Valency Classes in the World's Languages Volume 1: Introducing the Framework, and Case Studies from Africa and Eurasia [1] 3110332949, 9783110332940

Earlier empirical studies on valency have looked at the phenomenon either in individual languages or a small range of la

229 104 4MB

English Pages 873 [886] Year 2015

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Valency Classes in the World's Languages Volume 1: Introducing the Framework, and Case Studies from Africa and Eurasia [1]
 3110332949, 9783110332940

Table of contents :
Contents
Abbreviations
Acknowledgments
List of authors
Part I: The Leipzig Valency Classes Project: Introducing the Framework
1 Introduction
2 Leipzig Questionnaire on valency classes
3 Comparing verbal valency across languages
4 Valency classes and alternations: parameters of variation
5 Transitivity prominence
5a Assessing transitivity prominence from a statistical perspective: A commentary on Martin Haspelmath’s “Transitivity prominence”
6 Statistical observations on implicational (verb) hierarchies
Part II: Case Studies
Africa
7 Valency in Nllng
8 Valency properties of Mandinka verbs
9 Emai valency classes and their alternations
10 Valency classes in Yorùbá
11 Valency properties of verbs in Modern Standard Arabic
Eurasia
12 Icelandic valency classes: oblique subjects, oblique ambitransitives and the actional passive
13 Valency patterns in Italian
14 Valency classes in Eastern Armenian
15 Valency and valency classes in Bezhta
16 Valency classes in Even (North Tungusic) in a comparative Tungusic perspective
17 Valency properties of the Ket verb clause
18 Flexible valency in Chintang
19 Valency classes in Mandarin
20 Valency classes in Japanese
21 Valency classes in Ainu
Language index
Subject index

Citation preview

Valency Classes in the World’s Languages

Volume 1 Introducing the Framework, and Case Studies from Africa and Eurasia

Edited by Andrej Malchukov Bernard Comrie

DE GRUYTER MOUTON

Andrej Malchukov and Bernard Comrie (Eds.) Valency Classes in the World’s Languages Vol. 1

Comparative Handbooks of Linguistics

Edited by Edith Moravcsik and Andrej Malchukov

Volume 1.1

ISBN 978-3-11-033294-0 e-ISBN (PDF) 978-3-11-033881-2 e-ISBN (EPUB) 978-3-11-039527-3 ISSN 2364-4354 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2015 Walter de Gruyter GmbH, Berlin/Boston Cover image: Jupiterimages/PHOTOS.com/thinkstock Typesetting: Meta Systems Publishing & Printservices GmbH, Wustermark Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Volume 1 Abbreviations

ix x

Acknowledgments List of authors

xi

Part I: The Leipzig Valency Classes Project: Introducing the Framework Bernard Comrie, Iren Hartmann, Martin Haspelmath, Andrej Malchukov, and Søren Wichmann 1 Introduction 3 Andrej Malchukov and the Leipzig Valency Classes Project team 2 Leipzig Questionnaire on valency classes 27 Martin Haspelmath and Iren Hartmann 3 Comparing verbal valency across languages

41

Andrej Malchukov 4 Valency classes and alternations: parameters of variation Martin Haspelmath 5 Transitivity prominence

73

131

Damian E. Blasi 5a Assessing transitivity prominence from a statistical perspective: A commentary on Martin Haspelmath’s “Transitivity prominence” 149 Søren Wichmann 6 Statistical observations on implicational (verb) hierarchies

Part II: Case Studies Africa Martina Ernszt, Tom Güldemann, Alena Witzlack-Makarevich 7 Valency in Nǁng 185

155

vi

Contents

Denis Creissels 8 Valency properties of Mandinka verbs

221

Ronald P. Schaefer and Francis O. Egbokhare 9 Emai valency classes and their alternations Joseph Dele Atoyebi 10 Valency classes in Yorùbá

261

299

Csilla Kász 11 Valency properties of verbs in Modern Standard Arabic

327

Eurasia Jóhanna Barðdal 12 Icelandic valency classes: oblique subjects, oblique ambitransitives 367 and the actional passive Michela Cennamo 13 Valency patterns in Italian

417

Michael Daniel and Victoria Khurshudian 14 Valency classes in Eastern Armenian

483

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova 15 Valency and valency classes in Bezhta 541 Andrej Malchukov and Igor V. Nedjalkov 16 Valency classes in Even (North Tungusic) in a comparative Tungusic perspective 571 Edward J. Vajda 17 Valency properties of the Ket verb clause

629

Robert Schikowski, Netra Paudyal, and Balthasar Bickel 18 Flexible valency in Chintang 669 Bingfu Lu, Guohua Zhang and Walter Bisang 19 Valency classes in Mandarin 709 Hideki Kishimoto, Taro Kageyama and Kan Sasaki 20 Valency classes in Japanese 765

vii

Contents

Anna Bugaeva 21 Valency classes in Ainu 855

Language index Subject index

807

861

Volume 2 Andrej Malchukov and the Leipzig Valency Classes Project team Leipzig Questionnaire on valency classes xi

Part II: Case Studies (cont.) Austronesia and the Pacific Masayoshi Shibatani and Ketut Artawa 22 Balinese valency classes 877 Thomas Conners, John Bowden and David Gil 23 Valency classes in Jakarta Indonesian 941 Sebastian Nordhoff 24 Case frames in Sri Lanka Malay

987

Claire Moyse-Faurie 25 Valency classes in Xârâcùù (New Caledonia) Nicholas Evans 26 Valency in Nen

1015

1069

Eva Schultze-Berndt 27 Complex verbs, simple alternations: valency and verb classes in 1117 Jaminjung

Americas Osahito Miyaoka 28 Valency classes in Central Alaskan Yupik, an Eskimoan language

1165

viii

Contents

Richard Rhodes and Rand Valentine 29 Transitivity in Ojibwe 1205 Iren Hartmann 30 Valency Classes in Hoocąk (Ho-Chunk) Honoré Watanabe 31 Valency classes in Sliammon Salish

1265

1313

Zarina Estrada-Fernández, Mercedes Tubino and Jesús Villalpando 1359 32 Valency classes in Yaqui Eric Campbell 33 Valency classes in Zenzontepec Chatino Christian Lehmann 34 Valency classes in Yucatec Maya Frank Seifart 35 Valency classes in Bora

1391

1427

1481

Fernando Zúñiga 36 Valency classes in Mapudungun

1515

Part III: Theoretical Outlook Christian Lehmann 37 Situation types, valency frames and operations

1547

Tasaku Tsunoda 38 The hierarchy of two-place predicates: its limitations and uses Beth Levin 39 Verb classes within and across languages

1597

1627

Cliff Goddard 40 Verb classes and valency alternations (NSM approach), with special 1671 reference to English physical activity verbs Language index Subject index

1703 1709

Abbreviations 1 2 3 A ABL ABS ACC ADJ ADV AGR ALL ANTIP APPL ART AUX BEN CAUS CLF COM COMP COMPL COND COP CVB DAT DECL DEF DEM DET DIST DISTR DU DUR ERG EXCL F FOC FUT GEN IMP INCL IND INDF

first person second person third person agent-like argument of canonical transitive verb ablative absolutive accusative adjective adverb(ial) agreement allative antipassive applicative article auxiliary benefactive causative classifier comitative complementizer completive conditional copula converb dative declarative definite demonstrative determiner distal distributive dual durative ergative exclusive feminine focus future genitive imperative inclusive indicative indefinite

INF INS INTR IPFV IRR LOC M N NNEG NMLZ NOM OBJ OBL P PASS PFV PL POSS PRED PRF PRS PROG PROH PROX PST PTCP PURP Q QUOT RECP REFL REL RES S SBJ SBJV SG TOP TR VOC

infinitive instrumental intransitive imperfective irrealis locative masculine neuter non- (e.g. NSG nonsingular, NPST nonpast) negation, negative nominalizer/nominalization nominative object oblique patient-like argument of canonical transitive verb passive perfective plural possessive predicative perfect present progressive prohibitive proximal/proximate past participle purposive question particle/marker quotative reciprocal reflexive relative resultative single argument of canonical intransitive verb subject subjunctive singular topic transitive vocative

Acknowledgments The project on Valency Classes in the World’s Languages, some of whose results are reported in this volume, was funded by the Deutsche Forschungsgemeinschaft (DFG) as part of the DFG-Forschergruppe 742 (Grammar and Processing of Verbal Arguments), and was based at the Department of Linguistics of the Max Planck Institute for Evolutionary Anthropology, Leipzig. This volume partly originated from two workshops (2010, 2013) and a conference (2011) held in Leipzig and a workshop held at the 44th Annual Meeting of the Societas Linguistica Europaea at the Universidad de la Rioja, Logroño, Spain. We would like to thank the contributors to the volume and other participants in the conference and workshops for useful discussion. We further thank Max Planck staff members Claudia Bavero, Claudia Büchel, Julia Cissewski, and Peter Fröhlich for their administrative and technical support in organizing the Leipzig conference and workshops. We benefited greatly from the contribution of our highly motivated and reliable student assistants in editing and typesetting this volume; thanks are due especially to Diana Schackow and Tina Gregor. Finally, we thank Claudia Bavero for her invaluable work preparing the indexes and correcting the proofs. Andrej Malchukov and Bernard Comrie Leipzig, 15 May 2015

List of authors Ketut Artawa, Udayana University, E-mail: [email protected] Joseph Dele Atoyebi, Collège Boréal, E-mail: [email protected] Jóhanna Barðdal, Department of Linguistics, Ghent University, E-mail: [email protected] Balthasar Bickel, Universität Zürich, E-mail: [email protected] Walter Bisang, Johannes Gutenberg University, Mainz, E-mail: [email protected] Damian E. Blasi, Max Planck Institute for Mathematics in the Sciences & Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] John Bowden, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Anna Bugaeva, National Institute for Japanese Language and Linguistics, E-mail: [email protected] Eric Campbell, University of California, Santa Barbara, E-mail: [email protected] Michela Cennamo, University of Naples Federico II, E-mail: [email protected] Bernard Comrie, Max Planck Institute for Evolutionary Anthropology & University of California, Santa Barbara, E-mail: [email protected] Tom Conners, University of Maryland, E-mail: [email protected] Denis Creissels, Université Lyon 2, E-mail: [email protected] Michael Daniel, Higher School of Economics / Moscow State University, E-mail: [email protected] Francis O. Egbokhare, University of Ibadan, E-mail: [email protected] Martina Ernszt, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Zarina Estrada-Fernández, Universidad de Sonora, E-mail: [email protected] Nicholas Evans, Australian National University, E-mail: [email protected] David Gil, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Cliff Goddard, Griffith University, E-mail: [email protected] Tom Güldemann, Humboldt-Universität zu Berlin & Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Iren Hartmann, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Martin Haspelmath, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Taro Kageyama, National Institute for Japanese Language and Linguistics & Kwansei Gakuin University, E-mail: [email protected] Csilla Kász, Christian-Albrechts-Universität zu Kiel, E-mail: [email protected] Madzhid Khalilov, G. Tsadasa Institute for Language, Literature and Art of the Daghestan Scientific Center of the Russian Academy of Sciences & Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Zaira Khalilova, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Victoria Khurshudian, Centre national de la recherche scientifique, E-mail: [email protected] Hideki Kishimoto, Kobe University, E-mail: [email protected] Christian Lehmann, Universität Erfurt, E-mail: [email protected] Beth Levin, Stanford University, E-mail: [email protected] Bingfu Lu, Nanchang University, E-mail: [email protected]

