate). 3. ‘Anticausative’ refers to a derivation that decreases the valency of the verb and eliminates the agent, while the patient becomes a syntactic subject, e.g. se casser ‘break (intransitive)’ from casser ‘break (transitive)’ in French. I call this type of lability anticausative and not causative because its distribution is similar to that of anticausative markers. 4. See Letučij (2006a) on these classes. 5. Abbreviations: 1,2,3 – 1st, 2nd, 3rd person, A agent, ABS absolutive, ACC accusative, CAUS causative, DAT dative, DEF definite article, DIR directive, ERG ergative, F feminine, FOC focus, GEN genitive, INCOMPL incompletive, LOC localization, M masculine, NOM nominative, PART participle, PL plural, PRS present tense, PST past tense, PCT punctive, REFL reflexive, SUP superessive. 6. See Aikhenvald (in press). 7. The distinction between agent-preserving vs. patient-preserving lability was introduced in Dixon (1994). With patient-preserving lability, both uses of a labile verb have a patientive argument (I broke the stick / The stick broke). With agent-preserving lability both uses preserve the agentive argument (John eats meat / John eats). 8. The sole case that remains unexplained is the inability of some Godoberi verbs to take the causative marker in their transitive uses. I have no explanation for this fact, which is very rare typologically. 9. The term ‘autocausative’ was proposed by Geniušienė (1980) for agentive intransitive motion verbs (e.g. Russian podnjat’-sja ‘get up; go upstairs’ from podnjat’ ‘raise’). 10. I thank an anonymous reviewer for bringing up the question of other nonstandard correlations between the two uses of labile verbs. However, this question is beyond the scope of this paper. 11. Unfortunately, I do not have data on aspectual properties of non-labile verbs in Heerlen Dutch. 12. This may also be the case in Manambu (Ndu). According to Aikhenvald (in press: 16), causatives from object-preserving labile verbs focus the effort of

266

Alexander Letuchiy the causer, e.g. rali ‘untie’ (labile) vs. kay-rali (causative) ‘untie with special effort’.

References Aikhenvald, Alexandra Y. in press Causatives which do not ‘cause’: on non-valency-increasing effects of valency-increasing derivations. Arkadiev [Arkad’jev], Petr M., and Alexander B. Letuchiy [Letučij] 2007 Tipologičeski netrivial’nyje svojstva adygejskogo kauzativa [Typologically unusual properties of Adyghe causative constructions]. In Proceedings of the Third Conference on Typology and Grammar for Young Researchers. Saint-Petersburg, 2007. Bokarev, Aleksandr A. 1949 Sintaksis avarskogo jazyka [The syntax of Avar]. Moscow / Leningrad: Nauka. Căluianu, Daniela 2000 Emotion verbs in Romanian. In Modern Approaches to Transitivity, Ritsuko Kikusawa and Kan Sasaki (eds), 143–194. Tokyo: KuroshioPublisher. Chaker, Salem 1983 Un parler berbère d’Algérie (kabylie). Aix-en-Provence : Université de Provence. Chikobava, Arnold S. 1942 The problem of ergative construction in Caucasian languages: stabile and labile variants of this construction. Izvestija IYaIMK, XII. Tbilisi (in Georgian). Cornips, Leone, and Aalke Hulk 1996 Ergative reflexives in Heerlen Dutch and French. Studia Linguistica 50(1): 1–21. Curnow, Timoty J. 1997 A grammar of Awa Pit (Cuaiquer): An indigenous language of Western Columbia. Ph. D. diss., Australian National University. Dixon, Robert M. W. 1988 A Grammar of Boumaa Fijian. Chicago: Cambridge University Press. 1994 Ergativity. Cambridge: Cambridge University Press. Geniušienė, Emma Š. 1980 The Typology of Reflexives. Berlin, New York, Amsterdam: Mouton de Gruyter.

Towards a typology of labile verbs

267

Gišev, Nukh T. 1968 Glagoly labil’noj konstrukcii v adygejskom jazyke [Labile verbs in Adyghe]. Majkop: Adygejskoje knižnoje izdatel’stvo. Greimas, Algirdas J. 2001 Dictionnaire de l’Ancien Français. Paris: Larousse. Hale, Kenneth, and Keyser, Samuel J. 2002 Prolegomenon to a Theory of Argument Structure. Kluwer Academic Press. Haspelmath, Martin 1993a More on the typology of inchoative / causative alternations. In Causatives and transitivity, Bernard Comrie and Maria Polinsky (eds), 87–120. Amsterdam: Benjamins. 1993b A Grammar of Lezgian. Berlin: Mouton, Walter de Gruyter. Kemmer, Susanne 1993 The Middle voice. Amsterdam / Philadelphia: Benjamins. Kibrik, Aleksandr E., Sandro V. Kodzasov, Irina P. Olovjannikova, and Dzhalil’ S. Samedov 1977 Opyt strukturnogo opisanija arčinskogo jazyka. Tom 1. Leksika. Fonetika [A structural description of Archi. Volume 1. Vocabulary. Phonetics]. Moscow: Moscow State University. Kibrik, Aleksandr E., Sandro V. Kodzasov, and Irina A. Murav’jova 2000 Jazyk i fol’klor aljutorcev [Alutor language and folklore]. Moscow: Nasledie. Kibrik, Andrej A. 1996 Transitivity in Godoberi. In Studies in Godoberi, A.E. Kibrik (ed.), 107–146. München: LINCOM Europa. Kluge, Friedrich 1975 Etymologisches Wörterbuch der Deutschen Sprache, 21st ed. Berlin, New York: Walter de Gruyter. Letuchiy [Letučij], Alexander B. 2006a Tipologija labil’nyx glagolov: semantičeskije i morfosintačeskije aspekty [Typology of labile verbs: semantic and morphosyntactic aspects]. Ph.D. diss. Moscow: Russian State University for Humanities. 2006b Typology of systems of labile verbs. Handout of a talk at the conference ‘SWL II’. Lancaster. 2008 Labile verbs: looking for a cross-linguistic motivation. Handout of a talk at the Workshop on Transitivity, Köln. Available at www.zsm.uni-koeln.de/files/Letuchiy.pdf. Ljutikova, Ekaterina A. 2002 Russkije labil’nyje glagoly v tipologičeskoj perspektive [Russian labile verbs in a typological perspective]. Handout of a talk at the conference ‘Lomonosov readings’. Moscow.

268

Alexander Letuchiy

Ljutikova, Ekaterina A., Sergej G. Tatevosov, Mixail Ju. Ivanov, Anna G. Pazel’skaja, and Andrej B. Šluinskij 2006 Struktura sobytija i semantika glagola v karačaevo-balkarskom jazyke [Event structure and verb meaning in Karachay-Balkar]. Moscow: IMLI. Mattheson, Esther 1965 The Piro (Arawakan) language. Berkeley: University of California Press. Moyse-Faurie, Claire 1995 Le xârâcùù, langue de Thio-Canala (Nouvelle Calédonie), Eléments de syntaxe, (Coll. Langues et Cultures du Pacifique 10, SELAF 355.) Paris-Louvain: Peeters. Nedjalkov, Vladimir P. 1969 Nekotoryje verojatnostnyje universalii v glagol’nom slovoizmenenii [Some probabilistic universals in verbal derivation]. In Jazykovyje universalii i lingvističeskaja tipologija [Language universals and linguistic typology], Igor F. Vardul’ (ed.), 106–114. Moscow: Nauka. Nedjalkov, Vladimir P., and Georgij G. Sil’nickij 1969 Tipologija morfologičeskogo i leksičeskogo kauzativov [Typology of morphological and lexical causatives]. In Tipologija kauzativnyx konstrukcij [Typology of causative constructions], Alexander A. Xolodovič (ed.), 20–60. Moscow: Nauka. Nichols, Johanna, David A. Peterson, and Jonathan Barnes 2004 Transitivizing and detransitivizing languages. Linguistic Typology 8: 149–211. Rothemberg, Mira 1974 Les verbes à la fois transitifs et intransitifs en français contemporain. La Hague: Mouton. Schäfer, Florian 2003 The morphological patterns of anticausatives and their interpretations. Talk at the annual meeting ‘Linguistic representations and their interpretation’, Söllerhaus/Kleinwalsertal. Shibatani, Masayoshi (ed.) 2002 The Grammar of Causation and Interpersonal Manipulation. Amsterdam, Philadelphia: Benjamins. Zavala, Roberto 2000 Inversion and other topics in the grammar of Olutec (Mixean). Ph. D. diss., Department of Linguistics, University of Oregon. Vydrine, Valentin F., and Adama Coulibali 1994 Verbes réfléchis bambara, 1: pronoms réfléchis et groupements sémantico-syntaxiques des verbes non-réfléchis. Mandenkan 28: 1– 91.

Towards the typology of raising: A functional approach Natalia Serdobolskaya

1. Introduction1 There are a number of phenomena that are traditionally viewed as syntactic and are described primarily within formal linguistic frameworks. Semantic and pragmatic features of these phenomena often remain underestimated or ignored. Moreover, since formal syntactic theories tend to concentrate on a limited number of languages, the phenomena in question are often characterized on the basis of data from at most four or five languages. It may be that a theoretical explanation is proposed for just one syntactic pattern, and this may not be the most frequently attested pattern cross-linguistically. Raising2 seems to constitute an example of such a case. This phenomenon can be exemplified by the English sentences I believe him to be a linguist; He appears to be a good linguist. It has been argued that the noun phrases (NPs) marked with bold font are ‘raised’, because they show morphosyntactic properties of the matrix verb’s argument (direct object or subject), while semantically they belong to the embedded clause. The constructions analogous to English raising are cross-linguistically widespread, e.g. in Altaic, Caucasian, North American and other languages (see Serdobol’skaja 2005 for details). However, they do not show the same morphosyntactic properties as the English construction. In many languages, the raised NP does not have the morphosyntactic properties of the main verb’s direct object, even if it seems to belong to the matrix clause. The grammatical role of the raised NP (subject, direct object, or indirect object) within the embedded clause also differs among languages. The English syntactic pattern of raising thus seems not to be the only nor even necessarily the most frequent model; rather it seems to constitute one type of construction observed cross-linguistically. Accordingly, the discrepancy between the standard concept of raising and the constructions observed in other languages makes it difficult to formulate an exact definition of the term ‘raising’. Such a definition can only be proposed through a

270

Natalia Serdobolskaya

thorough analysis of the constructions occurring in languages of the world (widening the sample considered in this chapter). A tentative definition is, however, suggested in Section 2.2. Most researchers claim that the raised NP has no semantic association with the matrix verb (see the comments regarding example 3 in Section 2.1). The absence of a semantic distinction has been used as a diagnostic test to distinguish raising from infinitival control, e.g. I believe him to be a linguist vs. I persuaded him to be a linguist. However, this claim is not borne out even in English, as has been shown in Borkin (1973), Pesetsky (1991), etc.: the choice of raising construction or that-clause (e.g. I believe him to be a linguist / that he is a linguist) is highly influenced by the semantics of the matrix verb. In this chapter, data from various unrelated languages3 is considered, showing that in most cases the choice of the raising construction is determined by semantic and pragmatic factors. In this way, the chapter seeks to bring typological data to bear on the current views of raising. The chapter is organized as follows. In Section 2, syntactic patterns of raising are considered. Section 3 deals with the semantic and pragmatic properties of raising, and Section 4 concludes the chapter.

2. Syntax of raising 2.1. Syntactic properties of raising and long-distance agreement The term ‘raising’ has originally been used to refer to the constructions exemplified by the English sentence I believe him to be a linguist. The pronoun him in this example is, semantically, the subject of the embedded clause; however, it receives the object case from the matrix verb believe. This NP has been termed the ‘raised’ NP. It has been shown that the raised NP has the syntactic properties of the matrix verb’s direct object (see Postal 1974, Davies and Dubinsky 2004). For example, it becomes the subject if the matrix verb is passivized: English (Indo-European, Germanic; Postal 1974: 40) (1) Jack believed Joan to have been famous. – Joan was believed to have been famous by Jack. In addition, it can be replaced by reflexive and reciprocal pronouns coreferential with the antecedent in the matrix clause:

Towards the typology of raising

271

English (Indo-European, Germanic; Postal 1974: 42) (2) Jack believed himself to be immortal. Postal and others analyze a large number of properties associated with raising, such as quantifier scope, adverb position, etc. I propose to distinguish between two groups of tests: structural and constituency tests. Structural properties define the position occupied by the NP in question (the matrix verb’s direct object, subject or other), while constituency tests show whether the raised NP forms a constituent with the matrix or the dependent verb. Structural tests involve passivization, reflexivization, and reciprocality, as exemplified in (1–2). Constituency tests investigate the formation of clefts and pseudoclefts (What I believe is that Bill is intelligent / *Bill to be intelligent; Postal 1974: 132), the conjoining of the matrix clauses, the possibility of replacing the dependent clause with an anaphoric pronoun, the possibility of the dependent clause’s appearing as an independent utterance (as an answer to a question, e.g. What do you want? – *Nixon to win), the linear position of the raised NP, Pied-Piping effects, acceptance of particles modifying the whole dependent clause, etc. (See Testelec 2001 for the discussion of constituency tests in general, Postal 1974 for language-specific constituency tests applied to English raising, and Serdobol’skaja 2005 for the detailed discussion of these tests in raising constructions cross-linguistically.) Another group of tests (e.g. idioms’ test, dummy subjects’, dependent verb passivization test, etc.) may be employed to determine the difference between raising and control, i.e. constructions in which the NP in the matrix clause is raised, and those in which it originates as an argument of the matrix verb (see Serdobol’skaja 2005 for details). For example, dummy subjects are only possible in raising constructions: English (Indo-European, Germanic; Kuno 1976: 30, ex. 49–50) (3) a. I expected it to rain. b. * I persuaded it to rain. For the most part, the tests involved in this group are based on the following heuristics. The raised NP does not refer to a participant of the situation encoded by the matrix verb; hence, it can be a dummy subject (as in 3a), a part of an idiom etc. The matrix verb does not impose any selectional semantic restrictions on the raised NP (it can even have no denotatum, as in 3a). If the NP in question were an argument of the matrix verb,

272

Natalia Serdobolskaya

it would be sensitive to these restrictions. However, it is not. The raised NP is sensitive to the selectional restrictions imposed by the dependent verb. Hence, discussions of these tests (e.g. Postal 1974, Kuno 1976) typically presuppose that the matrix verb does not have any semantic association with the raised NP. However, this claim is not borne out even for English, see (17a, b). The tests discussed above are used by typologists and syntacticians as arguments for the existence or absence of raising in languages of the world. Usually, a construction is interpreted as involving raising if the NP in question acquires case marking from the matrix verb, and shows structural and constituent properties of the matrix verb’s direct object. Within the generative grammar paradigm, raising in English has been described in terms of the matrix verb assigning case to the lower clause’s subject: as the dependent verb is the infinitive, it is impossible for its subject to be assigned case in the lower clause.4 However, in many languages raising is possible from finite clausal complements (4) and from nominalizations (5). In both types of constructions the subject of the lower clause can be assigned case marking either by the dependent verb (nominative in [4]; nominative and genitive in [5]), or from the matrix verb (accusative in [4] and [5]):5 Japanese (Kuno 1976: 24) (4) Yamada wa Tanaka o / ga hannin Yamada TOP Tanaka ACC NOM criminal to dantei-si-ta. COMP sure-do-PST ‘Yamada was sure that Tanaka is a criminal.’ Uzbekh (Altaic, Turkic) (5) Olim-ni / Olim / Olim-ning Olim-ACC Olim(NOM) Olim-GEN men-ga shubhali ko’rinyapti. I-DAT doubtful seem ‘I doubt that Olim has (already) come.’

da COP.PRS

kel-gan-i go-PART.PST-3SG

In Uzbekh, as in other Turkic languages, the subject of a nominalization can occur either in genitive, nominative, or accusative case. The accusative case constructions show the properties of raising.

Towards the typology of raising

273

2.2. Cross-linguistic varieties of raising Raising in English and Japanese, the languages in which it has been most extensively documented, shows the following syntactic properties (the principles given below are taken from Postal 1974 [generative grammar] and Perlmutter and Postal 1983 [relational grammar]): (i) raising is possible from complement clauses only; (ii) only the subject of the lower clause can be raised; (iii) the raised NP occupies the syntactic position of one of the matrix verb’s complements (subject or direct object); (iv) the dependent clause does not occupy this position.

All these principles find their counterexamples in the languages of the world. In fact, in summarizing the properties of the constructions found in languages other than English, it should be concluded that languages in which raising exhibits all of the properties (i-iv) are much rarer than those in which it does not. For example, regarding (i) (raising from complement clauses only), raising is possible from adverbial clauses in Altaic languages (in a nonraising construction, the pronoun in (6) would appear in nominative case): Mongolian (Altaic, Mongolic; Sanžeev 1960: 74) (6) [Čam-ajg amralt-aas ire-x-ees] [you.OBL-ACC holiday-EL return-PART.FUT-EL] bi ene ažl-aa duusga-na. I this work-POSS.SJ finish-PRS ‘I’ll end up this work before you come after holiday.’

өmnө before

Similarly, Joseph (1990) argues for raising in Modern Greek constructions with the preposition me ‘with’, which function as temporal simultaneity clauses. Raising is observed in Irish temporal clauses with the conjunction i ndiaidh (Carnie and Harley 1997). LDA from adverbial clauses is likewise attested in Kashmiri (Hook and Kaul 1987: 56). With respect to (ii) (only the subject of the lower clause can be raised), in Kipsigis not only subjects can be raised (7b), but also direct objects (8), indirect objects (9) and non-argument NPs (10):

274

Natalia Serdobolskaya

Kipsigis (Nilo-Saharan, Nilotic; Jake and Odden 1979: 134–137) (7) a. mɔ́cè mù:sá [à-lápát]. want Musa(SJ) 1SG.SJ-run mù:sá [à-lápát]. b. mɔ́c-ɔ́:n wants-1SG.OBJ Musa(SJ) 1SG.SJ-run (a=b) ‘Musa wants me to run.’ (8)

mù:sá [kɔ̀-tɩ́l-an mɔ́c-ɔ́:n wants-1SG.OBJ Musa(SJ) 3SG.SJ-cut-1SG.OBJ ‘Musa wants Kiplangat to cut me.’

(9)

mɔ́cè mù:sá cì:tɔ́ [kɔ̀-tɩ́l-cí Kɩ́plàŋàt pè:ndɔ́]. wants Musa(SJ) man 3SG.SJ-cut-BEN Kiplangat(SJ) meat ‘Musa wants Kiplangat to cut the meat for the man.’

Kɩ́plàŋàt]. Kiplangat(SJ)

pè:ndɔ́]. (10) mɔ́cè mù:sá rô:twé:t [kɔ̀-tɩ́l-ɛ́:n wants Musa(SJ) knife 3SG.SJ-cut-INSTR meat ‘Musa wants Kiplangat to cut the meat with a knife.’

Similarly, in Niuean, A, S, and P arguments of the dependent verb can be raised (see Seiter 1983: 321). In Blackfoot and Quechua, raising of not only core arguments is attested, but also of obliques, and, in Quechua, of adverbials. Condition (ii) (only the subject of the lower clause can be raised) also appears to entail that only one NP could be raised at a time. However, some languages violate this constraint: Cuzco Quechua (Quechuan; Muysken and Lefebvre 1988: 146) merkadu-pi (11) Mariyacha Xosecha-q-ta platanu-ta Maria Jose-GEN-ACC banana-ACC market-LOC muna-n [ranti-na-n-ta]. want-3 buy-NMZ-3-ACC ‘Maria wants José to buy bananas in the market.’ In Quechua, the raised arguments are marked as follows: the A/S arguments get accusative marking (in a non-raising counterpart, Xosecha in (11) would be marked with genitive without accusative), while all other arguments preserve their original marking; raising affects their linear and structural position, as the position of the underlined NPs in (11) (see argumentation in Muysken and Lefebvre 1988). Some Irish constructions

Towards the typology of raising

275

headed by the verbal noun allow raising of both core arguments of the dependent verb (see Stenson 1981, Postal 1986 for discussion). We turn now to (iii) (the raised NP occupies the syntactic position of one of the matrix verb’s complements [subject or direct object]), and (iv) (the dependent clause does not occupy this position); let us investigate these properties in detail. According to the Relational Grammar analysis of raising, the raised NP forces the dependent clause out of its syntactic position, which is the position of the direct object of the matrix verb. The dependent clause then acquires chômeur status (i.e. it is a demoted element with the status assigned to agents in passive constructions; cf. the Motivated Chômage Law and Stratal Uniqueness Law in Relational Grammar, Perlmutter and Postal 1983). However, this rule does not always hold. In a number of languages the raised argument seems to belong to the matrix clause from the point of view of the constituency structure; yet it does not occupy the syntactic position of subject or direct object in the matrix clause. For example, in Tuvan, raising occurs from both nominalizations (12) and finite complements with compementizers (13): Tuvan (Altaic, Turkic) čedi-p kel-gen (12) Ajas-tǝ Ajas-ACC reach-CONV come-NMZ.PST ava-zǝ bil-ir. mother-POSS.3SG know-NMZ.FUT ‘Mother knows that Ajas has come.’

dep COMP

končužu-p (13) Ada-je-m-nǝ father-mother-my-ACC quarrel-CONV tur-gan-ǝn men dǝŋna-dǝ-m. stay-NMZ.PST-ACC.POSS.3 I hear-PST-1SG ‘I heard my parents quarrelling.’

The construction exemplified in (13) is intriguing, as both the nominalized verb and the raised NP get accusative marking from the matrix verb. If raising had taken place, the passivization of the matrix verb would make adajemnə ‘my parents’ the subject. However, such sentences are judged as unacceptable:

276

Natalia Serdobolskaya

Tuvan (Altaic, Turkic) (14) *Ada-je-m končužu-p tur-gan-ǝn father-mother-my quarrel-CONV stay-NMZ.PST-POSS.3 koža-lar-ga dǝŋna-l-gan. neighbour-PL-DAT hear-PASS-NMZ.PST (‘My parents have been heard quarrelling by the neighbours.’) This unacceptibility does not follow from any semantic restrictions on the passive, since the verb ‘hear’ with a nominal argument ada-je-m permits passivization. Other properties confirm that the raised NP does not occupy the position of direct object in the matrix clause; for example, constituency tests show contradictory results (Serdobolskaya 2006). In sum, it can be concluded that the raised NP in Tuvan is an element of the matrix clause; however, it does not occupy the direct object position there, even though it is marked with the accusative. This peculiarity of the Tuvan data seems to be due to the special status of the accusative case in this language, and in Turkic languages in general. Roughly speaking, the accusative in Tuvan tends to be used with definite and/or topical NPs, while other NPs tend to remain unmarked (see Muravyova 1992). The accusative is also used to mark adverbials of time: Tuvan (Altaic, Turkic) ulug balək-tə (15) Kəž-ən winter-ACC.POSS.3 big fish-ACC al-gan men. AUX-NMZ.PST I ‘In winter, I caught this big fish.’

tudu-p catch-CONV

In sum, the accusative in Tuvan does not only have a syntactic function, but also marks definiteness and topicality. Raising in Tuvan looks like a movement to a left periphery of the matrix clause, rather than a transformation to direct object. A similar argument seems to hold for Kalmyk (Mongolic). As shown in (6), Mongolic languages allow raising from adverbial clauses. In this case, the raised NP gets accusative marking, though the matrix verb does not have a direct object argument slot, and hence cannot assign accusative case. Instead, a new syntactic position seems to be created especially for the raised NP. This seems to follow from the special status of the accusative form in these languages – as in Turkic languages, the accusative in Kalmyk serves as a marker of topicality, definiteness and animacy, rather

Towards the typology of raising

277

than purely as a grammatical marker. Again, this leads to the conclusion that the accusative subject in these languages is not ‘raised’ to direct object position of the matrix verb, but occupies a position on the left periphery of the whole sentence. The movement to left periphery analysis was first proposed for longdistance agreement (LDA, by which an NP in the dependent clause controls the agreement of the matrix verb) in Tsez by Polinsky (2000). However, Tsez constructions differ from the Tuvan and Kalmyk cases in that the NP that triggers long-distance agreement in Tsez shows all the (structural and constituent) properties of the lower clause element. In Tuvan and Kalmyk, it demonstrates the properties of an element of the matrix clause (while still preserving some properties of the lower clause element). I conclude that the Tuvan and Kalmyk constructions are to be analyzed as raising to the left periphery of the whole sentence, while in Tsez raising to the left periphery of the lower clause is attested (Polinsky 2002). A similar phenomenon is observed in Komi-Zyrian and Mari. The subject NP associated with nominalized verbs receives genitive or nominative case marking. The subject in the genitive case shows a number of constituent and structural properties of an element of the matrix clause, as in the Tuvan example. However, the subject does not get case assignment from the matrix verb. The nominalization keeps its syntactic position as an argument of the matrix verb (see Serdobol’skaja 2005 for a detailed analysis.) A similar situation holds for some constructions in Irish (see Stenson 1981). The claim that the raised NP must force the dependent clause out of its syntactic position does not hold in many other languages of the world, besides Tuvan and Kalmyk. For example, in Turkish (Mulder 1976), Irish (Postal 1986), and in long-distance agreement constructions in Kashmiri (Hook and Kaul 1987), a new syntactic position is created especially for the raised NP. In Turkish, it is a nominative position that is created in the context of the verb görün ‘seem’ (which otherwise has neither nominative nor accusative arguments); in Irish the raised NP occurs with a preposition, associated with a matrix verb that otherwise has no prepositional arguments; and in Kashmiri intransitive matrix verbs can have an NP raised to direct object position (cf. the analysis of Cuzco Quechua in Muysken and Lefebvre 1988). To conclude this section, the constructions resembling raising that are attested cross-linguistically do not show the same syntactic properties as does raising in English. There are thus two alternative ways of dealing with the term ‘raising’: it can be used to refer to the English model of rais-

278

Natalia Serdobolskaya

ing only, or it can be extended to all the constructions examined in this section. I take the second alternative and propose to define raising as a construction in which one of the arguments of the dependent clause demonstrates morphosyntactic properties of an element of the matrix clause (namely, morphological marking of the raised NP, matrix verb agreement, or constituency properties). In fact, this definition can be applied to all those constructions described as involving long-distance agreement (LDA; see e.g. Polinsky 2000), the phenomenon by which an NP in the dependent clause controls the agreement of the matrix verb. This phenomenon appears to be closely related to raising; consider this example from Qunqi Dargwa: Qunqi Dargwa (East Caucasian, Dargwa) d=iχ-ub-ak:u (16) dammij pikri b=iχ-ub-ak:u / I.DAT think N=become-PST-NEG NPL=become-PST-NEG [gilad neʁ˳ d=erč:-ni]. child.PL.ERG soup NPL=eat-MSD ‘I didn’t notice that the children had eaten up the soup.’ In (16), the matrix verb pikri biχub ‘to notice’ can be marked as neuter singular (agreeing with the situation encoded by the lower clause). The marker of neuter plural is also possible (agreeing with the NP neʁ˳, as substances in Dargwa trigger neuter plural agreement). LDA is widely attested cross-linguistically, especially in head marking languages.6 It has been documented in languages of Daghestan: Dargwa, Tsakhur, Godoberi, Bagvalal, etc. (Kibrik 2003), Tsez (Polinsky 2000); in the Indo-Aryan languages Hindi (Butt 1993) and Kashmiri (Hook, Kaul 1987); in Itelmen (Bobaljik, Wurmbrand 2005); and in languages of North America, such as Seri (Hokan), members of the Algonquian family (Blackfoot, Passamaquoddy, Cree), the Wakashan family (Kwakwala), and others (cf. Bruening 2001). In Section 3, I show that LDA constructions show the same semantic and pragmatic properties as raising constructions. 2.3. Cross-linguistic parameters of raising On the basis of the patterns described in 2.2, I analyze the following parameters of raising constructions:

Towards the typology of raising

279

1. Raising effects. The raised status of the analyzed NP can affect either its case marking, as in English and Japanese, its agreement pattern (LDA), or its constituency properties. There are a number of constructions where no morphological evidence for raising can be observed, yet there is evidence that raising to the matrix clause has occurred (e.g. in Malagasy, Keenan 1976; Cuzco Quechua, Muysken and Lefebvre 1988; and KomiZyrian and Mari, Serdobol’skaja 2005). 2. Grammatical role of the raised NP. As shown in 2.2, in many languages it is not only the subject of the dependent clause that can be raised (as in English). In ergative languages, it is usually the NP in absolutive case that triggers various kinds of agreement, including LDA. 3. Dependent clause type. The dependent clause types that allow raising include infinitival clauses (English), nominalizations (Turkic), masdars (Tsez), and finite clauses with complementizers (Nieuan). 4. Syntactic function of the dependent clause. In most languages in the present sample, restrictions on raising are imposed, depending on the grammatical role of the lower clause within the main clause. For example, in many ergative languages, only dependent clauses that occupy the place of an absolutive argument can host LDA. 5. Number of NPs that can be raised at a time. 6. Availability of an alternative construction (a construction without raising/LDA). 7. Constituency tests (see 2.1) either show that the raised NP belongs to the matrix clause (M in the table below), or to the dependent clause (D); it may also be that different constituency test contradict each other (M/D). 8. Structural tests (see 2.1) that reveal the structural properties of the NP. For example, the passivization test (mentioned above) shows that the NP in question occupies the position of direct object of the main verb (the abbreviation DO in the table); the reflexivization (or reciprocal) tests only show that the NP occupies a position in a matrix clause, and it remains unclear which position this is (M in the table). The failure of these test shows that the NP remains in the lower clause (D). 9. Matrix verb type. In some languages, raising is restricted according to the matrix verb’s type (factives only, or mental verbs only, etc.). If no restrictions can be observed, the matrix verb type is marked as ‘various’. The cross-linguistic parameters of raising are summarized in Table 1 (see Serdobol’skaja [2005] for detailed discussion).

280

Natalia Serdobolskaya

agr., wo

Malagasy

case, wo/ subj wo case A, P, S

Niuean Passamaquoddy Tuvan

agr., wo case

Khakas

case, wo

Kalmyk

case, wo

Komi-Zyrian wo

Matrix verb type

Berber

Structural properties

case, wo

Constituency tests

Japanese

inf.

Altern. constructions

subj

Number of NPs

case, wo

Depend. clause position

Case/ LDA / word order

English

Depend. clause marking

Languages

Grammatical role of the raised NP

Table 1. Typological parameters of raising (based on Serdobol’skaja 2005)

7 DO/S mental M mental

DO/subj 1

yes M

subj, dat comp

DO, ?

1

yes M

subj

comp

DO/subj 1

yes M

DO

comp

DO/subj 1

yes ?

DO/S mental

comp

subj

1

yes ?

S

?ABS

1

yes D

DO

1

mental

aspectual, modal, phasal ?ABS mental

A, S, P, comp ? subj comp, nmz subj comp, nmz subj nmz, conv, comp subj nmz

DO, IndO various

1 1

yes M, D M, D perception, tell yes M D perception, non-factives yes M, D ? various

various

1

yes M, D M, D various

Mari

wo

subj

nmz

various

1

yes M, D M, D various

Tsez

agr.

abs

1

yes D

D

various

Dargwa

agr., wo

abs

comp, ABS nmz inf. various

1

yes/ ? 8 no

?

Tsakhur

agr.

abs

1

yes/ ? no

?

Bagvalal

agr., wo

abs

1

?

Blackfoot

agr.

1

?

Cree

agr., wo

A, S, P, nmz, ?ABS obl comp A, S subor- subj dinator

yes/ ? no yes D

1

yes ?

?

cause, phasal, wait, like, seem, modal want, forget, like, know, need, learn modal, want, phasal want, mental, perception non-factives

msd, ABS comp, paren. inf. ABS

Towards the typology of raising

Matrix verb type

Structural properties

Constituency tests

Altern. constructions

Number of NPs

Depend. clause position

Depend. clause marking

Grammatical role of the raised NP

Typological parameters of raising (based on Serdobol’skaja 2005)

Case/ LDA / word order

Languages

Table 1 continued.

281

Kipsigis

case, agr, A, S, P, comp, DO wo obl paren.

1

yes ?

?

Tagalog

case, wo

1

yes ?

?

Kurdish

sub- subj junct. inf. subj

1

yes ?

?

Hindi

case, P, S/ LDA, wo A, P, S agr., wo DO

believe, tell, perception, cause, want modal, want, avoid, mental mental

1

yes ?

?

modal

Kashmiri

agr., ?wo DO

inf.

various

1

?

Itelmen

agr.

DO, ?

1

yes ?D

Quechua

case

DO

>1 yes ?

Irish

case, wo

?DO aspect., modal, phase, cause, want, purp. ? cause, modal, forget, want M perception, mental, want ? various

Modern Greek case

ABS, obl comp TOP

DO, IO, inf. obl A, S, P, nmz obl subj, DO nmz subj

various

?M

>1 yes/ ? no sub- adverbial 1 yes ? junct.

?

preposition ‘with’

On the basis of the results given in Table 1, the following types of constructions can be identified: 1. Raising to subject/direct object (English, Japanese, Berber, Malagasy, Nieuan, and, possibly, Passamaquoddy). In the constructions of this type the raised NP has most of the morphological, constituency and structural properties of the matrix verb’s subject/direct object. 2. Raising to the left periphery of the matrix clause. The raised NP has structural and constituent properties of an element of the matrix clause. It receives morphological marking either in the matrix clause (Tuvan, Kalmyk, Turkish, and, possibly, Cuzco Quechua) or in the lower clause (Komi-Zyrian, Mari). 3. Raising to the left periphery position inside the dependent clause (Tsez, and possibly Irish, Itelmen; Bobaljik and Wurmbrand 2005). The

282

Natalia Serdobolskaya

NP in question has no properties of an element of the matrix clause – neither in structure nor constituency. 4. Clause reduction (Qunqi Dargwa, Kashmiri, and, possibly, Tsakhur, Bagvalal, and Hindi). Noonan (1985) defines clause reduction as a complementation construction in which the dependent verb has a reduced set of grammatical relations. These constructions are very similar to raising; however, they differ in the following way. In raising, it is presupposed that the ‘raised’ NP originates as an argument in the lower clause. Clause reduction is a construction where the ‘raised’ NP, in fact, originates as the matrix verb’s argument – which is coreferential to a participant that takes place in the situation encoded in the lower clause (e.g. in Nell made Dudley test the wort (Noonan 1985, ex. 137) the NP Dudley is not ‘raised’, but originates as the direct object of the matrix predicate). Clause reduction is mostly restricted to one-place predicates of phase, modality, aspect, and sometimes occurs with verbs of perception and volition. No conclusions can be drawn for a number of languages in the sample (Berber, Khakas, Blackfoot, Cree, Kipsigis, Tagalog, and Kurdish), due to the lack of information on constituency and structural tests in those languages.

3. Semantics and pragmatics of raising It is usually claimed that the raised NP is not semantically associated with the matrix verb (see, e.g., Postal 1974, Kuno 1976, Lasnik and Saito 1991). The absence of semantic shift via raising is used as a diagnostic that distinguishes raising from infinitival control. However, this claim has been disproven even for English (Kiparsky and Kiparsky 1971, Borkin 1973, Pesetsky 1991, Langacker 1995): the choice of the raising construction or that-clause may be strongly influenced by the semantics of the matrix verb: English (Germanic, Indo-European) (17) a. I find that this chair is uncomfortable. b. I find this chair to be uncomfortable. (Langacker 1995: 5) According to (Langacker 1995), (17a) denotes a situation in which the speaker bases his/her judgement on other people’s impressions, e.g. in a survey of customers’ polls, while (17b) is chosen if the speaker him/herself found the chair uncomfortable. Hence, it would be incorrect to

Towards the typology of raising

283

conclude that the semantics of the matrix verb plays no role in the choice of a raising/non-raising construction. In some languages, pragmatic properties of the raised NP determine the choice of the construction. In this section, I consider different factors that influence this choice in the languages of my sample. For many languages, however, the relevant data are lacking; thus the conclusions drawn here are based on a relatively small number of languages. 3.1. Matrix verb class The semantic type of the matrix verb is clearly relevant (see Table 1); many languages impose restrictions on raising with respect to matrix verb type. The following hierarchy of matrix verb types associated with raising can be proposed: mental verbs > want > perception, modal verbs > phasal, speech, emotive verbs > aspectual verbs, wait, causal verbs Figure 1. Hierarchy of verb classes

The probability of raising in the languages of the sample increases from right to left along this hierarchy. However, it is not an implicational hierarchy, but only an indication of the relative frequency of raising among languages: if a language accepts raising with, for example, the verb want, it does not follow that it accepts raising with mental verbs. The continuity condition does not hold, either: that a language accepts raising with want and phasal verbs does not indicate that it accepts raising with modal verbs (see Table 1). Hence, what we have here is a frequency pattern, which seems to reflect a complex set of factors. The types of matrix verbs that allow raising depend on the type of construction: clause union most often involves phasal and modal (or, more rarely, perception) verbs, while raising to direct object most often occurs with mental verbs. Apparently idiosyncratic lexical semantics of the verbs seem to play a role as well: for example, verbs that mean wait more often allow raising than hope, and notice is more likely to allow raising than see.