xii

List of authors

Andrej Malchukov, Institute for Linguistic Research, Russian Academy of Sciences, St. Petersburg & Johannes Gutenberg University, Mainz & Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Osahito Miyaoka, Kyoto University, E-mail: [email protected] Claire Moyse-Faurie, Centre national de la recherche scientifique, E-mail: [email protected] Igor Nedjalkov, St. Petersburg State University & Institute for Linguistic Research, Russian Academy of Sciences, St. Petersburg, E-mail: [email protected] Sebastian Nordhoff, Freie Universität Berlin, E-mail: sebastian.nordhoff@langsci- press org Netra Paudyal, Universität Leipzig, E-mail: [email protected] Richard A. Rhodes, University of California Berkeley, E-mail: [email protected] Kan Sasaki, Sapporo Gakuin University, E-mail: [email protected] Ronald P. Schaefer, Southern Illinois University Edwardsville, E-mail: [email protected] Robert Schikowski, Universität Zürich, E-mail: [email protected] Eva Schultze-Berndt, The University of Manchester, E-mail: [email protected] Frank Seifart, Universiteit van Amsterdam, E-mail: [email protected] Masayoshi Shibatani, Rice University, E-mail: [email protected] Tasaku Tsunoda, National Institute for Japanese Language and Linguistics, E-mail: [email protected] Mercedes Tubino, Western Michigan University, E-mail: [email protected] Edward J. Vajda, Western Washington University, E-mail: [email protected] J. Randolph Valentine, University of Wisconsin-Madison, E-mail: [email protected] Jesús Villalpando, University of Colorado Boulder, E-mail: [email protected] Honoré Watanabe, Research Institute for Languages and Cultures of Asia and Africa, Tokyo University of Foreign Studies, E-mail: [email protected] Søren Wichmann, Max Planck Institute for Evolutionary Anthropology, E-mail: [email protected] Alena Witzlack-Makarevich, Christian-Albrechts-Universität zu Kiel, E-mail: [email protected] Guohua Zhang, Research Center of Chinese Language History, Zhejiang University & School of Foreign Languages, Zhongnan University of Economics and Law, E-mail: [email protected] Fernando Zúñiga, Universität Bern, E-mail: [email protected]

Part I: The Leipzig Valency Classes Project: Introducing the Framework

Bernard Comrie, Iren Hartmann, Martin Haspelmath, Andrej Malchukov, and Søren Wichmann

1 Introduction 1 General information The present volume is one of the main products of the project “Valency classes in the languages of the world”, carried out by the authors of this chapter. This project grew out of an earlier project “Ditransitive constructions in the languages of the world”, which gave rise, among other publications, to Malchukov et al. (2010). The topic of the later project is clearly related to that of the earlier one, but it is also significantly different. The Valency Classes project has aimed to investigate the argument-structure properties of verbs of different valency classes in a typological perspective. It thus continued the line of research of the ditransitive project, but focused on lexical properties of verbs and extended the research questions to a variety of other valency classes, thus making a contribution to the nascent field of lexical typology. The present volume presents the results of an empirical study of a relatively large set of core verb meanings (70) for a relatively small set of languages (30);1 it can be viewed as a compromise between Levin’s (1993) classification and semantically based classifications proposed in the typological literature. The other main product of the project is the on-line database Valency Patterns Leipzig (Hartmann et al. 2013).

2 Background 2.1 Approaches to the study of valency classes All verbs in a language have different meanings, but with respect to their syntactic patterning, they show striking similarities and fall into a relatively circumscribed number of valency classes whose members behave alike. By “syntactic patterning”, we refer not just to the coding patterns (i.e. the ways in which the verbal arguments are flagged and indexed), but also to the behavior of verbs with respect to alterna-

1 The accompanying database increases this to 80 verb meanings and 36 languages. Some of the extra 10 verb meanings, and also further verb meanings selected as interesting by individual authors, are included in the chapters of Part II.

4

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

tions such as causative, passive, applicative and other constructions that are not fully productive for all verbs (e.g., secondary predicates, certain word order phenomena). It has often been noted that these syntactic verb classes typically correlate with semantic classes. The syntactic properties of verbs can be studied separately from the semantic classes, so the strength of the correlation is an empirical question. The literature on syntactic verb classes is vast, yet with few exceptions it is limited to relatively few languages, most of them European. A classic study in this area is Levin’s (1993) book on English verb classes, where she argues that verb classes are semantically based and can be identified in terms of their argument structures, possible argument structure alternations, as well as further syntactic diagnostics (such as middle alternation, unspecified object deletion, there-insertion, etc.). Unlike earlier studies, which divided the verbal lexicon into a few highly general classes (e.g., stative vs. active verbs, intransitive vs. transitive vs. ditransitive, or, for intransitives, unaccusatives and unergatives), Levin attempts a much more fine-grained classification, which is constructed in a bottom-up fashion, and where verb classes are defined in terms of their overall syntactic distribution. Levin’s study (as well as subsequent work with M. Rappaport Hovav; e.g., Levin & Rappaport Hovav 2005) has been highly influential not only in the theoretical work on lexical semantics, but also in computational linguistics, and underlies verb ontologies in WordNet (Fellbaum 1998) and extensions thereof such as VerbNet (see Kipper-Schuler 2005 for references) and FrameNet (see, e.g., Fillmore et al. 2003; see also Schulte im Walde 2003 for discussion of verb classes in German). But neither Levin’s study nor the pioneering study Experimental Investigation of the Russian verb by Apresjan (1969), where verb classes were also established on syntactic grounds, have been extended cross-linguistically. Jones (1994) is a small collection of working papers dealing with verb classes in English, German, Korean and Bangla, stemming from an MIT-based project explicitly aiming to extend the Levin-style classification to other languages. There are also occasional studies dealing with other languages (cf. Fukui et al. 1985 on Japanese; Vogel 2003 on Jarawara). Even contrastive studies devoted to a single verb type, such as the study of interaction verbs in English, German, Hungarian and Maori in Blume (1998), are rare. This lack of an extension to more languages does not seem to be accidental. In spite of its merits, Levin’s approach faces a number of problems which become evident once one attempts to extend it beyond English. First, since Levin’s classes are constructed on syntactic criteria, they are not always semantically coherent. This is a serious drawback for typological studies, where the phenomena to be investigated have to be defined in semantic terms to make a comparison possible. The main question is: Which aspects of this classification are universal and which are language-particular? Clearly, the details cannot be universal as the study refers to specific language forms (e.g., encoding of arguments through specific case

Introduction

5

forms, prepositions, etc.). Similarly, the encoding of alternative constructions, as well as syntactic diagnostics like the English Middle alternation are clearly not universal. Yet, it is expected that universal cross-linguistic patterns do exist, insofar as both cross-linguistically recurrent coding properties as well as the availability of certain alternations have a semantic motivation. This has already been anticipated in a work on transitivity alternation by Pinker (1989), whose approach is close to Levin’s, but additionally tries to provide explanations for encoding options in terms of semantic properties of verb classes on the one hand and the semantics of the alternation on the other hand (thus, for example, the middle alternation targets EFFECT verbs like break, not CONTACT verbs like hit). A different tradition in the research on verb classes (or, valency classes) takes its origin in the work on Case Grammar, different versions of which were developed in the 1970s by Fillmore, Gruber, Cook, J. Anderson, Jackendoff, and Chafe, among others. In this tradition, verb classes are identified in terms of the semantic roles of the verbal arguments. A related approach has been developed (particularly in France and Germany) in the work by Tesnière, Gross, Helbig, and others (see Ágel 2006 for a comprehensive bibliography of valency research and Herbst & GötzVotteler 2007 for a representative sample of contemporary approaches to valency research). In this approach, however, verbal valency types are defined more in terms of formal than semantic criteria (see, e.g., Somers 1987 for a comparative treatment). In the subsequent literature, the argument structure of verb classes has played an important role in linguistic theories of different persuasions (see, for instance, Levin & Rappaport Hovav 2005 and Butt 2006 for overviews and discussion), yet this research has rarely been carried out systematically. In the generative literature the issue of subcategorization frames of different verb classes has been present for a long time, but it has not been addressed systematically. Thus, while there is a large literature on individual verb classes found to be of particular theoretical interest (cf. the work by Grimshaw 1990; Pesetsky 1995 and others on emotion verbs, which present challenges for argument linking), comprehensive studies of verb classes in other languages have not been attempted. While the empirical basis of mainstream generative grammar has not been very broad until recently, other theories like Role and Reference Grammar (Van Valin & Lapolla 1997; Van Valin 2001) and Lexical Decomposition Grammar (Wunderlich 2006; Stiebels 2000) have developed a strong typological orientation. These studies pioneered systematic research into semantic argument types of languages of different alignment (in particular, the work by Van Valin has contributed to the study of argument alternations cross-linguistically). Yet, these studies operate in terms of broad valency-based classes rooted in aspectual properties and lexical decomposition, and never reach the level of granularity of Levin’s classification (see, for instance, the study of verb classes in Wunderlich 2006; cf. Joppen-Hellwig 2001). This is even true for valency studies, which have been specifically designed for capturing lexical variation in syntactic properties. Valency dictionaries remain few

6

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

and are mostly confined to European languages (see Schumacher 2006a, b; Busse 2006 for references). It is also instructive that the monumental handbook of dependency and valency edited by Ágel (2006), while containing some contrastive studies of valency, does not provide a systematic overview even for European languages. In typology, the cross-linguistic study of verb classes rooted in the Case Grammar and valency traditions was advanced especially by Lazard (1998), Lehmann (1991), and Drossard (1991). This research has contributed much to the understanding of how different verb types pattern in languages of different alignment, yet, like much of the research discussed earlier, it deals with rather broad classes. A partial exception is Lehmann’s work, which achieves a finer cross-classification through the use of different criteria: aspectual (Vendler-style) criteria, verbal valency, and properties of arguments. Another approach, which is typological in nature, although it has been applied to English, is that of Dixon (1991, 2005). Dixon’s taxonomy of verb classes, which predates Levin’s classification, is different inasmuch as Dixon’s is semantic in nature − the syntactic properties (even verbal valency) are secondary for his classification. The established classes are very general (although some are further subdivided) and include verbs of different valency; thus the MOTION type includes verbs such as run and take, CORPOREAL includes verbs such as eat and die, and so on. Yet, this classification, like Lehmann’s, is of obvious interest to typology because it is semantically based and thus can be more easily extended to the study of other languages. One general drawback of the typological work mentioned above is that it remains taxonomic in nature and has not led to broader generalizations. The only exception, to our knowledge, is the work by Tsunoda (1981, 1985), who proposed the following hierarchy of verb types that predicts the distribution of intransitive and transitive patterns in individual languages: Effective action > Perception > Pursuit > Knowledge > Feeling > Relation This hierarchy represents a scale stretching from the more transitive verb types on the left to the less transitive verb types on the right. It is called a hierarchy since it predicts that if a verb type lower in the hierarchy allows for a transitive case frame (NOM-ACC in accusative languages or ERG-ABS in ergative languages), so do verb types higher in the hierarchy. Tsunoda’s approach is generally considered an important contribution to the study of verb classes (cf. Lazard 1994; Lehmann 1991). Yet until recently, this line of research has not been pursued further in typological work, as it faces a number of empirical problems (see Malchukov 2005 for discussion and references). Malchukov (2005), however, suggests that counterexamples can be accounted for by decomposing Tsunoda’s hierarchy into the two dimensions of decreased patienthood of P (from ‘break’ to ‘look for’ and ‘go to’), and decreased agenthood of A (from ‘break’ to ‘like’ and ‘hurt’), recasting Tsunoda’s hierarchy as two-dimensional:

Introduction

effective action

contact

pursuit

motion

perception cognition

emotion

sensation

7

Fig. 1: Two-dimensional Transitivity Hierarchy (semantic map).

Importantly, this hierarchy can be seen as a part of a larger semantic map showing how different verb types pattern in the semantic space from transitive to intransitive verbs, as explained in 2.2 (see Haspelmath 2003 for a general discussion of the semantic map approach). Among recent contributions to the issues of verb classification and valency alternation one can mention the collection of papers Kratochvíl et al. (2011), focusing on the concept of transitivity and on transitivity mismatches (syntactic vs. morphological transitivity, etc.), another collection of papers on issues relating to verb classification from various theoretical perspectives (De Clerck et al. 2013), as well as Croft’s (2012) monograph providing a comprehensive discussion of argument structure and argument alternations from a cross-linguistic perspective. These studies are representative of the state of art of the research into the issues of verb classification and argument alternations, yet they do not attempt to offer a comprehensive typology of valency classes. As is clear from the brief discussion of the research on valency classes, even though each of the approaches has made important contributions to the study of verb classes, they all have certain drawbacks as well. The Valency Classes project selectively built on the advantages of different approaches, in particular the following desiderata: – the taxonomy should be fine-grained enough, as in Levin’s approach; – it should be semantically based such that it can be applied to other languages, as in the different typological approaches (Lehmann, Dixon, and others); – since the taxonomy is semantically based, the syntactic properties of these classes can be studied without the danger of circularity (cf. the work on ditransitive constructions reported in Malchukov et al. 2010); – the general purpose is not just taxonomic, but aims to uncover universal and language-particular properties of valency classes in terms of coding properties and behavior (as in Tsunoda’s hierarchies/semantic maps).

2.2 Semantic maps In the Valency Classes project, as in the earlier Ditransitives project, cross-linguistic similarities between valency classes are captured in the form of semantic maps.

8

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

contact

pursuit

motion

interaction effective action

affected Agent

reflexive

perception n

middle

cognition

spontaneous

emotion

i n t r a n s i t i v e

sensation

Fig. 2: Transitivity reduction on a semantic map.

The semantic map method (as developed by Anderson 1986; Haspelmath 1997, 2003; and Croft 2001; cf. also Cysouw et al. 2010) aims at uncovering semantic similarities among individual categories on the basis of cross-linguistically recurrent polysemy patterns. It is assumed − because of iconicity (Haiman 1985) − that recurrent similarity in form reflects similarity in meaning. The categories that are similarly encoded are therefore put contiguously in a semantic space. The central assumption of the semantic map methodology is that the semantic space is universal; thus a semantic map, once established empirically, makes universal predictions about possible and impossible (or rather, probable and improbable) polysemy patterns. Semantic maps have been applied to different domains, including verb types. Thus, Malchukov (2005) integrates Tsunoda’s hierarchy into the semantic map in Figure 2. This map has the immediate advantage, as compared to Tsunoda’s one-dimensional hierarchy, that it can incorporate different extensions of the transitive frame across different dimensions that are conflated by Tsunoda. For example, English differs from Japanese in extending the transitivity pattern further along the second subhierarchy (to emotion verbs like like or fear, which pattern intransitively in Japanese; cf. osorosii ‘fear(ful)’), but is more conservative with respect to the first subhierarchy (PURSUIT verbs in Japanese, unlike in English, pattern transitively; cf. matu ‘wait for’). This map also includes some other verb types proposed in the literature, which, however, differ in granularity. Thus, it additionally includes the categories of INHERENTLY REFLEXIVE, MIDDLE, and SPONTANEOUS verbs representing the reduction of transitivity along the dimension of distinguishability between subject and object (cf. Kemmer 1993). It also includes INTERACTION verbs discussed by Blume (1998), which are linked to PURSUIT and CONTACT verbs on the map. On the one hand, interaction verbs show similarities to PURSUIT verbs (both are treated as subclasses of AIMING verbs by Lazard 1998). Clearly, this hierarchy is still not fine-grained enough, as many classes may need further decomposition; for example, it is instructive that both Tsunoda’s PURSUIT verbs and

Introduction

Internal Possessor construction

External possession construction

Malefactive Source construction

9

BREAK him X

Patient Beneficiary construction SELL

BUILD him a house

Theme-Recipient construction GIVE SAY

SEND TELL FEED THROW

Patient Instrumental construction

Theme-Goal construction

(with HIT verbs) SPRAY/LOAD

(PUT,PULL)

Fig. 3: Semantic map for ditransitive verbs. The lines indicate an approximate range of verbs participating in the English Double Object Construction and to-NP Construction; their intersection delimits verb types participating in a dative alternation; DOC ------, to-NP .........

Blume’s INTERACTION verbs are distributed across several subclasses in Levin’s (1993) study of English verb classes. Yet the hypothesis embodied in the hierarchy clearly deserves to be explored cross-linguistically, as it constrains possible extensions of specific case patterns across the verbal lexicon in individual languages. Semantic maps also played an important role in the project on the typology of ditransitive constructions. Thus, for the ditransitive domain, the semantic map shown in Fig. 3 is proposed in Malchukov et al. (2010 b). For the present purpose it is important to note that such maps are well suited for the representation of specific argument patterns as well as for argument alternations. Thus, Fig. 3 shows extensions of two different strategies involved in a dative alternation across valency classes. The map is arrived at empirically, through the study of extensions of particular strategies across the verbal lexicon. It also incorporates the hierarchies proposed in the literature, for example the hierarchy of transfer verbs proposed by Croft et al. (2001): give > send > throw. In turn, the pattern underlying the semantic map is guided by semantic considerations, i.e.

10

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

the relative similarities between the verb types in terms of their argument structure. For instance, as claimed in earlier work (Levin & Rappaport Hovav 2005) SEND verbs are intermediate between transfer of possession verbs like GIVE and caused motion verbs like THROW. This ensures that − on iconicity assumptions − the extension of a coding strategy (construction) will occupy a contiguous region of the semantic map. In short, the semantic map methodology captures the patterns in diversity: even though different languages (and even a single language) may exhibit different constructions, the resulting configuration may make claims to universality. Similar maps can be established for other verbal domains; see, for example, Comrie & van den Berg (2006) for a semantic map of experiential verbs in Daghestanian languages. As for the domain of syntactic properties (alternations), it is also expected that syntactic behavior will largely respect the map, since the alternations themselves are sensitive to different functional properties that correlate with lexical properties. The semantic properties are not specifically discussed in Levin’s study, but have been extensively discussed elsewhere (e.g., by Goldberg 1995 and Pinker 1989). A further advantage of the semantic map methodology is that maps from different domains can be integrated as long as they deal with common types. Thus, the map for the transitive domain (in Fig. 2) can eventually be integrated with the map for the ditransitive domain (in Fig. 3) (Malchukov et al. 2010b: 52), and the two together would represent a part of a more comprehensive map for valency classes. Finally, the semantic maps method can be further refined through the application of statistical scaling methods, which are well suited for representing relative frequencies of individual patterns (cf. the semantic maps of motion verbs generated from parallel texts in Wälchli & Cysouw 2012). Within the present volume, the chapters by Malchukov, Haspelmath et al. and Wichmann all make use of semantic maps and hierarchies to represent cross-linguistic generalization, either conventional maps (“implicational maps” in terms of Haspelmath) as illustrated in Fig. 2 and 3 above, or automatically generated maps through the use of clustering techniques (see the NeighborNet graphs in the chapter by Wichmann).

3 Objectives and results 3.1 Objectives 3.1.1 General approach In the project that led to this volume, the participants studied the syntactic valency properties of verbs for a relatively large number of verbs, 70 (for the database, 80), in a relatively small sample of languages, 30 (for the database, 36). For each verb, they looked at two kinds of properties:

Introduction

– –

11

coding properties: flagging (= case and adpositional marking) and indexing (= “agreement/cross-referencing”); behavioral properties: agent demotion, agent (causee) addition, object rearrangement, indefinite object omission, reflexivization, reciprocalization (and possibly others).

Behavioral properties have not figured prominently in earlier discussions of European languages, which are predominantly dependent-marking, but they are indispensable in the discussion of head-marking languages. For example, different subject types in split intransitive languages (or languages with semantic alignment; Donohue & Wichmann 2008) are typically differentiated through indexing rather than flagging. The valency alternations considered are similar to those studied for ditransitive constructions: passivization, reflexivization, reciprocalization, as well as object rearrangement (applicativization), indefinite object omission and causativization (see Malchukov et al. 2010b for a list of other behavioral properties addressed). It should be noted that many alternations are of course limited to a subset of verbs and verb classes. This is also evident in Levin’s work, where many alternations are relevant only to certain valency classes. For example, the Unspecified Object alternation or the Middle alternation are relevant for some subgroups of transitive verbs, but not for intransitive ones. The syntactic alternations studied are independent of the degree of morphological synthesis of a language. As noted above, languages with a richer morphology than English tend to make use of morphological valency-changing operations (applicatives, causatives, anticausatives, etc.) instead of the uncoded alternations that are widely found in English (thus the dative alternation will involve the use of applicatives in Bantu languages; see, for instance, Van Valin 2001: 60–65). Thus in Even, a Tungusic language, the “middle alternation” is signaled by the Mediopassive marker, the “inchoative-causative alternation” by the Causative marker (in competition with the Mediopassive), and equivalents of English verbs allowing for a “reciprocal alternation” commonly involve a lexicalized Sociative marker (e.g., baka-lda [find-SOC] ‘meet’) (Malchukov 1995). For these languages the applicability of the valency-changing markers across the verbs must be established. Things get more complex when several markers compete for a certain domain, as in the case of causative and anticausative markers, which show particular preferences for specific verb types established by Haspelmath (1993a) and Nichols et al. (2004). Another complication concerns languages displaying several different constructions corresponding to one of our broad construction types. For example, Even distinguishes between Mediopassives and Adversative Passives (corresponding to our agent demotion construction), and Philippine languages display multiple passive-like voices targeting different roles (see Comrie 2008 for further discussion of passives). It is important to establish how these valency markers extend across the verbal