284

Natalia Serdobolskaya

3.2. Referential properties of the raised NP The referential properties of the NP to be raised appear to play an important role in determining whether raising can occur. For example, in Tuvan only specific NPs can be raised: Tuvan (Altaic, Turkic) košel’ok čidiri-p (18) a. Kǝm-nǝ lose-CONV who-ACC purse tur-gan-ǝn esker-di-ŋ? stand-NMZ.PST-ACC.POSS.3 notice-PST-2SG ‘Whom did you notice lose his purse?’ [It is known that the hearer did notice someone lose his purse.] The speaker presupposes that the hearer has seen the situation denoted by the lower clause. Hence, the subject of the dependent clause is specific and indefinite. When it is non-specific, on the other hand, raising does not occur, and the subject of the lower clauses is marked with genitive or nominative case: (18) b. Kǝm-nǝŋ košel’ok čidiri-p who-GEN purse lose-CONV tur-gan-ǝn esker-di-ŋ? stand-NMZ.PST-ACC.POSS.3 notice-PST-2SG ‘Did you notice anyone lose his purse?’ [No presupposition] The choice of the construction in Tuvan is thus determined by the specificity of the lower clause subject. In Khakas, it is definiteness that is relevant (E. Kalinina, p.c.): if the subject of the lower clause is definite, raising is preferred, and otherwise the non-raising construction is chosen. LDA constructions demonstrate the same pattern: in many languages, only specific/definite NPs can trigger LDA. Let us consider the following: Blackfoot (Algic, Algonquian; Frantz 1978: 102) (19) nits-íksstaa n-áxk-sskonak-ssi áattsistaai. 1-want(INTR) 1-able-shoot-COMP hare ‘I want to shoot rabbit(s).’ Non-specific NPs cannot trigger LDA, and they do not participate in agreement and transitivity marking: the verb in (19) is marked as intransi-

Towards the typology of raising

285

tive. Specific NPs can trigger LDA, whereas the verb takes the marker of transitivity and direct/obviative marking: Blackfoot (Algic, Algonquian; Frantz 1978: 90) (20) nit-wikIxtatw-a:-wa [n-oxkó-wa m-áxk-a’po’taki-xsi]. 3-might-work-COMP 1-want.TR-DIR-3 my-son-3 ‘I want my son to work.’ The same phenomenon is observed in Hindi (Butt 1993), Kashmiri (Hook and Kaul 1987), Itelmen (Bobaljik and Wurmbrand 2005), Tsez (Polinsky 2000), Cree (James 1984), and Passamaquoddy (Bruening 2001). However, for most languages such a restriction is not a property of LDA, but a restriction on agreement in general; e.g. in Cree and Passamaquoddy it is only the proxy that can trigger agreement on the verb. The same holds for specific NPs in Kashmiri and Blackfoot. Thus, the acceptability or probability of raising increases when moving from left to right on the following scale: definite >> indefinite specific >> non-specific9 Figure 2. Hierarchy of the raised NP properties

3.3. Raising and information structure of the sentence In many languages raising is triggered by topicality. For example, in Tuvan raising is unacceptable if the subject of the lower clause is focused: Tuvan (Altaic, Turkic) buzup-kan-ǝn (21) Ajas (??Ajas-tǝ) soŋga-nǝ Ajas Ajas-ACC window-ACC break-NMZ-ACC.POSS.3 men bodu-m kөr-dү-m. I RFL-1SG see-PST-1SG {The teacher is scolding Ajas because he has broken the window. Ajas says that he was not responsible. Another pupil gets up and says:} ‘It was Ajas whom I saw break the window.’ The accusative case (indicating raising) is preferred if the raised NP is the topic of sentence. The same kind of restriction is observed in Tagalog

286

Natalia Serdobolskaya

(Nakamura 2000), Passamaquoddy (Bruening 2001), and Tsez (Polinsky 2000). Creider (1979: 8) gives similar evidence for English. However, Kuno (1976) claims that not only the topical, but the focused status of an NP triggers raising in English. In focused contexts, raising is preferable to a that-clause, as in, for example, Mary, I expect to come, but Bill, I don’t and It is Mary that I expect to come, cf. (Kuno 1976: 20, ex. 7). Accordingly, Kuno (1976) argues that the motive of raising is “…to make the constituent subject an element that is movable to the position usually reserved for the topic or the focus of the sentence”. Such a conclusion might lead to a modification of the claim above: The NP in question is the topic of the sentence > is focused > belongs to the same domain (topic or focus) as the rest of the dependent clause.

However, it is topicality that is more relevant for raising/LDA in Tuvan, Tsez, Passamaquoddy and Tagalog. I argue that the information structure parameter works differently between English and those languages. It correlates with the difference in the syntactic construction: the Tsez constructions are analyzed as movement to the left periphery, as are those in Tuvan. This difference indicates that it is not the more fine-grained information structure itself that is crucial for raising in English, but rather the ‘prominence’ or ‘relevance’ of the NP in the discourse. The same factor has been proposed by Ljutikova and Bonč-Osmolovskaja (1999) to account for LDA in Tsakhur. 3.4. Raising and animacy Animacy is one of the most important factors determining the choice of the raising construction in Kalmyk, Tuvan, and Finno-Ugric languages. More precisely, it is not animacy, but the Animacy Hierarchy (Silverstein 1976) that is important. For example, in Kalmyk personal pronouns do not allow the non-raising construction for some matrix verbs: Kalmyk (Altaic, Mongolic) ir-s-i-n’ (22) Bi čamagə / ??či I you.ACC you(NOM) come-PART-ACC-POSS.3 med-sən uga-v. know-PART COP.NEG-1 ‘I didn’t know you had arrived. [When did you arrive?]’

Towards the typology of raising

287

The treatment of proper names, NPs denoting humans, and animate non-humans varies among native speakers of Kalmyk. Some speakers allow both accusative and nominative cases for these types of NPs, while some only allow accusative for proper names and NPs denoting humans. Animals and non-animate NPs most often occur in nominative case; however, the accusative is also possible: Kalmyk (Altaic, Mongolic) dala üsə ög-dgin’ (23) čini ükər / ükər-igə your cow cow-ACC so milk give-PART med-sən uga-v. know-PART COP.NEG-1 ‘I’m surprised (I didn’t know) that your cow gives so much milk.’ Accusative case with non-animates occurs most rarely in Kalmyk. Native speakers draw the line differently in what concerns the acceptability/preferability of the accusative; however, the Animacy Hierarchy is respected in all the variants (the acceptability of raising increases from left to right of this scale): pronouns > proper nouns > people > animals > inanimate Figure 3. Animacy Hierarchy

The Animacy Hierarchy also plays a crucial role in determining the distribution of genitive/nominative subjects in nominalizations in Mari, as shown in Table 2 (figures based on approx. 1200 examples; see Serdobol’skaja 2005 for data and discussion). Table 2. Frequency of genitive/nominative in Mari nominalizations Case GEN, % NOM, %

Personal pronouns, proper names > 95 5

NPs denoting humans > 93 7

Other animate NPs > 77 23

Inanimate NPs 43 57

The animacy of the noun to be raised also influences the choice of the raising construction in Komi-Zyrian and Tuvan (Serdobol’skaja 2005). The relevance of animacy to raising could be an areal or a genetic feature, since these languages belong to language families (Uralic and Altaic) that have a number of features in common. Similarly, animacy influences the

288

Natalia Serdobolskaya

choice of the construction in Blackfoot; however, it is a restriction on agreement in general. Therefore, it is not clear whether this parameter is important cross-linguistically; more information is needed for many of the languages discussed in this chapter. 3.5. Raising and idiosyncratic lexical semantics of the matrix verb We return here to the argument that raising may indeed have an effect on the semantics of the matrix verb. In English, many examples exist where the change of a that-clause to raising yields idiosyncratic semantic shifts in the matrix verb, cf. (17a) and (b) above. The following difference in evaluation is observed: by raising, the evaluation is understood to be made directly by the speaker, while in non-raising construction, mediated evaluation is implied. Similarly, in Japanese and Blackfoot, raising seems to be somehow connected to the speaker’s emotions and/or expectations, although the data are not sufficient to clarify this. Consider the following examples: Blackfoot (Algic, Algonquian; Frantz 1978: 96) (24) nít-ssksiniixpa kí’sa ot-áyo’kaa-xsi. 1-know(INTR) your-brother 3-sleeping-COMP ‘I know your (older) brother is sleeping.’ In Blackfoot, raising is preferred if the speaker wants the lower clause to evoke an emotion. Hence, it is more likely to occur in (25) than in (24): Blackfoot (Algic, Algonquian; Frantz 1978: 97) (25) nít-ssksino-a-wa kí’sa ot-oksiná’s-si. 1-know(TR.AN)-DIR-3 your-brother 3-cranky-COMP ‘I know that your (older) brother is cranky.’ We therefore conclude that in some languages the choice between raising and non-raising relates to idiosyncratic semantic properties of the matrix verb.

Towards the typology of raising

289

3.6. Generalizations The distribution of the semantic and pragmatic factors that trigger raising correlates with the syntactic types outlined in Section 2.3, as illustrated in Table 3. Table 3. Semantics and pragmatics of raising Types of raising

Languages

Matrix Information Referential verb structure properties semantics

Animacy

Raising to direct object Raising to left periphery of lower clause Raising to left periphery of matrix clause

English Japanese Tsez

+ + -

+ ? +

+ +

-

Tuvan Komi-Zyrian Mari Kalmyk Hindi Kashmiri Dargwa Tsakhur Tagalog Blackfoot

? ? +

+ + + + ? ? + + + ?

+ + + + + + + ? ? +

+ + + + +

Clause union

?

Idiosyncratic semantic properties of the matrix verb trigger raising to direct object/subject, as in English and Japanese. Pragmatic and discourse factors seem to be relevant for nearly all the types of raising considered. However, it is topicality proper that is most relevant for the second and third types of raising in the table, while for English and Tsakhur this factor works differently. The second and the third type are exactly the constructions for which raising to left periphery has been postulated. As for restrictions on referential properties of the raised NP, they often follow agreement/case marking rules in particular languages.

290

Natalia Serdobolskaya

4. Conclusions The chapter considers the phenomenon of raising, traditionally viewed as purely ‘syntactic’, from a functional-typological perspective. Taking crosslinguistic data into account leads to the conclusion that the definition of raising should be reformulated. As I have proposed here, at least three distinct types of constructions, attested cross-linguistically, are similar to raising as usually understood: these are raising to direct object/subject, raising to the left periphery of the lower clause, and raising to the left periphery of the matrix clause. Still another type, constructions involving clause reduction, should be analyzed separately from raising proper (although this demonstrates syntactic properties very similar to raising). Rather than being a purely syntactic phenomenon, raising can be triggered by the topicality, definiteness, or animacy of the NP to be raised, or the idiosyncratic lexical semantics of the matrix verb. Differences between syntactic types of constructions correlate with the types of semanticopragmatic factors that trigger raising.

Notes 1.

2.

3.

4. 5.

I would like to thank S. Toldova, E. Kalinina, E. Rudnickaja, M. Daniel, S. Minor, and the participants of the 4th and 5th Conferences in Typology and Grammar in Saint-Petersburg (2007, 2008) for the insightful discussion of different points of this chapter. I am grateful to Denis Creissels, Frank Seifart, and Pattie Epps for their accurate and helpful reviews. I use the term ‘raising’ (rather than ‘exceptional case marking’, etc.), following the terminology of typological works. This chapter focuses on raising from clauses; possessor raising is not considered. By ‘raising’ I mean both raising to object and raising to subject; however, for the sake of brevity, the data on raising to subject are omitted from this discussion. The data come partly from reference articles and grammars, partly from fieldwork (Komi-Zyrian, Mari (fieldtrips with the MSU team headed by A.I. Kuznecova, E. Kalinina, S. Toldova), Kalmyk (Saint-Petersburg University team headed by V. Vydrin, E. Perexval’skaja), Dargwa (a project with N. Sumbatova, D. Ganenkov, supported by the Russian Fund for Humanities, grant № 07-04-00266а), Tuvan (work with Chojgana and Ojuna Ojun), and Uzbek (work with Šerzod Tašpulatov). This analysis has been refined, cf. Yoon (2003), Ohta (1997). Abbreviations in glossing: NOM nominative, ERG ergative, ACC accusative, GEN genitive, DAT dative, OBL oblique, LOC locative, INSTR instrumental, EL

Towards the typology of raising

6.

7.

8. 9.

291

elative, BEN benefactive, SG singular, PL plural, SJ subject, OBJ object, PRS present, PST past, (I)PF (im)perfect, COP copula, PART participial, INF infinitive, CONV converb, MSD masdar, COMP complementizer, PTCL particle, ART article, NEG negative, TOP topic, AN animate, POSS possessive, AUX auxiliary, (IN)TR (in)transitive, DIR directive, RFL reflexive. It should be pointed out, as Polinsky (2002) has done, that many of these examples are not cases of LDA proper. Similar effects can arise due to infinitival control, raising (see the discussion of Passamaquoddy data in Bruening 2001), clause union (which probably explains patterns in Hindi, Godoberi and Bagvalal), etc. Polinsky (2002) argues that Tsez exhibits LDA proper: the noun phrase that triggers LDA is not an argument of the matrix verb, and there is evidence that LDA in Tsez is neither infinitival control, raising, nor clause union. M indicates that the raised NP shows the properties of an element of the matrix clause; S/DO/ABS – the properties of the matrix verb’s subject/DO/absolutive argument; D – the properties of an element of the dependent clause. The slash symbol here means that the situation differs for different matrix verbs. Generic NPs behave either as definite, or as non-specific NPs (cf. Frantz 1978).

References Bobaljik, Jonathan David, and Susi Wurmbrand 2005 The domain of agreement. Natural Language and Linguistic Theory, 23(4): 809–866. Borkin, Ann 1973 To be or not to be. Chicago Linguistic Society 9: 44–56. Bruening, Benjamin 2001 Syntax at the edge: Cross-clausal phenomena and the syntax of Passamaquoddy. Ph.D. Diss., MIT. Butt, Miriam 1993 A reanalysis of long distance agreement in Urdu. In Proceedings of the 19th Annual Meeting of the Berkeley Linguistic Society, 52–63. Berkeley, CA. Carnie, Andrew, and Heidi Harley 1997 PRO, the EPP and nominative case: Evidence from Irish infinitivals. Available at http://dingo.sbs.arizona.edu/carnie/publications/PDF/ UpennWP1paper.pdf.

292

Natalia Serdobolskaya

Creider, Chet A. 1979 On the explanation of transformations. In Discourse and Syntax (Syntax and Semantics, v. 12), J. M. Sadock (ed.), 3–23. New York, San Francisco, London: Academic Press. Davies, William D., and Stanley Dubinsky 2004 The Grammar of Raising and Control. London: Blackwell. Frantz, Donald G. 1978 Copying from complements in Blackfoot. In Linguistic Studies of Native Canada, E. Cook and J. Kaye (eds.), 89–109. University of British Columbia Press. Hook, Peter Edwin, and Vijay K. Kaul 1987 Case alteration, transitionality, and the adoption of direct objects in Kashmiri. Indian Linguistics 48: 52–69. Jake, Janice, and David Odden 1979 Raising in Kipsigis. Studies in the Linguistic Sciences 9(2): 131– 155. James, Deborah 1984 Raising to subject in Moose Cree: A problem for subjacency. Syntax and Semantics, 16: 205–213. New York, San Francisco, London: Academic Press. Joseph, Brian D. 1990 Is raising to prepositional object a natural language grammatical construction? In Studies in Relational Grammar, vol. 3, P. M. Postal and B. D. Joseph (eds.), 261–276. Chicago: University of Chicago Press. Keenan, Edward L. 1976 Remarkable subjects in Malagasy. In Subject and Topic, Charles Li (ed.), 247–301. New York: Academic Press. Kibrik, Aleksandr E. 2003 Konstanty i peremennyje jazyka. [Constants and Variables of Language]. Saint-Petersburg: Aleteja. Kiparsky, Paul, and Carol Kiparsky 1971 Fact. In Semantics: An Interdisciplinary Reader, L. Jakobovits and D. Steinberg (eds.), 345–369. Cambridge: Cambridge University Press. Kuno, Susumo 1976 Subject raising. In Japanese Generative Grammar, M. Shibatani (ed.), 17–49. (Syntax and Semantics, v. 5.) New York: Academic Press. Langacker, Ronald W. 1995 Raising and transparency. Language 71: 1–62.

Towards the typology of raising

293

Lasnik, Howard, and Mamoru Saito 1991 On the subject of infinitives. Chicago Linguistic Society 27: 324– 343. Ljutikova, Ekaterina A., and Anastasija A. Bonč-Osmolovskaja 1999 Aktantnye predlozhenija. In Elementy caxurskogo jazyka v tipologičeskom osveščenii, A.E. Kibrik (ed.), 481-537. Moscow: Nasledije. Mulder, Jean 1976 Raising in Turkish. In Proceedings of the Berkeley linguistic society (Annual meeting 2), 298–307. Berkeley, CA. Muravyova, Irina A. 1992 Unmarked noun form in Turkic languages: a typological point of view. In Proceedings of the 33rd Meeting of the Permanent International Altaistic Conference, 257–261. Budapest. Muysken, Pieter, and Claire Lefebvre 1988 Mixed Categories: Nominalizations in Quechua. (Studies in Natural Language and Linguistic Theory.) Dordrecht/Boston/London: Kluwer Academic Publishers. Nakamura, Masanori 2000 On raising in Tagalog. Lingua 110: 391–408. Noonan, Michael 1985 Complementation. In Language Typology and Syntactic Description (v. 2), T. Shopen (ed.), 42–140. Cambridge: Cambridge University Press. Ohta, Kaoru 1997 Tense in the subject raising construction. Japanese / Korean Linguistics 6: 353–369. CSLI Publications. Perlmutter, David M., and Paul M. Postal 1983 The relational succession law. In Studies in Relational Grammar, vol. 1, D. M. Perlmutter (ed.), 30–80. Chicago and London: University of Chicago Press. Pesetsky, David 1991 Zero syntax, vol. 2: Infinitives. Ms. Polinsky, Maria 2002 Non-canonical agreement is canonical. Talk given at Agreement Workshop, UMIST, Sept. 17, 2002. 2000 Variation in complementation constructions: Long-distance agreement in Tsez. In Complementation, K. Horie (ed.), 59–90. Amsterdam: John Benjamins. Postal, Paul M. 1974 On Raising. Cambridge, MA and London: MIT Press. Postal, Paul M. 1986 Why Irish raising is not anomalous. Natural Language and Linguistic Theory 4: 333–356.

294

Natalia Serdobolskaya

Sanžeev, Garma 1960 Sovremennyj mongol’skij jazyk. Moscow: Izdatel’stvo vostočnoj literatury. Seiter, William J. 1983 Subject–direct object raising in Niuean. In Studies in Relational Grammar, vol. 1, D.M. Perlmutter (ed.), 317–359. Chicago and London: University of Chicago Press. Serdobolskaya [Serdobol’skaja], Natalia V. 2005 Sintaksičeskij status aktantov zavisimoj nefinitnoj predikacii [Syntactic status of the arguments of non-finite dependent clauses]. Ph.D. diss., Moscow State University, Moscow. 2006 Accusative subjects in Tuvinian nominalizations: against subject-toobject raising analysis. Presented at the Workshop on Formal Approaches to Altaic Languages, Moscow State University, Moscow. Stenson, Nancy 1981 Studies in Irish Syntax. Tübingen: Günter Verlag. Testelec, Jakov G. 2001 Vvedenie v obščij sintaksis. Moscow: RSUH. Yoon, James H. 2003 Raising specifiers: a macroparametric account of SOR in some Altaic languages. Presented at the Workshop on Formal Approaches to Altaic Languages, MIT, Cambridge: MA.

Historical pathways in Northern Paiute verb formation Tim Thornes

1. Introduction1 Northern Paiute (Uto-Aztecan, Western Numic) has a rich verbal morphology that serves to inform grammaticalization theory well by demonstrating numerous interesting micro-level variations and historical developments. Two prevalent verb constructions in Northern Paiute, the instrumental prefix construction (IPC) and the secondary verb construction (SVC), have evolved to contribute much of the morphological complexity one finds in verbs. Both the IPC and the SVC serve to modify the core predicating element, but their pathways of historical development are quite different. This chapter traces several steps along these pathways through language-internal clues, considering comparative evidence and informed by typological expectations. The IPC is clearly patterned after compounding in the language, which, I argue, is its historical precursor. The frequent licensing of agents in the construction provides the context for the development of at least one morphological causative from among the inventory of instrumental prefixes. The SVC represents, at the very least, a distinct type of compounding, and it shares many typologically relevant features associated with verb serialization in other languages.2 It corresponds well with what others have named the nuclear serial verb construction (Foley and Olson 1985), coincident motion or posture serialization (Durie 1997: 336), and the asymmetrical serial verb construction (Aikhenvald 2006), on formal, functional, and developmental grounds. The Northern Paiute SVC is arguably the source construction for much of the verb’s directional and aspectual morphology. I also suggest that the SVC itself arose from a marked adverbial clause construction coding event simultaneity, through the loss of subordinate morphology. Although many of the properties of these construction types in Northern Paiute are neither unknown nor rare phenomena cross-linguistically, the data I bring to bear on their historical pathways of development elucidate

296

Tim Thornes

micro-level processes of grammaticalization that can only be fully appreciated through the intensive treatment and understanding of the workings of a single language. My aim is to contribute to the discussion of how descriptive fieldwork, particularly intensive work on one language, both informs typological generalizations and hones our understanding of processes and sub-processes in grammaticalization. For each of the constructions in question, I look both backward to evident precursors and forward to the ongoing developments of individual elements that have played a part in their evolution. Such work informs typology most obviously by shedding light on types and subtypes of language-specific grammatical phenomena by detailing their similarities and differences with respect to other, perhaps better known, phenomena. Sub-processes of grammaticalization can also be best understood within the context of a detailed study of variation in the properties of particular constructions in a single language. A detailed synchronic description of such variation in fact opens the door to diachronic explanation in ways that are, to some extent, fundamentally different from those available through comparative work across related languages. Of course, statements of historical development must be corroborated by looking, where possible, at similar phenomena in related languages. More specifically, the construction types I describe in detail here inform our understanding of the various ways languages develop complex verbal structures. The development of distinct complex structures can be best understood, and explained, by focusing on their different paths of historical development (Givón 2006). As such, the present study also contributes to an ongoing conversation in the field regarding the relationship between diachronic explanation and synchronic description. The modern instantiation of this conversation, perhaps, begins with the “Saussurean prohibition against mixing synchrony and diachrony, bolstered by the Chomskyan argument that the language-learning child must construct their grammar without reference to anything but synchronic facts” (Evans and Dench 2006: 19). And yet, children clearly acquire a dynamic system, one that undergoes changes even within their own lifetimes, and so many grammarwriters embrace diachronic explanations for facts of language that simply cannot be accounted for otherwise (Rankin 2006). With respect to language-internal variation and in exploring related construction types, the role of historical change becomes, perhaps, even more relevant, since it is in this sphere that one finds clues about how language change actually occurs.

Northern Paiute verb formation

297

In this study of Northern Paiute verb morphology, I hope to demonstrate some of the ways in which diachrony and synchrony are intertwined in coherent and detailed language description.

2. Language status and typological profile There are, at present, around 400 speakers of Northern Paiute, living mainly in and around scattered reservation communities throughout the northwest third of the Great Basin region of North America, including the states of Nevada, Oregon, Idaho and California. In the majority of these communities, the language is moribund, often spoken only by those over sixty years of age. The language is fairly uniform over much of this vast area, with a major south-north dialect division defined mainly by the presence versus absence of voiced fortis (or geminate) stops, respectively. Nichols (1974: 4) refers to the southern dialect area as Nevada Northern Paiute, and the northern dialect area as Oregon Northern Paiute. The data presented in this chapter come exclusively from Oregon Northern Paiute. Northern Paiute has most of the typical features of an SOV language, including postpositions and prehead modifiers, although there is significant word order flexibility that can be attributed to pragmatic factors (Thornes 2003: 258–262). The language has two core cases, marked on pronouns and noun phrase dependents (including determiner proclitics), following a nominative-accusative pattern (Thornes 2003: 135–143). It also makes extensive use of nominalizations and converbal constructions in complex clauses and as a way of presenting background information in discourse (Thornes 2003: 466–472). The verb is the most complex word class morphologically, and suffixes predominate over prefixes. An overview of Northern Paiute verb structure is represented in (1) and exemplified in (2– 4).3 (1) (2)

PRO= [Adverbial [[Valence [IP/[Root]] Valence] +DIR/ASP] +SUB/NMR] Object Manner THEME STEM Prefinal Final

na-

tɨ-

guhanni -kɨ -wɨnɨ -dɨ MM- APS- IP/fire- do -APL -CONT.SG -NMR4 ‘(s/he) who is cooking for (her/him)self’ (SL: 39)5

298 (3)

Tim Thornes

su=tɨɨtsi-ʔyu

naatsi pinoyo -katɨ -ga -wɨnɨ boy IP/butt- move -sit.SG -TRNSL -CONT.SG ‘The little boy is crawling away backwards on his butt.’

NOM=small-NOM

(4)

nɨ mɨ= misu -makwɨ -u -kwɨ I PL= can/easily -defeat/finish -PNC -FUT ‘I will easily defeat you all.’

The core, or stem layer, may be mono- or bi-morphemic, depending upon the absence or presence of instrumental prefixes (IP, in examples 2 and 3). The theme layer includes valence-changing operators. Two such operators are prefixes and perform typical detransitivizing functions (the middle marker [MM] and the antipassive [APS] in 2), and the others are suffixes and increase the number of arguments required by the verb (for example, the applicative [APL] in 2). A limited set of four manner adverbials can be incorporated into the verbal word as well, appearing immediately following the slot for object pronominal proclitics (misu in example 4). Secondary verbs, of the type described in detail below, follow the verb theme (-katɨ ‘sit.SG’ in example 3), as do the directional and aspectual suffixes they give rise to in the the Prefinal zone. Subordinating and nominalizing suffixes appear in final position, but do not preclude finite morphology like aspect-marking (Thornes 2006: 130). I begin in Section 3 with the instrumental prefix construction (IPC) by describing its synchronic properties and then discussing its historical developments. This is followed in Section 4 by a discussion of the secondary verb construction (SVC), presenting a general description followed by an illustration of the steps taken in the grammaticalization of specific verbs. I end with some intriguing evidence relating to a possible source construction for the SVC before concluding in Section 5.

3. Instrumental prefix construction 3.1. General properties The instrumental prefix construction (IPC) involves a verb root in combination with any of a closed set of around two dozen prefixes whose semantic force in the construction most commonly indicates something of the means or manner by which the action coded by the root is carried out. Table 1 is an inventory of instrumental prefixes in Northern Paiute.

Northern Paiute verb formation

299

Table 1. Northern Paiute instrumental prefixes PREFIX kɨ“ku(“)ku/ko(“)ma(“)mu(“)ni(“)nopa(“)pi(“)pusɨ(“)su(“)tata(“)tɨ(“)to(“)tsa “tsi(“)tso(“)wɨ(“)-

GLOSS ‘bite’ ‘fire’ ‘face’ ‘hand’ ‘nose’ ‘speech’ ‘load’ ‘water’ ‘butt’ ‘eye’ ‘cold’ ‘mind’ ‘sun’ ‘foot’ ‘rock’ ‘fist’ ‘grasp’ ‘sharp’ ‘head’ ‘long’

FULLER MEANING with the teeth; by biting; edge with heat; regarding fire, smoke face, neck, above, in front with the hand or arm; causative having a point; regarding the nose/beak, lips/mouth with speech; by talking egg / house / round; move / carry pertaining to water, moisture, liquid with respect to the hips or buttocks as location/direction with the eyes; by seeing cold; fear; shaking mental or emotional activity sun / day / light; appearing with or regarding the feet/leg; stepping with or regarding rock; by forceful activity with the fist; axial motion; out from the body with the fingers; grasping; toward the body with a sharp object; with the end or tip of a long object with or regarding the head or shoulders along or against the length of a long object; radial motion; by means of natural forces, especially wind

Most prefixes carry a final feature, as it is called in the Numic literature (Nichols 1974: 13), which triggers the mutation (gemination or fortition in Northern Paiute, nasalization and preaspiration elsewhere in Numic) of a following consonant. The presence of such a feature is indicated by double quotes (“) in the first column after the prefix. Where the effect appears to be inconsistent, the quotes are enclosed in parentheses. One striking aspect of the IPC is the occurrence of paradigmatic derivational sets like those in (5), all carrying the same core meaning, but whose semantics are narrowed or specified by the prefix: (5)

-kwonau ‘open’ a. kɨ-kwonau b. ma-kwonau c. pi-kwonau d. to-kwonau

‘open with the mouth or teeth’ ‘open by pushing with the hand’ ‘open with the hips; by backing into’ ‘open using outward force exerted axially’

300

Tim Thornes

e. f. g. h.

tsa-kwonau tsi-kwonau tso-kwonau wɨ-kwonau

‘open grasping with the hand and fingers’ ‘open with the end or tip of something’ ‘open with the head’ ‘open using a long instrument or by wind’

Speakers give a variety of possible interpretations or contexts for the use of one or another prefix. Due to this broad interpretive range, I adopt a strategy of providing a one-word interlinear gloss in examples, as listed in the middle column of Table 1, for simple cross-reference to the fuller semantic characterization in the last column of the table. Instrumental prefix constructions are found in numerous unrelated languages and families of languages in North America, including Siouan, Plateau Penutian, Pomoan, and many others (see Mithun 1999: 118–126 for an excellent general discussion). The instrumental prefix construction is a pervasive derivational property in only the Numic subfamily within UtoAztecan.6 The most common members of instrumental prefix inventories in the languages that have them include those that appear to classify an instrument in terms of physical properties, like shape, and those that indicate the use of a body part. Indeed, the most frequently occurring instrumental prefixes are exactly these. Their broad semantic range may even include an associated gesture, as opposed to classifying the nature of the instrument. Consider the following pair: (6)

a. tsi-kwiduʔi IP/sharp-stir ‘stir (soup)’

b.

wɨ-kwiduʔi IP/long-stir ‘stir (gravy)’

Here the contrast is not in the actual instrument used – one could use a wooden spoon for either task – but in the nature of the action. In (6a), the end of the instrument is involved and the spoon is held vertically, while in (6b), the preparation of gravy requires contact along the edge or length of the instrument, held more horizontally in order to blend the thickener and prevent scorching. An important formal characteristic of the IPC is that it often consists only of bound elements, each of which contributes semantically to the overall stem. In these cases, the verb root never occurs without an instrumental prefix. Miller (1972) and others in Numic studies refer to these formally dependent verb roots as instrumental verbs (and represent them with a hyphen to indicate their dependent status). Jacobsen (1980: 85) de-

Northern Paiute verb formation

301

scribes this phenomenon in Washo as “the bipartite formations of lexical prefixes plus dependent verb stems [that] are exactly equivalent grammatically to... verb stems”. Example (7) (as well as 6 above) illustrates this formal pattern in Northern Paiute. (7)

-kaʔa ‘to cut / sever / split’ a. kɨ-kaʔa ‘bite in two’ b. tsi-kaʔa ‘cut / slice’ c. tsa-kaʔa ‘pull apart’ d. wɨ-kaʔa ‘cut off / chop’

An inventory of Northern Paiute instrumental verbs from my corpus, with an illustrative example for each, appears in Table 2. Table 2. Northern Paiute instrumental verbs INSTR.VERB -kyota -kaʔa -kibuʔi -koba -kwonao -ma -mɨna -nɨʔya -pata -pɨʔma -poka -posa -sɨʔyɨgɨ -taʔni -tabui -tatsi -tsaga

*GLOSS *lift *cut *tear *break *open *touch *turn over *bump *spread *pile up *separate *pierce *shake *knock *fix/prepare *strike *kill/destroy

IP + VERB ROOT ta-kyota tsa-kaʔa wɨ-kibuʔi wɨ-koba ma-kwonao tsa-ma tsa-mɨna tso-nɨʔya wɨ-pada wɨ-pɨʔma tsi-poka tsi-bosa tsa-sɨʔyɨgɨ wɨ-taʔni ma-tabui ma-datsi ma-tsaga

MEANING ‘lift with the foot’ ‘pull apart’ ‘tear apart in the wind ‘break by swinging’ ‘open with the hand’ ‘touch with the fingers’ ‘flip over’ ‘bump the head’ ‘spread out’ ‘sweep / pile up’ ‘slice to bits’ ‘pierce with a knife’ ‘shake (transitive)’ ‘whack’ ‘get dressed’ ‘slap’ ‘kill by pushing’

Instrumental verbs typically encode a change of state. The presence of a set of dependent verb roots suggests that the instrumental prefix construction itself developed to encode just such event types. In other words, this may well be the conceptual source of the IPC or of bipartite stems more generally, since such dependencies are likely to have arisen earlier than the

302

Tim Thornes

expanded use of instrumental prefixes with independent verb stems described below. Several languages of North America (e.g. Washo and Klamath) have a more extensive system of bipartite stem types.7 These languages include at least two subtypes apart from the instrumental prefix plus change-of-state combinations I have described for Northern Paiute. DeLancey (1998), for example, describes one of these subtypes as having one part (the lexical prefix) that classifies an object that moves or is located somewhere, and another part (the locative-directive stem) that expresses some feature of its motion (direction or location). In Klamath, Washo, and other languages of the area, there always appears to be a subsystem of instrumental prefix plus change-of-state predication. This fact supports not only the great age of the IPC in North America, but also suggests that the IPC represents an historical precursor to these more elaborate bipartite systems.8 Another notable feature of instrumental prefixes is the fact that they often derive transitive stems from otherwise independent, intransitive verbs through the addition of an agent to the argument structure. We can see this in pairs like pɨnuyui ‘spin (intransitive)’ and u=ma-pɨnuyui (3=IP/handspin) ‘spin it (transitive)’ as well as in the following: (8)

a. su=tokano nɨ tsipugi-si NOM=night I emerge-SEQ ‘Once I escape in the night …’ b. i=tsa-tsipuki me=IP/grasp-emerge ‘Let/Pull me out!’

This causative-like pattern is described for Klamath by DeLancey (1996b). Talmy (1985) refers to the set of instrumental prefixes in Atsugewi as “cause satellites” by way of describing this effect. There are, however, many independently occurring, usually transitive, verb roots whose argument structure remains unaffected by the addition of an instrumental prefix, as in wɨnai ‘throw’ and ma-wɨnai (IP/hand-throw) ‘push aside’ as well as the following: (9)

a. nɨ u=homai I it=miss ‘I miss it (shooting).’

b.

u=wɨ-homai it=IP/long-miss ‘miss a guess (in a bone game)’

Northern Paiute verb formation

303

These examples demonstrate how the effect of an instrumental prefix in verb stem formation is mainly semantic. It is presumably the causative pattern coupled with the frequency of use of instrumental prefixes involving the hand that has led to the development of a morphological causative in Northern Paiute and in the unrelated languages Lakhota (Rood and Taylor 1996) and Haida (Hori 1998). This development in Northern Paiute is described in more detail in Thornes (1996). Pairs like pahonayaʔi ‘be tired’ and i=ma-pahonayaʔi (1=IP.hand/CAUS-tired) ‘make me tired’ illustrate this pattern, as do the following: (10) a. uuni-ʔyu nɨ sɨda nɨɨma thus-NOM I bad feel ‘That’s why I feel badly.’ b. ɨ pisa i=ma-nɨɨma you good me=CAUS-feel ‘You make me happy.’ The source context for this historical development appears to be one in which intransitive verbs coding internal states enter into the IPC, particularly where the literal involvement of the hand is ruled out.9 Jansen and Beavert (2008) note the possibility that one of the causative prefixes in Yakima Sahaptin may be related to an instrumental prefix referring to the hand as well. Unlike the Northern Paiute case, the Yakima causative is productive with human actors as causees – including some organically transitive contexts. In the examples above, however, the role of the human causee is one of experiencer or patient, not secondary actor. This suggests that the Northern Paiute causative is simply not as far along the grammaticalization path. Heine (2002) refers to these different stages as the difference between a ‘switch context’, where the source meaning of ‘with the hand’ is backgrounded in favor of the causative meaning for Northern Paiute, and full ‘conventionalization’, where only the target causative meaning remains, as in Yakima. One further development within the IPC in Northern Paiute is worthy of note, since I have not found it attested elsewhere. The language has developed a productive same-subject desiderative prefix related to the instrumental prefix su- ‘of or relating to the mind’. It seems a reasonable parallel of the pattern we have already seen for causatives, but at less cost, since the inherent argument structure of the verb root is unaffected by it, whether the verb is intransitive, as in (11), or transitive, as in (12).

304

Tim Thornes

(11) nɨ kai su-kima=paana, kimma I NEG DESID-come=DISJUNCT come.DUR ‘I don’t want to come, but (I am) coming.’ (12) nɨ pɨnoʔo u=su-hani I as_well 3=DESID-do ‘I want to do that, too.’ At this point in its development, its distribution is fully productive, since it has fairly predictable semantics and does not have the narrow selectivity of distribution with independent verb stems. As an instrumental prefix, su- can be found lexicalized with numerous verb stems relating to mental or emotional activity. I turn now to the question of the most likely historical precursor to the Northern Paiute instrumental prefix construction – stem compounding. 3.2. Historical source of the instrumental prefix construction It is often suggested that the instrumental prefix construction in Northern Paiute and elsewhere must have developed from noun incorporation or some other “specialized form of incorporation” (Sapir 1930: 101). Productive body part incorporation or prefixation is reported for many languages outside North America, for example, in the Panoan languages of South America (e.g. Fleck 2006). However, the connection between individual prefixes and noun stems, where they can be found, is mainly an historical one, and does not account for all cases, especially those cases where the etymology for a particular prefix leads to a former verb stem, rather than a noun. Still, there are clear parallels that would suggest noun incorporation of the compounding type (Mithun 1984) as a source for the IPC. In the case of Northern Paiute pa- ‘of or pertaining to water’ and the independent root paa ‘water,’ noun incorporation and instrumental prefixation are virtually indistinguishable. This is the exception, and not the rule, however, and the IPC is very clearly distinct from synchronic stem compounding. Still, the patterns of compounding that parallel the IPC in the language are instructive. As mentioned, the most likely historical sources for some prefixes are verb stems. Some examples include the highly productive prefix tsa- ‘with the fingers; grasping’ which can be linked to the verb tsaʔi ‘to hold,’ ni- ‘of or pertaining to speech’ with niʔa ‘to call or name,’ as well as

Northern Paiute verb formation

305

kɨ- ‘with the teeth’ which is likely historically related to Proto-Uto-Aztecan kɨʔi ‘to bite’ (Dayley 1989: 93). One property of compound stems in Northern Paiute is that they are semantically right-headed.10 In Verb + Verb compounds, the first member of the compound is often interpretable as adding a manner feature to the complex – distinguishing the construction semantically from the secondary verb construction I explore in Section 4. The following examples illustrate Verb + Verb compounding in Northern Paiute:

(13) a. watsi-mia hide-go.SG ‘sneak off’

b. naka-supidiaʔi hear-heed ‘pay attention’

These examples are composed of simple, independent verb stems, with the second stem serving as the semantic head and the first serving in a more dependent, modifying capacity, setting the stage, theoretically, for further developments in the direction of instrumental prefixation. Both watsi ‘hide’ in (13a) and naka ‘hear’ in (13b) are very common elements of compounds, but with a more abstract or adverbial meaning in compounds – watsi refers to any action performed ‘secretly’ and with naka, as in (13b), it is often difficult to discern just what its semantic contribution to the compound stem is. There are also examples where the first half of the compound no longer occurs as an independent stem, a morpho-syntactic dependency akin to instrumental prefixes, thus suggesting one more stage in the development of the IPC. (14) a. sogo-mia on_foot-go.SG ‘go on foot (SG)’

b. natsɨ-maŋi by_raft-cross_water ‘cross by raft’

The bound initial elements of these examples could be considered incipient instrumental prefixes, contributing, on the one hand, adverbial or manner semantics to the head verb while, on the other, developing the earmark morphosyntactic dependencies of true lexical prefixes. Here, however, they are lacking in the greater productivity and semantic broadening associated with full members of the category. The main differences between these and instrumental prefixes is that their distribution appears restricted due to their fairly narrow semantic range. In the case of natsɨ-, this is the only form in the corpus that contains it. In the case of ‘true’ in-

306

Tim Thornes

strumental prefixes, their more generalized meanings accompanied their more varied or productive distribution with different verb roots. We have seen some language-internal evidence for how one type of productive word formation process developed in Northern Paiute from the pattern of X + Verb compounding, and how semantic compatibility and effects on argument structure can provide the context for the further developments of specific prefixes, as with the morphological causative described above. I turn now to another verb formation process in Northern Paiute, the secondary verb construction (SVC), and explore its properties and distinctive history.