12

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

lexicon. In our research we aimed to establish not only availability of particular valency markers for certain verb types, but also their function in the case of polyfunctional markers. For example, the reflexive marker -sja in Russian (and other Slavic languages) is notoriously polysemous, and its meanings are sensitive to the verb type involved (see Geniušienė 1987). Similarly, in many languages the interpretation of the applicative markers depends on the verb type (cf. Gerdts & Kiyosawa 2005 on Salishan applicatives). Taking into account the function of polysemous markers allowed the identification of further clustering in the verbal lexicon. On the other hand, the verbal lexicon of languages of the more isolating type (like English) can be investigated through the study of (morphologically) uncoded valency alternations (as defined above). The study of such languages must be informed by the previous literature, not only in the descriptive tradition, but also in the generative tradition. Some other alternations must be established on a language particular basis. For instance, in many languages the coding patterns will be more informative for lexical typology than they are in English. Thus, Sauerland (1994) shows that German makes finer coding distinctions among verb classes than English, which is due to the availability of verb-specific dative and genitive cases in German. Case and adpositional patterns will be an important diagnostic for lexical splits in the project. For instance, in Tlapanec (an Otomanguean language of Mexico discussed in Wichmann 2009), verbs such as GET and RUN, which encode the single animate participant in the Dative case, contrast with verbs such as KILL and APPEAR, which encode the case-marked participant in the Absolutive (for a transitive verb such as KILL the case-marked participant will be the undergoer and for an intransitive such as APPEAR it will be the actor). Case is a sufficiently frequent phenomenon that it can be used as a consistent diagnostic across many languages, especially on a broad interpretation of case-marking which includes marking by adpositions as well (flagging). There will be other phenomena, however, that are language-specific to a higher degree. To continue with the Tlapanec example, this language has a distinction between agentive and patientive verbs which cross-cuts case distinctions, and by this criterion GET and APPEAR cluster in that they are treated as patientive, while RUN and KILL cluster in that they are agentive (see Wichmann 1996). The systematic study of certain diagnostic constructions across the lexicon can also provide important insights into the properties of the constructions in question. In explaining different syntactic preferences of arguments for certain alternations (e.g., preferential cross-coreferentiality of the recipient with the subject in reciprocal formation), the hypothesis to be pursued is the one advocated in Malchukov et al. (2010b) for the domain of ditransitive constructions, namely, that the preferences are motivated by functional (semantic and/or pragmatic considerations), as acknowledged in different frameworks (see for example Van Valin 1990; Kaufmann 1995; and Wunderlich 2006 on the semantic grounding of unaccusativity diagnostics).

Introduction

13

Thus, there are universal patterns for syntactic characteristics, just as there are universal patterns for coding properties. Different alternations target different regions in the semantic map. General patterns emerge even where languages differ with regard to coding properties as well as with regard to the availability of alternations. Thus, Russian differs radically from English in that it does not allow an uncoded inchoative-causative alternation (= ambitransitive verbs). Instead, the Reflexive-Anticausative marker -sja signals intransitive uses (lomat’ ‘break (tr.)’ vs. lomat’-sja ‘break (intr.)’). Yet the extension of the two constructions across verb classes is comparable in the two languages. Another point of contrast between English and Russian or German (Sauerland 1994) is that alternations involving rearrangement of two arguments remain uncoded in English, while they require prefixation in the other two languages (cf. Russian na-gruzit’ seno na telegu [PREFload hay on cart] ‘to load the hay on the cart’ vs. za-gruzit’ telegu senom [PREFload cart hay.INSTR] ‘to load the cart with hay’; cf. German laden vs. be-laden). This raises an interesting question regarding consequences of structural properties for the syntactic component (cf. Hawkins 1986 concerning a trade-off between morphological case and syntactic versatility). Thus, general patterns of valency classes can be predicted from their characteristics in terms of coding and behavioral properties. As noted above, such predictions hold insofar as the properties themselves are functionally grounded. Yet these functional motivations have their own limitations: A strategy may generalize beyond the core class to adjacent classes (e.g., extending the Allative pattern from caused motion verbs with a goal argument to beneficiaries with other verb types, as in Finnish). There may be other exceptions to the generalization which are rooted in functional, structural, and diachronic factors. As discussed in Malchukov (2005) and Malchukov (this volume), verb polysemy can be responsible for such exceptions. Thus in Lezgian (Haspelmath 1993b), the verb for ‘see’ retains its Dative-Absolutive pattern even when it is used as an active perception verb (‘look’). Another type of exception is due to structural factors. One example from the domain of ditransitive constructions, discussed in Malchukov et al. (2010a), is the use of a double object construction in Malayalam. In this language, ‘give’ takes a dative construction, while “less canonical ditransitives” like ‘entrust’ and ‘feed’ take a double object construction (Asher & Kumari 1997: 205). This goes against the general tendency for ‘give’ to be preferentially encoded in a double object construction (Kittilä 2006). This exception is due to a structural factor: The verbs occurring in the double object construction like ‘entrust’ and ‘feed’ are (lexicalized) causatives, and causatives of transitives in Malayalam regularly occur in the double object construction. Finally, as an example of a diachronic factor, consider the case of subject-experiencer constructions in ‘Standard Average European’ languages, discussed by Haspelmath (2001). Haspelmath suggests a historical explanation for the preference of subject-experiencer constructions in Standard Average European languages, noted by Bossong (1998). He shows that many emotion predicates his-

14

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

torically arise through metaphorical extension from verbs denoting a physical action; thus, worry derives its meaning from ‘strangle; seize by the throat’, stun from ‘deprive of consciousness with a blow’, etc. (Haspelmath 2001: 79). Thus, the verb has changed its meaning but retained the case frame. Diachronic explanations have also been invoked for other cases, for example, to explain idiosyncrasies of split intransitive languages, where the exceptional (minor) pattern of patientive verbs is derived historically from a transimpersonal (= transitive impersonal) construction (Malchukov 2008). At the synchronic level, however, the existence of idiosyncratic case patterns cannot be denied; these are often discussed under the label of ‘lexical case’ in the literature. All these exceptions to the majority pattern which have plagued the research on verb classes make generalizations in lexical typology subject to an important qualification: Implicational universals in terms of verb type hierarchies or semantic maps can only be formulated in existential terms (“for some member of the class X”), not in universal terms (“for every member of the class X”). Exceptions will often be found, due to the interference of other factors. In Malchukov (2005), such interfering factors were taken into account by identifying them as competing functional motivations in the domain of case marking (or competing constraints in terms of Optimality Theory). In other cases, deviant behaviors of certain verbs may be due to their structural make-up as derived (e.g., the Malayalam case mentioned in the previous paragraph) or periphrastic. In the accompanying database structural information is included which allows identification of such cases where this plays a role. To avoid this problem, whenever possible, contributors were asked to choose simplex equivalents, cf. the use of proxies in Johanna Nichols’s work (e.g., Nichols 2008). The questionnaire also includes information on verb polysemy to control for possible polysemy effects. As for the diachronic issues, these are addressed by specialists in particular languages in the chapters devoted to those languages.

3.1.2 Methods and data The general goal of the Valency Classes project was to approach the question of valency classes from a broad typological perspective, identifying general patterns in verb syntax through an empirical examination of a cross-linguistically comparable list of verbs. The procedure follows that of lexical typology as outlined by Nichols (2008), where translational equivalents of a sample of meanings are identified for all the sample languages. In actual practice, it included a lexical questionnaire containing sentences (“typical contexts”) rather than glosses, to provide more contextual support. As noted above, the characteristics of the present approach are:

Introduction

– – –

15

the ontology of valency classes is fairly fine-grained, like Levin’s; the ontology is grounded semantically, rather than syntactically; only in this way can the syntactic properties be studied without circularity; it is limited to verb meanings which are likely to have equivalents across languages.

The list of 70 verb meanings that underlie the chapters in Part II is given in Malchukov et al. (this volume), which should be consulted for discussion of further methodological issues. It can be viewed as a compromise between Levin’s classification and semantically based classifications (in particular those suggested by Dixon 1991 and Lehmann 1991). Verbs corresponding to these meanings were collected for a core sample of 36 languages from all parts of the world, for 30 of these languages leading to the chapters in Part II. The data for this project come from language experts who responded to the questionnaire (see Malchukov et al., this volume), which elicited information on both coding properties and behavioral properties of the verbs (such as occurrence in causative, passive and applicative constructions), as well as language-specific constructions that differentiate between different valency classes.

3.2 Results Other than this volume, the Valency Classes project has led to a single substantial product, the database Valency Patterns Leipzig (Hartmann et al. 2013), in addition to individual articles (Haspelmath 2014; Hartmann et al. 2014; Malchukov, forthc.; Wichmann forthc.), a special journal issue (Wichmann 2014), and an edited volume (Hellan et al. in prep). The present volume and the database should be viewed as two aspects of a single enterprise aiming to document variation in valency classification across languages, which complement and inform each other. Within the present volume the database information is represented most directly (even if partially) in the summary tables accompanying individual chapters which are exported from the database and provide information about coding frames and a selection of alternations for individual verb meanings. The results as presented in this volume are divided into three parts. Part I contains, in addition to this Introduction, other preparatory material essential to an understanding of valency classes and valency alternations cross-linguistically, in particular the Questionnaire and more detailed discussion of the cross-linguistic comparison of verbal valency and an overview of valency classes and alternations, followed by contributions outlining some of the generalizations that emerged (or that failed to be substantiated) in the course of the project. Part II contains 30 chapters dealing with individual languages of the sample. Geographically, these languages cover all six continents that are permanently inhabited by humans, while genealogically, they belong to the following families

16

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

(following the classification of Glottolog , consulted on 2014Sep-30): Tuu, Mande, Atlantic-Congo, Afroasiatic; Indo-European, Nakh-Daghestanian, Tungusic, Yeniseian, Sino-Tibetan, Japonic, Ainu; Mirndi, MoreheadWasur, Austronesian; Eskimo-Aleut, Algic, Salishan, Siouan, Uto-Aztecan, Otomanguean, Mayan, Boran, Araucanian. Finally, part III provides general perspectives on the results of the project in their broader scientific environment from four scholars (Beth Levin, Christian Lehmann, Tasaku Tsunoda and Cliff Goddard) who have made major contributions to our understanding of valency classes and valency alternations. Christian Lehmann presents − in a nutshell − his multilevel conception of argument structure, semantic roles and valency change. Beth Levin, whose work was one of the major sources of inspiration for the present project, reviews some recent developments in her study of valency classes, highlighting the role of the manner verbs vs. result verbs dichotomy for constraining argument alternations. Tasaku Tsunoda, whose work was another major influence, supplies new evidence for the role of the Transitivity Hierarchy (‘Hierarchy of Two-Place predicates’) in explaining cross-linguistic variation in verbal valency. Finally, Cliff Goddard, who along with Anna Wierzbicka is one of the most prominent advocates of the Natural Semantic Metalanguage (NSM) approach (e.g., Goddard & Wierzbicka 2002), demonstrates a potential of the NSMstyle analysis, relying on the radical decomposition of the verb meaning, for the study of argument alternations, and concludes with a plea for closer integration of syntactic and lexical typology in the study of verb classes and valency change. The following are some of the main general conclusions that emerge from the project, and more specifically from this volume. There is a certain functionally motivated comparability in distributionally defined verb classes identified through coding frames and alternations, marked or unmarked. The distinction between coding frames and alternations is not categorical because unmarked alternations may be also interpreted as multiple coding frames. Notably, those verbs which show an alternation in one language may show a pattern split in another language. Thus Japanese (Kageyama et al, this volume), Chatino (Campbell, this volume) and Even (Malchukov & Nedjalkov, this volume) permit only one of the frames allowed for load-verbs in English. Balinese (Shibatani & Artawa, this volume) by contrast, features (verb-marked) object alternations, but applicatives rather indicate a deviation from the basic construction; so constructional preferences are detected here as well. The generalizations in this domain can take the form of implicational hierarchies (Wichmann, this volume; Malchukov, this volume). These hierarchies would be distinct for different types of alternations, as the functions of alternations determine the preferential domain of application (i.e. to which verb classes an alternation preferentially applies) (Wichmann, this volume). Many alternations show a correlation which is motivated by the role of transitivity and the transitivity hierarchy as anticipated by Tsunoda (this volume, and passim) (Wichmann, this vol-

Introduction

17

ume). Verb types and coding frames can be arranged on a transitivity scale or on a transitivity map showing the relative propensity of individual verb meanings for transitive coding (see Haspelmath, this volume). Another general factor which defines relative propensity for subject alternation vs. object alternation is the result vs. manner verb dichotomy as proposed by Levin (this volume; cf. Malchukov, this volume). Our data also supports the frequently made observation regarding a tendency towards complementarity of marked and unmarked alternations: for example, for Japanese unmarked alternations are used more restrictedly, since marked alternations are used instead. Hoocąk (Hartmann, this volume), a radically head-marking language, does not feature unmarked alternations at all. On the other hand, this correlation is not deterministic: in some cases existence of an unmarked alternation does not exclude the use of the marked one; thus in Italian anticausatives and S=P labile verbs are rather in competition, which leads to aspectual differentiation (Cennamo, this volume). The hierarchies of coding (i.e. to which verbs an alternation preferentially applies) and the hierarchies of interpretation (preferential interpretation on the part of the polysemous form or multifunctional alternation) correlate with each other (Malchukov, this volume) Generally speaking, most languages seem to support the insight that syntactic distribution allows one to arrive at a semantic classification, moreover, this classification will be comparable across languages even though it may differ in granularity. In particular those chapters that enumerate verbs classes, such as the chapters on Icelandic, Eastern Armenian, Emai, Chintang and Ainu, are instructive in this respect. However, the structure of the language, namely availability of specific coding frames and alternations, will also determine the granularity of classification (see Malchukov, this volume). For example, Northern Tungusic languages with larger case inventories show more granularity in coding frames and valency classes, as compared to Southern Tungusic languages (Malchukov & Nedjalkov, this volume). Similarly in Xârâcùù (Moyse-Faurie, this volume), valency classes are more differentiated than in other Oceanic languages, due to grammaticalization of numerous oblique prepositions, some of which are specific to a small group of verbs. Similarly, for the domain of alternations the relative size of verb classes depends on availability/type of alternation. For example, in Chintang (Schikowski et al., this volume), which is outstanding for its rampant lability, alternations define only a few larger classes (including the preeminent class of labile verbs) as well as a host of minor verb classes with idiosyncratic behavior. Similar issues for the valency classification arise in Bora (Seifart, this volume), where some (marked) alternations are unproductive, while others apply regularly across the lexicon. Some languages stand out as having a set of verbs with a deviant behavior, with respect to coding or alternations or both. For example, for Jaminjung (Schultze-Berndt, this volume), the distributional method revealed predicate classes corresponding to

18

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

rather language-specific (possibly culture-specific) lexical fields, such as that of manner of application of heat (participating in a causative-inchoative pattern) or direction of gaze (participating in an S=A alternation). While application of distributional criteria (coding frames, alternations) produces a semantically coherent classification in many, if not most languages, in some languages it is less predictable. For example, in Ainu (Bugaeva, this volume), some verbs (particular verb meanings from the Leipzig list) seem to belong to the wrong class. Such inconsistencies are due to interference of structural factors (e.g., a verb may resist applicative or antipassive formation attested for a semantically related verb, if the verb itself is derived through the use of applicative or antipassive morphology). For marked alternations, the expression of the alternation may have repercussions on its applicability. If affixal, alternations may differ in productivity, while if expressed periphrastically they are generally less restrictive (and from that perspective of less interest for the issues of valency classification). At the other extreme there are languages that express valency change through stem alternation like Arabic; such types of alternation are more idiosyncratic, although one may try to achieve a more coherent classification by matching individual stem classes and controlling for the function of a particular alternation (see Kász, this volume, on Arabic, especially Appendix 1). Some other languages seem to be more challenging for valency classification: in particular, this holds for two Malay varieties, which are otherwise quite different in their grammatical make-up: Jakarta Indonesian and Sri Lanka Malay. Jakarta Indonesian (Conners et al., this volume) shows unusual flexibility, verb meanings being “promiscuous” with respect to the coding patterns they take. Sri Lanka Malay (Nordhoff, this volume), on the other hand, does not reveal an obvious argument/ adjunct distinction, so establishing verb classes beyond the transitive/intransitive distinction is problematic. Moreover, according to Nordhoff, this language hardly shows any alternations, whether marked or unmarked. As a result, application of the conventional criteria does not result in an obvious valency classification, so the role of valency as argument specification for a particular verb is also questionable (Nordhoff, this volume). This last conclusion, however, should be qualified in several respects. First, the distinction between argument and adjuncts may indeed be subtle for particular languages, so it will need additional tests (see, for example, the sophisticated diagnostics developed by Bisang for Mandarin Chinese to ascertain the status of a particular noun phrase as an argument or an adjunct; see Lu et al., this volume; cf. Bisang 2006). Second, absence of familiar types of alternations does not necessarily invalidate valency classification. Thus, Emai (Schaefer & Egbokhare, this volume), which generally lacks conventional alternations (either marked or unmarked) but makes use of valency extension (valency augmentation) through oblique arguments or through a serial verb construction, still shows a regimented division into valency classes by these distributional criteria. Third, even for languages which, like Indonesian, are exceptionally flexible in their distributional

Introduction

19

characteristics, frequency criteria may help to differentiate between different valency classes. It is instructive that Conners et al. (this volume) note that while Jakarta Indonesian fails to distinguish between conventional valency classes, it does make a distinction between valency preference classes, which are “sets of lexical items which differ with respect to the ease and frequency with which they enter into constructions associated with various valency patterns”. They note further: “As evident from the accompanying database, the valency preference classes and more frequent coding frames of Jakarta Indonesian fall within the range of variation of valency classes and grammaticalized coding frames observed in most other languages.” Yet, on the whole, it is true that in some languages the notion of valency seems to play a less important role than in other languages. In some languages the distinction between transitive and intransitive verbs or between core and oblique arguments is very clear (cf. Lehmann on Yucatec Maya and Creissels on Mandinka, this volume). Also for Ket, it is noted that verbal valency is fixed (as instantiated by a particular morphological pattern) and cannot be altered through an alternation (Vajda, this volume). In other languages, as illustrated by Jakarta Indonesian and Sri Lanka Malay, valency classes are underdifferentiated. It remains to be seen what is the explanation for these differences, but it is worth noting that both Jakarta Indonesian and Sri Lanka Malay are high-contact varieties, so the absence of valency specification (such as idiosyncratic coding of arguments or general selectivity of verbs with respect of certain patterns) may well be another trait of reduced complexity found in languages exposed to intensive language contact. With respect to alternations, the contributions to this volume have shown that in most cases these alternations can be grouped together into identifiable classes of valency-increasing, valency-reducing, etc. alternations as identified in the typological literature (see Lehmann, this volume; Malchukov, this volume, on the taxonomy of valency alternations). For example, many of the alternations identified by Levin and others for English find their parallels in Nǁng (Ernszt et al., this volume), including varieties of object/oblique alternations with a conative or partitive meaning, as well as analogs of a locative alternation. As a result the valency classes to which they apply are comparable across languages. Yet in some cases, alternations are more idiosyncratic; for example, “generalized voice alternations” (in terms of Gil) in Jakarta Indonesian may change verbal valency but need not do so, even with the same verb. In many cases such voice polyfunctionality has a principled basis: that is, a polyfunctional marker may change interpretation depending on the valency type of verb to which it applies (see Malchukov, this volume, on markedness patterns in voice ambivalence). An instructive example of principled preferences for different patterns of (S=A vs. S=P) lability is found in languages such as CAY Eskimo (Miyaoka, this volume) and Mandinka (Creissels, this volume), and is further discussed in Malchukov (this volume). For ambivalence in marked alternations, an instructive case is Bezhta (Comrie et al., this volume), where “most of the alternations (case-coded and verb-coded) have a semantic ef-

20

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

fect beyond mere valency change”. Generally, application of a morphological operation as such often yields very broad and incoherent verb classes. However, the more one takes into account the syntactic information associated with a particular morphological operation, the more coherent the valency classification becomes (see Malchukov & Nedjalkov, this volume, on Tungusic). The factors conditioning distributional patterns of verbs (coding patterns but also uncoded alternations) are variegated, but most of them have to do with semantic roles (see Lehmann, this volume, for general analysis of features contributing to semantic roles, and the chapter on Yucatec Maya, by Lehmann, for an illustration). Moreover, the manner/result dichotomy, which is held responsible by Levin for availability of object alternations, may be also related to differences in argument structure (insofar as only a result verb may have an incremental theme as its argument). In some languages, animacy is a relevant factor; in Lehmann’s approach animacy (degree of empathy) is also a contributing factor to semantic role identification. Ojibwe (Rhodes & Valentine, this volume), which shows lexical doublets depending on whether S of an intransitive verb or P of a transitive verb is inanimate or animate, is a particularly clear example of this sort. In some languages further factors may be of importance; the role of information structure has been held responsible for the dative alternation in English, and the same has been suggested for Nǁng. Finally, other factors such as verb polysemy or meaning change which may be accompanied by a persistence of the original argument structure may play a role here. Our investigations have confirmed once again the importance of the transitive/ intransitive dichotomy: The vast majority of languages, even those showing underdifferentiation in other domains, display the transitive/intransitive distinction. The privileged status of transitive verbs is investigated in the chapters by Haspelmath (this volume), and Blasi (this volume) on “transitivity prominence”; in particular, the latter chapter provides statistical evidence for the claim that the transitive pattern is the single major pattern among bivalent verbs. The privileged status of transitive verbs may be observed for the domain of alternations, as well. This is brought out especially clearly for some languages like Chintang, which also shows a privileged status of transitive verbs with respect to alternations (all but two of them apply to transitive verbs exclusively). Yet cross-linguistically, this distinction can be undermined on several counts (Malchukov, this volume): at the lexical level for languages showing rampant ambitransitivity (like Chintang); at the syntactic level for languages with inherent complements (like Yorùbá). Although it was not a focus of the volume, the languages examined differ with respect to the status of the transitive vs intransitive classes: some are intransitive based (intransitives as an open class), while others are transitive based. Ainu (Bugaeva, this volume), where most transitive verbs (except for statives) are derived (by causative and applicative affixes), belongs to the class of intransitive-based languages. Chatino (Campbell, this volume) and Mapudungun (Zuñiga, this vol-