4. Secondary verb construction 4.1. General properties The secondary verb construction (SVC) consists of a verb stem plus any of a restricted set of otherwise independent verbs that appear in the prefinal zone, following the theme layer of the overall verb structure presented in (1).11 As independent verbs, these include the basic posture verbs (sit, stand, lie), general motion verbs (travel, arrive), and a few others. As secondary verbs,12 they add postural semantics or associated motion to the complex, but also frequently carry the semantic force of aspectual or directional suffixes. Secondary verbs in Uto-Aztecan languages have been described as those which, “in addition to serving independently as primary verbs, may also be used almost like (optional) suffixes... in which capacity they indicate aspectual-like or adverbial-like meanings” (Crapo 1970: 182). Unlike the instrumental prefixes described in previous sections, secondary verbs are more difficult to categorize, since individual elements show differing degrees of grammaticalization. It is precisely this gradient property, however, that makes the SVC interesting for typology and grammaticalization theory and also informs the role of diachrony in synchronic description. The following examples illustrate forms of the independent verbs wɨnnɨ ‘stand’ and nɨmmi ‘travel; go around’, both singular forms (a subject taken up further below), and show these same forms functioning as aspectual (continuous) and directional (randomly directed motion) suffixes, respectively, in the secondary verb construction.

Northern Paiute verb formation

307

(15) a. toisu oʔo wɨnnɨ still DEM stand.SG ‘(He) is still standing out there.’ b. su=naatsi pida-wɨnɨ NOM=boy build_fire-CONT.SG ‘The boy is building a fire.’ (16) a. hanno ɨ nɨmmi ? where you travel.SG ‘Where have you been off to?’ b. tɨhoawai-nɨmi nɨ hunting-RNDM.SG I ‘I’ve been hunting around.’ In addition to verbs of these two major semantic types, posture and motion, there are a few others; e.g. tɨŋɨ which, as an independent verb, means ‘ask for; request’ (17a) and as a secondary verb means ‘tell to V’ (17b). (17) a. owitɨ

mooni na-tɨŋɨ-kwɨ money MM-request-FUT ‘(I) will get (request) money there (from a bank).’ b. ɨ i=tsi-kaʔa-tɨŋɨ-u you me=IP/sharp-cut-tell_to-PNC ‘You told me to cut it.’ DEM

Note that in the secondary verb capacity of (17b), the verb patterns syntactically like a causative by adding an agent to the argument structure of the main verb. One key formal criterion that differentiates the secondary verb construction from other types of Verb + Verb compounding in Northern Paiute is the position of the secondary verb relative to the applicative suffix -kɨ. Example (18) illustrates how the applicative suffix follows the second element of a typical compound – a property of the stem layer of verb structure. (18) ɨ=sakwa ka=toissapui i=sami-tɨkɨ-kɨ you=MOD OBL=chokecherry me=soak-put.SG-APL ‘You should put the chokecherries in to soak for me.’

308

Tim Thornes

In the SVC, however, we find the secondary verb following the applicative suffix. Example (19) demonstrates this for the secondary verb ‘tell to V’. (19) uuni-ku mɨ=wɨnɨ-kɨ-tɨŋɨ-yakwi that_kind-OBL PL=stand.SG-APL-tell_to-HAB ‘... in such a place (a corner, they) would tell them to stand.’ One way to view this morphosyntactic distinction is by considering the verb structure schema presented in (1). According to this schema, (18) above represents stem (or root) level compounding and (19) represents compounding at the level of the theme. My main point here is to lay out the distinctive properties of the secondary verb construction in order to clarify its unique historical developments. Semantic headedness also distinguishes the SVC from other types of compounding and from the IPC described in Section 3. Recall from Section 3.2 that compounds in Northern Paiute are generally right-headed – that is, the second member of the compound supplies the core semantics of the predicate, whereas the first qualifies these properties in some way. In the SVC, on the other hand, it is, I would argue, the first element that carries the core lexical-semantic content, with the secondary verb modifying that content.13 The composite semantics of the secondary verb construction also distinguishes it from typical compounding. Verbal compounds derive stems with lexical content that is more comparable to that of a simple verb. In the SVC, on the other hand, the subevents coded by the two verbs are coextensive and simultaneous, but their combination does not derive the expression of a new event type. The secondary verb construction may be best considered a subtype of serial verb construction, with respect to two key properties discussed in Aikhenvald’s (2006) typology: 1) it is assymmetrical; that is, it combines members from an open class with those of a closed class of verbs, and 2) the verbs in the construction are combined within what can properly be considered a single phonological word. More so than with the instrumental prefix construction, however, it is difficult to separate synchronic description from a discussion of diachronic developments in the case of the SVC, as we will see. The ensuing description of historical developments hinges on the fact that there is category gradience along the path from independent verb to inflectional suffix reflected in the secondary verb construction.

Northern Paiute verb formation

309

4.2. Historical developments in the secondary verb construction 4.2.1. Semantic bleaching In many cases, the core semantics of the secondary verb is retained in the SVC, as expected in the initial stage (Heine 2002) of grammaticalization. Examples (17b) and (19) above with -tɨŋɨ ‘request; tell to V’ illustrate this, as does (20) below with the suppletive postural form -tapi meaning ‘lie; do while lying down’. (20) su=nana iwi-tapi NOM=man sleep-lie.SPL ‘The man is lying (there) sleeping.’ Motion verbs also frequently retain their core semantics in the SVC, as with -kima ‘come’ and -pitɨ ‘arrive’ in the following: ɨni-kima (21) yaisi miu then QUOT say-come ‘... and so (she) came along calling out.’

(22) su=nɨmɨčozinna yotsi-pitɨ-u NOM=Nemechozinna fly.SG-arrive-PNC ‘Nemechozinna arrived flying.’ Examples like (23), however, provide a nice context for grammaticalization of ‘arrive’ into what Liljeblad (1966) calls an ‘inchoative’, but what may best be called prospective aspect, marking the starting point of an action, something like ‘be about to V’. (23) yo-kyoŋoona hɨkwa-pitɨ-ʔyakwi RE-evening wind-arrive-HAB ‘It begins to blow (the wind arrives) every evening.’ (SL: 79) Here, the inference provided by the appearance of this secondary verb allows for interpretation as a marker of aspect, creating an ambiguous or bridging context (e.g. Heine 2002). Example (24) shows how pitɨ ‘arrive’ still retains its status as the main lexical verb but may cooccur with its grammaticalized secondary verb counterpart.

310

Tim Thornes

(24) yaisi owitu nɨmmi na-tsa-pitɨ-kɨ-pitɨ-ga then DEM we MM-IP/grasp-arrive-APL-INCH-TRNSL ‘Then we were reaching (were about to arrive at) that place.’ The development of inflectional categories from secondary verbs is particularly clear when one compares (20) above, where -tapi ‘lie’ retains its postural semantics, with (25) below, where it clearly does not. (25) kutsu poo-do nɨmmi mia-tapi cow trail-LOC we go-lie.SPL ‘... we kept on going along a cow trail.’ In this example, the main verb to which -tapi attaches does not involve even the possibility of lying down, in keeping with Heine’s (2002) switch context where there is actual incompatibility with the source meaning. Rather, the semantic force of the form here is one of ‘keep on V-ing,’ a marker of verbal aspect. As with the development of a causative from an instrumental prefix (Section 3.3 above), these examples illustrate switch context quite well – again, a tipping point on the way to full conventionalization. It is exactly those contexts in which the original lexical semantics are ruled out that one finds expression of the new inflectional category. 4.2.2. Verbal number and grammaticalization Many of the independent verb sources of secondary verbs in Northern Paiute arise from a set of high-frequency verb stems whose forms are sensitive to verbal number (sometimes referred to as number suppletion).14 A threeway pattern of suppletion is present for the basic posture verbs plus a fourth, secondary verb form for ‘lie’, as we have just seen. The suppletive posture verb forms are given in Table 3. Table 3. Suppletion in Northern Paiute posture verbs sit Singular Dual Plural Secondary form

katɨ yɨgwi aataʔa

stand wɨnnɨ wammi konno

lie hapi kwapi pokwa / wakwapi 15 -tapi

Northern Paiute verb formation

311

The grammaticalization of the verb ‘stand’ is particularly interesting in that its pattern of number suppletion persists, even as a suffix coding continuous aspect, as illustrated in (26), from Liljeblad (1966: 78): (26) a. nɨga-wɨnɨ dance-CONT.SG ‘S/he is dancing.’ b. nɨga-wami dance-CONT.DL ‘They (2) are dancing.’ c. nɨga-gono dance-CONT.PL ‘They are dancing.’ The retention of number suppletion can be seen as a final conceptual link between the lexical verb source and its grammaticalized form in the context of the secondary verb construction. Evidence for the use of the singular as the default form exists, particularly among a slightly younger generation of speakers with whom I have worked. Although these developments could be argued to be one of the consequences of language obsolescence, they are still no different from what one would expect of natural diachronic change. Attested developments in the area of secondary verbs are incremental and uneven across Numic. All Numic languages exploit the SVC using, for the most part, the same set of verbs – especially the posture verbs. The degree to which conceptual links remain varies, however, with some forms retaining postural semantics in one language and not in another, some maintaining suppletion in the SVC but not elsewhere. Differences are even reported for other dialects of Northern Paiute. Garrett (2006) provides examples from the Mono Lake dialect (of Nevada Northern Paiute) where wɨnɨ ‘stand.SG’ retains its postural semantics in the SVC in cases where the context allows, while at the same time having lost the suppletive pattern demonstrated above. In Western Shoshone (Crum and Dayley 1993: 101– 105), we find suppletion and at least some postural semantics maintained with ‘stand’. In the Oregon Northern Paiute dialects that form the basis of this study, suppletion is maintained (if beginning to level somewhat) in the face of conventionalization. In short, the loss of suppletion and the semantic bleaching associated with grammaticalization are progressing independently in Northern Paiute and throughout Numic. Such unevenness in the grammatical development of the same forms across closely related lan-

312

Tim Thornes

guages, and even across dialects of the same language, sheds light on the micro-processes involved in grammaticalization. 4.2.3. Historical connections with full lexical verbs Table 4 summarizes the findings for the grammaticalization of secondary verbs in Northern Paiute, listing their suffixal forms, grammatical meanings, and hypothesized independent verb sources. Table 4. Grammaticalized secondary verbs and their likely source verbs SUFFIX -wɨnɨ -wami -kono -katɨ -tapi -nɨmi -moo -wɨnai -pidɨ -makwɨ -tɨŋɨ -mina -ki -[ʔ]yakwi -kwɨ -kɨ

GRAMMATICAL MEANING continuous.SG continuous.DL continuous.PL sit V-ing; stay V-ing lie V-ing; keep V-ing random motion.SG random motion.DL/PL move aside / away / against come V-ing; be about to V finish V-ing tell to V go and V, random focus motion toward speaker habitual / repetitive relative future applicative

LIKELY SOURCE wɨnnɨ ‘stand.SG’ wammi ‘stand.DL’ konno ‘stand.Pl’ katɨ ‘sit.SG’ -tapi ‘lie.SPL’ nɨmmi ‘travel.SG’ moo / moʔo ‘travel.DL/PL’ wɨnai ‘throw’ pidɨ ‘arrive.SG’ makwɨ ‘defeat; win a contest’ tɨŋɨ ‘request’ mia / miʔa ‘go’ kimma ‘come’ yakwi ‘carry’ kwɨ[h]ɨ ‘get; obtain’ kia / kɨa ‘give’

The relationship between some of the secondary verbs and their lexical verb sources is transparent in cases where there are clear formal similarities accompanying intact core semantics plus the maintenance of their suppletive patterns. For others, formal similarities combined with known cross-linguistic developments and processes are only suggestive. For example, ‘get’ verbs are attested (somewhat rarely, granted) as a source for the grammaticalization of a marker of future in a few languages (Bybee, Perkins, and Pagliuca 1994), as is ‘carry’ as a source for habitual aspect. Based on internal evidence and on typological grounds, I have proposed that the applicative suffix -kɨ, is likely to have developed historically from a verb meaning ‘give’ (Thornes 2003: 390–391 and 2007). This proposal is

Northern Paiute verb formation

313

only tentative, based largely upon the fact that its range of functions – from benefactive to sociative and indirect causative to direct causative to simple transitivizer – reflects well-documented developments for ‘give’ verbs in other languages, particularly those with extensive serial verb constructions (see especially Lord, Yap, and Iwasaki 2002). The proposal is marginally supported by formal similarities with Northern Paiute kia ‘give,’ but, barring supporting comparative evidence, remains speculative. Langacker (1977: 144) considers it possible to reconstruct an applicative suffix of the shape *kV for Proto-Uto-Aztecan. The great age of this suffix could account for its position within the theme layer. 4.2.4. Historical precursor to the secondary verb construction The Northern Paiute data provide an opportunity for the study of the extent of grammaticalization along several dimensions, and particularly of how independent forms develop into grammatical suffixes. What of the history of the secondary verb construction itself? The Northern Paiute SVC shares a great many properties in common with the category of assymmetrical serial verb construction, as defined by Aikhenvald (2006). Properties include: 1) the restricted nature of the verbs that can enter into the SVC, 2) the development of aspectual and directional properties, and 3) the lack of iconicity in the sequence of events coded by the open class main verb and that of the secondary verb (cf. Durie 1997). It is rarely attested, however, that such constructions arise historically from verb sequences that are in a marked dependent relationship with each other.16 Foley (1997: 384) describes the occurrence in Yimas (Papuan) of single word serial verb sequences whereby the temporal relationship holding between the verbs in the series is overtly (even repeatedly) marked when that relationship is sequential. Northern Paiute provides modest additional evidence of a similar phenomenon. Consider the following, from consecutive lines of a narrative, involving the posture verb aataʔa ‘sit.PL’: nɨmmi una aataʔa,... (27) a. ka=tɨppi-kuba OBL=rock-upon we.EXCL over_there sit.PL ‘... and on the rocks we sat up there,...’ b. mɨ=punni-na-aata-ʔyakwi. PL=see-CONV-sit.PL-HAB ‘would sit watching them.’

314

Tim Thornes

What is formally interesting about (27b) is the occurrence of the converbal suffix -na (CONV) on the first verb of the sequence, in spite of the fact that the sequence behaves prosodically like a single word.17 Adding to the curiosity of the form is the synchronic availability of the SVC. Of interest here is the fact that the verb aataʔa ‘sit.PL’ requires that the first verb be marked with -na, essentially as a syntactic dependent, whereas the singular (and dual) forms for ‘sit’ occur unambiguously in the SVC without such marking, as one would expect from a well-behaved (or at least unambiguous) serial verb construction. In (27b), the full lexical content of the would-be secondary verb is maintained. However, this is not always the case. I have found a certain amount of ambiguity regarding the semantic content of aataʔa in this context. Example (28) was translated both as the expression of two distinct subevents (sitting and eating) and alternatively as a single unitary event involving plural participants. (28) nɨmmi noo-ʔyu-na, tɨka-na aataʔa we.EXCL all-NOM-CONV eat-CONV sit.PL ‘We all were sitting eating’ ~ ‘We all were eating (severally).’ These facts appear to add yet another dimension to the grammaticalization of elements in the SVC, this time by illustrating traces of an historical source construction. The converbal suffix -na (CONV) marks one of the verbs as a syntactic dependent, suggesting that the SVC could have developed from the loss of subordinate (converbal) marking on one of the verbs. This possibility aids in explaining the form of one of the directional suffixes listed in Table 4, -mina ‘go and V’, a suffix that appears to have been grammaticalized as a fixed form with the converbal suffix. Aikhenvald (2006: 1) defines a serial verb construction as “a sequence of verbs which act together as a single predicate, without any overt marker of coordination, subordination, or syntactic dependency of any other sort”. What makes the data above interesting is that they provide some evidence that the SVC, in fact, may have developed from just such a marked relationship of syntactic dependency.

5. Summary and conclusion Two pervasive processes in the formation of Northern Paiute verbs have come about as the result of two distinct historical pathways. One, com-

Northern Paiute verb formation

315

pounding, gave rise to the instrumental prefix construction (IPC) and its extensions into the domain of causation and desideration. The other, converbal subordination, may have given rise to the secondary verb construction (SVC), a type of serial verb construction, and provided the context for the development of a wide range of inflectional affixes in Northern Paiute. The variation and irregularity one finds in Northern Paiute provide clues to these processes. Both the IPC and the SVC constitute assymetrical construction types, whereby one closed morphological class (which defines the construction) combines with a theoretically open class of verbs. But that is where the similarities end. Although I hypothesize that there are forms in synchronic Northern Paiute that can be viewed as incipient members of the class of instrumental prefixes, the IPC can practically be considered a closed system at this point in its history. Although I hypothesize that the IPC likely served as the source context for more elaborate bipartite stem systems such as that described for Washo by Jacobsen (1980) in his seminal paper, how or why these more elaborate systems developed falls well outside the scope of this study. Likely, for these languages, there must have been a time when the system was not closed, as I suggest the IPC is now in Northern Paiute. Broader typological studies in the area of complex predicates may provide clues as to what a more open system entails, or whether stemming from an IPC carried restrictions with it. The SVC, on the other hand, shows the variation one expects of a system that is still, in many ways, developing. In comparing these two constructions, perhaps even more striking than their formal and functional differences are their developmental ones. Even though the SVC is defined by a restricted set of forms relating mainly to simple posture and motion verbs, the construction does not appear completely closed to the possibility that new members may enter into it and, having done so, develop more grammatical properties of their own, constrained only by the semantic features they carry. Developmental projections from semantic bases are not without controversy, but the growing typological data are suggestive – familiar basic concepts having to do with basic motion, orientation, transfer, manipulation and the like, develop into relatively familiar, if not predictable, grammatical categories, given the right constructional contexts. I hope that this brief look at verb formation in Northern Paiute sheds some light on the processes that enter into such developments and does its part to encourage in-depth descriptive work as a major contribution to the science of linguistic typology and to the ongoing conversation regarding the role of diachronic explanation in synchronic description.

316

Tim Thornes

Notes 1.

2.

3.

4.

5. 6.

7.

8.

9.

Institutional support for this work has been provided by a grant from the National Science Foundation (#0418453). A special thanks to my language teachers, especially Nepa Kennedy, Rena Beers, Justine Louie-Brown and Ruth Lewis. Thanks to Lloyd Louie, Patricia Miller, Shirley Tufti, Ken Barney, Phyllis Harrington, and Yolanda Manning as well for engaging me in my study over the years. I would also like to acknowledge the late Irwin Weiser (1909–1996), the late Maude Washington Stanley (1913–2000), and the late Myrtle Peck (1934–2006) for their many contributions to my understanding of Northern Paiute. Thanks also to Lynn Burley and Joana Jansen for comments on an earlier draft, and to Pattie Epps, Mark Van de Velde and an anonymous reviewer for extensive suggestions on how to strengthen this chapter. Any errors are, of course, my own. For this reason, I embrace the fact that SVC may remind readers of ‘serial verb construction’, particularly as defined by Aikenvald (2006), while remaining true to the term ‘secondary verb’ as it is known in the Uto-Aztecan literature. The Northern Paiute examples in this chapter will be represented as something near to a practical orthography. Consonant gradation and mutation, particularly at morpheme boundaries, will be preserved in their surface forms. Abbreviations: APL applicative, APS antipassive, ASP aspect, CAUS causative, CISL cislocative, CONT continuous, DEM demonstrative, DESID desiderative, DIR directional, DISJUNCT disjunction, DL dual, DUR durative, EXCL exclusive, FUT future, HAB habitual, INCH inchoative, INCL inclusive, IP instrumental prefix, MM middle marker (reflexive/passive), MOD modal, NEG negation, NMR nominalizer, NOM nominative case, OBL oblique/non-nominative case, PL plural, PNC punctual aspect, PRO pronoun, CONV converbal, RE reduplication, RNDM random motion, SEQ sequential, SG singular, SPL suppletive form, SUB subordinate, TRNSL translocative, = clitic boundary. SL stands for data from Liljeblad (1966). All other data are from my field notes. Traces of it appear elsewhere in the family, particularly in the Piman subgroup, leading most scholars to assume it to be a trait that predates Proto-UtoAztecan (Langacker 1977: 133) By ‘extensive’, I mean not only in terms of distinct subtypes, but also in terms of frequency. “The striking fact about Klamath is that a majority of all verb stems are of this bipartite pattern” (DeLancey 1998: 60). Jacobsen’s (1980) analysis of Washo concurs, based upon language-internal evidence, as does DeLancey’s (1998) discussion of the relative age of the various lexical prefix types in Klamath, based largely upon their phonological properties. Rood and Taylor (1996: 464) report this development with verbs of “physical state or valuatives”.

Northern Paiute verb formation

317

10. I use the term ‘head’ in a more traditional sense to refer to the element that provides the core lexical semantics, as opposed to serving in a more secondary, modifying capacity. 11. This zone consists of no fewer than four position classes. That these layers may reflect historical accretions of individual morphemes via the SVC lies beyond the scope of this chapter. 12. This term is attributed to Crapo (1970) by scholars of Uto-Aztecan languages. 13. Of course, as pointed out by an anonymous reviewer, compounds may be multi-headed or variously right or left-headed in a single language. This does not detract greatly, in my view, from using semantic headedness as one property that distinguishes the SVC from lexical compounding, particularly since the relationship between the secondary and primary verbs is otherwise distinctive, as discussed below. 14. The issues surrounding whether or not a ‘verbal number pair/triple’ in Veselinova’s (2007: 140) terms counts as suppletion or ought to more properly be considered the lexical expression of distinct event types is not entirely relevant to the discussion I present here. Langacker (1977: 127) describes what appears, on the one hand, to be a clear case of participant number that is widespread in Uto-Aztecan, but many of the formal patterns reflect event number – more properly a category of verbal aspect (Corbett 2000: 246). 15. The reduplication of the dual could be a recent regularization of the paradigm. 16. Rose (2003) describes an interesting case of a serial verb construction arising from a gerundive construction in the Tupí-Guaraní language, Emerillon. 17. What I term the ‘converbal suffix’ (-na) covers a wide range of subordinating functions, from marking event nominalizations, non-subject relative clauses, and adverbial ‘while’ clauses, to backgrounding events in narrative.

References Aikhenvald, Alexandra Y. 2006 Serial constructions in typological perspective. In Serial Verb Constructions: A Cross-linguistic Typology. Alexandra Y. Aikhenvald and R. M. W. Dixon (eds.), 1–68. Oxford: Oxford University Press. Bybee, Joan L., Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Corbett, Greville G. 2000 Number. Cambridge: Cambridge University Press. Crapo, Richley 1970 The origin of the directional adverbs in Uto-Aztecan languages. International Journal of American Linguistics 36: 181–190.

318

Tim Thornes

Crum, Beverly, and Jon Dayley 1993 Western Shoshoni Grammar. Boise, Idaho: Boise State University. Dayley, Jon P. 1989 Tümpisa (Panamint) Shoshone Grammar. University of California Publications in Linguistics, vol. 115. Berkeley: University of California Press. DeLancey, Scott 1996 Argument structure of Klamath bipartite stems. SSILA annual meeting, San Diego, California. 1998 Lexical prefixes and the bipartite stem construction in Klamath. International Journal of American Linguistics 65: 56–83. Durie, Mark 1997 Grammatical structures in verb serialization. In Complex Predicates, Alex Alsina, Joan Bresnan, and Peter Sells (eds.), 289–354. Stanford: CSLI Publications. Foley, William A. 1997 Polysynthesis and complex verb formation: the case of applicatives in Yimas. In Complex Predicates, Alex Alsina, Joan Bresnan, and Peter Sells (eds.), 355–395. Stanford: CSLI Publications. Foley, William A., and Mark Olson 1985 Clausehood and verb serialization. In Grammar Inside and Outside the Clause Johanna Nichols and Anthony Woodbury (eds.), 17–60. Cambridge: Cambridge University Press. Evans, Nicholas, and Alan Dench 2006 Introduction. In Catching Language: the Standing Challenge of Grammar Writing, Felix K. Ameka, Alan Dench, and Nicholas Evans (eds.), 1–39. Berlin/New York: Mouton de Gruyter. Fleck, David W. 2006 Body-part prefixes in Matses: derivation or noun incorporation? International Journal of American Linguistics 72: 59–96. Garrett, Andrew 2006 The evolution of complex predicates. Seminar on Complex Predicates. Rice University, Houston. Givón, Talmy 2006 Multiple routes to clause union: the diachrony of syntactic complexity. Seminar on Complex Predicates. Rice University, Houston. Heine, Bernd 2002 On the role of context in grammaticalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 83– 101. Amsterdam/Philadelphia: Benjamins.

Northern Paiute verb formation

319

Hori, Hirofumi 1998 Instrumental prefixes in Skidegate Haida. In Languages of the North Pacific Rim, Osahito Miyaoka and Minoru Oshima, (eds.), 33–46. Kyoto: Kyoto University. Jacobsen, William H., Jr. 1980 Washo bipartite verb stems. In American Indian and Indoeuropean Studies: Papers in Honor of Madison S. Beeler, Kathryn Klar, Margaret Langdon, and Shirley Silver (eds.), 85–99. The Hague: Mouton. Jansen, Joana, and Virginia Beavert 2008 Bipartite causative constructions in Yakima Sahaptin. SSILA annual meeting. Chicago, Illinois. Langacker, R.W. 1977 Studies in Uto-Aztecan Grammar, vol. 1: An Overview of UtoAztecan Grammar. Arlington, TX: SIL Publications. Liljeblad, Sven 1966 Northern Paiute manual: grammatical sketch of the northern dialects. Ms. Lord, Carol, Foong Ha Yap, and Shoichi Iwasaki 2002 Grammaticalization of “give:” African and Asian perspectives. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 217–235. Amsterdam/Philadelphia: Benjamins. Miller, Wick 1972 Newe Natekwinappeh: Shoshoni Stories and Dictionary. University of Utah Anthropological Papers, 94. Salt Lake City. Mithun, Marianne 1984 The evolution of noun incorporation. Language 60: 847–893. 1999 The Languages of Native North America. Cambridge: Cambridge University Press. 2002 An invisible hand at the root of causation: the role of lexicalization in the grammaticalization of causatives. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 237–257. Amsterdam/Philadelphia: Benjamins. Nichols, Michael J. P. 1974 Northern Paiute historical grammar. Ph.D. diss., Department of Linguistics, University of California, Berkeley. Rankin, Robert L. 2006 The interplay of synchronic and diachronic discovery in Siouan grammar-writing. In Catching Language: The Standing Challenge of Grammar Writing, Felix K. Ameka, Alan Dench, and Nicholas Evans (eds.), 527–547. Berlin/New York: Mouton de Gruyter.

320

Tim Thornes

Rood, David S., and Allan R. Taylor 1996 Sketch of Lakhota, a Siouan language. In Handbook of American Indians vol. 17: Languages, William C. Sturtevant (ed.), 440–482. Washington D. C.: Smithsonian. Rose, Françoise 2003 Morphosyntaxe de l’émérillon. Une langue tupi-guarani de Guyane française. Ph.D. diss., Université Lumière Lyon II. Sapir, Edward 1911 The problem of noun incorporation in American languages. American Anthropologist 13: 250–282. [Reprinted in The Collected Works of Edward Sapir, William Bright (ed.), vol. 5, American Indian Languages, 27–60. Berlin: Mouton, 1990.] Talmy, Leonard 1985 Lexicalization patterns. In Language Typology and Syntactic Description vol. 3: Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 57–149. Cambridge: Cambridge University Press. Thornes, Tim 1996 Yahooskin-Northern Paiute verb morphology. M.A. thesis, Department of Linguistics, University of Oregon. 1999 ‘Secondary’ verbs in Northern Paiute. In Proceedings of the First Workshop on American Indian Languages: Santa Barbara Working Papers in Linguistics, vol 8. University of California: Santa Barbara. 2003 A Northern Paiute grammar with texts. Ph.D. diss., Department of Linguistics, University of Oregon. 2007 Causation as functional sink in Northern Paiute. LSA annual meeting. Anaheim, California. Veselinova, Ljuba N. 2007 Suppletion from a typological perspective. In New Challenges in Typology: Broadening the Horizons and Redefining the Foundations, Matti Miestamo and Bernhard Wälchli (eds.), 127–151. Berlin/New York: Mouton de Gruyter.

Part V. Class struggle: Erasing borderlines

Reference and predication in Movima Katharina Haude

1. Introduction1 The issue of whether nouns and verbs are universally distinct syntactic categories comes up again and again in the typological literature (cf. the recent discussion in Linguistic Typology 9). Languages or language families in which this distinction has been a matter of debate include, for instance, Salishan (cf. e.g. Jelinek and Demers 1994; Kinkade 1983), Tongan (cf. Broschart 1997), Tagalog (cf. e.g. Himmelmann 2005a), Mundari (cf. Evans and Osada 2005), and Tupi-Guaranían (Queixalós 2001). This article demonstrates that Movima, an endangered, unclassified language of Amazonian Bolivia (Haude 2006), should be added to this list. Movima nouns and verbs are distinguished on morphological grounds, and the criteria for a noun-verb distinction in accordance with propositional act functions (Croft 2003) clearly hold: verbs typically function as predicates and nouns typically form part of referring expressions. However, nouns can also function as predicates, and verbs can form part of referring expressions, without any morphological modification. Argument encoding is similar with both nominal and verbal predicates: one of the arguments of a transitive clause is encoded in the same way as a nominal possessor, and the other is encoded in the same way as the single argument of an intransitive clause. Thus, on the syntactic level there is only a weak distinction between nouns and verbs. While this is nothing new cross-linguistically, Movima furthermore has an apparently unique alignment system, whose explanation may lie in the syntactic similarity of lexical categories. Transitive clauses are organized according to the relative position of the arguments in a referential hierarchy. The argument with the lower-ranking referent has syntactic argument status, while the argument with the higher-ranking referent is encoded like a nominal possessor, i.e. as a phrase-internal modifier. The syntactic flexibility of nouns and verbs, together with the association of possessors with high-ranking entities, may form the historical basis of this unusual pattern. This chapter is structured as follows. Section 2 describes the characteristics of affirmative main clauses (for an account of embedded clauses, see

324

Katharina Haude

Haude, forthcoming a): argument encoding and alignment (2.1), referential elements (2.2), and the morphological distinction of nouns and verbs (2.3). Section 3 describes the syntactic flexibility of nouns and verbs, i.e., nouns functioning as predicates (3.1) and verbs occurring in determiner phrases (DPs) (3.2). An interpretation of the findings, which characterize Movima as an ‘omnipredicative’ language (Launey 1994), is proposed in Section 3.3.2

2. The clause and its parts 2.1. Argument encoding and alignment The basic constituents of a Movima clause are predicate and argument. The predicate consists of a content word (verb or noun). An argument is expressed by a non-oblique pronoun or determiner phrase (DP), the latter formed by a determiner preceding a content word, as schematized in (1). Constituent order is largely fixed and predicate initial. (1)

Predicate content word

Argument [determiner + content word]

Intransitive clauses are defined by the fact that they may contain maximally one overt argument; in (2), it is represented by a DP (in square brackets; the phonetic representation in the third line will become relevant below). (2)

chinaɬa] kat-pit [is break-half ART.PL manioc [kap͡ʔmˈpitt͡ʔn ʔis t͡ʃinaɬa] ‘The manioc roots are broken in halves.’3

When the argument of an intransitive clause is represented by a bound pronoun, this pronoun is cliticized to the predicate through what I term ‘external cliticization’ (represented by a double dash), as in (3). (3)

kat-pit--as break-half--3N.AB [ˈkap͡ʔmpittas] ‘It was broken in halves.’

Reference and Predication in Movima

325

External cliticization (in contrast to ‘internal cliticization’, which occurs in transitive clauses and will be described below) is characterized by the fact that when the host ends in a consonant, this consonant forms the syllable onset of the vowel-initial cliticized pronoun (cf. Haude 2006: 101–103). This can be observed by comparing the phonetic representations in (2) and (3) above: in coda position (2), /t/ is realized as a complex glottalized consonant, [tʔn], which attracts stress; in onset position (3), /t/ is realized as [t]. The vowel-initial referential element, furthermore, is preceded by a glottal stop when not cliticized, as in (2). The argument of an intransitive clause can also be represented by a free pronoun, which is not cliticized and typically occurs in topic position, i.e. preceding the predicate: (4)

i’ne

dejal PRO.F cook ‘She cooks.’

When the argument is known from the context, it can be omitted, as in (5): (5)

kat-pit ja’a break-half just ‘(They, i.e. the palm leaves) simply break.’

In sum, intransitive clauses have the property that they take maximally one overt argument, which can be expressed as a DP, as an independent pronoun, or as an externally cliticized bound pronoun, and which may be omitted. Transitive clauses, in contrast, are characterized by the fact that they may contain two overt arguments, which typically both follow the predicate. Their linear order corresponds to the position of their referents in the referential hierarchy in (6), which basically involves person and topicality: the higher-ranking participant is encoded first, with the lower-ranking participant encoded in second position after the predicate. (6)

1 > 2 > 3 topic (given) > 3 nontopic (new)

Semantic roles are indicated through direct or inverse marking on the predicate: when the actor outranks the undergoer in a two-participant event, the verb is marked as direct (7); when the undergoer outranks the actor, the verb is marked as inverse (8).

326

Katharina Haude

(7)

[os ɬow-na=y’ɬi pull-DR=1PL ART.N.PST ‘We pulled the oxcart.’

(8)

ju:-kay-a=y’ɬi [is bito’] scold-INV-LV=1PL ART.PL old_person ‘The old people scolded us.’

kare:ta] oxcart

Due to the hierarchy-based constituent order and the ‘remapping’ effect of the direct/inverse opposition (cf. Zúñiga 2006: 62), it is difficult to assign uncontroversial labels to the nominal constituents in a Movima transitive clause. Based on their position in the referential hierarchy, I will refer to the argument encoded in first position after the predicate as ‘Proximate Argument’ (short: PROX) and to the argument encoded in second position after the predicate as ‘Obviate Argument’ (short: OBV).4 Apart from linear order, the formal differences between PROX and OBV are as follows. PROX is obligatorily realized. The absence of a pronominal PROX enclitic from the predicate of a transitive clause indicates the first person singular, as shown in (9). PROX is also expressed after the predicate when there is a coreferential free pronoun in topic position, as in (10). (9)

jiwa-ɬe:-na=∅ [is come-CO-DR=1SG ART.PL ‘I brought motacú nuts.’

chujat-di] motacú-CL.seed

(10) i’ne

jiwa-ɬe-na=’ne come-CO-DR=3F ‘She (was the one who) brought (it).’

PRO.F

PROX is phonologically attached to the predicate through ‘internal cliticization’. This process creates a structure whose stress properties are those of a prosodic word: when a monosyllabic element is internally cliticized, this causes stress shift, stress moving one position to the right. When the host has an open penultimate syllable, this syllable loses its original lengthening. The phonetic representations in (11) illustrate the stress shift and the shortening of the vowel.

Reference and Predication in Movima

(11) a. aya:-na=∅--us wait_for-DR=1SG--3M.AB [aˈja: ˈja:naʔus] ˈja: ‘I wait for him.’

327

b. aya-na=us wait_for-DR=3M.AB [ajaˈna ˈnaʔus] ˈna ‘He waits for (him/her/it/them).’

Internally cliticized elements furthermore require a preceding vowel: when the host ends in a consonant, the linking vowel -a is inserted, as illustrated in (12). Example (12) also demonstrates that internal cliticization (unlike external cliticization) involves determiners in the same way as pronouns. (12) kay-a-poj-a=[us eat-DR-CAUS-LV=ART.M ‘The man fed the dog.’

itila:kwa] man

[os ART.N.PST

pa:ko] dog

With these properties, the encoding of PROX is identical to the encoding of a nominal possessor. This is illustrated in (13), where a pronoun is attached through internal cliticization to a consonant-final noun augmented by the linking vowel. (13) as

powol-a=us straw_mat-LV=3M.AB ‘his straw mat’

ART.N

Obligatorily possessed nouns (e.g. kinship terms, body-part terms) without an internally cliticized element are interpreted as having a first person singular possessor (14), in the same way as unmarked bivalent verbs are interpreted as having a first person singular PROX argument (see [9] above).5 majni=∅ (14) kinos ART.F.AB offspring=1SG ‘my daughter’

In contrast to PROX, OBV has the same formal properties as the single argument of an intransitive clause, described above. When expressed as a DP, as in (7) and (8) above, it is phonetically independent; when realized as a bound pronoun, it is attached to the preceding constituent (the predicate with the PROX enclitic) through external cliticization, as in (15) below. It is not obligatorily realized, as shown in (16); and OBV can also be expressed as a free pronoun preceding the predicate, as in (17).

328

Katharina Haude

(15) dewaj-na=n--is see-DR=2--3PL.AB ‘You see them.’ (16) jayna

tikoy-na=us kill-DR=3M.AB ‘Then he killed (it, i.e. the tapir).’

DSC

jayna yey-na=is (17) a’ko PRO.N DSC want-DR=3PL.AB ‘This, then, is what they want.’

Table 1 lists the formal properties that distinguish the PROX argument and the possessor, on the one hand, from OBV and the single argument of an intransitive clause, on the other. Table 1. Formal properties of argument and possessor encoding PROX and possessor Internal cliticization ( = ): stress shift, epenthetic /a/ Pronouns and articles are cliticized Obligatory

OBV and argument of intransitive clause External cliticization ( -- ): resyllabification, no stress shift, no epenthetic /a/ Only pronouns are cliticized Not grammatically obligatory

To sum up, the argument of a transitive clause that refers to the participant higher in the referential hierarchy is encoded in the same way as a nominal possessor; the argument with the lower-ranking referent, in contrast, is encoded in the same way as the single argument of an intransitive clause. Moreover, as I have demonstrated elsewhere in more detail (Haude, in press; Haude, forthcoming b), OBV is syntactically privileged, whereas there is no evidence that PROX has access to any syntactic operations to which OBV does not have access. In particular, as illustrated by the following elicited examples (based on [12] above), OBV can be relativized (18), while in order to relativize the participant encoded as PROX in a transitive clause, a valency-decreasing operation is needed, marked by the particle kaw (19). Here, the former PROX becomes the single argument of the now intransitive clause, while the former OBV is demoted to adjunct status (marked as oblique).

Reference and Predication in Movima

(18) [os

pa:ko] di’ dog REL ‘the dog that the man fed’

ART.N.PST

(19) [us

itila:kwa] di’ man REL [n-os pa:ko] OBL-ART.N.PST dog ‘the man who fed the dog’ ART.M

kay-a-poj-a=[us eat-DR-CAUS-LV=ART.M

kaw kaw

329

itila:kwa] man

kay-a:-poj eat-DR-CAUS

Interestingly, this operation works with nouns as well (see Haude, forthcoming b): a noun combined with the particle kaw denotes the possessor of an entity, not the entity itself, and cannot be marked for a possessor anymore by an internal enclitic: kinos kaw majni ‘his/her/their mother’ (cf. [14] above). 2.2. Referential elements The class of referential elements comprises articles, free pronouns, bound pronouns, and demonstratives.6 These all indicate the referent’s gender, number, and presence at or absence from the speech situation. Articles (see Table 2) additionally indicate when an absent referent has ceased to exist (cf. Haude 2004). Table 2. Articles7 Presential/generic ART.M ART.F ART.N ART.PL

us i’nes as is

Absential kus kinos kos kis

Past (ceased to exist) us isnos os is

Articles are obligatorily followed by a content word, together with which they form a DP, as illustrated in (20). They do not distinguish between definite and indefinite reference. (20) i’nes kwe:ya kinos kwe:ya isnos kwe:ya

‘the/a woman (present)’ ‘the/a woman (absent, in existence)’ ‘the/a woman (absent, deceased)’

330

Katharina Haude

Plural number is used for more than one countable entity as well as for liquids and masses (cf. Haude 2006: 150–151). Gender is semantically determined: humans are distinguished according to natural gender; nonhumans are classed as neuter (21). (21) masculine: feminine: neuter:

us dichi:ye i’nes dichi:ye as wa:ka

‘the/a boy’ ‘the/a girl’ ‘the/a cow’

Presence at the speech situation implies that the referent is more or less within calling distance, for example in the same house or compound. The presential forms of the referential elements are also used for entities whose location is common knowledge, e.g. places: (22) as Kochawamba ‘Cochabamba (town in Bolivia)’ With these properties, the article clearly has the function of establishing reference, be it specific or generic, definite or indefinite. 2.3. Nouns and verbs: morphological characteristics As the point of departure for examining the formal differences between nouns and verbs in Movima, I define nouns as belonging to the class of words whose meaning is prototypically associated with entities, and verbs as words whose meaning is prototypically associated with events. The main formal criteria for distinguishing nouns from verbs are morphological. They are quite subtle, due to the fact that the typical word-class defining morphological categories (cf. e.g. Schachter 1985) are lacking: nouns are not morphologically marked for gender, number, or case, and the typical verbal categories tense, aspect, and mood are not morphologically marked on verbs. The distinction between verbs and nouns can be recognized through morphological tests: there are morphemes that can be combined with nouns but not with verbs (cf. Haude 2006: 106–111). In particular, nouns undergo reduplication in order to form the predicate of an embedded clause, as illustrated in (23) with the noun tolkosya ‘girl’. Verbs, in turn, need to take the suffix -wa to function as predicates of embedded clauses, as shown in (24). (All embedded predicates are obligatorily possessed action/state nominals.)

Reference and Predication in Movima

331

(23) n-os tolkosya=sne OBL-ART.N.PST girl=3F.AB ‘when she was a girl’ (lit.: ‘at her former being a girl’) (24) n-os iloni-wa=sne OBL-ART.N.PST walk-NMZ=3F.AB ‘when she walked’ (lit.: ‘at her former walking’) According to this criterion, property-denoting words (semantic adjectives) can be identified as morphological nouns, since they undergo reduplication in embedding or negation; in (25), this is illustrated with the word tochik ‘small’. (25) n-os tochik-a=sne OBL-ART.N.PST small-LV=3F.AB ‘when she was small’ (lit.: ‘at her former being a small one’) The distinction of bivalent verbs from nouns and monovalent verbs is more straightforward, since bivalent verbs are always overtly marked as either direct or inverse. In those cases in which nouns contain the direct morpheme, syntactic criteria distinguish them from bivalent verbs, as will be illustrated in (35) below.

3. Nouns and verbs: distributional characteristics 3.1. Predicate nominals There is no copula in Movima affirmative clauses, and nouns can function as predicates without morphological modification. A rough text count showed that predicate nominals constitute only about 10% of the total number of predicates. However, unless morphological tests are carried out (cf. 2.3 above), it is often impossible to identify a monovalent predicate on formal grounds as either a verb or a noun. In particular, semantic adjectives occur more often as predicates (26) than inside a DP (27), even though morphological tests identify them as nouns (see [25] above). [is majni=∅] (26) ka:w-e much-CL.person ART.PL child=1SG ‘I have many children.’ (lit.: ‘My children are many.’)

332

Katharina Haude

(27) jiwa [us come ART.M ‘The old (man) came.’

bi:jaw] old

Clauses with unpossessed predicate nominals are formally identical to clauses with monovalent verbal predicates. The argument of such a clause can be expressed as a phonologically independent DP (28); as an externally cliticized pronoun (29); as a free pronoun in topic position (30); and it can be omitted when it is known from the context (31). (In [28–31], the predicate nominals are rendered in bold print.) e:ɬ-a=is ney wu’tu] (28) tomo:re [is tomore ART.PL name-LV=ART.PL here pot ‘The name of those pots (was) tomore.’

(29) tolkosya--sne girl--3F.AB ‘She (was a) young woman.’ (30) a’ko lopa:vos PRO.N manioc_stem ‘This (is a) manioc stem.’ (31) jayna mo’incho:but DSC manioc_mass ‘(It was) already fermented manioc mass.’ As the examples show, predicate nominals denote the state of being the entity denoted by the noun. I am aware of only one type of case in which the predicative meaning of a noun clearly differs from its meaning inside a DP. This involves words denoting languages, such as kaste ‘Spanish’ or chonsineɬ ‘native language’ (see Evans and Osada 2005 for a similar example). When occurring inside a DP, as in (32a), these words denote the language itself; when functioning as predicate, in contrast, they denote the act of speaking the language, as in (32b). (32) a. iwa:ni--sne [n-as chonsineɬ] speak--3F.AB OBL-ART.N native_language ‘She speaks in the native language.’

Reference and Predication in Movima

333

b. chonsineɬ--isne native_language--3F.AB ‘She speaks the native language.’ As was illustrated in 2.1 above, the possessor of a noun is encoded in the same way as PROX. Therefore, it can be expected that clauses with a possessed predicate nominal are identical to transitive clauses with a bivalent verbal predicate. However, this is not entirely the case: the argument of a possessed predicate nominal can only be expressed by a free pronoun in topic position, as in (33). (33) kay-ak, a’ko eat-IRR PRO.N ‘May (it) eat, it’s your pet!’

nono=n pet=2

The elicited example (34b), contrasted with (34a), shows that it is ungrammatical to express the argument of a clause with a possessed predicate nominal as an externally cliticized pronoun: pa:ko=us (34) a. asko PRO.N.AB dog=3M.AB ‘It is his dog.’

b.

*pa:ko=us--k-as dog=3M.AB--OBV-3N.AB (‘It is his dog.’)8

This syntactic restriction on argument encoding in clauses with possessed predicate nominals serves as a criterion for distinguishing obligatorily possessed nouns from bivalent verbs in those cases where these are morphologically similar. On inherently monovalent verbal bases (cf. Haude 2006: 340–344), the direct marker -na serves as a nominalizer that derives a location-denoting noun. When this noun functions as a predicate, its argument must be encoded by a free pronoun in topic position, as in (35a). The ungrammatical example (35b) was elicited to confirm this pattern. (35) a. a’ko as-na=[kus ya:yak=∅], [as Kachweli:ta] PRO.N sit-DR=ART.M.AB uncle=1SG ART.N Cachuelita ‘This is where my uncle lives, Cachuelita.’ (lit. ‘This is my uncle’s sitting place, Cachuelita.’) b. *as-na=[kus ya:yak=∅] [as Kachweli:ta] sit-DR=ART.M.AB uncle=1SG ART.N Cachuelita (intended meaning: ‘My uncle lives in Cachuelita,’ lit.: ‘Cachuelita is my uncle’s sitting place.’)

334

Katharina Haude

As is summed up in Table 3, monovalent verbs and nonpossessed nouns have identical syntactic properties, which is in line with the fact that they are also difficult to distinguish morphologically. Bivalent verbs and possessed nouns differ from the former two in that they are combined with an internally cliticized referential element. Unlike the case of bivalent verbal predicates, however, the argument of a clause with a possessed predicate nominal has restricted distributional possibilities: it cannot be expressed as an externally cliticized pronoun. Table 3. Argument encoding in nominal and verbal clauses Predicate type Monovalent verb / nonpossessed noun Bivalent verb Possessed noun

Takes internal clitic (=) no

Can take external clitic (--) yes

Can take free pronoun in topic position yes

yes yes

yes no

yes yes

3.2. Verbs in DPs When a DP contains a verb instead of a noun, as is the case in approximately 5% of my text corpus, it refers not to an event, but to a participant in an event. The role the participant has in the event is specified by the verb’s argument structure. When the verb is bivalent, the DP refers to the participant that is encoded as OBV when the verb functions as predicate. This means that a DP containing a direct-marked verb refers to the undergoer in the event, as in (36), and a DP containing an inverse-marked verb refers to the actor, as in (37). Gender, number, and spatio-temporal properties of the referent are indicated by the article. [is (36) nokowa rimeɬ-na=∅ right_now buy-DR=1SG ART.PL ‘Now I’ll buy (the things) you want.’

yey-na=n] want-DR=2

(37) jayna jiwa [us rey yey-kay-a=n] DSC come ART.M MOD want-INV-LV=2 ‘The one who loves you, you know, has come already.’9 Similarly, when the verb inside a DP is monovalent, the DP refers to the participant that is encoded as the single argument when the same verb

Reference and Predication in Movima

335

functions as a predicate (although see below). The DP in (38) refers to the actor in the event, which is encoded as OBV when the verb deja:jal is a predicate (cf. (4) above); the DP in (39) refers to the undergoer, in line with the role of OBV when the verb katpit functions as predicate (cf. (3) above). (38) kiro’ [kis dejal] DEM.PL.AB ART.PL.AB cook ‘There are (people who) cook.’ rey (39) jayna ɬ vel-na=∅ [os DSC 1 look_at-DR=1SG ART.N.PST MOD ‘Then I looked at the broken (part), you see.’

kat-pit] break-half

In contrast to bivalent verbs, there are a few exceptional cases with respect to the interpretation of monovalent verbs inside DPs. In particular, these include all verbs that contain an incorporated argument (often represented by a classifier-like element, see Haude 2006: 283–286). These verbs are composed of a transitive root, an incorporated nominal element, and the direct morpheme. As predicates, they are monovalent, and the argument represents the actor, as illustrated in (40a). However, when they occur inside a DP, the DP refers to the undergoer, and the actor is encoded as PROX/possessor, as in (40b).10 (40) a. loj-a:-’oj--is wash-DR-CL.clothes--3PL.AB ‘They wash clothes.’ b. kis loj-a-’oj-a=is ART.PL.AB wash-LV-CL.clothes-LV=3PL.AB ‘their laundry’ (not: ‘the [ones who] wash clothes’) The most important fact to note is that a DP containing a verb never refers to an event, but to a participant in an event, whose characteristics are specified by the determiner. Reference to events, in contrast, is carried out with derived forms. The embedded clause in (41) and the negated clause in (42) illustrate an event-denoting expression derived from a verb (an example of a state-denoting predicate nominal was given in (23) above):

336

Katharina Haude

(41) no-kos joyaj-wa=sne ney OBL-ART.N.AB arrive-NMZ=3F.AB here ‘when she arrived here (earlier today)’ (lit.: ‘at her earlier arriving here’) (42) kas

joyaj-wa=sne NEG arrive-NMZ=3F.AB ‘She didn’t arrive.’ (lit.: ‘Her arriving is/was not.’)

Unlike verbs, event- and state-denoting forms can only be combined with the neuter article, as in (41) and (23) above. Furthermore, with these forms, the article only has a temporal, not a spatial interpretation: as illustrated in (41), the absential article refers to a time on the same day, but before the moment of speaking (cf. Haude 2004: 83; 2006: 168–171). Verbs, in contrast, can be combined with all forms of the article, in accordance with the referential properties of the event participant. 3.3. Interpreting the syntactic flexibility of nouns and verbs With the properties described above, Movima fulfils the requirements of an ‘omnipredicative’ language (Launey 1994; 2002: 114–118). In a language of this type, while nouns and verbs can be morphologically distinct categories, both word classes “should be able to function directly as predicates, and should have equal potential to form referring expressions through […] at least the addition of some sort of determiner” (Evans and Osada 2005: 361).11 A characteristic property of omnipredicative languages is that nouns functioning as predicates are interpreted as ‘be X’, and that the addition of a determiner to a noun or verb produces “no further semantic increment than that accompanying relativization in English” (Evans and Osada 2005: 362), i.e., ‘(the) one that (is) X’. As could be seen in Sections 3.1 and 3.2, this corresponds to the interpretations of nouns in predicate function and verbs inside DPs, respectively, in Movima. The question is, however, whether a semantic increment is involved at all when verbs occur in a referential expression in Movima. It may also be possible to interpret all verbs as nominal expressions that characterize a participant in an event rather than denoting the event itself (cf. Sasse 1993: 655 on Salishan and Tagalog). A monovalent verb like deja:jal (cf. (4), (38) above) may have the meaning ‘one who cooks’ rather than ‘(to) cook’; a verb like katpit in (2–3), (5), and (39), may have the meaning ‘broken

Reference and Predication in Movima

337

(one)’ rather than ‘break’. In the same way as nouns, then, their occurrence as a predicate may trigger an equational reading, i.e., a verb in predicate function has the meaning ‘be (the) one who does/undergoes X’. For bivalent verbs, this interpretation is equally possible, since PROX is encoded in the same way as a possessor. A direct bivalent verb can be interpreted as an undergoer-oriented participle, i.e. yeynan in (36) as ‘your wanted X’; and an inverse-marked verb can be interpreted as an actororiented participle, i.e. yeykayan in (37) as ‘your wanting/loving X’. As with predicate nominals, when functioning as predicates, these verbs create equational clauses meaning ‘your wanted X is’ and ‘your wanting/loving X is’, respectively. Further research may reveal that the direct and inverse markers originate from nominalizers that derive oriented participles (cf. Sasse 1993: 660–661 on Semitic), as is suggested by the nominalizing function of the direct marker on certain verbal bases (see (35) above). This interpretation of verbs helps to understand the cross-linguistically unusual fact that in Movima, the argument representing the lower-ranking participant (OBV) aligns with the single argument of an intransitive clause and is syntactically privileged: OBV can be interpreted as the only syntactic argument of an intransitive, equational clause, while PROX has the status of a phrasal modifier. In this way, all Movima clauses can be characterized as being of the Aristotelian type, consisting of a predicate and a subject (cf. Sasse 1991). Examples (7) and (8) above can accordingly be paraphrased by something like ‘the oxcart was our pulled one’ and ‘the old people were our scolding ones’, respectively. The fact that PROX is associated with the participant high in the referential hierarchy can perhaps be explained by its formal analogy to a possessor in a DP: a prototypical possessor ranks higher than the entity possessed by it. While diachronically, this scenario may indeed be responsible for the syntactic patterns of Movima, it cannot be entirely maintained for the synchronic situation. The syntactic differences between possessed nouns and bivalent verbs in predicate function (cf. Table 3 above) prove that synchronically, nominal and verbal predicates do not have an identical syntactic status. Furthermore, semantic differences in the interpretation of certain content words depending on their syntactic position (cf. [32], [40] indicates that to a certain degree, conversion does exist in Movima, which would not be the case if the semantics of content words were independent of syntactic position (on this issue see Croft 2001: 70–75; Croft 2005: 432–434; Evans and Osada 2005: 367–375; Himmelmann 2005b: 131).

338

Katharina Haude

4. Conclusion In Movima, all content words can function as predicates and, in combination with a determiner, form part of referential expressions. According to my interpretation, and apart from very few exceptions, both nouns and verbs retain their meaning independently of their function in the clause. This phenomenon has also been observed in other areally and genetically unrelated languages, where it has given rise to concepts such as ‘omnipredicativity’ (Launey 1994), ‘flexibility’ (Hengeveld 1992), or ‘syntactic uniformity’ (Himmelmann 2008). Movima seems to represent a particularly strong example of this language type. The determiner, which indicates semantic and spatio-temporal properties of the referent, provides clear evidence that verbs in referring expressions denote event participants and not events, an interpretation that can be extended to verbs in predicate function. A further feature of Movima, briefly mentioned above, is that both bivalent verbs and possessed nouns undergo the same valency-decreasing operation, which emphasizes their similar syntactic status. The existence of a nominalizing function of the direct marker on certain verbal bases points to a possible origin of synchronically bivalent verbs from oriented nominalizations. Movima syntax furthermore displays a feature that has so far not been described for any other language: the expression of the arguments in a transitive clause is determined by the position of the event participants in a referential hierarchy, and semantic roles are indicated through direct and inverse morphology on the predicate. This leads superficially to an unusual alignment split, with direct clauses patterning ergatively and inverse clauses patterning accusatively. The present analysis, under which synchronic transitive clauses are interpreted as basically intransitive, equational clauses with a possessor-like modifier on the predicate, may contribute to an explanation of this apparently unique pattern.

Notes 1.

This article was prepared within the Movima documentation project financed by the DoBeS programme of the Volkswagenstiftung. The presented data were collected within the Spinoza program ‘Lexicon and Syntax’ at the Radboud University Nijmegen. I am very much indebted to the Movima speakers for teaching me their language. For discussions on the topic presented here I owe special thanks to Werner Drossard, Spike Gildea, Theresa Hanske, Nikolaus

Reference and Predication in Movima

339

Himmelmann, Francesc Queixalós, and Hans-Jürgen Sasse. I thank Eva van Lier and Loretta O’Connor, as well as the reviewers Peter Arkadiev and Hein van der Voort and the editors of the volume, for their detailed comments on the manuscript. Any remaining errors are my own responsibility. 2. The data on which the study is based were collected in Santa Ana del Yacuma between 2001 and 2007. Unless otherwise indicated, all examples stem from spontaneous discourse. Symbols (partly adapted from the Leipzig Glossing Rules) are: = internal cliticization; -- external cliticization; < > infix; ~ reduplication. Abbreviations in glosses are: 1 first person, 2 second person, 3 third person, AB absential, ART article, CAUS causative, CL classifier, CO coparticipant, CTF counterfactual, DEM demonstrative, DET determiner, DP determiner phrase, DR direct, DSC discontinuous, EV evidential, F feminine, INV inverse, IRR irrealis, LV linking vowel, M masculine, MD middle, MOD modal, N neuter, NEG negative particle, NMZ action/state nominalization, OBL oblique, OBV obviative marking, PST past, PL plural, PRO free pronoun, REL relativizer, SG singular. 3. Tense is not always overtly encoded in Movima; in the translations of the examples, tense is chosen according to the original context. 4. See Bickel (in press) for first using the labels ‘proximative’ and ‘obviative’ for the nominal constituents in Movima. These labels stem from the description of Algonquian inverse systems and refer to the formal distinction of third-person arguments based on their discourse status. The capital letters in ‘Proximate Argument’ and ‘Obviate Argument’ used here are intended to show that, while semantically/pragmatically based, they refer to formal categories (cf. Haspelmath 2007: 125). 5. Nouns that are not obligatorily possessed are marked for a first-person singular possessor by a proclitic element J; this marker is optional with obligatorily possessed nouns and, as in (39), with bivalent predicates (cf. Haude 2006: 234; 271–272). 6. Only articles will be treated here in detail. For the system of demonstratives, see Haude (2006: 174). Free pronouns are glossed as PRO, bound pronouns are not glossed for their category, but simply for the person they mark. 7. Some presential and past forms of the article are homophonous with absential forms of bound pronouns. 8. The ‘obviative’ marker k- on a third-person bound OBV pronoun occurs whenever PROX is or includes a third person, and would therefore be expected here. 9. The occurrence of TAM- or discourse particles inside a DP is common, also when the DP contains a noun (cf. Haude 2006: 510). 10. It is possible that inside a DP, these verbs are interpreted as right-headed nominal compounds. Alternatively, it may be the direct marker that is responsible for the changed ‘argument structure’ of these verbs when occurring in a DP. Two other exceptions, which behave like argument-incorporating verbs,

340

Katharina Haude

are the monovalent verbs samna ‘weave’ and ya:lo:we ‘drink’. The verb samna is problematic in that it is the only verb with the form of a direct bivalent verb (root + DR) that behaves like a monovalent verb when functioning as predicate. The irregularity of the verb ya:lo:we may have to do with its historical complexity, which may also have involved incorporation: ya:- ‘under(?)’, -lo ‘CL.liquid’ and -we ‘CL.person’. 11. See also Queixalós (2006) on Tupi-Guaranían, including a discussion of the diachronic instability of omnipredicative systems.

References Bickel, Balthasar in press Grammatical relations typology. In The Oxford handbook of linguistic typology, Jae-Jung Song (ed). Oxford: Oxford University Press. Broschart, Jürgen 1997 Why Tongan does it differently: Categorial distinctions in a language without nouns and verbs. Linguistic Typology 1: 123–165. Croft, William C. 2001 Radical Construction Grammar. Oxford: Oxford University Press. 2003 Typology and Universals (2nd edn). Cambridge: Cambridge University Press. 2005 Word classes, parts of speech, and syntactic argumentation. Linguistic Typology 9 (3): 431–441. Evans, Nicholas and Toshiki Osada 2005 Mundari: The myth of a language without word classes. Linguistic Typology 9 (3): 351–390. Haspelmath, Martin 2007 Pre-established categories don’t exist: Consequences for language description and typology. Linguistic Typology 11 (1): 119–132. Haude, Katharina 2004 Nominal tense marking in Movima: Nominal or clausal Scope? In Linguistics in The Netherlands 2004, Cornips, Leonie and Jenny Doetjes (eds.), 80–90. Amsterdam/Philadelphia: Benjamins. 2006 A grammar of Movima. Ph. D. diss., Radboud Universiteit Nijmegen. in press Hierarchical alignment in Movima. International Journal of American Linguistics. forthc. a Embedded clauses in Movima. In Subordination in South American Languages, Rik van Gijn, Katharina Haude, and Pieter Muysken (eds.). forthc. b The intransitive basis of Movima clause structure. In Ergativity in Amazonia, Spike Gildea and Francesc Queixalós (eds.).

Reference and Predication in Movima

341

Hengeveld, Kees 1992 Nonverbal Predication. Theory, Typology, Diachrony. Berlin/New York: Mouton de Gruyter. Hengeveld, Kees and Jan Rijkhoff 2005 Mundari as a flexible language. Linguistic Typology 9 (3): 406–431. Himmelmann, Nikolaus P. 2005a Tagalog. In The Austronesian Languages of Asia and Madagascar, Adelaar, Alexander and Nikolaus P. Himmelmann (eds.), 350–376. London/New York: Routledge. 2005b The Austronesian Languages of Asia and Madagascar: Typological Characteristics. In The Austronesian Languages of Asia and Madagascar, Alexander Adelaar and Nikolaus P. Himmelmann (eds.), 110–181. London/New York: Routledge. 2008 Lexical categories and voice in Tagalog. In Voice and Grammatical Functions in Austronesian Languages, Peter Austin and Simon Musgrave (eds.), 247–293. Stanford: CSLI. Jelinek, Eloise and Richard A. Demers 1994 Predicates and pronominal arguments in Straits Salish. Language 70: 697–736. Kinkade, M. Dale 1983 Salish evidence against the universality of ‘noun’ and ‘verb’. Lingua 60: 25–40. Launey, Michel 1994 Une Grammaire Omniprédicative. Paris: CNRS Editions. 2002 Compound nouns vs. incorporation in Classical Nahuatl. In Problems of Polysynthesis, Nicholas Evans and Hans-Jürgen Sasse (eds.), 113– 134. Berlin: Akademie Verlag. Payne, Thomas E. 1997 Describing Morphosyntax. A Guide for Field Linguists. Cambridge: Cambridge University Press. Queixalós, Francesc 2006 The primacy and fate of predicativity in Tupi-Guaraní. In Root Classes and Lexical Categories in Amerindian Languages, Lois, Ximena and Valentina Vapnarsky (eds.), 249–287. Vienna: Peter Lang. Queixalós, Francesc (ed.) 2001 Des noms et des verbes en tupi-guarani: état de la question. Munich: LINCOM EUROPA. Sasse, Hans-Jürgen 1991 Predication and sentence constitution in universal perspective. Semantic universals and universal semantics, Dietmar Zaefferer (ed.), 75–95. Berlin, New York: Foris.

342

Katharina Haude

1993

Syntactic categories and subcategories. In Syntax. An International Handbook of Contemporary Research (Vol. 1), Jacobs, Joachim; Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), 646–686. Berlin/New York: de Gruyter. Schachter, Paul 1985 Parts-of-speech systems. In Language Typology and Syntactic Description, Vol. 1: Clause Structure, Shopen, Timothy (ed.), 3–61. Cambrigde: Cambridge University Press. Zúñiga, Fernando 2006 Deixis and alignment. Inverse systems in indigenous languages of the Americas. Amsterdam/Philadelphia: Benjamins.

All typologies leak: Predicates of change in Lowland Chontal of Oaxaca Loretta O’Connor

1. Introduction In his (1921) book, Language, Sapir made the famous observation, “All grammars leak” (38). By this he meant that within the systematic paradigms, rules and routinized patterns of any grammar, we always find a few irregularities and surprises. The same can be said for linguistic typologies. Typological theories are critical tools for linguists, for exploring differences and similarities among languages, for learning about the cognitive factors and social practices that make languages the way they are, and for making predictions about other properties of languages that are members of a certain type. So what do we do when a typology leaks? This paper follows the spirit of such work as Aske (1989) on path types and Mithun and Chafe (1999) on grammatical relations types to understand the grammatical and functional motivations of language-internal typological diversity: that is, why and how a single language uses patterns and constructions of more than one type. In doing so, it offers support for calls in the literature to broaden our focus to consider typologies of constructions and of usage patterns of language-specific resources (e.g. Bickel 2007, Slobin 2004, 2008). The data are from my in-depth study of predicates of change in Lowland Chontal of Oaxaca, an endangered language of southern Mexico.1 I approached the study of expressions of change in Chontal using Talmy’s well-known typology of lexicalization patterns (1985, 1991, 2000) but found that no single type provided a satisfying way to characterize the language as a whole. The solution that emerges here is to evaluate the language-internal lexicalization typology, identifying grammatical and functional usage patterns of each predicate type.2 Section 2 serves as introduction to the language, the lexicalization typology, and the rich variety of structural patterns in Chontal. In Section 3 I present the three change predicates in the context of particular grammatical and discourse functions afforded by each type. Section 4 summarizes the problems with an assessment of language type, and the final section illus-

344

Loretta O’Connor

trates the advantages of investigating Chontal expressions of change as a set of construction types.

2. Chontal and the lexicalization typology of languages Lowland Chontal of Oaxaca is an endangered indigenous language spoken near the Pacific coast in southeastern Mexico. Together with sister language Highland Chontal it forms the Chontal of Oaxaca family, a small and unclassified language family often linked to the proposed Hokan stock of languages that are or were spoken throughout southwestern United States and northwestern Mexico, and/or to the Jicaque languages of Honduras. Today there are probably fewer than 100 fluent first-language speakers of Lowland Chontal and perhaps another few hundred fluent first-language speakers of Highland Chontal. This paper deals exclusively with the lowland variety. Typologically, Chontal is a verb-initial, head-marking language with variable constituent order and no case marking. It has a complex aspectual system but no tense marking. Nominal morphology is mostly prefixing, and verbal morphology is mostly suffixing. The major person-marking paradigm is an agentive system responsive to the perceived volition or intention of the participant with respect to change. Person-marking morphology in the agentive (AGT) series can occur as free pronouns but frequently occur as clitics. Third persons have no AGT markers but can be expressed lexically or as polyclitic pronouns.3 The patientive (PAT) series of person markers are verbal affixes that reference non-agentive participants, defined for this language as undergoers of involuntary or unintentional change. Third person singular has no PAT marker, neutralizing the agentive / nonagentive distinction for this person. The main points of this paper are drawn from the larger qualitative study of the expression of change in Chontal, using data collected from narratives, naturally-occurring speech, response to non-verbal stimuli, and response to elicitation. Chontal has three formal predicate types to depict change events, defined as any event in which an undergoer moves to a new location, shifts to a new position, or transforms to a new state, with respect to an endpoint. The structures are illustrated in Table 1, with the formal structure that expresses ‘change’ in bold type. In a simple predicate, change is encoded in the verb root. In a compound stem predicate, the semantics of change are expressed by a typically two-part construction formed of an initial verbal element (V1) and a mor-

Predicates of change in Lowland Chontal

345

pheme of direction or topological relation (DTR) that may or may not be independent verbs and may or may not independently encode change. And finally, in an associated motion (AM) predicate, change is encoded in verbal morphology that follows all other derivational morphology and can occur verb-finally as inflection (hence, the parentheses in the table). The main verb of an associated motion predicate is most often a process or stative root that does not encode change. Table 1. Formal predicate types in Chontal Predicate type simple compound stem associated motion

Form – (derivation) – inflection V1-DTR – (derivation) – inflection VERB – (derivation) – AM – (inflection) VERB

I use concepts and terminology from Talmy’s seminal work on lexical and constructional typology to describe the locus of change semantics in expressions of change in Chontal. In Talmy’s approach, languages are classified into two framing types by locating the core schema, or characterizing semantics, of any event description. The core schema is identified as, for example, the path of a motion event, or the changed property of a state change event (Talmy 1991: 480). If the core schema is found in the verb root, then the language is classified as verb-framed, and if it is expressed outside the verb root, e.g. in verbal morphology, adpositions, or adjuncts, it is what he calls a satellite-framed language. The claim is that other factors such as manner or cause will be found where the core schema is not and that the framing designation holds for the primary or most frequent means of expressing each event type in a given language. I also adopt a third category of framing, drawing on work by Slobin (2004) and Zlatev and Yangklang (2004), among others. In a language with equipollent framing, the core schema is located in two or more elements of equal syntactic status, such as the multiple verbal elements of a serial verb construction. My claim in this analysis is that the core schema of a change event is the fact of change, located in a simple or complex verbal structure. Adapting the definitions of framing type to the structural patterns in Chontal, the three change predicate types in Chontal represent all three language types in the Talmy-Slobin typology:4 simple predicates are verb-framed; compound stem predicates are equipollently-framed, and associated motion predicates are satellite-framed. The next section examines each predicate type and highlights the functions of each one.

346

Loretta O’Connor

3. Predicates of change in Lowland Chontal 3.1. Simple predicates: verb-framed change Simple predicates are formed by an inflected verbal root or stem without special morphology of ‘change’. There are 73 simple predicates in the current corpus that express caused and uncaused change with respect to an endpoint. The semantics of change are in the verbal root, as in the following examples of change of location (1), change of position (2), and change of state (3).5 (1)

may-pa=ya’ el pana’ f’aj-pa=ya’ go-PFV=1S.AGT DET river ascend-PFV=1S.AGT ‘I went to the river and climbed a tree.’

(2)

pang-na-pa sa=ya’ sit-TERM-PFV.SG CLIT=1S.AGT ‘Then I sat down there, crying.’

lya’ there

(3)

iya=sa jas-p-ola’ 1S.AGT=CLIT tear-PFV-3P.PAT ‘I tore the papers.’

lan-je’e DET.PL-paper

el DET

’ej tree

joo-gi cry-DUR.SG

Simple predicates of change of position and change of state include both transitive and intransitive verbs, while simple predicates of change of location are all intransitive. If the change is perceived as non-agentive, this is indicated by non-agentive person marking, as in (4). (4)

joypa xwij-ko-p-ola’ lan-naranja already turn_yellow-APPL-PFV-3P.PAT DET.PL-orange tye-p-ola’ lamats’ fall-PFV-3P.PAT earth ‘The oranges ripened and fell to the ground.’

3.1.1. Semantic typology of the path of change The notion of change with respect to an endpoint is a critical criterion in distinguishing predicates that express change. Typological studies of the concepts of motion and path (Aske 1989, Slobin 1996, 1997, Levinson and

Predicates of change in Lowland Chontal

347

Wilkins 2006: 531–536) point out that there are multiple types of path encoded in the languages of the world and that path type often correlates with other patterns within a language. Change of location verbs in Chontal show two types of path. One type, called translational or translocational in the studies cited above, asserts a change over a trajectory and implies an endpoint. Simple predicates with a translational path all orient to Goal in Chontal; these were exemplified in (1). Another path type, called boundary-crossing, state change, or nondurative change of location, asserts a change at endpoint and implies a trajectory. In Chontal this path type is identified by the inflectional patterns of certain roots. In (5) and (6), the change of location is expressed as ongoing, following a perfective departure from Source. (5)

jaape=ma’ ay-pa where=2S.AGT depart_away-PFV.SG ‘Where are you going (where have you departed toward)?’

(6)

pa’-pa lakwi come-PFV.SG rain ‘It’s going to rain (rain is coming, has departed toward here).’

Chontal has manner of motion verbs (e.g. run, walk, swim) and a series of verbs that depict ‘transport’ which can depict localized motion but which do not imply direction or endpoint. The root pe- means ‘transport a small thing’, and pepa with perfective inflection could mean ‘he took, he brought, he carried’ or even ‘he held’. Manner verbs typically recruit a change verb to supply the notion of change with respect to an endpoint. (7)

ñulye-jlay-pa’ la’way’ ma+la+y-pa’ tyuwaj-lado, run-PL-PFV.PL children go+PL+-PFV.PL other-side tyan-lado el parke that-side DET park ‘The kids ran across the park (ran, went to that side of the park).’

In example (7), the manner verb ñulye- ‘run’ depicts localized manner of motion, and the path verb may- ‘go’ encodes the change of location to the other side of the park.6

348

Loretta O’Connor

3.1.2. Summary of simple predicates Simple predicates, with semantics of change in the verb root, constitute the verb-framed portion of the grammar of change. Change of location predicates orient to a specific endpoint in space, either the Source (the starting point) or Goal (the ending point) of motion, while manner of motion verbs and transport verbs orient to neither and are not considered change predicates. The endpoint of change of position or change of state is the position or state expressed by the verb. The primary functions of simple predicates are to depict spontaneous change of location and position and caused and uncaused change of state. 3.2. Compound stem predicates: equipollently-framed change In addition to simple predicates, Chontal has a second formal predicate type that encodes change in a typically two-part compound stem construction. Each element in the compound stem makes a specific contribution to the meaning of the predicate, yet it is the construction itself that consistently depicts ‘change’ (see O’Connor 2007: 20–25, 156–165). The initial element of the compound stem, here called a V1, falls into one of four notional classes according to the semantics of the element and to the patterns of combination with the second element. The labels were chosen to capture broadly the various senses within each category. – A means V1 describes the manner in which or means by which the change takes place. – A classificatory V1 identifies the shape, size, or identity of the undergoer of change (semantic Theme) or, more rarely, describes the Goal participant. – A dispositional V1 identifies the posture, spatial disposition, or configuration of the undergoer participant with respect to the endpoint of change. – A trajectory V1 depicts the specific shape of the path taken by the undergoer in motion.

There are approximately 100 V1 elements in the current corpus, some of which are stand-alone verbs but many of which are not. These combine with one of a dozen morphemes that encode direction or topological relation (DTR), such as ‘up’, ‘down’, ‘in’, ‘inside’, ‘out’, ‘up on’, ‘down on’, ‘down in’, ‘across’, and ‘on the edge’. A few of these fundamentally spatial elements are verb roots, capable of standing alone and taking inflection

Predicates of change in Lowland Chontal

349

to form a complete verbal word, but most are bound elements that only occur within a compound stem construction. Not all combinations of V1 and DTR are allowed: there are about 175 such predicates in the current corpus. In a given compound stem, either, both or neither element might be an independent verb, and either, both or neither might encode change. In my analysis, the name of the V1 notional class also characterizes the compound stem construction. Most change of state compound stems are means constructions, and most change of position compound stems are dispositional constructions. A change of location might be encoded by either of these, a classificatory construction, or a trajectory construction. The endpoint of change is the spatial orientation, topological relation, or result state depicted by the DTR element. 3.2.1. Means predicates: adverbial function Unlike English, with the suffix -ly, or Spanish, with -mente, Chontal does not have a productive process for forming adverbs that describe the manner or means in which an event occurs. Instead, this detail can be provided with a means construction. The following examples describe changes at or to an endpoint ‘inside’, achieved by ‘digging’ (8) or ‘stabbing’ (9). (8)

awe’ el pana’ pu-k’oy-pa lamats’ big DET river dig-inside-PFV.SG earth ‘The river is big, it eroded (dug inside) the earth.’