Introduction

21

ume) also belong to the transitivizing type. On the other hand, Nen (Evans, this volume), where most intransitives are derived through the use of (detransitivizing) middle morphology, belongs to the second class. Another issue related to transitivity and valency in general is that different constructions appear with different frequency; whenever corpus data was available (see the chapters on Nǁng, Chintang, Jakarta Indonesian) it indicates that the distinctions between valency classes are scalar rather than gradual, and even for labile verbs there is a cline showing different predisposition of different verb meanings for different kinds of lability (see in particular the chapter on Chintang). The gradualness of such distinctions becomes still more evident once one tries to generalize across languages, as captured by alternation hierarchies in Wichmann (this volume). On the other hand, as already mentioned, frequency data can not only blur somewhat the distinction between valency subclasses, but also help to discern a general picture, as in the case of Jakarta Indonesian, which otherwise shows exceptional distributional flexibility, but still allows identification of verb classes once frequency information (about preferred frames) is included. In addition to structural factors accounting for exceptions to the functionally based hierarchies in coding and behavior, there are some deeper rooted discrepancies related to the lexical meaning of individual verbs, that is to differences in how certain events are lexicalized in particular languages. In Lehmann’s approach (Lehmann, this volume; Lehmann et al. 2004), the variation here concerns the mapping between the universal referential level and the lexical/semantic level, rather than a mapping from semantics to morphosyntax. These issues pertaining more to matters of lexical typology have not been systematically investigated within the project, except for the observation that periphrastic expression of verb meaning is a frequent source of idiosyncratic coding patterns. Such cases are briefly discussed in terms of “excorporation” in Malchukov (this volume), and “apotaxis” by Haspelmath (this volume). In some cases such mismatches have a consistent character; Lehmann (this volume and passim; see Lehmann et al. 2004) speaks in this connection of typological strategies. A good illustration of the latter is a predisposition to encode arguments as possessors, characteristic of certain languages such as Yucatec Maya (Lehmann, this volume). Finally, there are obviously idiosyncratic patterns of meaning which are responsible for particular deviations from the expected valency pattern. This aspect is underscored in Goddard’s chapter which argues that a predictive theory should take into account the full analysis of verbal meaning (rather than only putative syntactically relevant parts of it), as well as provide an explicit semantic analysis of different constructions (including marked alternations, like the passive construction), in order to achieve a better understanding of the applicability of certain operations (constructions) to particular verb meanings. This approach, which is close to Construction Grammar approaches, needs to be pursued typologically to spell out its predictions and check these predictions against a cross-linguistic set

22

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

of data. Goddard’s in-depth analysis implies that verbal behavior with respect to clustering is also more complex than suggested in the literature, as there will be constructions sensitive to other components of meaning. The predictions seem plausible in the light of the data assembled in this volume2 but would need a broader follow-up study to corroborate them. In conclusion, we would like to stress that the Leipzig Valency Classes Project is conceived as a first systematic large-scale effort to construct a comprehensive typologically informed typology of valency classes that paves the way for future cross-linguistic studies of verb classes and argument marking. Some issues could only be treated tangentially within this project, and will have to be pursued in follow-up projects. In fact, apart from the Leipzig Valency Classes Project, there are two other current projects concerned with typology of verb classes and argument marking: the Zurich-based project dealing with role clustering and variation in argument structure within the AUTOTYP research program (e.g., Bickel et al. 2014), and the St. Petersburg-based typological project primarily concerned with syntactic clustering of bivalent verbs in the languages of Europe (e.g., Say 2014). The projects differ somewhat in scope and also in certain methodological assumptions, but all share the general goal of making an empirical contribution to establishing a crosslinguistically valid typology of valency classes. Another project which should be mentioned in this context is an ongoing project dealing with argument alternations at the National Institute for Japanese Language and Linguistics (Tokyo) and coordinated by Taro Kageyama, which is mostly focused on Japanese, but also has typological extensions (see, in particular, contributions to Kageyama & Jacobsen (forthcoming), as well as an online typological database of transitivity alternations WATP edited by Prashant Pardeshi (http://watp.ninjal.ac.jp/en/). Clearly, some broader issues brought up in the Leipzig Valency Classes Project can only be resolved through the concerted efforts of a linguistic community involving different subfields of the discipline (typology, semantics, and corpus linguistics, to name a few); yet we hope that the results of the Leipzig Valency Classes Project in the form of the present volume and the ValPaL database will remain a valuable resource informing future studies of this topic.

References Ágel, Vilmos (ed.). 2006. Dependenz und Valenz: ein internationales Handbuch der zeitgenössischen Forschung. Berlin: Mouton. Anderson, Lloyd B. 1986. Evidentials, paths of change, and mental maps: typologically regular asymmetries. In Wallace Chafe & Johanna Nichols (eds.), Evidentiality: The Linguistic Encoding of Epistemology, 273–312. Norwood: Ablex. Apresjan, Jurij D. 1969. Èksperimental’noe issledovanie russkogo glagola. Moskva: Nauka. 2 This view is also compatible with the studies of verbal valency in the Construction Grammar tradition such as Boas (2006), as well as with findings of some corpus research (see Faulhaber

Introduction

23

Asher, Ronald E. & T. C. Kumari. 1997. Malayalam. London: Routledge. Bickel, Balthasar, Taras Zakharko, Lennart Bierkandt & Alena Witzlack-Makarevich. 2014. Semantic role clustering: an empirical assessment of semantic role types in non-default case assignment. Studies in Language 38(3). 485–511. Bisang, Walter. 2006. Widening the perspective: Argumenthood and syntax in Chinese, Japanese and Tagalog. In Daniel Hole, André Meinunger & Werner Abraham (eds.), Datives and Other Cases. Between Argument Structure and Event Structure, 331–381. Amsterdam: Benjamins. Blume, Kerstin. 1998. A contrastive analysis of interaction verbs with dative complements. Linguistics 36(2). 253–280. Boas, Hans C. 2006. A frame-semantic approach to identifying syntactically relevant elements of meaning. In Petra Steiner, Hans C. Boas & Stefan Schierholz (eds.), Contrastive Studies and Valency: Studies in Honor of Hans Ulrich Boas, 119–149. Frankfurt: Peter Lang. Bossong, Georg. 1998. Le marquage différentiel de l’objet dans les langues d’Europe. In Jack Feuillet (ed.), Actance et valence, 259–294. Berlin: Mouton de Gruyter. Busse, Winfried. 2006. Valenzlexika in anderen Sprachen. In Vilmos Ágel (ed.), Dependenz und Valenz: ein internationales Handbuch der zeitgenössischen Forschung, 1424–1435. Berlin: Mouton. Butt, Miriam. 2006. Theories of Case. Cambridge: Cambridge University Press. Comrie, Bernard. 2008. What is a passive? In Zarina Estrada-Fernández, Søren Wichmann, Claudine Chamoreau & Albert Álvarez González (eds.), Studies in Voice and Transitivity, 1–18. Munich: Lincom Europa. Comrie, Bernard & Helma van den Berg. 2006. Experiencer constructions in Daghestanian languages. In Ina Bornkessel, Matthias Schlesewsky, Bernard Comrie & Angela D. Friederici (eds.), Semantic Role Universals and Argument Linking: Theoretical, Typological, and Psycholinguistic Perspectives, 127–154. Berlin: Mouton de Gruyter. Croft, William. 2001. Radical Construction Grammar. Oxford: Oxford University Press. Croft, William. 2012. Verbs: Aspect and Causal Structure. Oxford: Oxford University Press. Croft, William, Jóhanna Barðdal, Willem Hollmann, Maike Nielsen, Violeta Sotirova & Chiaki Taoka. 2001. Discriminating verb meanings: The case of transfer verbs. Handout, LAGB Autumn Meeting, Reading. Cysouw, Michael, Martin Haspelmath & Andrej Malchukov (eds.). 2010. Semantic maps: theory and applications. [special issue] Linguistic Discovery 8(1). De Clerck, Bernard, Timothy Colleman & Dominique Willems (eds.). 2013. Thoughts on Typologies: Different Perspectives on Verb Classifications. [special issue] Linguistics 51(4). Dixon, Robert M. W. 1991. A New Approach to English Grammar, on Semantic Principles. Oxford: Clarendon Press. Dixon, Robert M. W. 2005. A Semantic Approach to English Grammar. Oxford: Oxford University Press. Donohue, Mark & Søren Wichmann (eds.). 2008. The Typology of Semantic Alignment. Oxford: Oxford University Press. Drossard, Werner. 1991. Verbklassen. In Hansjakob Seiler & Waldfried Premper (eds.), Partizipation: das sprachliche Erfassen von Sachverhalten, 150–182. Tübingen: G. Narr. Faulhaber, Susen. 2011. Verb Valency Patterns: A Challenge for Semantics-based Accounts. Berlin: Mouton De Gruyter.

2011). Moreover, typological studies operating with a larger set of verb meanings such as Say (2014) and Bickel et al. (2014) suggest that more caution is required for generalizations in terms of semantic classes, and, further, that certain verb meanings may be more representative for individual classes than some other meanings.