(9)

k’ane-duy=ya’ lay-pich’ale extend-DUR.SG=1S.AGT my-clothes sk’wi-k’oy-yuy-ya’ jaape lukwix stab-inside-DUR.SG=1S.AGT where rope ‘I’m hanging clothes (by) sticking the tips into the clothesline.’

A frequent pattern with change of state verbs is the use of the element -ñi ‘across’ to signify ‘apart, separate’. Compare the earlier example (3) with the constructed sentence below. (10) iya=sa jas-ñi-p-ola’ 1S.AGT=CLIT tear-across-PFV-3P.PAT ‘I tore the papers apart.’

lan-je’e DET.PL-paper

350

Loretta O’Connor

The difference in (10) is that the endstate is explicitly described as being ‘apart’, an earlier whole now separated into more than one piece. 3.2.2. Classificatory predicates: referent introduction and tracking Classificatory V1 elements index some feature of the undergoer of change (size or shape) (11), identify the type of undergoer (as liquid, grain, or animate) (12), or identify the semantic Goal (as a container or a human) (13).7 p-ayj-pa (11) tyinchi maa=yma’ why NEG=2S.AGT small-down-PFV.SG ‘Why don’t you take off your backpack?’

lo-bolsa your-bag

(12) ch’uj-’mi-’ma=yma’ ten sa=yma’ maj-ko-da grain-in-IPFV=2S.AGT what CLIT=2S.AGT cook-APPL-SNJCT.SG ‘You put in whatever grain you want to cook.’ (13) sa=yma’ wa-s-’mi-’ma jaape CLIT=2S.AGT container-AND-in-IPFV.SG where ‘You put it in your little purse.’

lo-na’wa-bolsa your-DIM-bag

The primary discourse functions of classificatory constructions involve referent introduction and tracking. In (14), an excerpt from instructions for cooking fish stew, the cooking pot is introduced and tracked twice as a ‘container’ before the first lexical mention in (d). (14) a. wa-s-’mi-’ma sa=ya’ container-AND-in-IPFV.SG CLIT=1S.AGT ‘I put in a banana (in the pot),’ b. wa-s-’mi-’ma sa=ya’ container-AND-in-IPFV.SG CLIT=1S.AGT ‘I put in salt (in the pot),’ c. wa-s-’mi-’ma sa=ya’ container-AND-in-IPFV.SG CLIT=1S.AGT ‘I put in onion (in the pot).’ d. joypa sa pulya sa lapixu already CLIT boiling CLIT pot ‘Once the pot is boiling...,’

lankiña, banana li-ju’e, its-salt li-seboya. its-onion

Predicates of change in Lowland Chontal

351

The ability to introduce and disambiguate among multiple referents is particularly important in a language like Chontal, which has no morphology of nominal classification and uses a set of semantically general clitics to track third-person referents in discourse. 3.2.3. Dispositional predicates: spatial precision and referent tracking Dispositional predicates describe in rich detail the spatial disposition of the undergoer of change at endpoint, as for example ‘mouthward upon’ (15) or ‘sitting down in’ (16). (15) k’o-ma-f’i-yuy iya’ li-tapadera mouth-X-up_on-DUR.SG 1S.AGT its-lid ‘I put/am putting a lid on my well.’

lay-k’ejwa’ my-well

(16) jojl-may-pa lapixu’ maj-lixpantalek-’ej sitting-down_in-PFV-SG pot LOC-forked_branch-tree ‘The pot sat/is sitting in the crotch of the tree.’ In addition, dispositional predicates play a role in referent tracking. The narrative excerpt (17) is from a description of the rituals for the dead. The body of the deceased is mentioned lexically in (a), tracked by the classificatory element le- ‘animate’ in (b) and (c), and then referenced with a dispositional element ñaj- ‘lying’ in (d). (17) a. chee-duy le-f’-kix-pa lemayñe go_return-DUR.SG anim-up-AND-PFV.SG corpse ‘He goes and lifts up the deceased.’ b. ñoy-yuy lamats’ para naa=sa=‘le lay-DUR.SG earth for REF=CLIT=thing le-ñi-wa lamats’ anim-across-PROG.SG earth ‘He lays the body on the ground so that the earth receives its spirit.’ c. joyya sa le-f’-na-’me’ sa after CLIT anim-up-TERM-IPFV.PL CLIT ‘And then they lift the body.’

352

Loretta O’Connor

d. tyijpe sa ñaj-’mi-yay’ jaape li-ñejwijma there CLIT lying-in-DUR.PL where his-mat ‘There they place him on his sleeping mat.’ Example (18) from the same story demonstrates that some compound stems combine a series of two DTR elements to depict an elaborated trajectory of motion, here, ‘up’ and then ‘in’. (18) tyijpe sa ñaj-f’-’mi-pa’ jaape li-caja there CLIT lying-up-in-PFV.PL where his-coffin ‘There they picked him up and put him in his coffin.’ 3.2.4. Trajectory predicates: elaborated path of motion Only a few compound stem predicates in my corpus depict an elaborated trajectory by stacking DTR elements as a series of endpoints. More commonly, details of path shape are expressed by the initial element of a trajectory predicate, such as a flat arc (19), a high arc (20), or a straight line (21). (19) ‘oy-f’i-’ma sa=ya’ flat_arc-up_on-IPFV.SG CLIT=1S.AGT ‘I will saddle my burro.’

lay-buru my-burro

(20) xpa-gi-’ma sage=l kasi para sa fa-’ma high_arc-out-IPFV 3S=DET chili to CLIT plant-IPFV ‘He would transplant the chili to plant it (in the field).’ (21) joypa sa=yma’ ki-ñi-pa el puente de Piña already CLIT=2S.AGT straight-across-PFV the_bridge_of_Piña ‘Once you crossed the Piña bridge...’ Trajectory predicates constitute the smallest group of compound stem predicates. They sometimes play a backgrounding role in discourse by directing the focus away from the undergoer. Example (22) shows one way to describe putting a breadsheet into the oven, with the classificatory predicate wask’oy- ‘put inside a container’.

Predicates of change in Lowland Chontal

(22) wa-s-k’oy-’ma=yma’ container-AND-inside-IPFV.SG=2S.AGT ‘You put your bread in the oven.’

lo-’i your-bread

353

maj-horno LOC-oven

In a different story about bread-making, the speaker describes making the dough, letting it sit, and, when it has risen, putting it in the oven. (23) joypa sa cho-lyu-pa ’oy-k’oy-ta already CLIT rise-CLOC-PFV.SG flat_arc-inside-DLOC.SG sa=yma’ con el pala CLIT=2S.AGT with the paddle ‘Once it has risen, you go off and put it in with the paddle.’ In (23), the activities involved in the process of bread-making are the salient topics, and the speaker backgrounds both bread and oven with a trajectory predicate that highlights the pathshape of the final activity. 3.2.5. Summary of compound stem predicates The compound stem predicate type is defined as a change construction, or more accurately, a family of change constructions, each one a meaningful structure composed of two elements of equal syntactic status. The semantics of change are in the combination of elements in the multi-part construction, so these predicates arguably constitute the equipollently-framed portion of the grammar of change, in terms of the Talmy-Slobin typology. Compound stem predicates express most caused change of location, both caused and uncaused change of position, and play a minor role in depicting the realization state or location of a change of state. 3.3. Associated motion/change predicates: satellite-framed change A third type of change predicate in Chontal is composed of an inflected verbal root or stem with additional morphology of associated motion and associated change, here abbreviated as AM. In the literature (e.g. Wilkins 1991), AM is described as a semantic category that associates a subevent of motion prior to, simultaneous with, or subsequent to the action of a main verb. The AM paradigm in Chontal is comprised of only four suffixes, yet the set is internally diverse in seman-

354

Loretta O’Connor

tics, distribution, and function. The three suffixes to be discussed here8 depict a displacement in space or time prior to the action or state of main verb, supplying a boundary-crossing or non-durative path of change. While simple and compound stem predicates situate some participant as an undergoer of change, most AM predicates instead situate an event or state in a certain place or time. 3.3.1. Locate an event in another place Three of the AM morphemes occur with virtually any process verb in the language and with certain change verbs to signal that the action of the main verb takes place in the context of a ‘go and do’ or ‘come and do’ subevent. In (24), an andative suffix meaning ‘go from Source’ indicates that the ‘buying’ took place after motion away from the speech event. (24) ’ña-s-pa naa ñulyi montón lampityu’ buy-AND-PFV.SG REF one pile whistles ‘He went and bought a pile of whistles.’ Andative morphology also occurs in the imperative, with a special suffix that also signifies ‘motion’, as a command to do the verb action elsewhere. (25) sma-s-ki sleep-AND-IMPV.MOV.SG ‘Go and sleep!’ Similarly, the venitive suffix of ‘come from Source’ means the ‘asking’ event in (26) took place after someone had come from somewhere else. (26) jl-nu’ee-way-pa naa=sa 1S.PAT-ask_for-VEN-PFV.SG REF=CLIT ‘She came and asked for me.’ Both andative and venitive encode a departure from Source and occur with perfective or imperative inflection. In contrast, a third AM morpheme, the dislocative, appears to have fused with imperfective aspect and occurs verb-finally. This suffix contributes semantics of ‘go to Goal’ and signals

Predicates of change in Lowland Chontal

355

that the action of the main verb takes place after arrival at a Goal location (27). (27) joola sa o-melyu ’na-ta=yma’ panela if CLIT your-money buy-DLOC.SG=2S.AGT sugar ‘If you have money, you go and buy some sugar.’ Another typical function of dislocative inflection is to specify an activity as the Goal of a motion event, in a motion-cum-purpose construction, as in (28). (28) lay-pekwe may-pa x’wi-tya my-husband go-PFV.SG chop_bush-DLOC.SG ‘My husband went to clear his land.’ The endpoint of the simple predicate may- ‘go’ is an unspecified Goal location. In a motion-cum-purpose construction, the AM predicate casts the event of the main verb as the Goal, the ‘endpoint activity’ of the preceding change of location event. In effect, ‘My husband went to do some bushchopping.’

3.3.2. Locate a state along a path of change In a temporal sense, ‘go and do’ morphology affects the positioning of a stative main verb event in time. Andative derivation with perfective inflection changes a stative root into an inchoative predicate (29), while dislocative inflection depicts the result of a state change (30). (29) joypa k’ej-’mi-pa=yma’ manj-s-pa lapixu’ already water-in-PFV=2S.AGT be_full-AND-PFV.SG pot ‘You poured the water, and the pot is filling.’ (30) spojl-’ma sa=yma’ manj-ta nestle-IPFV.SG CLIT=2S.AGT be_full-DLOC.SG ‘You nestle them (inside the pot) until it’s full.’

sa CLIT

When suffixed to a stative root, AM morphology exploits the path semantics of a boundary-crossing type of path, encoding the departure to-

356

Loretta O’Connor

ward the new state, with andative plus perfective, and the arrival at the new state, with dislocative. Phasal verbs, too, are usually derived and inflected with andative and dislocative morphology. 3.3.3. Frame change of position as state change The role of AM morphology in encoding change of position is a minor one and can best be described as framing a change of position as a change of state. An interesting example of this occurs with the verb root pang-, which means ‘sit’ or ‘live’. An AM construction with pang- denotes ‘being born’, a non-agentive state change (31). (31) a. pang-ix-p-ola’ sit-AND-PFV-3P.PAT ‘They are/were being born.’

b. pang-t-o’ sit-DLOC-2S.PAT ‘You were born.’

3.3.4. Summary of associated motion predicates In AM predicates, the semantics of change are in the AM morphology, making these predicates the satellite-framed portion of the grammar of change. The primary function of AM predicates is to situate not a participant but an event or state in a specific place or time. With a process or change root, the event of the main verb is situated at a particular point in space, sometimes casting the Goal itself as an activity more than an identifiable location. With a stative root, the state of the main verb is situated at a particular moment in time, as the inception of change or as the result state of change. Although they play an important role in the language ecology of Chontal, AM predicates are not the primary or most frequent means for expressing a change of location, position, or state, and therefore this framing type cannot be argued to characterize the language as a whole.

4. Language framing type and predictive power This analysis of expressions of change in Chontal presents a mixed picture for assessing language type. The grammar of change involves three verbal constructions of distinct morphosyntactic make-up and with distinct functional roles.

Predicates of change in Lowland Chontal

357

The primary function of a change event, in which an undergoer moves to a new location, shifts to a new position, or transforms to a new state, is divided between simple predicates and compound stem predicates. Simple predicates mostly encode uncaused change of location, uncaused change of position, and caused and uncaused change of state. Compound stem predicates mostly encode caused change of location, caused change of position, and caused change of state with a specified realization as ‘apart’. The subevent of change in an AM construction serves an important complementary role, often working in combination with other predicate types, to express change of location, change of position, and change of state. In the larger study cited previously (O’Connor 2007), I evaluated the predictive powers of the Talmy-Slobin typology through an analysis of narrative style in Chontal. I devised expectations for narrative style, involving factors such as the type of path, the number of endpoints, and the quality of manner description, extrapolating primarily from Slobin’s (1997) predictions for the types of patterns expected in motion event descriptions. The results of narrative style analysis in Chontal were again mixed in terms of language framing type, and they highlighted the value of a constructionbased approach that takes language ecology, grammatical function, and discourse function into account.

5. Construction type and predictive power A better solution is to explore the typologically interesting questions that arise from a consideration of the individual constructions. The comparative study of constructions involves examination of the types of meaning encoded, the types and combinations of structural elements that express these meanings, and the discourse-functional roles and language-internal properties that influence the usage patterns of particular constructions, all within the context of the individual genetic inheritance and contact history of a given language. The typological study of meaning investigates the types of semantic distinctions made within a particular domain, at ever-finer degrees of detail, identifying semantic parameters and dependencies as units and patterns that can be compared across languages. This means, for instance, breaking an abstract construct like ‘manner of motion’ into subcomponents such as velocity, gait, medium, effort, posture of the undergoer, and psychological state of the undergoer. My own work on Chontal has involved methods and tools of semantic typology developed at the Max Planck Institute in Ni-

358

Loretta O’Connor

jmegen (e.g. Levinson and Wilkins 2006: 1–23) for analysing semantic domains of motion, location, placement (‘put and take’), and object separation (‘cut and break’). A typological approach at this level of detail addresses cross-linguistic similarities and differences in the categorizing and encoding of finely grained semantic distinctions. From the perspective of morphosyntactic typology, we know that meaningful elements such as the semantics of space can be distributed throughout the clause (Sinha and Kuteva 1995, Grinevald 2006), in verbs, nouns, particles, adpositions, affixes, and constructions. Investigations by Grinevald and colleagues in the Trajectory Research Group at Lyon explore the semantic structures in lexical and grammatical forms of path expressions in languages throughout the world. Many languages have inventories of spatial elements similar to the paradigms I treat separately in Chontal, as associated motion morphology and elements of direction and topological relation. It may be useful to type languages according to the size and membership of these paradigms, to the ways in which elements can be combined, and to the semantic classes of verbs with which they occur. For example, Grinevald (2006) shows that in Jacaltek Popti’, spatial elements can be divided into three distinct sets, with fixed position in combinations, and that these directionals can even attach to verbs of perception and locution, to stipulate the fictive path of gaze or speech. The topic of associated motion itself portends promising ground for typological investigation. The AM paradigm in some languages is quite extensive, such as the fourteen suffixes of deictic and directional motion in the Central Australian language Mparntwe Arrernte, described by Wilkins (1991). Cavineña, a Tacanan language of Bolivia, boasts eleven suffixes that qualify the direction and aspectual properties of motion as well as the precise location and stability or permanence of the target of motion (Guillaume 2007). In other languages, morphology of associated motion is used not only for spatial deixis but can also serve aspectual functions, be used in anaphora, and/or play a role in participant marking (Hooper 2002 on Tokelauan, Polynesian; Margetts 2002 on Saliba, Oceanic). The extended role of associated motion morphology cross-linguistically underlines the value of investigating usage patterns of constructions within the context of language-specific resources. We may be more likely to find complex predicates such as the Chontal means construction in languages without a productive process of adverb formation. Similarly, we may find that languages without referent-individuating pronominal operators are more likely to rely on structures like a classificatory or dispositional compound stem construction for referent introduction and management (e.g.

Predicates of change in Lowland Chontal

359

Hellwig 2003 on the classificatory properties of posture verbs in Goemai, West Chadic). Typological investigation at the level of the construction can be applied to topics of language classification. Chontal has the resources it does for reasons of genetic inheritance and of language contact. These topics go well beyond the scope of this paper, but a few observations are in order. The two-part compound stem constructions arguably represent a semantic and syntactic subset of the bipartite stem construction, providing a possible link to other candidate Hokan languages of the California linguistic area (see Mithun 2007 for a good discussion of the difficulties in characterizing the link as genetic or the result of contact). From an areal perspective, space is a well-known preoccupation of Mesoamerican languages, and indeed many languages of the area have multiple and interacting sets of formal structures that encode spatial meanings. In conclusion, in terms of the Talmy-Slobin typology of language types, this paper has shown that a classification of Chontal within a single framing type obscures the rich variety of formal and semantic structures that Chontal speakers use to express the notion of change. The analysis presented here suggests that we should take the illuminating insights from Talmy’s work on typologies of languages, expand greatly such basic notions as ‘path’, ‘manner’, and ‘changed property’ to reflect the nuanced distinctions that languages make, and use the tools of semantic and syntactic typology to investigate typologies of constructions, as products of genetic and areal development, and as used by speakers in specific grammatical, discourse and cultural contexts.

Notes 1.

This study was the basis of my 2004 doctoral thesis, later revised and published (O’Connor 2007). I acknowledge my intellectual debt to the Chontal community in and around San Pedro Huamelula, Oaxaca, for their insights and patience, and to the linguist Viola Waterhouse, for the first descriptions of this marvellous language. I am also grateful for generous funding from University of California Santa Barbara, the Humanities and Social Sciences Research Council, UC-Mexus, the Max Planck Institute for Psycholinguistics, and the Volkswagen Foundation DoBeS Initiative. I thank Alice Gaby, Antoine Guillaume, Katharina Haude, the editors, and one anonymous reviewer for insightful comments that greatly improved an earlier draft of this paper.

360 2.

3.

4.

5.

6.

7.

Loretta O’Connor Other studies have pointed out that typological tendencies can also be overridden by cultural factors (cf. Wilkins 1996 on Arrernte of Central Australia) and by dialectal differences (cf. Berthele 2004 on varieties of German). These strands will not be developed here but are important parts of the explanation of ‘leaks’ in language typologies. Chontal has a rich inventory of clitics that function in reference, deixis and interaction. Some clitic morphemes can occur alone, but mostly these attach to nouns, verbs, and to each other, forming polyclitic words. To illustrate, in the data for this paper there are many instances of sa, a clitic which occurs (1) alone, (2) with person markers, as in sa=ya’ or iya=sa for first person singular and sa=yma’ for second person singular, (3) with =ge ‘person’ or =jne ‘persons’ to form sa=ge and sa=jne, used as pronouns for third person singular and plural, respectively, and (4) with naa, a clitic that indexes a known referent, in naa=sa and naa=sa=’le. See O’Connor (2007: 55–56) for a preliminary but more extensive description of clitics in Chontal. Slobin’s analysis has evolved in recent years, departing from the basics of the Talmy-Slobin framework to take a more finely-grained approach. For example, with respect to motion events, Slobin now talks about languages with PIV and PIN constructions (path in verb vs. path in non-verb) rather than verbframed and satellite-framed languages. Only words pertinent to the discussion were fully glossed. Glossing and orthographic conventions include the following: 1S-3P person, AGT agentive, AND andative, APPL applicative, CLIT clitic, CLOC cislocative, DET determiner, DLOC dislocative, DUR durative, IMPV imperative, IPFV imperfective, LOC locative, NEG negation, PAT non-agentive, PFV perfective, PL plural, REF given referent, SG singular, SJNCT subjunctive, TERM terminative, VEN venitive, X unknown morpheme. At morpheme boundary, a hyphen (-) marks derivation or inflection, an equal sign (=) marks a clitic, and a plus sign (+) marks an infix. Special graphemes are {j} for the glottal fricative /h/ and the velar fricative /x/, {x} for the alveopalatal fricative /š/, and the apostrophe {’} for the glottal stop and for glottalization as secondary articulation. There is rampant palatalization of alveolars, conditioned environmentally, and in a regular morphophonemic process, verb-initial alveolar consonants are palatalized when the verb occurs with a third person subject (n~ñ, t~ty, s~x, l~ly). One reviewer asked if the example in (7) could be considered an instance of equipollent framing (as will be discussed in 3.2). The answer in this analysis is no, but a sequence of verbs as in (7) illustrates the likely diachronic path to the formation of both compound stems and associated motion constructions. The specific sequence of elements and the permissible semantic combinations in each of these constructions are introduced in 3.2 and 3.3, respectively. The stems was’mi- ‘put into container’ and wask’oy- ‘put inside container’ are highly unusual in that they have what appears to be an andative morpheme inside the stem. No other compound stems, including others with wa- ‘container’,

Predicates of change in Lowland Chontal

8.

361

show this pattern. One explanation is that the extra morpheme distinguishes movement to a container, as a Goal, from movement of a container, as a Theme. The remaining suffix, the cislocative, occurs exclusively with certain change of location roots, as in (23). It seems to depict simultaneous motion, supplying a durative or translational path of ‘coming from Source’. The cislocative is infrequent in my corpus and unusual within the small AM paradigm, with senses or homophonous forms not yet well understood.

References Aske, Jon 1989

Path predicates in English and Spanish: A closer look. In Proceedings of the 15th Annual Meeting of the Berkeley Linguistics Society, 1–14. Berkeley: University of Calfornia. Berthele, Raphael 2004 The typology of motion and posture verbs: a variationist account. In Dialectology Meets Typology, B. Kortmann, (ed.), 93–126. Berlin/New York: Mouton de Gruyter. Bickel, Balthasar 2007 Typology in the 21st century: Major current developments. Linguistic Typology 11 (1): 239–251. Grinevald, Colette 2006 The expression of static location in a typological perspective. In Space in Languages: Linguistic Systems and Cognitive Categories. M. Hickman and S. Robert (eds.), 29–58. Amsterdam/Philadelphia: Benjamins. Guillaume, Antoine 2007 The system of “associated motion” in Cavineña. Paper presented at the Reunion “Trajectoire”, Institut des Sciences de l’Homme, Lyon, November. Hellwig, Birgit 2003 The Grammatical Coding of Postural Semantics in Goemai (a West Chadic Language of Nigeria). Nijmegen: MPI Series in Psycholinguistics 22. Hooper, Robin 2002 The Tokelauan directional particles mai and atu. Studies in Language 26 (2): 283–313. Margetts, Anna 2002 The linguistic encoding of three-participant events in Saliba. Studies in Language 26 (3): 613–636.

362

Loretta O’Connor

Levinson, Stephen C., and David Wilkins (eds.) 2006 Grammars of Space: Explorations in Cognitive Diversity. Language, Culture and Cognition Series 6. Cambridge: Cambridge University Press. Mithun, Marianne 2007 Grammar, contact and time. Journal of Language Contact – THEMA 1: 144–167. Mithun, Marianne, and Wallace Chafe 1999 What are S, A, and O? Studies in Language 23 (3): 579–606. O’Connor, Loretta M. 2007 Motion, Transfer, and Transformation: The Grammar of Change in Lowland Chontal. (Studies in Language Companion Series 95.) Amsterdam/Philadelphia: Benjamins. Sapir, Edward 1921 Language. New York: Harcourt Brace. Sinha, Chris, and Tania Kuteva 1995 Distributed spatial semantics. Nordic Journal of Linguistics 18: 167– 199. Slobin, Dan I. 1996 From “thought and language” to “thinking for speaking”. In Rethinking Linguistic Relativity, J. J. Gumperz and S. C. Levinson (eds.), 70–96. Cambridge: Cambridge University Press. 1997 Mind, code and text. In Essays on Language Function and Language Type, J. Bybee, J. Haiman and S. A. Thompson (eds.), 437–468. Amsterdam/Philadelphia: Benjamins. 2004 The many ways to search for a frog: linguistic typology and the expression of motion events. In Relating Events in Narrative: Typological and Contextual Perspectives. S. Strömqvist and L. T. Verhoeven (eds.), 219–257. Mahwah, NJ: Lawrence Erlbaum Associates. 2008 Relations between Paths of Motion and Paths of Vision: A Crosslinguistic and Developmental Exploration. In Routes to Language: Studies in Honor of Melissa Bowerman, V. M. Gathercole (ed.). Mahwah, NJ: Lawrence Erlbaum Associates. Talmy, Leonard 1985 Lexicalization patterns: Semantic structure in lexical forms. In Language Typology and Syntactic Description. Vol. III: Grammatical Categories and the Lexicon, T. Shopen (eds.), 57–149. Cambridge: Cambridge University Press. 1991 Path to realization: A typology of event conflation. In Proceedings of the 17th Annual Meeting of the Berkeley Linguistics Society, 480– 519. Berkeley: University of California.

Predicates of change in Lowland Chontal 2000

363

Toward a Cognitive Semantics: Typology and Process in Concept Structuring. Cambridge: MIT Press. Wilkins, David P. 1991 The semantics, pragmatics and diachronic development of ‘associated motion’ in Mparntwe Arrernte. Buffalo Papers in Linguistics: 207–257. 1996 The verbalization of motion events in Arrernte Central Australia. In Proceedings of the 28th Annual Meeting of the Child Language Research Forum, 295–308. Stanford, CA: Stanford University. Zlatev, Jordan, and Peerapat Yangklang 2004 A Third Way to Travel: The Place of Thai in Motion Event Typology. In Relating Events in Narrative: Crosslinguistic and Crosscontextual Perspectives. S. Strömqvist and L. T. Verhoeven (eds.), 159–190. Mahwah, NJ: Lawrence Erlbaum Associates.

Multidimensional typology and Miraña class markers Frank Seifart

1. Introduction Nominal classification systems are complex in many languages, maybe particularly so in Amazonian languages like Miraña.1 Nominal classification systems are also notorious for taking a variety of forms across languages, which is amply illustrated in, e.g., Aikhenvald (2000, ed. 2004) and Senft (ed. 2000). Current approaches to the typology of nominal classification focus on establishing relatively few and broad universal types of classification systems, each defined by a relatively small number of morphosyntactic criteria. These types are conceived as prototypes, acknowledging that most actual systems found in individual languages are deviant from these broad types in one way or another. Some systems, like that of Miraña, are almost impossible to characterize in term of these types. This chapter argues for the need to shift the focus away from broad types defined by relatively few characteristics as the basis for typological characterization and comparison towards a model which uses a larger number of more detailed and more varied parameters. These may not only do better justice to the complexities of individual systems, but they may also be more adequate for the main purpose of typology, namely to reveal correlations and implications between logically independent linguistic features. This enterprise is in the line with a number of recent typological approaches which acknowledge that “linguistic diversity is captured by large sets of fine-grained variables, not by grand type notions” (Bickel 2007: 245). It is also related to recently articulated doubts about the existence of pre-established categories (Haspelmath 2007). Based on a brief introduction to the typology of nominal classification (Section 2), Sections 3–5 discuss Miraña data with respect to this general framework, pointing out where its application is problematic and why. Section 6 suggests how some of the types of classification systems can be more rigorously defined, excluding Miraña as one example of them and therefore yielding more coherent types. Section 7 proposes a number of

366

Frank Seifart

additional parameters for typological comparison, each representing one dimension in a multidimensional typology. Section 8 concludes this chapter.

2. Broad types of nominal classification systems Current approaches to the typology of nominal classification systems (Dixon 1986, Aikhenvald 2000, Grinevald 2000) use two main parameters to define two basic types of systems and a number of subtypes of one of these. The first main parameter is the presence vs. absence of agreement: Noun class systems are defined by the realization of noun classes in agreement, and classifier systems (as a cover term for numeral classifiers, verbal classifiers, etc., see below) by the absence of agreement. Closely related to this distinction, noun classes are defined additionally by the requirement that (almost) all nouns belong to one class only.2 The following characteristics are said to be contingent, typically correlating characteristics of noun classes: They are closed systems consisting of a smallish number of classes (2–20), their marking can be fused with other grammatical categories and there is no variation in register or according to individuals. Prototypical examples of noun classes are the well-known classes of Bantu languages (see, e.g., Katamba 2003: 111 for examples). Since agreement is the major defining criterion for this type of system, gender systems, such as those of European languages, also fall within this type (Dixon 1986; Seiler 1986: 111; Corbett 1991: 5, 105, 136; Creissels 1999; Aikhenvald 2000: 19; Grinevald 2000: 55). Correlating characteristics of classifiers, on the other hand, are said to be the following: nouns may be associated with more than one classifier, the set of classifiers is usually large and may be open, and the use of classifiers may be variable. The second main parameter in this typology applies only within systems with absence of agreement, i.e. classifier systems. It distinguishes individual classifier types by the morphosyntactic locus in which classifiers occur:3 numeral classifiers, defined by occurrence in numeral phrases; verbal classifiers, defined by occurrence in verbs; and noun classifiers, defined by occurrence next to nouns, independent of other constituents. Each classifier type is said to have correlating characteristics (in addition to the ones that are attributed to classifiers in general). For instance, numeral classifiers, of which prototypical examples are found in Mandarin Chinese (see example 13, below) and Yucatec Maya (see, e.g., Lucy 1992), typically have an

Multidimensional typology and Miraña class markers

367

individuating function and the shape of objects is prominent in their semantics.

3. The Miraña inventory of class markers The system of nominal classification in Miraña consists of over 70 class markers.4 The set of Miraña class markers is defined distributionally by exclusive occurrence in a number of specific morphosyntactic contexts, including suffixed to nouns, numerals, and verbs. The set of class markers includes six ‘general class markers’ (GCMs), which encode distinctions of animacy, sex and number, e.g. -dʒɛ (GCM.FEM.SG), -mɯtsi (GCM.MASC.DL), and -nɛ (GCM.INAN). The remaining class markers are mostly shapedenoting ‘specific class markers’ (SCMs) (examples 1). These are a diverse set ranging from a core subset of about ten frequently used monosyllabic forms with relatively broad semantics (examples 1a-c are three of these) to rarely used polysyllabic forms with relatively specific semantics (examples 1i–k). In addition to general and specific class markers, a set of about fifty nouns can be used as ‘repeaters’ (RPs) in contexts that are otherwise filled by class markers, such as numerals, where the form of the noun is repeated for agreement marking, as in tsá-bahkɯ báhkɯ (one-RP.bone bone) ‘one bone’. (1)

a. b. c. d. e. f. g. h. i. j. k.

-ko -hɨ -ʔo -ha -hɯ -ʔaːmɨ -hpajko -paːhɨ -tsoːʔo -roʔdʒo -tsaːragwa

‘one-dimensional, pointed object’ ‘two-dimensional, round object’ ‘three-dimensional, oblong object’ ‘cover’ ‘tube-shaped’ ‘leaf; two-dimensional, flexible object’ ‘liquid ‘cavity’ ‘medium-sized palm tree’ ‘completely twisted, long and thin object’ ‘unordered fibres with an upward orientation’

The set of Miraña class markers has a number of properties typically associated with classifier systems, in particular the size of the inventory and its ‘openness’ (through repeaters). The prominence of shape-related semantics, dimensionality among them, of the specific class markers is typical of numeral classifiers as one type of classifiers.

368

Frank Seifart

Miraña class markers fulfil two basic functions in the language: one is the derivation of new noun stems (described in the following section), the other is agreement marking on a variety of targets, including numerals, demonstratives, and relative clauses, and subject cross-reference on verbs (described in Section 5).

4. Derivational use of Miraña class markers on nouns A major function of Miraña class markers is to form new noun stems from bare noun roots or derived noun stems. These uses account for the majority of occurrences of class marker tokens in texts. Derivation via class markers is a productive pattern in the language, often allowing for the derivation of different nouns from the same root, using different class markers (examples 2a–f). Sometimes various class markers can be stacked on one and the same noun (examples 2e–f). (2)

a. ɯ́ hɨ banana ‘banana(s), banana bunch(es), banana plant(s), etc.’ b. ɯ́ hɨ-ʔo banana-SCM.3DIM.oblong ‘banana fruit’ c. ɯ́ hɨ-ʔi banana-SCM.bunch ‘bunch of bananas’ d. ɯ́ hɨ-ko banana-SCM.1DIM.pointed ‘banana plant’ e. ɯ́ hɨ ́-ko-ʔáːmɨ banana-SCM.1DIM.pointed-SCM.leaf ‘leaf of a banana plant’ f. ɯ́ hɨ-dʒíːhɯ-ro banana-SCM.powder-SCM.bottle ‘bottle of banana powder’

In some cases, class markers are fused to the root (which does not occur without it) (examples 3), or even incorporated into the noun stem with other morphology following the class marker (examples 4). In this case, the

Multidimensional typology and Miraña class markers

369

only evidence for the noun class is the use of the corresponding class marker in agreement (as further discussed in Section 5, below). (3)

a. áːnɛ-hto / *áːnɛ needle-SCM.spine ‘needle’ b. ɨ ́ːbɯ-ɯ / *ɨ ́ːbɯ heart-SCM.3DIM.round ‘heart’

(4)

hɯːβa a. tsa-hɯ one-SCM.tube path ‘one path’ b. tsa-ha gwájhamɯ one-SCM.cover clothes ‘one (piece of) clothing’

In the examples given so far, class markers make a semantic contribution to the derived noun, most clearly so in examples (2a–f). In these cases, noun class assignment is semantically motivated in the sense that the noun belongs to a class with a coherent semantic basis, which is signalled by the class marker. In some cases (less than 20% of nouns), however, there is no semantic relation between the meaning carried by the class marker and that of the derived noun. Class marker assignment is semantically opaque in these cases (example 5). All of these cases involve one of the ten class markers from the core subset of specific class markers. (For a distributional test showing the difference between motivated and opaque noun class assignment, see Seifart 2005: 206.) (5)

a. kaʔgɯ́ nɯ-ko cahuana-SCM.1DIM.pointed ‘cahuana (a thick drink made from manioc starch)’ (Not: a pointed object made from or related to cahuana) b. pájkóːmɯ-gwa manioc-SCM.2DIM.straight ‘manioc plant’ (Not: a flat object with a straight edge made from or related to manioc)

370

Frank Seifart

An important function of class markers suffixed to noun roots is individuation, i.e. the formation of grammatically countable nouns. Bare noun roots usually denote substances, masses or unspecified numbers of objects (see example 2a, above). When used on their own, these noun roots (as well as some other stems, such as nominalized verbs) are grammatically uncountable, i.e. they cannot combine with dual or plural markers (examples 6a, b). Nouns that are derived with class markers from these roots denote singular objects (see example 2b, above) and obligatorily combine with dual and plural markers when non-singular in reference (examples 6c–d). (6)

a. *ɯ́ hɨ-ːkɯ banana-DL Intended meaning: two bananas b. *ɯ́ hɨ-ːnɛ banana-PL Intended meaning: bananas c. ɯ́ hɨ-ʔó-ːkɯ banana-SCM.3DIM.oblong-DL ‘two bananas’ d. ɯ́ hɨ-ʔó-ːnɛ banana-SCM.3DIM.oblong-PL ‘bananas’

The use of class markers suffixed to nouns (as in examples 2a–f) displays some important parallels to classifier systems: the use of the classifying morphemes is in most cases not constrained by a morphosyntactic rule but is semantically based, and they contribute semantic content to the resulting, derived noun. Strikingly similar examples – in some cases also involving the root for ‘banana’ – are found in different classifier languages, e.g. in Yucatec Maya (Lucy 1992: 74). The individuating function is typical of numeral classifiers as one type of classifiers (see further discussion in Section 6).

5. Agreement = noun classes? As mentioned above, the notion of agreement is used in the typology of nominal classification to distinguish the two major types of systems, noun classes and classifiers. Therefore, the nature of noun class agreement in

Multidimensional typology and Miraña class markers

371

Miraña will be discussed in some detail here. Following Corbett (2006), this discussion is based on the “canonical approach” to agreement, which acknowledges that individual constructions may be more or less “canonical” instances of agreement. The most canonical instance is identified as gender agreement between a head noun and a modifier within a noun phrase, in languages such as Italian (Corbett 2006: 9). The ways in which agreement phenomena extend outwards from the canonical cases tends to cluster around three principles: canonical agreement is (i) redundant rather than informative, (ii) syntactically simple and (iii) inflectional. Noun class agreement in Miraña is obligatorily marked on practically every modifying nominal expression, including relative clauses, as well as on pro-forms and one type of verbal predicate. The following examples illustrate noun class agreement on a relative clause and a numeral (example 7a) and on a demonstrative (example 7b). Example (7b) additionally illustrates the use of a class marker as subject cross-reference marker on a verb. (7)

tsa-hɨ a. ó-ʔdi íhka-hɨ 1S-POS COP.SUB-SCM.2DIM.round one-SCM.2DIM.round ɯgwáː-hɨ metal-SCM.2DIM.round ‘I have one axe.’ (lit. ‘What is to me (is) one axe.’) b. áːkítɛ-hɨ ɛː-hɨ fall-SCM.2DIM.round DIST-SCM.2DIM.round ɯgwáː-hɨ metal-SCM.2DIM.round ‘It fell, that axe.’