24

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

Fellbaum, Christiane (ed.). 1998. WordNet: an Electronic Lexical Database. Cambridge, Mass.: MIT Press. Fillmore, Charles J., Christopher R. Johnson & Miriam Petruck. 2003. Background to FrameNet. International Journal of Lexicography 16(3). 235–250. Fukui, Naoki, Shigeru Miyagawa & Carol Tenny. 1985. Verb classes in English and Japanese: a case study in the interaction of morphology, syntax and semantics. Lexicon Working Papers, 3. Center for Cognitive Science, MIT. Geniušienė, Emma. 1987. The Typology of Reflexives. Berlin: Mouton de Gruyter. Gerdts, Donna & Kaoru Kiyosawa. 2005. Halkomolem psych applicatives. Studies in Language 29(2). 329–362. Goddard, Cliff & Anna Wierzbicka (eds.). 2002. Meaning and Universal Grammar – Theory and Empirical Findings. Vols I and II. Amsterdam: John Benjamins. Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. Grimshaw, Jane. 1990. Argument Structure. Cambridge, Mass.: MIT Press. Haiman, John (ed.) 1985. Iconicity in Syntax. Amsterdam: Benjamins. Hartmann, Iren, Martin Haspelmath & Bradley Taylor (eds.). 2013. Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://valpal.info. Hartmann, Iren, Martin Haspelmath & Michael Cysouw. 2014. Identifying semantic role clusters and alignment types via microrole coexpression tendencies. Studies in Language 38(3). 463– 484. Haspelmath, Martin. 1993a. More on the typology of inchoative/causative verb alternations. In Bernard Comrie & Maria Polinsky (eds.), Causatives and Transitivity (Studies in Language Companion Series 23), 87–120. Amsterdam: Benjamins. Haspelmath, Martin. 1993b. A Grammar of Lezgian. (Mouton Grammar Library, 9). Berlin: Mouton de Gruyter. Haspelmath, Martin. 1997. Indefinite Pronouns. (Oxford Studies in Typology and Linguistic Theory). Oxford: Clarendon Press. Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages. In Alexandra Y. Aikhenvald, Robert M. W. Dixon & Masayuki Onishi (eds.), Non-Canonical Marking of Subjects and Objects, 53–83. Amsterdam: Benjamins. Haspelmath, Martin. 2003. The geometry of grammatical meaning: semantic maps and crosslinguistic comparison. In Michael Tomasello (ed.), The New Psychology of Language, vol. 2, 211–243. New York: Lawrence Erlbaum Associates Publishers. Haspelmath, Martin. 2014. Arguments and adjuncts as language-particular syntactic categories and as comparative concepts. Linguistic Discovery 12(2). 3–11. Hawkins, John. 1986. A Comparative Typology of English and German: Unifying the Contrasts. Austin: University of Texas Press. Hellan, Lars, Michela Cennamo & Andrej Malchukov (eds.). in prep. Contrastive studies in verbal valency. Herbst, Thomas & Katrin Götz-Votteler (eds.). 2007. Valency: Theoretical, Descriptive and Cognitive Issues. Berlin/New York: de Gruyter. Jones, Douglas (ed.). 1994. Working Papers and Projects on Verb Class Alternations in Bangla, German, English, and Korean. MIT AI Memo 1517. Joppen-Hellwig, Sandra. 2001. Verbklassen und Argumentlinking. Tübingen: Niemeyer. Kageyama, Taro & Wesley M. Jacobsen (eds.). Forthcoming. Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: De Gruyter Mouton. Kaufmann, Ingrid. 1995. O- and D-predicates: a semantic approach to the unaccusative-unergative distinction. Journal of Semantics 12. 377–427.

Introduction

25

Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam: Benjamins. Kipper-Schuler, Karin. 2005. VerbNet: a broad coverage, comprehensive verb lexicon. Ph.D. dissertation, University of Pennsylvania. Kittilä, Seppo. 2006. The anomaly of the verb ‘give’ explained by its high (formal and semantic) transitivity. Linguistics 44(3). 569–612. Kratochvíl, František, Alexander R. Coupe & Randy J. LaPolla (eds.). 2011. Studies in Transitivity: Insights from Language Documentation. [special issue] Studies in Language 35(3). Lazard, Gilbert. 1994. L’Actance. Paris: Presses Universitaires de France. Lazard, Gilbert. 1998. Actancy. Berlin: Mouton de Gruyter. Lehmann, Christian. 1991. Predicate classes and PARTICIPATION. In Hansjakob Seiler & Waldfried Premper (eds.), Partizipation: das sprachliche Erfassen von Sachverhalten, 183–239. Tübingen: G. Narr. Lehmann, Christian, Yong-Min Shin & Elisabeth Verhoeven. 2004. Direkte und indirekte Partizipation. Zur Typologie der sprachlichen Repräsentation konzeptueller Relationen, 2 nd revised edn. Erfurt: Seminar für Sprachwissenschaft der Universität (ASSidUE, 13). Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chicago Press. Levin, Beth & Malka Rappaport Hovav. 2005. Argument Realization. Cambridge: Cambridge University Press. Malchukov, Andrej. 1995. Even (Languages of the World/Materials, vol. 12). München: Lincom Europa. Malchukov, Andrej. 2005. Case pattern splits, verb types, and construction competition. In Mengistu Amberber & Helen de Hoop (eds.), Competition and Variation in Natural Languages: The Case for Case, 73–117. Amsterdam: Elsevier. Malchukov, Andrej. 2008. Split intransitives, experiencer objects and ‘transimpersonal’ constructions: (re-) establishing the connection. In Mark Donohue & Søren Wichmann (eds.), The Typology of Semantic Alignment, 76–101. Oxford: Oxford University Press. Malchukov, Andrej. Forthcoming. “Ambivalent voice”: markedness effects in valency change. In Taro Kageyama & Wesley M. Jacobsen (eds.), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: De Gruyter Mouton. Malchukov, Andrej, Martin Haspelmath & Bernard Comrie (eds.). 2010a. Studies in Ditransitive Constructions: A Comparative Handbook. Berlin: De Gruyter Mouton. Malchukov, Andrej, Martin Haspelmath & Bernard Comrie. 2010b. Ditransitive construction: a typological overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions: A Comparative Handbook, 1–64. Berlin: De Gruyter Mouton. Nichols, Johanna. 2008. Why are stative-active languages rare in Eurasia? A typological perspective on split-subject marking. In Mark Donohue & Søren Wichmann (eds.), The Typology of Semantic Alignment, 121–139. Oxford: Oxford University Press. Nichols, Johanna, David A. Peterson & Jonathan Barnes. 2004. Transitivizing and detransitivizing languages. Linguistic Typology 8(2). 149–211. Pesetsky, David. 1995. Zero Syntax. Cambridge, Massachusetts: The MIT Press. Pinker, Steven. 1989. Learnability and Cognition. The Acquisition of Argument Structure. Cambridge, Massachusetts: The MIT Press. Say, Sergej. 2014. Bivalent verb classes in the languages of Europe: A quantitative typological study. Language Dynamics and Change 4. 116–166. Sauerland, Ulrich. 1994. German diathesis and verb morphology. In Douglas Jones (ed.), Working Papers and Projects on Verb Class Alternations in Bangla, German, English and Korean, 37– 92. MIT AI Memo 1517. Schulte im Walde, Sabine. 2003. Experiments on the automatic induction of German semantic verb classes. PhD Thesis. University of Stuttgart.

26

B. Comrie, I. Hartmann, M. Haspelmath, A. L. Malchukov, and S. Wichmann

Schumacher, Helmut. 2006a. Deutschsprachige Valenzwörterbücher. In Vilmos Ágel (ed.), Dependenz und Valenz: ein internationales Handbuch der zeitgenössischen Forschung, 1396– 1424. Berlin: Mouton. Schumacher, Helmut. 2006b. Kontrastive zweisprachige Valenzwörterbücher. In Vilmos Ágel (ed.), Dependenz und Valenz: ein internationales Handbuch der zeitgenössischen Forschung, 1435– 1447. Berlin: Mouton. Somers, Harold L. 1987. Valency and Case in Computational Linguistics. Edinburgh: Edinburgh University Press. Stiebels, Barbara. 2000. Linker inventories, linking splits and lexical economy. In Barbara Stiebels & Dieter Wunderlich (eds.), Lexicon in Focus, 211–245. Berlin: Akademie Verlag. Tsunoda, Tasaku. 1981. Split case-marking in verb types and tense/aspect/mood. Linguistics 19. 389–438. Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21. 385–396. Van Valin, Robert D. 1990. Semantic parameters of split intransitivity. Language 66(2). 221–260. Van Valin, Robert D. 2001. An Introduction to Syntax. New York: Cambridge University Press. Van Valin, Robert D. & Randy Lapolla. 1997. Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Vogel, Alan R. 2003. Jarawara verb classes. Ph.D. dissertation, University of Pittsburgh. Wälchli, Bernhard & Michael Cysouw. 2012. Lexical typology through similarity semantics: Toward a semantic map of motion verbs. Linguistics 50(3). 671–710. Wichmann, Søren. 1996. The degrammaticalization of agentivity in Tlapanec. In Elisabeth Engberg-Pedersen, Michael Fortescue, Peter Harder, Lars Heltoft & Lisbeth Falster Jakobsen (eds.), Expression and Structure: Studies in Danish Functional Grammar (Studies in Language Companion Series), 343–360. Amsterdam/ Philadelphia: John Benjamins. Wichmann, Søren. 2009. Case relations in a head-marking language: verb-marked cases in Tlapanec. In Andrej Malchukov & Andrew Spencer (eds.), The Oxford Handbook of Case, 797– 807. Oxford: Oxford University Press. Wichmann, Søren (ed.). 2014. Arguments and Adjuncts Cross-Linguistically. [Special issue]. Linguistic Discovery 12(2). Wichmann, Søren. Forthcoming. Quantitative tests of implicational verb hierarchies. In Taro Kageyama & Wesley M. Jacobsen (eds.), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: De Gruyter Mouton. Wunderlich, Dieter. 2006. Towards a structural typology of verb classes. In Dieter Wunderlich (ed.), Advances in the Theory of the Lexicon, 57–166. Berlin: Mouton de Gruyter.

Andrej Malchukov and the Leipzig Valency Classes Project team

2 Leipzig Questionnaire on valency classes Preliminaries The present questionnaire was compiled by Andrej Malchukov with feedback from the other Leipzig Valency Classes Project members (Bernard Comrie, Martin Haspelmath, Iren Hartmann and Søren Wichmann) at an early stage of the project and distributed to contributors to the volume. Since 2010 the conception of the project as reflected in the questionnaire has undergone certain changes, as described in the Database Manual (Haspelmath & Hartmann 2013, available at: http://www.eva.mpg.de/ lingua/valency/files/database_manual.php). The Database Manual can be seen as a follow-up to the questionnaire but is intended for contributions to the database rather than for book chapters. Moreover, it differs somewhat in scope (see in particular the advanced part of the Leipzig Questionnaire, which addresses a broader set of questions not implemented in the database) and also provides the database contributors with a succinct introduction to the framework and terminological conventions used in the project. The present questionnaire does not pursue this latter goal, which is largely fulfilled within the present volume by the chapter by Haspelmath and Hartmann (this volume). Thus, the Leipzig Questionnaire largely follows the original format reflecting its use to guide contributions to the volume rather than database contributions, except for one important update. It includes an updated list of 70 core meanings rather than the original list of 64 verbs. The definitive list was agreed on the basis of the input from the project participants, as well as the feedback from the contributors.

Introduction The present questionnaire deals with a typology of valency classes, or verb types, in terms of Levin (1993). Levin (1993) is a seminal study of syntactic classes of verbs in English, which shows that a semantic classification of verbs can be achieved through applying syntactic diagnostics. Yet, this study, as well as an earlier study by Apresjan (1969) on Russian, has not been followed up cross-linguistically, which leaves open the question of which aspects of these classifications are universal and which are language particular. Similarly, valency dictionaries are few in number and mostly deal with European languages, thus they cannot fill the gap. The questionnaire has been compiled by participants of the DFG funded project on valency classes1 and is designed to obtain a consistent set of data from a representative set 1 See http://email.eva.mpg.de/~haspelmt/ValencyClasses.pdf for the project description.