For nouns that are derived with a specific class marker, agreement can be marked either with that same specific class marker or with a general class marker, which reflects the animacy of the noun (compare example 8 with example 7). This alternation in agreement marking may occur on any agreeing expression. It is not subject to syntactic restrictions, but rather to discourse factors (see Seifart 2005: 247–306). (8)

tsa-nɛ a. ó-ʔdi íhka-nɛ 1S-POS COP.SUB-GCM.INAN one-GCM.INAN ɯgwáː-hɨ metal-SCM.2DIM.round ‘I have one axe.’ (lit. ‘What is to me [is] one axe.’)

372

Frank Seifart

ɛː-nɛ b. áːkítɛ-nɛ fall-GCM.INAN DIST-GCM.INAN ‘It fell, that axe.’

ɯgwáː-hɨ metal-SCM.2DIM.round

With nouns that are not derived with specific class markers, such as most animate nouns, agreement can only be marked with the corresponding general class marker (example 9). No specific class marker can be used for agreement marking in these cases. (9)

a. ó-ʔdi íhka-ːbɛ tsa-pi 1S-POS COP.SUB-GCM.MASC.SG one-GCM.MASC.SG oːʔí-ːbɛ dog-GCM.MASC.SG ‘I have one dog.’ (lit. ‘What is to me (is) one dog.’) b. áːkítɛ-ːbɛ aː-di fall-GCM.MASC.SG DIST-GCM.MASC.SG oːʔí-ːbɛ dog-GCM.MASC.SG ‘It fell down, that dog.’

Agreement in examples (7–9) is syntactically relatively simple (second principle of canonical agreement), as there are no special rules, except for the possible alternative agreement marking. Agreement is also inflectional (third principle), at least in the sense that it is obligatory and marked on the edge of words (notwithstanding the clearly derivational function of class markers on nouns). The question of how ‘informative’ agreement is in terms of the first principle of canonical agreement is somewhat more difficult to settle and will be considered next. Class markers contribute semantic content to the nouns that they derive (see Section 4) and thus they are certainly informative there. However, once formed with a specific class marker, agreement with a derived noun can only be marked with that same class marker, not any other specific class marker (examples 10–11). This is true even if the referent may happen to semantically match better with another class marker, e.g. if a particular avocado happens to be saliently oblong (examples 10b) or if a particular banana happens to be particularly flat (example 11b). Thus noun class agreement marked with specific class markers is never informative, in that it never conveys any information that would not be expressed in the controller of the agreement. This is also true for general class markers and the information they encode, namely animacy.

Multidimensional typology and Miraña class markers

373

kóːhɯ-ba (10) a. tsa-ʔba one-SCM.3DIM avocado-SCM.3DIM ‘one avocado (fruit)’ b. *tsa-ʔo kóːhɯ-ba one-SCM.3DIM.oblong avocado-SCM.3DIM Intended meaning: one avocado (fruit) ɯ́ hɨ-ʔo (11) a. tsa-ʔo one-SCM.3DIM.oblong banana-SCM.3DIM.oblong ‘one banana’ b. *tsa-hɨ ɯ́ hɨ-ʔo one-SCM.2DIM.flat banana-SCM.3DIM.oblong Intended meaning: one banana

However, noun class agreement must be considered informative when an agreement controller is missing, i.e. when an expression that usually functions as a modifier agreeing with a head noun or as a pro-form agreeing with an antecedent is used on its own, independently establishing reference. In example (12) there are two such instances, one as subject crossreference marker and one in a numeral. These occurred during an experiment where Miraña speakers were asked to arrange particularly shaped wooden objects which have no conventional name in the language (see Seifart 2005: 341). mɨ-hɨ ́-ːkɯ (12) aːhá ɛʔdɯ́ íhka-hɨ ́-kɯ́ ITJ also COP-SCM.2DIM.round two-2DIM.round-DL ‘Yes, there are also two flat and round ones.’

It is also worth addressing here the question of whether Miraña nouns belong to more than one class, since this is taken to be another definitional criterion for noun classes. Recall that Miraña noun roots can combine with different class markers. But it is at the level of nouns (not noun roots) that noun classes are defined, namely through the agreement pattern that a noun takes (Corbett 1991: 5, Aikhenvald 2000: 20), and once a Miraña noun is derived with a particular class marker, its agreement pattern is fixed, i.e. it cannot be ‘re-classified’ in agreement marking. The (highly constrained and systematic) possibility of using either a specific or a general class marker with some nouns may be considered as part of the one invariable agreement class of that noun. For instance, the noun kɯ́ ːhɯ-gwa (fire-

374

Frank Seifart

SCM.2DIM.straight) ‘fire’ straight, GCM.INAN}.

belongs to the agreement class {SCM.2DIM.

In sum, the use of Miraña class markers on numerals, relative clauses, verbs, etc. follows a uniform pattern that is close to canonical agreement. If the agreement criterion (together with the one-class-per-noun criterion) is taken as decisive, Miraña should clearly be a noun class language. However, such an analysis poses a serious problem for the typology, since Miraña does not display any of the other characteristics that the typology predicts for this type of system, but rather it shares many of the characteristics that classifier systems are predicted to have. These include the large number of class markers, the fact that the set is somewhat open and the variable use of class markers as well as the individuating function. Thus, the typology would have a very reduced predictive power, at least for Miraña. An additional problem is that by grouping Miraña with prototypical noun class systems, such as those of Bantu languages and gender systems of European languages, the nature of this taxon would be radically enlarged.

6. Multiple classifiers? For an alternative analysis of Miraña as a classifier system, the question arises which type of classifier system Miraña should be considered to have: each classifier type is defined as occurring only (or at least mainly) in one context, such as numerals, verbs or next to nouns, but Miraña class markers are used in all of these contexts in addition to many more. Of these contexts, none (except perhaps the use on nouns) can be considered more basic that the others. Based heavily on data from Amazonian languages, Aikhenvald (2000) accounts for such a situation with the conception of ‘multiple classifier systems’. Under the assumption of ‘multiple classifiers’, occurrences of one and the same set of classifying morphemes in different morphosyntactic contexts are treated as if they were instances of different types of nominal classification systems. For example, one and the same set of “classifiers” of Eastern Tucanoan languages (a group of languages in contact with Miraña and with very similar classification systems, see, e.g. Morse and Maxwell 1999: 73–76) is presented alternatively as “noun class” (Aikhenvald 2000: 80), “noun classifiers” (93), “numeral classifiers” (100, 110, 123), “possessed classifiers” and “deictic classifiers” (207). Bora (a close dialectal variant of Miraña) is cited as an example for “multiple classifiers” (221, 246), including “numeral classifiers” (123).5 Strictly

Multidimensional typology and Miraña class markers

375

speaking, as many classifier types as there are morphosyntactic contexts for classifying morphemes have to be recognized under such an analysis. For Miraña this would include noun classifiers, numeral classifiers and verbal classifiers. But even then, the uses of class markers for which there is no corresponding classifier type are still unaccounted for, e.g. their use in relative clauses. The concept of ‘multiple classifier’ may thus be missing descriptive generalizations by interpreting the use of one and the same set of forms in a number of different contexts as instances of different types of systems, obscuring what is a clearly uniform pattern such as the use of Miraña class markers on numerals, verbs, demonstratives, etc. But the conception of ‘multiple classifiers’ is also problematic from a typological perspective since the inclusion of one and the same system in a number of different categories seriously jeopardizes the identification of each of these types as a coherent set of systems. Introducing more definitional criteria which exclude deviant cases like Miraña may be able to enhance the internal coherence of these types to some extent. For instance, a well-known characteristic of numeral classifiers is that they form a more tightly integrated unit (morphological or syntactic, according to the language type) with the numeral than with the noun (Greenberg 1977: 293, Aikhenvald 2000: 104–105).6 This is illustrated with the following example (13) from Mandarin Chinese (see also Li and Thompson 1981: 104). Mandarin Chinese (Sino-Tibetan, Chinese; Sackmann 2000: 423) (13) a. [[yī fēng] xìn] one CL letter(s) ‘one letter’ b. [[sān běn] shū] three CL book(s) ‘three books’ In contrast to Mandarin Chinese, in Miraña (and in Eastern Tucanoan languages, for that matter) the enumerated noun must already include a class marker, which is suffixed to the noun root, before it can be enumerated with a numeral (example 14a). The numeral also includes a class marker, but it is used for marking agreement with the enumerated noun. The individuation takes place through class marker suffixation directly on nouns (see Section 4), independently of the numeral. Consequently, a class marker used on a separate word, which can be a numeral, cannot individu-

376

Frank Seifart

ate an uncountable noun root (compare examples 14b with examples 13a, b). There is thus a crucial difference in the constituent structure of numeral constructions between languages like Mandarin Chinese and Miraña (see further discussion below). [ɯ́ hɨ-ʔo] (14) a. [tsá-ʔo] one-SCM.3DIM.oblong banana-SCM.3DIM.oblong ‘one banana’ b. *tsa-ʔo ɯ́ hɨ one-SCM.3DIM.oblong banana Intended meaning: one banana

Verbal classifiers are another example of how a type can be more narrowly defined in order to exclude Miraña, which would be a deviant case within this category. Verbal classifiers usually cross-reference the subject noun phrase of intransitive verbs or the object noun phrase of transitive verbs (Mithun 1986, Aikhenvald 2000: 149), while Miraña class markers used on main clause predicates always cross-reference the subject noun phrase, as illustrated in the following example (15). Also, verbal classifiers are typically incorporated in the verb stem, or at least they appear very close to the verb root (see Mithun 1986, Aikhenvald 2000: 149–171, Grinevald 2000: 67–69), while Miraña class markers are suffixed in the outermost position of inflected verb forms (example 16). Excluding Miraña from the type ‘verbal classifiers’ thus yields a more coherent type. It also preserves the validity of generalizations about the diachronic development of verbal classifiers from noun incorporation (Mithun 1986), which would not hold for Miraña. (15) a. áːkítɛ-dʒɛ fall-GCM.FEM.SG ‘She fell down.’ b. ɨ ́ːtɛ-dʒɛ see-GCM.FEM.SG ‘She saw a tapir.’

okáhi-kɛ tapir-ACC

(16) níːtɛ́-kóː-i-tɯ́ -ro-ːbɛ go_down-PF-FUT-NEG-FRUS-GCM.MASC.SG ‘He was already at the point of not going down.’

Multidimensional typology and Miraña class markers

377

Incidentally, nominal classification in Miraña has at least two characteristics that do not match what Aikhenvald (2000) claims to be properties of multiple classifier systems, and which are said to distinguish these from noun class systems. First, she claims that “the choice of a classifier in a multiple classifier language is always semantically based” (Aikhenvald 2000: 229). This is not true for the use of class markers for agreement marking, in particular when noun class assignment is opaque, as in example 5, above). Second, “if classifiers of any type are fused with the head noun, or with a modifier, they cannot be used in multiple classifier environments” (Aikhenvald 2000: 205). This is not true for the nouns in Miraña where class markers are fused or incorporated in the noun stem (as in examples 3 and 4, above). In sum, a characterization of the Miraña system within the typological approach sketched in Section 2 is highly problematic. Section 5 showed that Miraña class markers are involved in agreement, the definitional characteristic of noun class systems, but that they do not display practically any of the other characteristics of this type. This section showed that an analysis of the system as being composed of ‘multiple classifiers’ turns out to be at least equally problematic because important characteristics of individual classifier types are not met. Additionally, characteristics that are used to define the complex type ‘multiple classifiers’ are not displayed by Miraña either.

7. Towards a multidimensional typology The preceding section argued for more narrowly defined types, which are relatively coherent in themselves, providing at least a more consistent terminology. From a typological perspective, the importance is that the characteristics used to define these more narrow types can be more confidently predicted to correlate than the definitional and contingent characteristics of broad types. For instance, for the narrowly defined type ‘verbal classifiers’ the following correlations can be predicted: classifiers occur in verbs, cross-reference objects and intransitive subjects, and occur close to the root. Such predictions are considerably less robust for a more broadly defined type, which would include Miraña. For this reason, Miraña was excluded from the types ‘numeral classifiers’ and ‘verbal classifiers’. A recently proposed new type (‘multiple classifiers’) was shown to be internally inconsistent and not applicable to Miraña either.

378

Frank Seifart

Does this mean that nominal classification in Miraña defies typological characterization and comparison even though it shares important characteristics with other systems? This section argues that this is not necessarily the case if the focus is shifted away from broad, universal types that are defined by few characteristics as the basic units of the typology towards comparing a larger number of more detailed characteristics directly, following an approach that has been applied fruitfully to other grammatical systems (see Bickel 2007: 245 for a summary). Each of these characteristics corresponds to one dimension in what could eventually become a multidimensional typology of nominal classification. Taken together, these dimensions aim at capturing the full range of particularities of nominal classification systems of individual languages with no need to view one or the other characteristic as deviant from a pre-established type. As a comparative tool, such a multidimensional typology could systematically describe the variability of nominal classification systems, fully acknowledging their particularities. In such a multidimensional space, correlations between individual parameters would become visible and, ultimately, implicational relations between individual parameters could be formulated, based on the ways in which individual parameters correlate crosslinguistically. This may apply not only to phenomena treated under the heading of nominal classification but also to the relation of individual parameters with characteristics of other grammatical subsystems. As an example of how such an approach could be applied to nominal classification, let us consider again the definition of numeral classifiers. This type of system can be reconstructed as the correlation of a number of independent parameters, among them the occurrence of a classificatory element in numeral constructions and a certain constituent structure of these constructions. Viewed as such, this cluster of characteristics can be related to other grammatical properties, such as obligatory plural marking. As Sanches and Slobin (1973: 6) have shown (see also Greenberg 1977: 290, Lucy 1992: 46), there cannot be obligatory plural marking in constructions with numeral classifiers. There is thus a network of at least three logically independent parameters: (i) classificatory element in numeral constructions, (ii) a certain constituent structure of numeral constructions and (iii) obligatory plural marking. This cluster of parameters could be further related to Gil’s (1987) typological generalization involving numeral classifiers, definiteness, noun phrase configurationality and the count-mass distinction. Miraña does not enter into these networks of correlating parameters, but it does appear to have something resembling numeral classifiers as defined

Multidimensional typology and Miraña class markers

379

by other parameters, namely the individuating function and the prominence of shape in the semantic profile, which is typical of numeral classifiers (Croft 1994: 152, Aikhenvald 2000: 286–293, Grinevald 2000: 72–73). What the Miraña data suggest is that this semantic profile is not directly related to the occurrence of classifying expressions in numerals (as implied by the traditional conception of numeral classifiers), but rather to the individuating function of classifying elements. This is thus an example of two partially intersecting clusters of parameters: one is the correlation of shape semantics with individuation (displayed by Miraña and, e.g., Mandarin Chinese), the other is the cluster of ‘narrowly defined numeral classifiers’ mentioned above. These examples show how in a multidimensional typology a number of parameters may independently vary and converge to form clusters. Some of these clusters correspond to systems that fall within traditional types of classification systems, e.g. numeral classifiers, while other clusters crosscut these traditional types. A full-fledged multidimensional typology would require the identification of a much larger number of parameters and their application to a much larger sample of languages, ideally backed by statistical methods. The parameters derived from the analysis of Miraña and their application to a few other, well-known languages should have at least shown that this is a worthwhile enterprise. Before concluding this section, it is worth mentioning here how the perspective of a multidimensional typology on Miraña data may also contribute to resolving some analytical difficulties in other nominal classification systems. From the perspective of the traditional distinction between noun classes and classifiers, maybe the most striking fact about Miraña is that its class markers perform derivational and agreement functions to the same extent (see Seifart 2005: 310–315, 324–329 for extensive argumentation). This is striking because the derivational function, where the use of class markers is semantically based, and where class markers make a semantic contribution to the resulting noun phrase and may individuate it, is so reminiscent of classifiers, while the morphosyntactically constrained and semantically redundant agreement function is the hallmark of noun classes. The traditional ‘noun class vs. classifier’ distinction is difficult to apply to Miraña because it is based on the assumption that these two functions are mutually exclusive. From the perspective of a multidimensional typology, the suggestion is to simply view them as often negatively correlating, but in principle independent parameters. This may be of significance both from the perspective of noun class systems, viewed as agreement systems, and from the perspective of classifier

380

Frank Seifart

systems, viewed as systems whose main function is semantic contribution to noun phrase reference (including in some cases individuation), be it through morphological derivation or in syntactic constructions like numeral phrases. From the perspective of noun classes, this finding suggests taking more seriously the derivational functions of the noun class markers of noun class languages, which are defined by agreement. These derivational functions are often treated as marginal, or sometimes they are simply defined out by treating the shared stems of the derived items as belonging to different lexemes (see also Lucy 2000: 330–331). For instance, Corbett (1991: 44, 67) analyses instances such as Spanish hijo ‘son’ vs. hija ‘daughter’ or Swahili (Bantu) ki-ti (class.1-wood) ‘(wooden) stool’ vs. m-ti (class.7wood) ‘tree’ as “two separate nouns sharing a similar stem but with different inflections” (Corbett 1991: 67), thus denying the derivational function of noun class (and gender) markers. This analysis may be justified when the function of class markers on nouns is restricted to, e.g., specifying the natural gender of nouns with human referents (as in Spanish). However, such an analysis becomes increasingly problematic when a given noun root combines productively with a number of different class markers (see Mufwene 1980 for more examples from Bantu languages) and it becomes impossible in Miraña. From the perspective of classifier systems, the analysis of Miraña shows that forms that look like classifiers and behave like classifiers in some constructions can be involved in morphosyntactically constrained and semantically redundant agreement marking in other constructions. This finding may be of particular relevance for systems of nominal classification of Amazonian languages. For these languages, the general descriptive strategy has been to simply label classifying morphemes as “classifiers” (presumably based on criteria such as the size of the inventory of classifying morphemes and possibly their occurrence in numeral constructions) without considering an analysis as an agreement system.7 To further explore these issues, it would be interesting to couple a multidimensional approach with recent discussions in typology about getting away from classifying whole languages as representing one type or another, and instead typologizing constructions, both within and across languages (see, e.g., O’Connor, this volume).

Multidimensional typology and Miraña class markers

381

8. Summary, conclusion and outlook This chapter analysed the complex nominal classification system in Miraña from the perspective of a multidimensional typology. Traditionally, nominal classification systems are conceived as prototypes defined by relatively few definitional characteristics, which are said to typically correlate with a number of contingent characteristics. It was shown that Miraña is almost impossible to characterize in terms of these types. Certainly Miraña is not alone in not fitting well into existing typologies (of various kinds) and after all, the conception of prototypes allows for deviant and transitional cases. However, this chapter suggested that a more promising approach is to shift the focus away from the broad types defined by relatively few characteristics towards a larger number of more detailed characteristics, each corresponding to one parameter in a multidimensional typology. Some parameters that such a typology could use were proposed and it was shown how these vary and cluster independently of the traditional types, resolving some problematic issues in the analysis and typology of nominal classification systems. This suggests that in a fully worked-out multidimensional typology of nominal classification, further clusters of individual parameters and implicational associations across dimensions could become apparent, describing more accurately the actual cross-linguistic variation and giving the typology more predictive power.

Notes 1.

2.

Nominal classification in Miraña is described in detail in Seifart (2005), on which this chapter is based, although this chapter offers some more widereaching conclusions. Miraña is spoken in the Colombian Amazon region by less than 50 speakers now. It has a close dialectal variant, Bora, which is spoken mainly in Peru and described in Thiesen (1996) and Thiesen and Weber (forthcoming). Together with their sister language Muinane (Vengoechea 2005), Miraña and Bora probably belong to the Witotoan family (Aschmann 1993, but see Kaufman 1994 and Seifart 2007). I would like to thank Colette Grinevald, Steve Levinson, Nick Enfield, Nikolaus Himmelmann, Pattie Epps and two anonymous reviewers for their contributions to the development of the ideas expressed in this chapter. Arguably, this characteristic is part of what defines a grammaticalized agreement system (Corbett 2006), i.e. it is not an independent characteristic of noun class systems.

382 3.

4.

5.

6.

7.

Frank Seifart Other, minor types of systems identified by some authors, e.g. Aikhenvald (2000), include deictic classifiers, defined by occurrence in demonstratives, and locative classifiers, defined by occurrence in locative constructions. We do not need to take these into account for the purpose of this paper. The choice for the term ‘(noun) class marker’ is based on their agreement function (see Section 5) through which class markers set up disjunctive noun classes. It is not intended to prejudice the typological characterization of the system. Abbreviations are as follows: 1DIM one-dimensional shape, GCM general class marker, 1S first person singular, INAN inanimate, 2DIM twodimensional shape, ITJ interjection, 3DIM three-dimensional shape, MASC masculine, ACC accusative case, NEG negation, CL classifier, PF perfective, COP copula, PL plural, DIST distal demonstrative, POS possessor, DL dual, RP repeater, FEM feminine, SG, S singular, FRUS frustrative, SCM specific class marker, FUT future tense, SUB subordinate clause. Bora-Witoto languages are also cited as examples of “gender” (Aikhenvald 2000: 246) and “two noun class systems for different types of modifiers” (Aikhenvald 2000: 80), but it is not clear whether these are supposed to be different uses of the same set of “multiple classifiers” or separate systems. It is also not clear which modifiers are meant. Closely related to this is the word order universal according to which the numeral classifier and the numeral are always adjacent, never separated by the enumerated noun in a numeral construction (Greenberg 1963; Dryer 1992). A notable exception is D. L. Payne’s (1990) study that shows that the use of classifying morphemes in numerals and demonstratives in Yagua (PebaYaguan) has to be analysed as inflectional agreement, while their use in other modifiers of the noun may be analysed as derivation. Aikhenvald’s work on North Arawakan languages is not conclusive on the question whether the use of classifying morphemes in contexts such as demonstratives, numerals, and verbs is determined by agreement or not. For instance, Tariana is characterized by Aikhenvald (2000: 235–240) as a “multiple classifier language”, which – by definition – should “involve lexical selection rather than agreement” (Aikhenvald 2000: 229). At the same time, the classifying morphemes in Tariana are described as “noun class agreement markers” (Aikhenvald 2000: 93) and as having the function of marking “predicate-argument agreement” (Aikhenvald 2000: 201).

References Aikhenvald, Alexandra Y. 2000 Classifiers. A Typology of Noun Categorization Devices. Oxford: Oxford University Press.

Multidimensional typology and Miraña class markers

383

Aikhenvald, Alexandra Y. (ed.) 2004 Nominal Classification. Berlin: Akademie Verlag. (Special issue of Sprachtypolgie und Universalienforschung 57, 2/3) Aschmann, Richard P. 1993 Proto Witotoan. Dallas: The Summer Institute of Linguistics and the University of Texas at Arlington. Bickel, Balthasar 2007 Typology in the 21st century: Major current developments. Linguistic Typology 11 (1): 239–251. Corbett, Greville G. 1991 Gender. Cambridge: Cambridge University Press. 2006 Agreement. Cambridge: Cambridge University Press. Creissels, Denis 1999 ‘Genres’ indo-européens et ‘classes nominales’ Niger-Congo. In Faits de Langue: Classification, M. A. Morel (ed.), 177–184. Paris: OPHRYS. Croft, William C. 1994 Semantic universals in classifier systems. Word 45 (2): 145–171. Dixon, Robert M. W. 1986 Noun classes and noun classification in typological perspective. In Noun Classes and Categorization, C. Craig (ed.), 106–112. Amsterdam/Philadelphia: Benjamins. Dryer, Matthew S. 1992 The Greenbergian Word Order Correlations. Language 68 (1): 80– 138. Gil, David 1987 Definiteness, Noun Phrase Configurationality, and the Count-Mass Distinction. In The Representation of (In)definiteness, E. J. Reuland and A. G. B. ter Meulen (eds.), 255–269. Cambridge, MA: MIT Press. Greenberg, Joseph H. 1963 Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Human Language, J. H. Greenberg (ed.), 73–113. Cambridge, MA: MIT Press. 1977 Numeral classifiers and substantival number: Problems in the genesis of a linguistic type. In Linguistics at the Crossroads, A. Makkai, V. Becker Makkai and L. Heilmann (eds.), 276–300. Padova: Liviana Editrice. Grinevald, Colette 2000 A morphosyntactic typology of classifiers. In Systems of Nominal Classification, G. Senft (ed.), 50–92. Cambridge/New York: Cambridge University Press.

384

Frank Seifart

Haspelmath, Martin 2007 Pre-established categories don’t exist: Consequences for language description and typology. Linguistic Typology 11 (1): 119–132. Katamba, Francis 2003 Bantu nominal morphology. In The Bantu Languages, D. Nurse and G. Philippson (eds.), 103–120. London: Routledge. Kaufman, Terrence 1994 Review of Proto Witotoan by Richard P. Aschmann, Arlington, TX: Summer Insititute of Linguistics and University of Texas at Arlington, 1993. Language 70 (2): 379. Li, Charles N., and Sandra A. Thompson 1981 Mandarin Chinese: A Functional Reference Grammar. Berkeley, CA: California University Press. Lucy, John A. 1992 Grammatical Categories and Cognition. A Case Study of the Linguistic Relativity Hypothesis. Cambridge: Cambridge University Press. 2000 Systems of nominal classification: A concluding discussion. In Systems of Nominal Classification, G. Senft (ed.), 326–341. Cambridge/New York: Cambridge University Press. Mithun, Marianne 1986 The convergence of noun classification systems. In Noun Classes and Categorization, C. Craig (ed.), 379–397. Amsterdam/Philadelphia: Benjamins. Morse, Nancy L., and Michael R. Maxwell 1999 Cubeo Grammar. Arlington, TX: Summer Institute of Linguistics. Mufwene, Salikoko S. 1980 Bantu Class Prefixes: Inflectional or Derivational? In Papers from the Sixteenth Regional Meeting Chicago Linguistic Society, J. Kreiman and A. E. Ojeda (eds.), 246–258. Chicago: Chicago Linguistic Society. O’Connor, Loretta this vol. All typologies leak: Predicates of change in Lowland Chontal of Oaxaca. Payne, Doris L. 1990 The Pragmatics of Word Order. Typological Dimensions of Verb Initial Languages. Berlin/New York: Mouton de Gruyter. Sackmann, Robin 2000 Numeratives in Mandarin Chinese. In Approaches to the Typology of Word Classes, P. M. Vogel and B. Comrie (eds.), 421–477. Berlin/New York: Mouton de Gruyter.

Multidimensional typology and Miraña class markers

385

Sanches, Mary, and Linda Slobin 1973 Numeral classifiers and number marking: An implicational universal. Working Papers in Languages Universals, Stanford University 11: 1–22. Seifart, Frank 2005 The structure and use of shape-based noun classes in Miraña (North West Amazon). Nijmegen: Max Planck Institute for Psycholinguistics. (Ph.D. diss., Radboud University Nijmegen) 2007 The prehistory of nominal classification in Witotoan languages. International Journal of American Linguistics 73 (4): 411–445. Seiler, Hansjakob 1986 Apprehension. Language, Object, and Order. Part III: The Universal Dimension of Apprehension. Tübingen: Gunter Narr Verlag. Senft, Gunter (ed.) 2000 Systems of Nominal Classification. Cambridge/New York: Cambridge University Press. Thiesen, Wesley 1996 Gramática del Idioma Bora. Yarinacochoa, Pucallpa (Peru): Instituto Lingüístico de Verano. Thiesen, Wesley, and David J. Weber forthc. A Grammar of Bora. SIL International. Vengoechea, Consuelo 2005 Morphosyntax of Muinane: Typological Remarks. Amerindia 29/30: 119–140.

Part VI. New challenges in methodology

Steps toward a grammar embedded in data Nicholas Thieberger

1. Introduction1 Inasmuch as a documentary grammar of a language can be characterized – given the formative nature of the discussion of documentary linguistics (cf. Himmelmann 1998, 2008) – part of it has to be based in the relationship of the analysis to the recorded data, both in the process of conducting the analysis with interactive access to primary recordings and in the presentation of the grammar with references to those recordings. This chapter discusses the method developed for building a corpus of recordings and time-aligned transcripts, and embedding the analysis in that data. Given that the art of grammar writing has received detailed treatment in two recent volumes (Payne and Weber 2006, Ameka, Dench, and Evans 2006), this chapter focuses on the methodology we can bring to developing a grammar embedded in data in the course of language documentation, observing what methods are currently available, and how we can envisage a grammar of the future. The process described here creates archival versions of the primary data while allowing the final work to include playable versions of example sentences, in keeping with the understanding that it is our professional responsibility to provide the data on which our claims are based. In this way we are shifting the authority of the analysis, which traditionally has been located only in the linguist’s work, and acknowledging that other analyses are possible using new technologies. The approach discussed in this chapter has three main benefits: first, it gives a linguist the means to interact instantly with digital versions of the primary data, indexed by transcripts; second, it allows readers of grammars or other analytical works to verify claims made by access to contextualized primary data (and not just the limited set of data usually presented in a grammar); third, it provides an archival form of the data which can be accessed by speakers of the language. The great benefit for typologists is that they can access annotated corpora from fieldwork-based research in order to address topics not considered by the original analyst. Well-structured linguistic data will allow a number of outputs to be created, including a grammatical description writ-

390

Nicholas Thieberger

ten as a book, but with other potential ways of visualizing the data. The citable archival form of the data can be derived in more ephemeral multimedia or online forms suited to ‘mobilizing’ the data (Nathan 2006). For example, archival files are typically very high resolution and are accordingly very large. To deliver them via the web or on a CD or DVD attached to a grammar requires that they be converted to a lower resolution. Similarly, it is now common to create lexical databases from which a dictionary can be derived. This separation of underlying forms of the data from delivery forms is central to the methods discussed in this chapter. Writing a grammar of a previously undescribed language is a major undertaking, typically the endpoint of fieldwork-based linguistic research, of which the methodology has recently undergone significant changes with the introduction of new tools for digital recording, transcription and analysis. The process of recording such a language has, until recently, relied little on an empirical dataset and more on the genius of the researcher – observing and writing notes while they live in a community of speakers. The resulting work, the grammar, is a crafted collection of these observations and, more often than not, it is only fieldnotes and written texts that are recorded, with perhaps a few tapes to confirm phonological claims (cf. Dixon 2006). The innovative approach discussed in this chapter does not supersede the linguist’s role, it enhances the scientific basis of the linguist’s work by providing ready access to primary data. The decisions about what to record, how to record it, and all the normal elicitation and experimental techniques that make up fieldwork today are still implemented by the linguist, but it is in the methods of recording, naming and transcribing the field materials and consequent access to data that novel outcomes can be achieved. An opportunistic corpus cannot answer all questions asked by a grammar writer and there will always be a need to elicit forms, especially for paradigms. Such forms do, of course, need to be marked as being elicited rather than naturally occurring so that their status is clear. The typical grammar of the past few decades makes no reference to the source of its data nor to how to access further data on the language than is included in the grammatical description. For example, was the data all elicited or was it recorded and transcribed? If it was recorded, then who was it recorded with – are the speakers old or young, male or female? If texts are the source of example sentences, then where in the text does the example come from? Where is the data itself stored? A sample of some thirty grammars from that period found one (Heath 1984, discussed below) that provided sufficient data to allow verification

Steps toward a grammar embedded in data

391

of the author’s analysis with textual data readily available to investigate features of the grammar not addressed by the grammar writer. None of them provided recordings and none provided links from example sentences to the context of the sentence, neither by provision of the media, nor by the more arduous path of spelled out timecodes to media files that may be available somewhere in the world (but not collocated with the grammatical analysis). It is appropriate to focus on the past few decades because it has been possible to provide access to textual and dynamic media recordings for most of that time, but it has not been part of normal linguistic methodology to take advantage of this possibility. Thus it is not a criticism of any one grammar to observe that it does not consider the corpus on which it is built to be a relevant part of its construction so that the whole corpus or a suitable presentation version of it should be provided to the reader. Rather, one has a sense of wonder that field linguistics as a discipline has kept going as long as it has in willful ignorance of the availability of new methods for recording, transcribing, concordancing, annotating, and presenting the data on which it bases its generalizations. These are methods with which it should be completely engaged, relishing the opportunity to access recorded textual material instantly, and to account for any small inconsistency in the data by reference to that data rather than by sidestepping one or two seemingly aberrant forms in the transcripts because of the difficulty of locating the primary media in an analog tape. Not only do these methods improve the linguist’s work, but they also create the kinds of records that speakers of the languages can reasonably expect to result from fieldwork, and that funding bodies are increasingly coming to demand of publicly funded research.

2. The art of grammar writing in recent literature Two recent works on grammar writing summarize the state of the art, but neither considers the possibilities offered by new technologies for access to primary data. Thus, in a collection of work which details many aspects of grammar writing, Ameka, Dench, and Evans (2006) briefly discuss the issue of new technological methods for accessing data, but conclude that it means that data should be made available by a digital archive (Evans and Dench 2006: 25). Archiving data ensures its longevity; however, it is the relationship of the grammar to the data that ideally forms the basis of the analysis engaged in by the grammar writer. Archived data provides the foundation for this relationship, and archiving is a necessary but not suffi-

392

Nicholas Thieberger

cient activity to ensure both that linguistic analysis is embedded in the data, and that there can be long-term access to the data. Mosel (2006), in the same volume, hopes that every grammar would include a text collection which: consists of annotated digitalized recordings of different language genres (e.g. myths, anecdotes, procedural texts, casual conversation, political debates and ritual speech events), accompanied by a transcription, a translation and a commentary on the content and linguistic phenomena. (Mosel 2006: 53)

New technologies provide the means for creation of digital records in the course of linguistic fieldwork and analysis, and this requires a change in linguistic methodology, as discussed below. Another recent collection of papers on grammar writing (Payne and Weber 2006) makes no reference to the potential of a new kind of grammatical description interoperating with its source data via the use of new technologies, despite two chapters touching on technology in grammar writing (Weber 2006a, 2006b). Weber’s (2006b) discussion of the linguistic example likens a grammatical description to a museum of fine art, with galleries exhibiting the features of the language, so there could be a gallery of relative clauses, a gallery of noun classes and so on. In these galleries the example sentences form the exhibits. He points out, however, that a museum is not a warehouse – suggesting that data collection on its own, as advocated by some proponents of language documentation, simply results in a warehouse of recorded material, in contrast to a museum in which each item needs to be provided with interpretive material. Similarly, I suggest, a grammatical description must be based on a corpus (the warehouse) of catalogued items (time-aligned transcripts, recordings, and so on), but use examples to illustrate given points within the description. To go further with the analogy, a museum provides a catalogue of the huge warehouse collection that underlies the few items displayed in a particular gallery. In most grammars to date, the example is the only language data provided, and there is no catalogue of the rest of the data, nor an indication of the relationship of the example to that data. As Weber notes, we must “keep in mind that some day the examples may be appreciated more than the author’s fine words giving some clever analysis or theory” (Weber 2006b: 446). In a similar vein, Dorian observes that:

Steps toward a grammar embedded in data

393

The only real certainty about data is that you never have quite as much of it as you’ll someday wish you had. If you have an inconvenient or puzzling datum that you don’t know what to do with, it’s wise to put it in an extended footnote or into an appendix, where it remains accessible and at least marginally on the record, for your own future use or others’. (Dorian n.d.: 18)

How much more useful will it be when grammars of little-known languages provide not just decontextualized and unsourced examples, but annotated corpora linked to media? In this way the grammar provides a point of entry to a set of data and, it is to be hoped, allows new analyses to be made by other researchers who are able to locate unusual examples that eluded the initial recorder. The data on which a grammatical description of a language are based are necessarily a partial set of recordings, observations and elicited forms. They are partial because we are sampling whatever we can within a short timeframe, all the more so when the work is part of a Ph.D. and so constrained to the period of a student’s candidacy. Nevertheless, the analysis of this set of data can be replicable if it is clear which elements of the data form the basis of each analytical claim. Thus, an example sentence, the archetypal ‘proof’ used in a contemporary grammar, has to bear a considerable burden; not only will it provide the authority for the current claim, but it will inevitably then be reused in other work. It is decontextualized first by the original linguist who has some knowledge of the language and the frame from which the example was taken, but is then decontextualized further by other researchers for whom it is an example of a phenomenon, regardless of the context from which it was taken. These example sentences then take on a life of their own as tokens of authenticity of a particular theoretical point, a problem that could be ameliorated if the example was properly provenanced to source data in a digital repository. Why is there this lack of engagement with the possibilities offered by new technologies for grammatical descriptions? A major reason is the lack of very simple-to-use tools and hence the slow uptake of existing tools among the community of researchers. I suggest that several other reasons conspire to keep data out of grammatical descriptions. They include: inertia, the reluctance of academia to change the way it has been conducting itself; a so-called ‘theoretical linguistics’ that is satisfied with minimal data selected to prove foregone conclusions and which hence requires no large datasets (see Beaugrande 2002); a perception that the creation of a corpus on which to base claims is too time consuming and therefore cannot be part of a normal fieldwork-based linguistic investigation; a general antipathy to

394

Nicholas Thieberger

technology among humanities researchers (but linguists have often been more technically engaged than the typical humanities scholar); a fear of presenting evidence on which claims are made and hence exposing one’s analysis to scrutiny; and, related to the last point, a desire to be the unquestioned authority whose word must be accepted in the absence of any corroborating evidence. The recent increasing interest in the use of corpora (witness the use of the World Wide Web as a corpus from which Natural Language Processing applications opportunistically harvest material with commercial implications) and in language documentation methodology (see Himmelmann 1998, several contributions in Gippert, Himmelmann and Mosel 2006, EMELD conferences 2001–2006,2 Austin (ed.) 2003, 2004, 2005, 2006, Woodbury 2003), has led to an opening up of discussion of a new analytical approach for field linguistics.

3. What could an embedded grammar be? How could a grammar embedded in data be conceived? Grammars can be characterized as a range of types, beginning with those for which the source data is not mentioned through to a complex dataset in which the grammatical description forms just a part of the whole representation of the language. An ideal embedded grammar would allow the reader to move between the apparatus of the grammar and the source data, using either as a point of entry. So, from watching a video of the performance of a story and seeing its orthographic representation on the screen, the reader may want to find out about the meaning of a particular word. They will be able to link to and view the dictionary entry. They may want to know about the role of a participant and link to a discussion of arguments, or of case roles, or of the pronominal system, depending on the context of the departure point in the text. Constructing such links is hugely time consuming, but allowing them to fall out naturally from a well-structured set of data is a goal that is worth pursuing. However, such data structures are not yet developed, so there are no tools made easily available to the ordinary linguist. In the meantime, we operate on the assumption that the more explicitly we can structure the data we create, the more likely it is to survive and be interoperable with emerging systems. Thus the basic desiderata for a grammar embedded in data must also conform to the principles of portability set out in Bird and Simons (2003). The process of creating such data will be discussed below.

Steps toward a grammar embedded in data

395

At the more recognizable end of the spectrum of grammars, those that have been produced more recently usually include a number of example sentences. Examples are the main currency of a descriptive grammar and the usual form of argumentation is to provide some analysis of a phenomenon and then the example that illustrates its existence and its usage. However, if examples are not provenanced and their status not made clear to the reader (Is the sentence elicited? Was it part of a larger text? Who was it spoken by?) then they provide poor data for others who may want to test the analysis. A first step towards an embedded grammar could be to provide links from the grammar to sources of data, establishing that examples exist in a corpus of texts, as shown by Heath in his Nunggubuyu Grammar (1984), Dictionary (1982) and Texts (1980). Cross-references in these volumes constructed by hand (that is, without using a computer to generate links in the way that we now can do) allow the reader to locate contextual information for example sentences. Heath notes that: The standards of accuracy and documentation which I have set for myself in preparing this volume have been high, though I may not have lived up to them uniformly. In essence this is a corpus-based grammar, and my ideal has been to account for all or nearly all instances in the texts of each morpheme or other feature under consideration. (Heath 1984: 4)

Heath accordingly presents references to many examples of any morpheme discussed in his grammar, with the result that, as noted by Musgrave (2005: 113), the complexity of referencing in Heath’s typescript presentation makes it difficult to read, with a dense use of superscripts and visual references that could today be replaced by less obtrusive hyperlinks. Heath (1984: 5) discusses the need for documentation because of his own “sad experiences as a reader of other linguists’ grammars, which have almost never provided me with the information I wanted to undertake my own (re)analysis of the language in question.” Heath’s work cited but did not provide taped recordings, but an early example of the provision of textual material with a recording is Brandenstein’s (1970) text collection, which included a 45 rpm vinyl disk containing the stories which were provided in interlinear format in the book. He clearly thought it essential for a modern comprehensive linguistic work that a record should be attached to each copy of the book (Thieberger 2008: 327). Last century, when we were dealing with analog data, it could be linked from texts but it took far more effort than is required for digital recordings today. A further, but unpub-

396

Nicholas Thieberger

lished, example of textual material accompanying a description is Nicholas Reid’s (1990) dissertation on Ngan’gityemerri, which included four audiocassettes. Citation of data from primary media requires that the media have persistence, both in location and naming, so that when the speakers want access to recordings they can find them, and when a researcher wants to listen to an example sentence to find out if there is a feature there that the original work ignored, they can do that. However, such persistence is difficult, if not impossible, for the individual researcher to achieve (hence the need for linguistic digital repositories such as the DOBES archive,3 PARADISEC,4 or AILLA).5 The researcher may also create a website or a multimedia package in which the media is available for public use. Websites created to provide access to primary data typically do not have longevity, and issues related to delivery of media over the web mean that high-resolution primary data is not suited to web-delivery. Multimedia packages as they are currently being made have a very short life, usually relying on software that is updated every year or so, leaving orphaned earlier versions unreadable. Musgrave (2005) presents three case studies of multimedia representations of language,6 each of which includes little grammatical information, concentrating instead on texts linked to lexicons and media files. The presentation of linked texts, media and lexica is to be welcomed; however, it seems that, of the three case studies, it is only for Nahuatl (Amith n.d.) that the underlying data is produced in an archival form. This is important when creating multimedia products, as otherwise there is a risk that the only available data for the language will be in a form that cannot easily be archived and thus may not be available in the long-term. From the perspective of writing a grammar embedded in data and conforming to the principles of portability (mentioned earlier), these three projects are not descriptions presenting primary data but are representations of the data, an important distinction that must be understood in order to create wellformed primary data in an archive, with derived versions being used in multimedia representations. Morey’s (2004) grammar of Tai includes a CD-ROM which contains the text of the grammar, with example sentences linked to audio files and to details about the characteristics of the speaker. In this work the links are all produced by hand and each link is to a single media file (each target utterance corresponds to a single file). Morey, in a postscript to the work (and echoing Weber’s sentiments quoted above) notes that, “in a hundred years time, though aspects of my analysis may not have stood the test of

Steps toward a grammar embedded in data

397

time, I am confident that the richness of the corpus of texts will ensure that this work is still useful” (2004: 402). With similar sentiments, I built a corpus of South Efate texts on which my analysis (Thieberger 2006a) of the language was based.7 The field recordings were not segmented into utterance-length files as a matter of principle – if they were providing contextualized data for my analysis, it would defeat the purpose to cut them into potentially decontextualized units, and, from an archival point of view, there should not be too many objects (files) to keep track of and to describe. The focus was to develop a means by which a collection of transcripts could maintain links back to the media they transcribed, allowing any subsequent mentions of parts of the texts (as in example sentences) to be cited to that media file. This then is a second step toward an embedded grammar: using a media corpus as the source, and providing the media files with names that will endure over time (also known as persistent identification), as provided, for example, by their being lodged in an archive. This means that any reference made to the media files will be resolvable by readers of the grammar into the future, as per Morey’s and Weber’s desiderata quoted earlier. The paper grammar, in the form of a book, has references to citable data, and may include a DVD of data (that is, an mp3 or similarly compressed form of the file suitable for delivery) derived from the archival form that is stored in an archive with a persistent identifier (a name that will not change). What was required to create a corpus with examples cited to the level of words or sentences? To summarize, in my work with South Efate I created archival digital media files, paying attention to accepted standards in filenaming (in order that files have persistent identification and thus persistent location over time) and file formats (so that the files can be read over time), and depositing these files with a digital repository (with sufficient metadata to describe the content of the file). These files were then transcribed with time-alignment (using stand-off timecoding) (the process outlined here is discussed in more detail in Thieberger 2004, 2006b) facilitating links to be instantiated between any textual chunk and its media representation. Any sentence in some twenty hours of data can be clicked and heard, and accessed by a concordance (a listing of each word in the data in context). While the software for transcribing with time-alignment has been available since the late 1980s (see MacWhinney 1996), it was necessary to devise a method to access the audio and video data as a corpus in order to facilitate its analysis, as no similar approach had been developed when it was needed for this research. Happily this is no longer the case and tools are becoming easier to use and more widely adopted by lin-

398

Nicholas Thieberger

guists (in fact, my students now ask how it could have been possible to do this work in the past without these tools). I wrote a tool called Audiamus8 to allow me to work on the transcripts of all media files (produced, for example, by Transcriber) as a corpus, and to search and browse through any part of them while maintaining the ability to access the primary media associated with them by mouseclick. Audiamus also allows me to quote relevant chunks of the data and to paste these quotes directly into my manuscript and to export it to a number of formats. Thus the references between the textual and media representations of the data are maintained through the process of data analysis. A problem with the current model is that data are not normalized, that is, the same data appear in various parts of the corpus (in the transcript, texts, dictionary and example sentences) and so any change to one version needs to be reflected in the others. A third step into the future of grammar/data entangling will be the creation of grammatical descriptions whose structure is encoded according to an accepted standard, perhaps using elements determined by the Text Encoding Initiative (Burnard and Sperberg-McQueen 2002) or some similar schema to be determined by linguistic practice as it develops. This would create a document in which all structural elements were explicitly declared – normal and generic elements such as chapters, paragraphs and so on, as well as elements more particular to a linguistic grammar. While some parts of a grammar could use the same markup standards as other texts (essays, theses and so on), there are some particular datatypes in linguistic descriptions that will need standards agreed to by the community who will use them. Work has been done to identify some of these datatypes and strategies for encoding them, in particular paradigms (typically used in representation of pronouns) (Penton and Bird 2004) and interlinear glossed text (IGT) (Hughes, Bird, and Bow 2003, Bird, Bow, and Hughes 2003). An immediate benefit of adopting a standard encoding for these datatypes would be to increase their archivability as underlying data rather than as images of data on a page. This would also mean that a typological search for all members of a given class (first person pronouns for example) could be executed via the structure of the data. Another example using interlinear glossed text would allow searches over form/meaning pairs within texts, currently difficult if not impossible to do due to the lack of an explicitly encoded relationship between lines of interlinear glossed text. A similar focus on conceptualizing the structure of a grammar can be found in Good (2004) and Nordhoff (2007), but neither has yet resulted in a workable

Steps toward a grammar embedded in data

399

framework, let alone replicable methods or tools for construction of interoperating grammars and data. A further motivation for building an explicit structure for a grammatical description is that it will provide automatically comparable data across grammatical descriptions in addition to an archival form of the text. At the moment our expectations of an archival linguistic text are very low, as there is a long way yet to go in convincing the linguistic community of the necessity to create functioning digital versions of textual documents in addition to paper outputs. A well-formed and schema-based XML version of a grammar will allow remote (that is, internet-based) searching over specific types of data, of the kind that typologists would like to do but are currently finding difficult due to the lack of a standard encoding schema for grammars. At present a grammar, if it is available on the internet, is an undifferentiated collection of textual material, most likely to be a pdf file, and with no distinguishing characteristics that allow it to be identified as a grammar by a search engine except that its title may include the word ‘grammar’. If there were a schema associated with the document type ‘grammar’ then a search on that type would locate only grammars, and only those with a predictable structure able to be searched remotely. The remaining issue of an agreed ontology for linguistic terms is currently being investigated.9 With the lack of data structure in current written grammatical descriptions it is not possible to automate links between various parts of a linguistic corpus and annotation set. However, a schema-based grammatical description could provide a normalized data structure in which a change to an item in one place (for example, the lexicon) would be reflected in all mentions of that form throughout the corpus, texts and grammar. A schema for the representation of IGT has been developed, following both the work discussed above and an XML-based presentation system developed by Michel Jacobson at LACITO.10 To begin, this project established an XML schema for a minimal five tier interlinear text, with provision for the text line, morphemic and gloss lines, a free translation, and a reference to a media file (together with a small set of metadata fields to assist in discovery of the text). This is a schema-based XML system for making explicit the relationship between parts of interlinear texts together with links to the source media, and has resulted in an online tool (called EOPAS, see Schroeter and Thieberger 2006) for presentation of interlinear texts, which allows searching and concordancing linked directly to the media. EOPAS also provides the environment into which linguists can upload files in a Toolbox11 export format for online viewing and playback

400

Nicholas Thieberger

of media linked to transcripts. The resulting tool is portable, allowing other initiatives to capitalize on the work and potentially develop it in different directions. The ultimate aim of this approach is to allow new perspectives on the data itself, provided by contextualized access to primary data, and then to allow new research questions to be asked, and richer answers to be provided, all in a fraction of the time that it would have taken with analogue data. We can envisage a set of XML-based tools that operate over a richly annotated corpus, using schemas for various datatypes to understand how to read and interact with them. Thus when a snippet of IGT is located in a grammar, the browser uses the schema definition of IGT to present it and to instantiate the link to media encoded in it. The schema also allows us to browse all IGT in the grammar and to search for texts in which a target structure occurs, as we can be confident that the formats in use are all validated against a schema. A fourth step to an embedded grammar would be one in which each hypothesis could be computationally tested by the reader – assuming that the particular theoretical model of analysis employed is computationally tractable. For example, PARGRAM (Butt, Flickinger, and Oepen 2002) implements a Lexical Functional Grammar while Bender (n.d.) and Bender et al. (2002) describe a system they call the LinGO Grammar Matrix which includes an HPSG computational grammar, and Bender (n.d.) and Bender et al. (2004) summarize various grammar engineering projects and the development of a proposed model called Montage (see also Good 2004) in which they “hope to bring the hypothesis-testing power of grammar engineering to linguists engaged in primary linguistic documentation” (Bender n.d.: 15). From the point of view of primary grammatical descriptions of endangered languages, written as most are in a fairly eclectic and certainly not mono-theoretical model (as discussed for example in Dryer 2006), it is unlikely that this kind of grammar will appear anytime soon. It is more likely that it will be derivative of a more narrative form of grammar, one that I hope is of at least the second or third steps described earlier.

4. Building documentation into normal field methods A major objection to creating even the most minimal dataset as the basis for a grammatical description is that it takes too long and so detracts from the analysis (this is especially the case for a student writing a Ph.D. thesis). As long as it is not rewarded by the linguistic community, the preparation

Steps toward a grammar embedded in data

401

of archival data linked in to a grammatical analysis will only be carried out by those who have sufficient technological skills, who recognize its value to posterity and are committed to the scientific method in their own work. We clearly need more accessible tools and methods, but there are sufficient tools now being used by new researchers and the methods are established well enough that it does not add significantly to the linguist’s workload to create time-aligned transcription, adequate filenaming, metadata, and to archive the primary materials (if we accept that creation of a representative corpus of recorded material is a minimal outcome of our work, as argued by proponents of language documentation, e.g. Himmelmann 1998, Woodbury 2003). In fact, it has a number of benefits that reflect a substantial return on the initial investment, including an improved analysis and an ability to keep working on the same dataset over time because of the ease of access to the material. The process of creation of good field records reinforces one’s knowledge of the language and enriches the analysis, first because it requires interaction with the data through the initial transcription and annotation, and second because it provides instant access to the media and so allows us to locate examples and to correct mistranscriptions incrementally as we work with the data. As Beaugrande (2002: 105) points out, we are ‘seeing’ phenomena in language which only become visible through the technology. For example, my Ph.D. thesis is available in an eprints repository,12 which means that the text can be searched via Google or a similar search engine. Robert Eklund, studying ingressives in the world’s languages, located a reference to the ingressive in the thesis and emailed a query. He had doubts that it actually was a pulmonary ingressive, especially as the example included [s], a phone he thought cannot be produced ingressively, at least not without considerable acoustic changes. The segments in my sound files were easily located and within ten minutes he had a clip of the sound so that he could confirm that it was indeed a pulmonary ingressive (see Eklund 2007). Another example of the research benefit of investing effort in creating an accessible dataset is from the pre-verbal benefactive phrase in South Efate, which can be represented by a single morpheme in the form of a possessive pronoun, or by a complex prepositional phrase (Thieberger 2006c). The possessive pronouns in some cases differ minimally in form from the subject pronouns (e.g., ag ‘2sgS’, gag ‘2sgPOS’).13 In the initial transcriptions, and before it was clear to me that this was a benefactive construction, both I and my South Efate colleague had occasionally mistranscribed these forms. The resulting transcript thus provided ambiguous data about the form of the pronoun in this position. The ability to instantly

402

Nicholas Thieberger

access the recordings and to confirm that the pronoun is a possessive form allowed the analysis to be more robust and the transcript to be incrementally corrected. Thus, with an improved analysis we are able to update the transcription via the digital media, rather than being constrained by a onceonly analog transcription system in which the media are no longer accessed and the transcript is taken to be the primary authority. Another motivation for creating digital data using these methods is that it is easier to repatriate or to return copies of this material to the speakers. Recently, while at my fieldsite in Erakor village, I was able to respond to a request from the widow of my main language teacher to provide an audio CD of fifteen tracks, including stories and songs from a range of villagers, some now deceased. It took a few hours to select and locate the data and then produce ten copies of the CD with cover notes. I subsequently placed a collection of narratives, songs and hymns into an ITunes14 installation on the village school’s computer so that others in the village can burn their own selections to CD. This was only possible because of the wellstructured nature of the data and metadata and illustrates the way in which new methods can make us more responsive to community requests for information; especially valuable at a time when academia is increasingly being called to account by its funding sources. The process of discovery of linguistic facts proceeds best by constant interaction with the corpus. Noting that certain features exist in the data, by means of tagging the textual annotation, allows those examples to be retrieved later. The nature of our language learning varies from individual to individual but typically includes achieving plateaux of understanding of structures in the language, which then become an operating framework. New data can challenge these frameworks and allow us to supersede previous working hypotheses if we can interact with the data in real time. New insights can thus be confirmed and incorporated into the corpus and the analysis.

5. Conclusion Having established a methodology for recording and linking media and transcriptions as part of normal fieldwork, our corpus will allow all derived work, including texts, example sentences or dictionaries to be verified by reference to primary recordings. An academic grammar of a language that is created using the methods described in this chapter will have multiple possible outputs in addition to the grammar, and the data created using

Steps toward a grammar embedded in data

403

these methods will be available for others to access in the future. A book written using these methods will allow the reader to access media, either on an accompanying disk or via web delivery. The underlying data will be safely stored in an archive that guarantees longevity of the unique recordings made as part of the grammar-writer’s effort.

Notes 1.

2. 3. 4. 5. 6.

7.

8. 9. 10. 11. 12. 13. 14.

An early version of this paper was presented at the Australian Linguistic Society in July 2008 and has benefited from discussion with members of that audience. It has also been improved by comments from Pattie Epps, David Nash and an anonymous reviewer. Work reported on in this chapter was supported by the following Australian Research Council grants: SR0566965 – Sharing access and analytical tools for ethnographic digital media using high speed networks; DP0450342 – New methodologies for representing and accessing resources on endangered languages: a case study from South Efate. http://www.emeld.org/events/index.cfm http://www.mpi.nl/DOBES http://paradisec.org.au http://www.ailla.utexas.org/site/welcome.html The three exemplars are: Amith, Jonathan D. (n.d.), Nahuatl Learning Environment ; Csato, Eva A. and David Nathan (2003), Spoken Karaim (CD-ROM); Nikolaeva, Irina and Thomas Mayer (2004), Online Documentation of Kolyma Yukaghir, . South Efate is an Austronesian Oceanic language of central Vanuatu. My fieldwork began in the mid 1990s and has so far resulted in a grammar, dictionary and texts in the language, as well as an archived corpus of recordings. http://www.linguistics.unimelb.edu.au/thieberger/audiamus.htm See, for example, preliminary results at http://www.linguistics-ontology.org/ http://lacito.vjf.cnrs.fr/archivage/index.htm Toolbox is the best current software for entering and working with structured interlinear data (http://www.sil.org/toolbox). http://repository.unimelb.edu.au/10187/1318 South Efate orthography uses /g/ to represent the velar nasal. ITunes is commercial but free software produced by Apple Inc. (http://www. apple.com/itunes/).

404

Nicholas Thieberger

References Ameka, Felix, Alan Dench and Nicholas Evans (eds.) 2006 Catching Language: The Standing Challenge of Grammar Writing. Berlin: Mouton de Gruyter. Amith, Jonathan D. n.d. Nahuatl learning environment. Available at http://nahuatl.ldc.upenn. edu. Austin, Peter (ed.) 2003 Language Documentation and Description. 1. London: SOAS. 2004 Language Documentation and Description. 2. London: SOAS. 2005 Language Documentation and Description. 3. London: SOAS. 2006 Language Documentation and Description. 4. London: SOAS. Beaugrande, Robert de 2002 Descriptive linguistics at the millennium: corpus data as authentic language. Journal of Language and Linguistics. 1 (2): 91–131. Bender, Emily M. n.d. Grammar engineering for linguistic hypothesis testing. Manuscript. Bender, Emily M., Dan Flickinger, Jeff Good and Ivan A. Sag 2004 Montage: Leveraging advances in grammar engineering, linguistic ontologies, and markup for the documentation of underdescribed languages. In Proceedings of the Workshop on First Steps for Language Documentation of Minority Languages: Computational Linguistic Tools for Morphology, Lexicon and Corpus Compilation, LREC 2004. Accessed at http://faculty.washington.edu/ebender/ papers/Montage_LREC.pdf on 14/9/2008. Bender, Emily M., Dan Flickinger, and Stephan Oepen 2002 The Grammar Matrix: An open-source starter-kit for the rapid development of cross-linguistically consistent broad-coverage precision grammars. In Proceedings of the Workshop on Grammar Engineering and Evaluation at the 19th International Conference on Computational Linguistics, Taipei, Taiwan. J. Carroll, N. Oostdijk, and R. Sutcliffe (eds.), 8–14. Bird, Steven, and Gary Simons 2003 Seven dimensions of portability for language documentation and description. Language 9 (3): 557–582. Bird, Steven, Catherine Bow, and Baden Hughes 2003 Towards a general model of interlinear text. In Proceedings of the EMELD Language Digitization Project Conference 2003: Workshop on Digitizing and Annotating Texts and Field Recordings. Accessed at http://www.emeld.org/workshop/2003/bowbadenbird-paper.pdf on 14/9/2008.

Steps toward a grammar embedded in data

405

Brandenstein, Carl Georg von 1970 Narratives from the North-West of Western Australia in the Ngarluma and Jindjiparndi Languages. (Volumes1–3 + audio-disc.) Canberra: Australian Institute of Aboriginal Studies. Burnard, Lou, and C. Michael Sperberg-McQueen 2002 TEI Lite: an introduction to text encoding for Interchange. Accessed at http://www.tei-c.org/Guidelines/Customization/Lite/ teiu5_split_en.html on 23/3/2008. Butt, Miriam, Helge Dyvik, Tracy Holloway King, Hiroshi Masuichi, and Christian Rohrer 2002 The Parallel Grammar Project. Proceedings of COLING-2002 Workshop on Grammar Engineering and Evaluation. 1–7. Csato, Eva A., and David Nathan 2003 Spoken Karaim. (CD-ROM) Dixon, Robert M. W. 2006 [Dixon accepts Bloomfield Award] SSILA Newsletter Number 236: April 10, 2006: [6]. Dorian, Nancy n.d. Getting it right: some conscious and unconscious obstacles to getting a language on the record realistically. (Manuscript.) Dryer, Matthew S. 2006 Descriptive theories, explanatory theories, and basic linguistic theory. In Catching Language: The Standing Challenge of Grammar Writing. Felix Ameka, Alan Dench, and Nicholas Evans (eds.), 207– 234. Berlin: Mouton de Gruyter. Eklund, Robert 2007 Pulmonic ingressive speech: a neglected universal? In Proceedings of Fonetik 2007, May – 1 June 2007, Stockholm, Sweden, TMHQPSR, 50 (30): 21–24. Evans, Nicholas, and Alan Dench 2006 Introduction: catching language. In Catching Language: The standing challenge of grammar writing, Felix Ameka, Alan Dench and Nicholas Evans. (eds.), 1–39. Berlin: Mouton de Gruyter. Gippert, Jost, Nikolaus P. Himmelmann, and Ulrike Mosel. (eds.) 2006 Essentials of Language Documentation. Berlin: Mouton De Gruyter. Good, Jeff 2004 The descriptive grammar as a (meta)database. Paper presented at the EMELD Workshop, June 2004. Available at http://www.emeld.org/ workshop /2004/jcgood-paper.html. Heath, Jeffrey 1980 Nunggubuyu Myths and Ethnographic Texts. Canberra: Australian Institute of Aboriginal Studies.

406

Nicholas Thieberger

1982

Nunggubuyu Dictionary. Canberra: Australian Institute of Aboriginal Studies. 1984 Functional Grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies. Himmelmann, Nikolaus P. 1998 Documentary and descriptive linguistics. Linguistics. 36: 161–195. 2008 Reproduction and preservation of linguistic knowledge: Linguistics’ response to language endangerment. Annual Review of Anthropology 37. Hughes, Baden, Steven Bird, and Catherine Bow 2003 Encoding and presenting interlinear text sing XML technologies. In Proceedings Australasian Language Technology Workshop. Alistair Knott and Dominique Estival (eds.), 105–113. Melbourne, Australia. Accessed at http://eprints.unimelb.edu.au/archive/00000455 on 23/3/2008. MacWhinney, Brian 1996 The CHILDES System. American Journal of Speech-Language Pathology. 5:5–14 Morey, Stephen 2004 The Tai Languages of Assam: A Grammar and Texts. Canberra: Pacific Linguistics, Research School of Pacific and Asian Studies. Musgrave, Simon 2005 Multimedia presentation of grammatical description: design issues. In Proceedings of the Australasian Language Technology Workshop. Timothy Baldwin, James Curran and Meno van Zaanen (eds.), 113– 119. Accessed at http://www.alta.asn.au/events/altw2005/cdrom/pdf/ ALTA200517.pdf on 14/9/2008. Mosel, Ulrike 2006 Grammaticography: the art and craft of writing grammars. In Catching Language: The Standing Challenge of Grammar Writing. Felix Ameka, Alan Dench, and Nicholas Evans (eds.), 41–68. Berlin: Mouton de Gruyter. Nathan, David 2006 Proficient, permanent, or pertinent: aiming for sustainability. In Sustainable Data from Digital Fieldwork, Linda Barwick and Nicholas Thieberger (eds.), 57–68. Sydney: Sydney University Press. Nikolaeva, Irina, and Thomas Mayer 2004 Online Documentation of Kolyma Yukaghir. Accessed at http://ling.uni-konstanz.de/pages/home/nikolaeva/documentation/ intro.html on 14/9/2008. Nordhoff, Sebastian 2007 Growing a grammar with Galoes. Paper presented at the DobeS workshop, June 14, 2007.

Steps toward a grammar embedded in data

407

Payne, Thomas E., and David Weber (eds.) 2006 Perspectives on grammar writing. Studies in Language 30:2. Penton, David, and Steven Bird 2004 Representing and Rendering Linguistic Paradigms. In Proceedings of the Australasian Language Technology Workshop, 123–130. Sydney, Australia. December 2004. Accessed at http://www.alta.asn.au/ events/altw2004/publication/04-26.pdf on 14/9/2008. Reid, Nicholas 1990 Ngan’gityemerri: a language of the Daly River region, Northern Territory of Australia. Ph.D. diss., Department of Linguistics, Arts, Australian National University. Schroeter, Ronald, and Nicholas Thieberger 2006 EOPAS, the EthnoER online representation of interlinear text. In Sustainable Data from Digital Fieldwork. Linda Barwick and Nicholas Thieberger (eds.), 99–124. Sydney: Sydney University Press. Thieberger, Nicholas 2004 Documentation in practice: developing a linked media corpus of South Efate. In Language Documentation and Description, Peter Austin (ed.), 2: 169–178. London: SOAS. 2006a A Grammar of South Efate: An Oceanic Language of Vanuatu. Oceanic Linguistics Special Publication, No. 33. Honolulu: University of Hawai’i Press. 2006b Computers in field linguistics. In Encyclopedia of Language & Linguistics, Second Edition. Keith Brown (Editor-in-Chief), 780–783. Volume 2. Oxford: Elsevier. 2006c The benefactive construction in South Efate. Oceanic Linguistics 45 (2): 297–310. 2008 Language is like a carpet: Carl-Georg von Brandenstein and Australian languages. In Encountering Aboriginal Languages: Studies in the History of Australian Linguistics. William M. McGregor (ed.), 321–335. Canberra: Pacific Linguistics. Weber, David 2006a Thoughts on growing a grammar. In Perspectives on Grammar Writing. Thomas Payne and David Weber (eds.), Studies in Language, 30 (2): 417–444. 2006b The linguistic example. In Perspectives on Grammar Writing. Thomas Payne and David Weber (eds.), Studies in Language, 30 (2): 445–460. Woodbury, Anthony C. 2003 Defining documentary linguistics. In Language Documentation and Description. Peter K. Austin (ed.), 1: 35–51. London: SOAS.

Index of languages

Acehnese, 224 Adyghe, 175, 252, 254, 256, 258– 260, 265 Agul, 127–130, 132–133, 137–139, 141, 143–145, 175, 183–184, 187, 189, 192–193 Algonquian languages, 74, 97, 278, 339 Altaic languages, 241, 269, 273, 287 Alutor, 248 Andi, 130, 136–137 Andic languages, 130, 145 Arabic, 248, 252 Arawakan languages, 241, 382 Archi, 224, 233 Arrernte (Mparntwe), 236, 358, 360 Atsugewi, 302 Australian languages, 97, 238 Avar, 130, 145, 249 Awa Pit, 252 Bagvalal, 130, 181–182, 242, 278, 280, 282, 291 Bambara, 248–249, 251, 265 Bantu languages, 3, 35–38, 41, 44, 49, 51, 53–58, 155, 366, 374, 380 Bantu languages (Northwestern), 35, 38, 40–41, 54–55 Basque, 158, 175, 192–193, 206, 236, 241–242 Berber, 280–282 Beti-Bulu-Fang languages, 41, 56 Bezhta, 130–134, 136, 139–140, 144 Blackfoot, 274, 278, 280, 282, 284– 285, 288–289 Bukiyip, 200

Bulgarian, 11–12, 30, 156, 175, 247–248, 250–252, 265 Bunuba, 236 Burushaski, 116 Caribe, 71 Cavineña, 358 Chadic languages, 177 Chali, 115–116 Chamalal, 130 Chinese (Hakka), 242 Chinese (Mandarin), 155, 175, 192– 193, 265, 366, 375–376, 379 Chinese (Min), 232 Chontal of Oaxaca (Highland), 344 Chontal of Oaxaca (Lowland), 6, 343–349, 351, 353, 356–360 Chontal of Oaxaca languages, 344 Chukotko-Kamchatkan languages, 117, 241 Chuvash, 210, 213, 218 Cree, 72, 278, 280, 282, 285 Czech, 110–112, 122, 175 Danish, 157, 175, 192 Dardic languages, 105, 117 Dargwa, 130, 145, 278, 280, 289– 290 Dâw, 91, 99 Djabugay, 238–239, 243 Dutch, 128–129, 145, 155 Dutch (Heerlen), 259–260, 265 Dyirbal, 97 East Caucasian languages, 3, 127, 129–131, 135, 137, 140–141, 143–145, 241

410

Index of languages

Emerillon, 3, 63–64, 67, 69–71, 73– 79, 317 Enets, 175 English, 4–5, 7, 12, 30, 37, 58, 117, 119, 127–128, 164, 166, 177– 178, 181–183, 188–190, 192– 193, 200–201, 216, 227, 229– 230, 232–234, 237–238, 249, 251–252, 258, 263–264, 269– 273, 277, 279–282, 286, 288– 289, 336, 349 English (Old), 113–114, 117, 119 Eshtehardi, 116 Estonian, 248, 251–252 Eton, 2, 35–36, 40–42, 44, 48–58 Ewe, 37–38 Ewondo, 42, 48, 56 Fijian, 254 Finno-Ugric languages, 87, 286 Fongbe, 206 Fox, 73, 74 French, 4, 7, 12, 19, 21, 58, 151, 155–157, 160–161, 164, 166– 168, 175, 192, 230, 241, 248– 249, 252–253, 255–257, 260, 265 French (Old), 11, 118–119, 155, 256 Fulfulde (Maasina), 232 Georgian (Old), 109–110 German, 117, 128–129, 145, 155– 156, 164, 166–167, 229, 253– 257, 260, 265, 360 German (Old High), 112–113, 117 Germanic languages, 112–115, 120, 155 Giramay, 97 Godoberi, 130, 137–140, 254–256, 265, 278, 291 Goemai, 359 Gothic, 117 Greek (Ancient), 252–253

Greek (Modern), 157, 260, 273, 281 Haida, 303 Hausa, 235–236, 242 Hebrew, 88, 175 Hindi, 88, 278, 281–282, 285, 289, 291 Hinukh, 130, 134–135, 142 Hokan languages, 344, 359 Hua, 206, 213 Hunzib, 130 Hup, 3, 85–86, 88–100 Icelandic (Old), 112–113 Indo-European languages, 87, 109, 116–117, 120–121, 155, 241, 265 Indo-Iranian languages, 117–120 Indonesian, 232 Indonesian (Riau), 226 Ingush, 200, 206, 216, 217 Inuttut (Labrador), 204, 206–208, 213 Iranian languages, 98, 115 Irish, 273–274, 277, 281 Italian, 12, 19, 21, 155, 260, 371 Itelmen, 208–209, 214, 217, 278, 281, 285 Jacaltec, 234 Jacaltek (Popti’), 358 Japanese, 206–207, 209, 212, 217, 265, 272–273, 279–281, 288– 289 Kabyle, 248–249, 265 Kalasha, 117–118 Kalmyk, 276–277, 280–281, 286– 287, 289–290 Karachay-Balkar, 175, 182, 192–193 Karata, 130 Kashmiri, 273, 277–278, 281–282, 285, 289

Index of languages Kati, 117–118 Ket, 206, 217 Khakas, 265, 280, 282, 284 Kham, 200, 206–207 Khinalug, 130, 145 Khwarshi, 130, 134–135, 142, 248, 251–252 Kiowa, 72, 76 Kipsigis, 273–274, 281–282 Klamath, 302, 316 Kobon, 235 Komi-Zyrian, 180, 277, 279–281, 287, 289–290 Korean, 175 Korowai, 206 Koryak, 117 Koyo, 39 Kurdish, 122, 281–282 Kurmanci, 117–118, 122 Kwa, 37–38, 51, 54–56 Kwakwala, 26, 278 Lahu, 213, 218 Lak, 130, 145 Lakhota, 303 Latin, 145, 151, 153, 168 Lezgian, 130, 138, 142–143, 248, 252–253 Lezgic languages, 130, 138, 142– 143, 145 Limbu, 265 Lithuanian, 175, 177, 183, 185, 192 Maasai, 181 Macedonian, 11, 26, 30 Makwe, 37, 49 Malagasy, 279–281 Manambu, 265 Maninka, 179–180 Maori, 175 Margi, 242 Mari (Meadow), 175, 235, 277, 279– 281, 287, 289–290

411

Marind, 91 Miraña (Bora), 6, 365, 367–368, 371, 373–381 Mixe, 72, 249 Mongolian (Khalkha), 240, 273 Mongolic languages, 276 Mon-Khmer, 36 Mosetén, 204, 206 Movima, 6, 323–324, 326, 330–331, 336–339 Muinane, 381 Mukri, 122 Munda, 36 Mundari, 323 Nadahup (Makú) languages, 88, 91, 99 Nadëb, 99 Nahuatl, 396, 403 Nakh languages, 130 Nakh-Daghestanian languages, see East Caucasian languages Nenets, 175, 178–180, 182, 193 Nganhcara, 26 Ngiyambaa, 26, 231 Niger-Congo languages, 3, 35–38, 41, 56–58, 241 Niuean, 274, 280 Nivkh, 186, 187 non-Indo-European languages, 109, 155 Numic languages, 299–300, 311 Nunggubuyu, 72, 395 Olutec, 249 Paiute (Northern), 5, 295, 297–299, 301–308, 310–316 Pamir languages, 115 Panoan languages, 304 Passamaquoddy, 278, 280, 281, 285–286, 291 Pilaga, 215

412

Index of languages

Piman languages, 316 Piro, 252, 265 Polish, 97, 110–112, 122 Portuguese, 156–157, 160, 170 Portuguese (European), 2, 12–15, 19–23, 27–31 Proto-Bantu, 49–51, 53–54 Proto-Indo-European, 120–121 Proto-Indo-Iranian, 120 Proto-Niger-Congo, 38, 54 Proto-Tupí-Guaraní, 68, 71, 80 Proto-Uto-Aztecan, 305, 313, 316 Quechua, 156, 159, 161–162, 165, 170, 241 Quechua (Cuzco), 274, 277, 279, 281 Quechua (Imbabura), 203–204, 215, 217, 228–229 Qunqi Dargwa, 127, 278, 282 Romance languages, 11, 22, 26, 30, 153, 156–157, 160, 164, 168 Romanian, 248 Russian, 7, 97, 107, 110–111, 114, 116, 122, 127–129, 175, 182– 185, 228–231, 234–236, 251– 252, 257–259, 263, 265, 290 Rutul, 130, 135, 137–138, 142–143, 145 Saliba, 358 Salishan languages, 323, 336 Sarykoli, 115, 120 Selkup, 187, 189 Semelai, 207 Semitic languages, 87, 98, 337 Serbo-Croatian, 11–12, 21, 26, 111, 122 Seri, 278 Shipibo-Konibo, 210, 215 Shoshone (Western), 311 Sinhalese, 88

Slave, 91 Slavic (Common), 110–111, 122 Slavic languages, 105, 110–111, 114–115, 120, 122 Slovene, 110–111 Songhay, 265 South Efate, 397, 401, 403 Spanish, 12, 19, 151–154, 156–170, 235, 332, 349, 380 Spanish (Peruvian), 4, 159, 161, 165 Suleimani, 122 Susu, 175, 185–186, 192 Swahili, 189, 380 Tabasaran, 130, 136–139, 143, 145 Tagalog, 281–282, 285–286, 289, 323, 336 Tai, 241, 396 Tai-Kadai languages, 241 Tajik, 175, 192 Takestani, 116 Tangut, 72 Tariana, 382 Telugu, 201 Tepehua, 72 Tetun, 233–234 Thai, 206 Tibeto-Burman languages, 72, 265 Tiene, 38–39 Tindi, 130 Tirahi, 119 Tokelauan, 358 Tongan, 323 Trumai, 265 Tsakhur, 130, 137, 139, 141–143, 278, 280, 282, 286, 289 Tsez, 130, 134–135, 140, 142, 277– 281, 285–286, 289, 291 Tsezic languages, 130, 134–135, 142, 145 Tucano, 91–92, 156 Tucanoan languages (Eastern), 88, 91, 374–375

Index of languages Tukang Besi, 206, 209 Tupí-Guaraní languages, 63–64, 70– 74, 77, 80, 323, 340 Turkic languages, 87, 98, 228, 272, 276, 279 Turkish, 156, 277, 281 Tuvan, 204, 275–277, 280–281, 284–287, 289–290 Tuvinian, see Tuvan Udi, 2, 12, 23–27, 29–31 Udmurt, 175, 188, 192–193 Uralic, 178–180, 188–189, 235, 241, 248, 287 Uto-Aztecan, 241, 300, 306, 316 Uzbekh, 272

413

Wakashan languages, 278 Wakhi, 115 Wanano, 91 Warekena, 249 Washo, 301–302, 315–316 Xârâcùù, 248, 251, 265 Yagua, 203, 206, 213, 218, 382 Yakima, 303 Yaqui, 109 Yimas, 313 Yucatec Maya, 366, 370 Yuhup, 88, 91, 99 Yukaghir (Kolyma), 190, 241, 403 Yurok, 186, 202, 217

Index of authors

Aaron, Jessi Elana, 153–154, 170 Abdoulaye, Mahamane L., 235–236, 242 Aikhenvald, Alexandra Y., 88, 108, 265, 295, 308, 313–314, 365– 366, 373–377, 379 Aissen, Judith, 87–88, 107 Aldai, Gontzal, 152, 158 Alonso, Amado, 161 Ambrazas, Vytautas, 177 Ameka, Felix, 389, 391 Amith, Jonathan D., 396, 403 Anderson, Gregory D. S., 217 Anderson, John M., 227, 241 Anderson, Stephen R., 11–13, 21, 23, 26–27, 30 Anderssen, Jan, 122 Andrews, Edna, 106 Arkadiev, Peter M., 3, 105, 264 Arkad’jev, Petr M., see Arkadiev, Peter M. Arkhipov, Alexandre, 1, 5, 223, 233–234, 241 Aronoff, Mark, 31 Arxipov, Aleksandr V., see Arkhipov, Alexandre Aschmann, Richard P., 381 Aske, Jon, 343, 346 Austin, Peter, 394 Backhouse, Anthony E., 217 Baerman, Matthew, 105 Bakker, Dik, 200 Barnes, Jonathan, 253 Bastin, Yvonne, 50 Battistella, Edwin L., 105–106, 108 Beaugrande, Robert de, 393, 401 Beavert, Virginia, 303

Béjar, Susana, 108 Beliën, Maaike, 146 Bender, Emily M., 400 Berger, Hermann, 116 Bermúdez-Otero, Ricard, 30 Berthele, Raphael, 360 Bertinetto, Pier Marco, 155, 157– 158, 160–161, 164 Bickel, Balthasar, 71, 339, 343, 365, 378 Bierwisch, Manfred, 107 Bird, Steven, 394, 398 Blake, Barry J., 88, 241 Bobaljik, Jonathan David, 278, 281, 285 Boeder, Winfried, 105, 108 Bokarev, Aleksandr A., 247 Bok-Bennema, Reineke, 206–207, 213 Bonč-Osmolovskaja, Anastasija A., 286 Borkin, Ann, 270, 282 Bossong, Georg, 85–88, 95 Bow, Catherine, 398 Bowerman, Melissa, 128–129, 146 Brandenstein, Carl Georg von, 395 Broschart, Jürgen, 323 Brown, Dunstan, 105 Brown, Penelope, 69 Bruening, Benjamin, 278, 285–286, 291 Brugger, Gerhard, 151, 161 Brugmann, Karl, 120 Bryan, Margaret, 51 Bulygina, Tat’jana V., 178 Burnard, Lou, 398 Butt, Miriam, 278, 285, 400

Index of authors Bybee, Joan L., 152–157, 160, 166, 168–170, 184, 312 Căluianu, Daniela, 248 Camacho, José, 235 Carlson, Gregory N., 178, 188 Carnie, Andrew, 273 Cartagena, Nelson, 170 Cèdèndamba, C., 240 Chafe, Wallace, 343 Chaker, Salem, 249 Chappell, Hilary, 232, 242 Chierchia, Gennaro, 178 Chikobava, Arnold S., 247, 250 Choi, Soonja, 128–129 Cole, Peter, 203, 217, 228–229 Comrie, Bernard, 73, 85, 87, 91, 97– 98, 145, 152, 156, 158–160, 164 Conrad, Robert J., 200 Corbett, Greville G., v–vi, 29, 91, 105, 317, 366, 371, 373, 380– 381 Cornips, Leone, 259–260 Couchili, Ti’iwan, 63, 71 Coulibali, Adama, 249 Cowie, A. P., 226 Crapo, Richley, 306, 317 Creider, Chet A., 286 Creissels, Denis, 58, 366 Croft, William C., 85, 87, 107, 323, 337, 379 Crum, Beverly, 311 Crysmann, Berthold, 19–20 Csato, Eva A., 403 Cukerman, Isaak I., 117 Curnow, Timoty J., 252 Cusic, David D., 177 Cuyckens, Hubert, 142, 146 Dahl, Östen, 152, 189–190 Daniel, Michael, 145, 233 Davies, John, 235, 270

415

Davies, William D., 235, 270 Dayley, Jon P., 305, 311 De Bray, R. G. A., 110–112 de Hoop, Helen, 100 de Swart, Henriette, 181 de Swart, Peter, 85, 87, 95, 100 de Vries, Lourens, 206 DeLancey, Scott, 71–72, 75–76, 302, 316 DeMello, George, 151, 161, 162, 168 Demers, Richard A., 323 Dench, Alan, 296, 389, 391 Devos, Maud, 37 Diesing, Molly, 178 Dixon, Robert M. W., 97, 107, 108, 122, 247, 250, 254, 265, 366, 390 Donegan, Patricia, 36, 38, 57 Donohue, Mark, 206, 209, 217 Dorian, Nancy, 392–393 Drabbe, P., 91 Dressler, Wolfgang, 177 Dryer, Matthew S., 57, 382, 400 Dubinsky, Stanley, 270 Durie, Mark, 224, 295, 313 Dybo, Vladimir A., 111 Dyvik, Helge, 405 Eckman, Fred R., 106 Edelman, Dzhoi I., 116–117, 119 Èdel’man, Dzhoj I., see Edelman, Dzhoi I. Eklund, Robert, 401 Ellington, John, 38 Epps, Patience, 1, 3, 85, 88, 91, 96 Escobar, Anna María, 156, 161 Essono, Jean-Marie, 42 Evans, Nicholas, 296, 323, 332, 336–337, 389, 391 Filimonova, Elena, 97–98 Fleck, David W., 304

416

Index of authors

Flickinger, Dan, 400 Foley, William A., 295, 313 Fortescue, Michael, 206, 217 Foulet, Lucien, 118 Frantz, Donald G., 284–285, 288, 291 Ganenkov, Dmitry, 4, 127, 145 Garrett, Andrew, 311 Gbegble, Nada, 37–38 Geniušienė, Emma Š., 262, 265 Gil, David, 226, 378 Gildea, Spike, 71 Gippert, Jost, 394 Gišev, Nukh T., 247 Givón, Talmy, 72, 74–75, 106–108, 188, 296 Gnalibouli, Boureima Amadu, 232 Goldberg, Adele, 201 Good, Jeff, 398, 400 Greenberg, Joseph H., 105, 107– 108, 375, 378, 382 Greimas, Algirdas J., 255 Grinevald, Colette, 358, 366, 376, 379 Gromova, Nelli V., 189–190 Gruškina, Elena V., 187, 189 Guillaume, Antoine, 358 Güldemann, Tom, 54 Gwynn, John Peter Lucius, 201 Haiman, John, 206, 213 Hale, Kenneth, 262 Hall, Daniel Currie, 108 Halpern, Aaron, 11–12, 21, 30 Hamilton, W. S., 122 Harley, Heidi, 273 Harris, Alice C., 12, 23–26, 31 Harris, Martin, 156–157, 160, 162, 164, 168–170 Harrison, Carl, 63, Harrison, David K., 217

Haspelmath, Martin, 57, 87, 91, 96, 106, 108–109, 143, 146, 234, 236, 242, 247, 250, 252–254, 259, 262, 339, 365 Haude, Katharina, 6, 323–325, 328– 330, 333, 335–336, 339 Heath, Jeffrey, 71, 75, 390, 395 Heine, Bernd, 170, 303, 309–310 Helimski, Eugene, see Xelimskij, Evgenij A. Hellwig, Birgit, 359 Hengeveld, Kees, 338 Henríquez Ureña, Pedro, 161 Hernández, José Esteban, 159 Himmelmann, Nikolaus P., 237–238, 241–242, 323, 337–338, 389, 394, 401 Hjelmslev, Louis, 107 Hook, Peter Edwin, 273, 277, 278, 285 Hooper, Robin, 358 Hopper, Paul J., 95, 151, 153, 154, 168 Hori, Hirofumi, 303 Hornby, A. S., 226 Howe, Chad, 4, 151–152, 159–166, 168 Hughes, Baden, 398 Hulk, Aalke, 259–260 Hyman, Larry, 35, 37–40, 54, 57 Ichikawa, Yasuko, 219 Idiatov, Dmitry, 48 Ivanov, Mixail Ju., 268 Iwasaki, Shoichi, 206, 217, 313 Jacobsen, William H., Jr., 300, 315, 316 Jacobson, Michel, 399 Jacobson, Steven, 206, 217 Jäger, Gerhard, 87 Jake, Janice, 274 Jakobson, Roman, 105

Index of authors James, Deborah, 285 Janda, Laura, 105 Jansen, Joana, 303 Janssens, Baudouin, 50 Jara Yupanqui, Ileana Margarita, 161, 170 Jelinek, Eloise, 323 Jensen, Cheryl, 63–64, 68, 80 Johns, Alana, 206–207, 213 Johnson, David E., 224 Jolivet, Alfred, 112 Joseph, Brian D., 273 Kaiser, Stefan, 217 Kakumasu, James, 80 Kalinina, Elena Ju., 284 Kaplan, Lawrence, 206, 217 Katamba, Francis, 366 Kaufman, Terrence, 381 Kaul, Vijay K., 273, 277–278, 285 Kazenin, Konstantin I., 242 Keenan, Edward L., 279 Keita, Boniface, 179 Kemmer, Susanne, 252 Keyser, Samuel J., 262 Khanina, Olesya, 4, 199–201, 210, 213, 216–218 Kibrik, Aleksandr E., 107, 145, 181, 182, 224, 247–248, 254, 278 Kibrik, Andrej A., 255 King, Tracy Holloway, 405 Kinkade, M. Dale, 323 Kiparsky, Carol, 282 Kiparsky, Paul, 282 Kisseberth, Charles, 44 Klaiman, M. H., 74 Klavans, Judith, 11–12, 20–22, 26, 29–30 Klimov, Georgij A., 116 Kluge, Friedrich, 255 Kobayashi, Noriko, 219 Kodzasov, Sandro V., 247–248, 267 König, Christa, 122

417

König, Ekkehard, 228, 233 Kratzer, Angelica, 178 Kroeber, Alfred Louis, 202 Kruspe, Nicole, 207 Kučera, Henry, 181 Kuno, Susumo, 271–272, 282, 286 Kuteva, Tania, 170, 358 Kuznecova, Ariadna I., 187, 189 Lai, Huei-ling, 242 Langacker, Ronald W., 282, 313, 316–317 Lasersohn, Peter, 177 Lasnik, Howard, 282 Launey, Michel, 324, 336, 338 Lazard, Gilbert, 85, 87–88, 95 Lefebvre, Claire, 206, 274, 277, 279 Legendre, Géraldine, 12 Lehmann, Christian, 223 Leite, Yonne, 63, 80 Letuchiy, Alexander B., 5, 247, 252, 256, 261, 265 Letučij, Aleksandr B., see Letuchiy, Alexander B. Levinson, Stephen C., 69, 128–129, 146, 346, 358 Li, Charles N., 375 Liljeblad, Sven, 297, 309, 311, 316 Lindenfeld, Jacqueline, 109 Ljutikova, Ekaterina A., 247–248, 253, 256, 262, 286 Lord, Carol, 233, 313 Lucy, John A., 366, 370, 378, 380 Luís, Ana R., 2, 11–12, 18–19, 23, 27, 29, 31 MacKenzie, David Neal, 122 MacWhinney, Brian, 397 Maisak, Timur, 25, 31 Manninen, Satu, 178 Margetts, Anna, 358 Martins, Ana Maria, 19–20, 22 Martins, Silvana, 91

418

Index of authors

Maslova, Elena, 190, 224, 231, 242 Masuichi, Hiroshi, 405 Matisoff, James A., 213 Mattheson, Esther, 252 Maurel, Didier, 63, 71 Maxmudova, Svetlana M., 142 Maxwell, Michael R., 374 Mayer, Thomas, 403 McConvell, Patrick, 20 Meeussen, Achilles, 49–50 Meillet, Antoine, 110 Meira, Sérgio, 80, 128–129, 146 Merdanova, Solmaz R., 184, 189 Mereshkov, Sultan, 200, 206 Miller, Philip, 12, 21, 26 Miller, Wick, 300 Mithun, Marianne, 300, 304, 343, 359, 376 Molčanova, E. K., 115, 118, 120 Monachesi, Paola, 12, 21, 26 Montserrat, Ruth, 71, 80 Moravcsik, Edith A., 106 Morey, Stephen, 396–397 Morse, Nancy L., 374 Mosel, Ulrike, 392, 394 Mossé, Fernand, 112 Moyse-Faurie, Claire, 248 Mpaayei, J. Tompo Ole, 181 Mufwene, Salikoko S., 380 Mulder, Jean, 277 Murav’jova, Irina A., see Muravyova, Irina A. Muravyova, Irina A., 248, 276 Musan, Renate, 178 Musgrave, Simon, 395–396 Muysken, Pieter, 274, 277, 279 Nakamura, Masanori, 286 Nathan, David, 390, 403 Nedjalkov, Vladimir P., 231–232, 242, 247, 252, 262 Newman, Paul, 176

Nichols, Johanna, 71–72, 75–76, 79, 216, 253 Nichols, Michael J. P., 297, 299 Nikolaeva, Irina, 403 Noonan, Michael, 215, 282 Nordhoff, Sebastian, 398 Nüse, Ralf, 146 Nzang Bie, Yolande, 50 Odden, David, 44, 274 Oepen, Stephan, 400 Ohori, Toshio, 236 Ohta, Kaoru, 290 Olovjannikova, Irina P., 267 Olson, Mark, 295 Onishi, Masayuki, 206–208, 209 Osada, Toshiki, 323, 332, 336–337 Ospina Bozzi, Ana María Otoguro, Ryo, 19, 29 Oxotina, Natal’ja V., 189, 190 Øvrelid, Lilja, 87 Paducheva, Elena V., 181 Padučeva, Elena V., see Paducheva, Elena V. Pagliuca, William, 152–157, 160, 166, 168–170, 184, 312 Panfilov, Vladimir Z., 186–187 Pardeshi, Prashant, 242 Patz, Elizabeth, 238–239, 243 Paulian, Christiane, 40 Paxalina, Tatiana N., 115 Payne, Doris L., 63, 72, 73, 74, 80, 203, 206, 213, 214, 382 Payne, John, 30 Payne, Thomas E., 74, 389, 392 Pazel’skaja, Anna G., 268 Pederson, Eric, 128–129 Pelletier, Francis J., 188 Penny, Ralph, 151, 161 Penton, David,, 398

Index of authors Perkins, Revere, 152–157, 160, 166, 168–170, 184, 312 Perlmutter, David M., 273, 275 Pesetsky, David, 270, 282 Peterson, David A., 253 Plank, Frans, 36, 113, 117, 119 Plungian, Vladimir, 184 Plungjan, Vladimir A., see Plungian, Vladimir Polinsky, Maria, 145, 277–278, 285– 286, 291 Pope, Mildred K., 118–119 Postal, Paul M., 270–273, 275, 277, 282 Preeya, Ingkaphirom, 206 Pullum, Geoffrey, 12, 24 Queixalós, Francesc, 63, 71, 323, 340 Ramirez, Henri, 91, 99 Rankin, Robert L., 296 Redden, James, 42 Reid, Nicholas, 396 Rice, Keren, 91 Rijkhoff, Jan, 200 Robins, Robert H., 186, 202 Rodrigues, Aryon Dall’Igna, 63 Rohrer, Christian, 405 Rood, David S., 303, 316 Rose, Françoise, 3, 63, 69–70, 80, 317 Rothemberg, Mira, 247, 256 Sackmann, Robin, 375 Sadock, Jerrold M., 11 Sag, Ivan A., 12, 21, 26 Saito, Mamoru, 282 Sakel, Janet, 206, 217 Samedov, Dzhalil’ S., 267 Sanches, Mary, 378 Sancho, Iker, 206 Sankoff, David, 154

419

Sanžeev, Garma, 273 Sapir, Edward, 304, 343 Sasse, Hans-Jürgen, 336–337 Sauerland, Uli, 122 Schachter, Paul, 330 Schadeberg, Thilo, 50–51 Schäfer, Florian, 260 Schanidse, Akaki, 109 Schlesinger, Izchak M., 225 Schroeter, Ronald,, 399 Schultze-Berndt, Eva F., 237–238, 241–242 Schwenter, Scott A., 159–160, 162– 165, 168, 170 Seifart, Frank, 6, 365, 369, 371, 373, 379, 381 Seiler, Hansjakob, 366 Seiter, William J., 274 Seki, Lucy, 63, 80 Senft, Gunter, 365 Serdobol’skaja, Natalja V., see Serdobolskaya, Natalia Serdobolskaya, Natalia, 5, 269, 271, 276–277, 279–281, 287 Serebrennikov, Boris A., 180 Shibatani, Masayoshi, 242, 259, 262 Shin, Yong-Min, 223 Shluinsky, Andrey, 4, 175, 177, 268 Siewierska, Anna, 75 Sil’nickij, Georgij G., 262 Silverstein, Michael, 70–71, 85, 97, 100, 107, 286 Simons, Gary, 394 Singer, Ruth, 236 Sinha, Chris, 358 Slobin, Dan I., 343, 345–346, 353, 357, 359–360 Slobin, Linda, 378 Smeets, Catharina J., 200 Smirnickij, Aleksandr I., 113 Smith, Lawrence R., 91, 107, 206– 207, 213 Smith-Stark, T. Cedric, 91, 107

420

Index of authors

Smyth, David, 206 Soares, Marília Faco, 71, 80 Spencer, Andrew, 11–13, 21, 23, 26–27, 30–31 Sperberg-McQueen, C. Michael, 398 Squartini, Mario, 155, 157–158, 160–161, 164 Stampe, David, 36, 38, 57 Stassen, Leon, 223–225, 233, 234 Steblin-Kamenskij, Mixail I., 112 Stenson, Nancy, 275, 277 Stenzel, Kristine, 88, 91 Stilo, Donald, 98 Stolz, Thomas, 223, 226–227, 241 Streitberg, Wilhelm, 117 Stroh, Cornelia, 223, 227, 241 Stump, Gregory T., 27 Šalbuzov, Kurban T., 143 Šluinskij, Andrej B., see Shluinsky, Andrey Šmeljov, Aleksej D., 228 Talmy, Leonard, 302, 343, 345, 353, 357, 359–360 Tatevosov, Sergei, 186, 268 Taylor, Allan R., 303, 316 Testelec, Jakov G., 146, 271 Thibault, Pierrette, 154 Thieberger, Nicholas, 6, 389, 395, 397, 399, 401 Thiesen, Wesley, 381 Thompson, Sandra A., 95, 375 Thornes, Tim, 5, 295, 297–298, 303, 312 Toivonen, Ida, 22 Torres Cacoullos, Rena, 159, 163, 165, 170 Traugott, Elizabeth C., 151, 153, 154 Trubetzkoy, Nikolay, 105 Tucker, Archibald N., 181 Urdze, Aina, 223, 227, 241

Vajda, Edward, 206 Valenzuela, Pilar M., 210 Van de Velde, Mark L. O., 35, 40, 56 van der Auwera, Johan, 184 van Geenhoven, Veerle, 177 van Klinken, Catharina, 233–234 van Staden, Miriam, 146 Vengoechea, Consuelo, 381 Verhelst, Mariet, 146 Veselinova, Ljuba N., 317 Vidal, Alejandra, 215 Vigário, Marina, 31 Volodin, Alexander P., 208–209, 217 Vydrine, Valentin F., 249 Walmsley, John B., 227 Watters, David E., 200, 206–207 Weber, David, 381, 389, 392, 396– 397 West, Birdie, 156 Westermann, Dietrich, 51 Wilkins, David P., 129, 146, 347, 353, 358, 360 Wirth, Jessica R., 106 Wogiga, Kepas, 200 Wood, Esther J., 177 Woodbury, Anthony C., 394, 401 Wurmbrand, Susi, 278, 281, 285 Xalilov, Madžid Š., 132, 142 Xanina, Olesja, see Khanina, Olesya Xanmagomedov, Bejdullax G.-K., 143 Xelimskij, Evgenij A., 187, 189 Xrakovskij, Viktor S., 177 Yamamoto, Hilofumi, 219 Yangklang, Peerapat, 345 Yap, Foong Ha, 313 Yar-Shater, Eshan, 115–116 Yatsushiro, Kazuko, 122

Index of authors Yoon, James H., 290 Yu, Alan C. L., 177 Zaliznjak, Anna A., 228 Zavala, Roberto, 249 Zeevat, Henk, 87

421

Zlatev, Jordan, 345 Žukova, Alevtina N., 117 Zúñiga, Fernando, 75–76, 79, 88, 97, 326 Zwicky, Arnold M., 11–12, 24

Index of subjects

accent, see prominence accompaniment, 226, 227 adjectives, 17, 107, 119, 122, 217, 237, 239, 331 affix phrasal, 2, 11, 12, 21–23, 28–30 word-level, 12–13, 15, 18, 21, 24, 26, 28 agentivity, 259, 261 agreement, 2, 3, 6, 12, 15, 17, 23, 39, 55, 57–58, 61, 70, 72, 74– 75, 77, 79, 92, 218, 235, 270, 277, 279, 284–285, 288–289, 291, 366, 375, 377, 379–380 alignment, 6, 63, 72, 75–76, 78–79, 85, 98, 323–324, 338 split (split-ergative), 338 allomorphy, 2, 12, 17, 23, 25, 27, 30, 31, 35 ‘reciprocal’, 14–16 stem, 23, 25, 26, 27, 30 Amazonia, 3, 6, 86, 99, 323, 340, 365, 374, 380 andative, 354, 356 Animacy Hierarchy, see hierarchy anterior, 155, 170 anticausative, 52 anticausative lability, 5, 247–249, 251–254, 256–257, 259–265 aoristic drift, 157–158 applicative, 36–37, 49–50, 53, 55– 56, 58, 298, 307–308, 312– 313 Archive of the Indigenous Languages of Latin America (AILLA), 396 argument structure, 5, 199, 212, 213, 302, 303, 306, 307, 334, 339

aspect, 4, 15, 152, 167, 175, 176, 179, 184, 259–261, 280–283, 295, 298–299, 306, 309–313, 317, 330, 344, 354, 358 associated motion, 306, 345, 353, 356, 358, 360 asymmetry, 23, 224, 226–227, 231, 234, 240, 242, 308, 313 attachment of clitics, 13, 15, 21, 22, 23, 25, 26, 27, 28, 29, 30 spatial, 4, 127–135, 137, 139– 141, 143–146 tonal, 47 attrition, 41, 49, 55, 56, 58, 214 Audiamus (software), 398 benefactive construction, 53, 313, 401 bipartite stems, 301–302, 315–316, 359 capacitive, 4, 176, 184–186, 190– 192, 194 cardinality, 234, 235 case case frame alteration, 204, 201– 211, 215 case-marking, 3–4, 70, 85–86, 88, 90, 97–98, 101, 107, 238, 272, 277, 279, 289–290, 344 causative, 36–37, 39, 49–50, 52–53, 55, 213, 239, 242, 247–248, 251–265, 295, 299, 302–303, 306–307, 310, 313 change expressions of, 6, 343–351, 353– 357

Index of subjects classification, 2, 6, 58, 129, 247– 249, 251–252, 264, 351, 359, 365–367, 370, 374, 377–381 classifiers, 6, 366–367, 370, 374– 380, 382 multiple, 374–375, 377, 382 noun, 366, 374, 375 numeral, 366–367, 370, 374– 375, 377–379 verbal, 366, 375–377 clause union, 283, 289, 291 clitic, 2, 11–31, 36, 43–45, 47, 57, 77, 80, 235, 316, 324–328, 332–334, 339, 344, 351, 360 placement, 11, 13, 15, 17, 19–23, 25, 27, 29, 31 enclitics, 2, 12, 15–23, 29, 31, 58, 91, 326–327, 329 proclitic triggers, 19, 20 proclitics, 2, 12, 18–23, 29, 45– 48, 53, 80, 297–298, 339 Second position, 11, 22 special, 2, 11, 12, 13, 30 Tobler-Mussafia, 11 verbal, 11, 18, 20–21, 30 Wackernagel clitic, see ~ Second position cliticization, 2, 11–13, 20, 80, 324– 328, 332–334 comitative, 5, 117, 223–243 case, 233 constructions, 223, 225, 229– 230, 234, 238, 240–243 coordinating comitative, 234, 236 verbs, 233 complementation, 209, 216, 279, 282, 291 complex predicate, 5, 315, 358 complexity cognitive, 106–107 structural, 2, 5, 106–107, 295 tonal, 42

423

compound predicates, 5, 295, 304– 308, 317, 344–345, 348–349, 352–354, 357–360 configuration (spatial), 128–145, 237, 239, 263, 348 constraints maximality constraints, see maximality on neutralization, 107–109, 115– 117, 121 construction benefactive, see benefactive comitative, see comitative desiderative, see desiderative typology of, 343, 348, 357, 359 control, 257–258, 270–271, 277, 282, 291 controller of agreement, 278, 373 converse lability, see lability coordination, 1, 225, 226, 229, 234, 235, 240, 242, 243, 314 comitative coordination, see comitative corpus, 159, 162, 389–395, 397, 399, 401–403 cross-reference, 63, 202, 300, 368, 371, 373, 376–377 data longevity of access, 391, 396, 403 metadata, see metadata portability of, 394, 396 declension, see inflection class definiteness, 3, 85, 86, 87, 88, 90, 95, 276, 284, 290, 378 depictives, 237–238, 240–241 derivation, 2–3, 5, 29, 37–38, 40–42, 49, 53–55, 221, 247–249, 253–257, 260–265, 299–300, 345, 355, 360, 368, 372, 379– 380, 382

424

Index of subjects

descriptive linguistics, 199, 201, 203, 205, 207, 209, 211, 213, 215, 217 desiderative construction, 4, 199– 202, 204–207, 209, 211–212, 214–218 detachment, 141 determiner, 21, 30, 297, 324, 327, 335–336, 338–339, 360 differential object marking, 3, 85–91, 93, 95–100 different-subject wanting, 201, 204, 205 different-subject, see wanting digital archive, 391 digital recording, 390, 395 digital repository, 393, 397 direct (vs. inverse), 72–73, 78, 325– 326, 331, 333–335, 337–340 case (vs. oblique), 109, 115–120 object, 87, 224, 250, 269–273, 275–277, 279, 281–283, 289– 290 direction and topological relation (DTR), 345, 348–349, 352, 358 dislocative, 354–356, 360 DoBeS (Dokumentation bedrohter Sprachen), 338, 359 documentary grammar, 389 elicitation, 95, 344, 390 epistemic modality, 151, 154, 161, 165, 167 EUROTYP, 152 event-external pluractional marker, 4, 177–180, 182–183, 185– 186, 188–189, 191 event-external pluractional marker, see pluractional evidentialilty, 153, 156, 161, 168 example sentence as ‘proof’ , 393

field recording, 397 fieldwork-based research, 389–390, 393 Figure (vs. Ground), 127, 130–137, 139–144 floating tones, 41–44, 49, 55, 57 framing type, 345, 356–357, 359– 360 frequency, 52, 96, 106–107, 109, 119, 151–152, 154, 158–162, 165, 167–170, 252, 283, 287, 303, 310, 316 of use, 52, 167, 169, 303 gender, 44–47, 53, 56–58, 107–108, 114–118, 122, 170, 193, 329– 330, 334, 366, 371, 374, 380, 382 generic, 4, 69–70, 79, 93, 184, 188– 190, 193–194, 224, 291, 329– 330, 398 grammatical categories, 98, 108, 120, 315, 366 dominant, 107–109, 112, 116, 121 grammatical description, 4, 179, 199, 204–205, 209–212, 215, 389– 390, 392–394, 398–400 grammaticalization, 2, 4–5, 56, 71, 100, 151–154, 157–158, 160, 162, 165–169, 191, 208, 295– 296, 298, 303, 306, 309–314 gravity, 131, 134–135 Ground (vs. Figure), 4, 127, 130– 137, 139–146 habitual, 177, 179, 181–184, 188– 190, 192, 194, 312 hierarchy animacy, 85–86, 88, 90, 93–94, 96–100, 107, 286–287 grammatical roles, 63, 66–67, 69–72, 75, 79

Index of subjects markedness, see markedness hierarchy, 107, 108 of the raised NP properties, 285 of verb classes, 283 person, 3, 63, 65–67, 70–73, 76, 78, 107 referential, 323, 325–326, 328, 337–338 relational, 224 Hokan hypothesis, 344, 359 holistic approach, see typology homophony, 112 IGT (interlinear glossed text), 398– 399, 400 imperfective, 181, 190, 192–194, 354, 360 inclusory constructions, 235–236, 243 incorporation, 25, 70, 298, 304, 335, 339–340, 368, 376–377 individual-level meaning, 4, 175, 193 predicate, 175–176, 178–181, 188, 193 state, 176, 180, 182–185, 190– 192, 194 individuation, 85, 87, 95, 98, 370, 375, 379–380 infinitive, 30, 48, 56, 58, 193, 272, 291 inflection class, 110–115, 119, 121– 122 inflectional formatives, 27–28, 30 ingressives, 401 instrumental, 38, 111, 122, 227 verbs, 300–301 instrumental prefix construction (IPC), 5, 295, 298–306, 308, 310, 315 inverse, 3, 63, 71–75, 78–80, 97, 325–326, 331, 334, 337, 338– 339

425

lability, 2, 5, 247–265 converse, 247–249, 251–252, 262–264 language archiving, 389–391, 396– 399, 401, 403 language change, 3, 106, 120–121, 296 language contact, 98, 118, 156, 161, 165, 167, 170, 357, 359, 374 language sample, 206, 207 layering, 154 lexical categories, see word classes lifetime effect, 178–179, 189 localization, 3, 61, 127–128, 130– 145 long-distance agreement (LDA), 270, 273, 277–281, 284–286, 291 markedness, 3, 29, 105–110, 114– 117, 119–121 correlates of, 106 hierarchies, see hierarchy minimization of, 108 theory of, 105–106, 110, 115, 119, 121 masdar, 279, 291 maximality constraints, 35, 37, 39– 42, 49, 51, 54–56, 58 mental verbs, 279–281, 283, 304 Mesoamerica, 359 metadata, 397, 399, 401, 402 Mexico, 6, 109, 162, 343–344 modal verbs, 154, 185, 194, 280– 283 morphosyntactic features and values, 3, 29, 108, 121 motion, 127, 131, 141, 239, 252, 257–258, 265, 295, 299, 302, 306–307, 309, 312, 315, 345– 348, 352–358, 360–361 motivation, 87, 95, 98–99, 105, 120– 121, 136, 139, 140, 227, 230, 248, 261–262, 343, 399, 402

426

Index of subjects

neutralization, 3, 14, 40, 69, 71, 105–119, 121 complete, 112 new technologies, 389, 391–393 nominalization, 43, 272, 275, 277, 279, 287, 297–298, 317, 333, 338–339, 370 noun class, 6, 58, 366, 369–375, 377, 379–382, 392 Oaxaca, 343, 344, 359 object markers, 94, 232 obviate, see obviative obviative, 285, 326, 339 omnipredicativity, see word class flexibility opposition, 4, 40, 58, 69, 106, 115– 116, 132, 178, 224, 234, 247, 248, 250, 251, 257, 259, 261, 264, 326 paradigm, 13, 27–29, 31, 42, 76–78, 80, 110, 113–114, 117–121, 179, 206, 212, 217, 262, 272, 299, 317, 343–344, 353, 358, 361, 390, 398 paradigmatic relationship, 212 paradigmatic structure, 114, 119 PARADISEC (Pacific and Regional Archive for Digital Sources in Endangered Cultures), 396 PARGRAM (software), 400 participant, 5, 65–67, 70–72, 74–75, 80, 162, 166, 177, 183–184, 201–203, 205, 215, 223–234, 237–240, 242, 247, 249–250, 257–259, 271, 282, 290, 314, 317, 325, 328, 334–338, 344, 348, 354, 356, 358, 361, 394 participant set, 224–225, 228, 230–231, 239, 242 pluralization, 5, 223, 225, 240 parts of speech, see word classes

passé composé (French), 4, 151, 157, 160–161, 164 passive, 36–37, 49–52, 55, 275–276 passive lability, 248–249, 251–252, 262, 264 path, 343–348, 352–355, 357–361 perception verbs, 252, 280–283, 358 perfect, 4, 151–170 durative, 157, 161 experiential, 156–157 types, 156, 159 perfective, 4, 42, 48–49, 151–153, 155–158, 160–161, 164–170, 182, 347, 354–356 periphrastic (compound) forms, 151, 153–161, 164, 166–167, 169– 170, 190 periphrastic constructions, 41, 49– 50, 54, 205, 208 persistence of location and naming of data files, 396 person hierarchy, see hierarchy person indexation, 3, 63–65, 69–70, 72–76, 78–80 person markers, 3, 12, 23–27, 29, 31, 64, 70, 74, 78, 80, 344, 360 phasal verbs, 280, 283, 356 pluractionality, 4, 175–194 polyadic, 225, 230–232, 242 polysemy, 225, 262–264 portability of data, see data possession, 96, 138, 144, 233, 240 possessor, 6, 64, 74, 77–78, 138, 290, 323, 327–329, 333, 335, 337–338 posture, 237, 295, 306–307, 309– 311, 313, 315, 348, 357, 359 predicate nominal, 331–335, 337 predicates individual-level, see individuallevel mental, see mental verbs modal, see modal verbs

Index of subjects of change, see change of perception, see perception verbs phasal, see phasal verbs polyadic, see polyadic stage-level, see stage-level predication (vs. reference), 323–324, 330–332, 334–339 preterit(e), 151–152, 157–160, 162– 170 pretérito perfeito composto (Portuguese), 156–157 prominence discourse, 55–56, 85, 286 syllabic, 35, 40–41, 49, 55–57 pronouns, 11–12, 14–19, 21, 56, 68, 85–86, 88, 91, 94, 97, 115– 116, 119, 270, 286–287, 297, 327–329, 339, 344, 360, 398, 401 property, 4, 176, 182–184, 190–192, 194 proprietive, 239 prosody, 3, 35–36, 40, 42–44, 47, 52–58, 111, 314, 326 proximate, 326–328, 335, 337, 339 qualitative, 4, 176, 186–188, 190– 192, 194 raising, 5, 269–291 reciprocal lability, 248–249, 251– 252, 264 reciprocals, 39, 51–52, 225, 230– 233, 242, 270–271, 279 ‘reciprocal allomorphy’, see allomorphy reduplication, 186–187, 317, 330– 331 reference, 87–88, 188, 323, 329, 335, 373, 380 non-singular, 230–231, 235, 370

427

temporal, 151–152, 154–156, 160–166, 181, 189 referential hierarchy, 6, 323, 325– 326, 328, 337–338 relative chronology, 120 relative clauses, 44, 317, 368, 371, 374–375 relative markedness, 108, 120–121 relativization, 55, 328, 336 repatriation of recordings, 402 resultative, 56, 153, 157, 237 same-subject, see wanting secondary verb construction (SVC), 5, 295, 298, 305–316 semantic map, 131, 144–146, 176, 190–191 semantic space, 153–156, 159–160, 168 semantic transitivity, 252, 256, 263 serialization, 37–38, 57, 209, 224, 232–233, 295, 308, 313–317, 345 simple forms, see synthetic forms simple predicates, 315, 344–348, 354–355, 357 sociative, 223, 228–230, 232–233, 240–242, 313 space, 138, 348, 354, 356, 358–359 spatial configuration, see configuration specificity, 86–88, 90, 96, 284 split ergativity, 71, 97 split intransitivity, 80 split plurality, 3, 85, 90–91, 93, 95– 96, 100 spontaneity, 252, 257, 348 stage-level predicate, 175–176, 178– 181, 193 state, 4, 159, 176, 178–185, 190– 192, 194, 237, 303, 316, 332, 335–336, 357

428

Index of subjects

change of, 301–302, 344–349, 353, 356–357 final (=endstate), 261, 350 stress, 36, 111, 325–326, 328 structural rank, 224–225 subordination, 1, 5, 20, 226, 237, 242, 280, 295, 298, 314–315, 317 suppletion, 39, 253, 309–312, 316– 317 surface, 128, 130–137, 139–140, 142, 144–146 symmetry, 231, 234, 242 syncretism, 3, 13–14, 105, 107, 110– 111, 113–122 synthetic (simple) forms, 151, 153– 155, 157, 160, 164, 170, 181, 183, 185, 189–190, 194 TAM, 2, 26, 42, 58, 152, 158, 169, 179, 339 TEI (Text Encoding Initiative), 398 tense-aspect-modality, see TAM time-aligned transcripts, 389, 392, 397, 401 togetherness, see sociative tone, 35, 40–47, 49, 52, 55–58, 212 topic , 55, 72, 74–75, 90, 93, 96, 227, 276, 285–286, 289–290 Transcriber (software), 398 transcript, 389–392, 397–398, 401– 402 typological generalization, v–vi, 1–3, 86, 98, 105, 145–146, 199, 211, 214–215, 296, 376, 378 typology, v–vi, 1–8, 35 family-internal, 2, 4, 35–37, 74, 127, 129, 131, 145 holistic approach, 35–36, 57 of attachment marking, 127–128 of clitics, 13 of constructions, see constructions

of desideratives, 199 of neutralization, 105 of perfect types, 156, 159 of verb agreement, 75 semantic, 346, 357 underspecification, 108, 240 unification (morphological), 114, 120–121 uniform clitic systems, 21 uniformity morphosyntactic, 213 syntactic, see word class flexibility universal (patterns, structures, etc.), 182, 201, 205, 210–211, 214, 234, 323, 365, 378 universals, 1, 105, 107, 382 implicational, 107, 253 valency, 250 valency(-)changing, see valency derivation valency derivation, 5, 36–37, 49, 241, 247–249, 257, 261–265, 297–298, 328, 338 venitive, 100, 354, 360 verb formation, 295, 306, 315 verbal number, 226, 310, 317 wanting, 200–201, 204–205, 209, 211, 217 different-subject, 201, 203–205, 208–209, 211, 213–214 same-subject , 201, 203–205, 208–209, 303 word classes, 291, 297 word class flexibility, 6, 323–324, 336, 338 word order flexibility, 76, 297 XML (Extensible Markup Language), 399, 400