28

Andrej Malchukov and the Leipzig Valency Classes Project team

Tab. 1: The 70 verb meanings. meaning label

role frame

typical context

RAIN BE DRY BURN SINK ROLL BE A HUNTER BE HUNGRY BE SAD DIE FEEL COLD FEEL PAIN

(it) rains S is dry S burns S sinks A rolls S is a hunter E is hungry E is sad S dies S is cold E feels pain in X

SCREAM LAUGH PLAY LIVE LEAVE GO SING JUMP SIT DOWN SIT RUN CLIMB COUGH BLINK SHAVE DRESS WASH EAT HELP FOLLOW MEET HUG SEARCH FOR THINK

S screams S laughs S plays S lives somewhere (L) A left L S goes somewhere (L) S sings A jumps S sits down (somewhere (L)) S sits somewhere (L) A runs A climbs (up L) S coughs S blinks A shaves (his beard/hair) A dresses P A washes P A eats P A helps X A follows X A meets X A hugs P A searches for X A thinks about X

KNOW LIKE FEAR FRIGHTEN SMELL LOOK AT SEE TALK

A knows P E likes X E fears X A frightens P E smells X A looks at P E sees X A talks (to X) (about Y)

It rained yesterday. The ground is dry. The house is burning. The boat sank. The ball is rolling. This man is a hunter. The baby is hungry. The little girl was sad. The snake died. I’m cold. My arm is hurting. = I’m feeling pain in my arm. The man screamed. The little girl laughed. The child is playing. The old people live in town. The boy left the village. The woman went to the market. The boy sang (a song). The girl jumped. The children sat down on the bench. The children sat on the floor. The horse is running. The men climbed (up) the tree. The old man coughed. I blinked (my eyes). The man shaved his beard/cut his hair The mother dressed her daughter The mother washed the baby. The boy ate the fruit. I helped the boys. The boys followed the girls. The men met the boys. The mother hugged her little boy. The men searched for the women. The girl thought about her grandmother yesterday. The girl knew the boy. The boy liked his new toy. The man feared the bear. The bear frightened the man. The bear smelled the boy. The boy looked at the girl. The man saw the bear. The girl talked to the boy about her dog.

Leipzig Questionnaire on valency classes

29

meaning label

role frame

typical context

ASK FOR SHOUT AT TELL SAY NAME BUILD BREAK KILL BEAT HIT TOUCH CUT

A A A A A A A A A A A A

asks (X) for Y shouts at X tells (X) Y says “...” (to X) name X (a) Y builds P (out of X) breaks P (with I) kills P (with I) beats P (with I) hits P (with I) touches P (with I) cuts P (with I)

TAKE TEAR PEEL HIDE SHOW GIVE SEND CARRY THROW TIE

A A A A A A A A A A

takes P (from X) tears P (from X) peels (X off) P hides T (from X) shows T (to R) gives T to R sends T (to X) carries T (to X) throws T somewhere (L) ties P (to L) (with I)

PUT POUR COVER

A puts T somewhere (L) A pours T somewhere (L) A covers P (with X)

FILL LOAD

A fills P (with X) A loads T (onto L)

The boy asked his parents for money. The woman shouted at the children. The girl told the boy a funny story. They said “no” to me. The parents called the baby Anna. The men built a house of wood. The boy broke the window with a stone. The man killed his enemy with a club. The boy beat the snake with a stick. The boy hit the snake with a stick. The boy touched the snake with a stick. The woman cut the bread with a sharp knife. The man took the money from his friend. The girl tore the page from the book. The boy peeled the bark off the stick. The boy hid the frog from his mother. The girls showed pictures to the teacher. We gave the books to the children. The girl sent flowers to her grandmother. The men carried the boxes to the market. The boy threw the ball into the window. The man tied the horse with a rope to the tree. I put the cup onto the table. The man poured water into the glass. The woman covered the boy with a blanket. The girl filled the glass with water. The farmer loaded hay onto the truck. = The farmer loaded the truck with hay.

of languages to be described in contributions to the edited volume. It starts with the study of a list of 70 verb meanings (Vs, for short) taken as representative of the verbal lexicon, as well as Levin’s taxonomy (but see below, in particular § 6)2. Since Vs might allow for different uses, the meanings are narrowed down through the use of example sentences to be translated into the respective languages (see the reference to ‘typical contexts’ in Table 1 above). The subsequent parts of the questionnaire address coding and syntactic properties of Vs. It begins with questions about coding properties in constructions formed by a V (in particular, case-marking of arguments) in order to determine the basic valency pattern. In

2 This study focuses on lexical verbs rather than auxiliary verbs (with modal, aspectual and other uses).

30

Andrej Malchukov and the Leipzig Valency Classes Project team

accordance with a conventional usage the valency of a V is understood here as the list of its arguments with their coding properties (referred to as coding frames; see Haspelmath & Hartmannn this volume, for definitions of terms). § 3 deals with case alternations, which do not involve voice morphology (cf. different ‘transformations’ used to cross-classify the English verbal lexicon in Levin’s work). § 4 addresses diathetic alternations (or verb-coded alternations), asking about availability of certain valency operations (like passives and applicatives) for Vs, as well as for the meanings expressed by the valency changing markers with Vs, in case they are polysemous. § 3 and § 4 can be seen as largely complementary, as what is a case alternation in one language will be coded as a diathetic alternation in another language (for example, many case alternations in English will be expressed by different valency operations in languages with richer morphology). The last (advanced) part of the questionnaire explores to what extent Vs are representative of lexical classes, i.e., which other verbs belong to the same valency class. This latter part cannot be fully reflected in individual contributions to the volume, which due to size limitations will just offer a summary of verb taxonomies starting from these 70 verb meanings.

I Basic Questionnaire 1 Valency patterns basic examples Please provide glossed examples of sentences containing the relevant Vs (see the prompt typical contexts in the Table 1 above).

The examples sentences exemplifying typical contexts are intended to elicit the verb meanings (Vs) introduced above. The author of a chapter is asked to provide either translational equivalents of the sentences above or other comparable constructions found with the Vs (possibly extracted from corpora). In either case, of special interest is the coding of verbal arguments, for this reason the arguments should be overtly expressed (at least in dependent-marking languages). For headmarking languages, constructions with pronominal (1st/2nd person) subjects and objects should also be considered, as 3rd person arguments are often not indexed on the verb. Alternatively the corresponding constructions with pronominal arguments should be described in § 2.2 under indexing. Examples might be somewhat modified to reflect cultural realities. In cases where a V has different translational equivalents, please choose the verbal lexeme which is more basic (i.e. more frequent and/or morphologically less complex); in the case of several basic items, please include all.

Leipzig Questionnaire on valency classes

31

The following sections will provide further explanation of the glossed examples with respect to coding and behavioral properties of Vs.

2 Coding properties of valency patterns Coding properties involve the following techniques (Haspelmath 2005; Malchukov et al. 2010): – flagging (case or adposition marking) – indexing (agreement, cross-referencing) – word order (in the absence of other kinds of marking)

2.1 Flagging How are the arguments of the verbs flagged (by a case or adposition)?

NB degree of differentiation will naturally depend on the number of cases available. Thus, German makes finer coding distinctions among verb classes than English, which is due to the availability of verb-specific dative and genitive case selection in German (Sauerland 1994). Further questions about flagging, which may be addressed in case they interact in an interesting way with verb classification. Does flagging differ for different kinds of nominals (animate/inanimate, definite/indefinite)?

Some languages show variation in case marking depending on nominal features such as animacy and definiteness; most commonly it has been observed for (direct) objects (cf. differential object marking in languages like Hindi; where P is marked if human, or definite in the case of inanimates). Other arguments may also differ in case marking depending on nominal features; cf. different cases for animate/ inanimate locations, as in Dyirbal. Do free pronouns show the same valency pattern? (Bound pronouns are considered in § 2.2).

This need not be the case, as witnessed by split-ergative languages of the Australian type. What are other relevant factors affecting argument marking here?

In some languages, alignment patterns further depend on TAM-features, as is familiar from split-ergative languages like Hindi and Newari which have ergative

32

Andrej Malchukov and the Leipzig Valency Classes Project team

alignment in the perfective/past and accusative in the imperfective. Such alternations are relevant for the project to the extent they are sensitive to verb classes. Thus, in Newari, availability of an ergative pattern depends on tense, on the one hand, and on lexical class of the verb, on the other hand, so that 2-argument verbs deviating from the transitive prototype can take this pattern optionally.

2.2 Indexing How are the arguments of Vs indexed (by agreement/cross-referencing)?

Some languages (head-marking languages), may distinguish valency patterns through indexing rather than flagging. Thus, Tlapanec has 4 different patterns: ergative vs. absolutive indexing patterns are used for canonical transitives, and the ‘pegative’-dative alternation is used for less canonical transitives (like ‘fear’ and ‘meet’). Note that the discussion of indexing (conjugation) markers that also signal voice distinctions (like the middle voice in Greek) should be postponed to § 4 dealing with diathetic alternations. Further questions about indexing; which need to be addressed to the extent these patterns reveal verb classification. Does indexing depend on the features of the nominal (see above)?

Indexing, like flagging, may depend on nominal features. For example, in many languages with object agreement only prominent (animate/definite) objects are indexed. Again, such cases will be relevant for our project insofar as these features further interact with the verb type. For example, in the Austronesian language Manam some experiencer verbs (‘like’, ‘know’, ‘be bad at’) use object indexing only when the object is prominent, while canonical transitives (like ‘break’) invariably index the object. Thus, here we observe effects of differential object marking (i.e. differential object indexing) for verb-types deviating from canonical transitives.

2.3 Word order What are the word order patterns associated with Vs?

Most often word order depends on syntactic transitivity, but some languages make further distinctions depending on the valency class. Thus, in (Gao) Songhai, canonical transitives (‘break’, ‘kill’), have SOV order, while less canonical transitives (‘see’, ‘follow’, ‘love’), have SVO order. Questions of word order need to be ad-

Leipzig Questionnaire on valency classes

33

dressed to the extent word order interacts in an interesting way with verb classification.

3 Argument alternations In this section only argument alternations (or uncoded case alternations) are considered; discussion of verb-marked diathetic alternation is postponed to the next section. The two sections should be seen as largely complementary, as alternations will be coded by dedicated markers in some languages with a richer morphology and left uncoded in other languages (like English). Do Vs allow for an alternative construction (valency pattern)?

For example, for English, Levin (1993) mentions, in particular, the following alternations. (The list below mentions only fairly productive alternations; and does not include verb-coded diathetic alternations like the passive alternation): a) the inchoative-causative alternation (John broke the stick ~ The stick broke) b) the “middle” alternation (John cut the bread ~ The bread cuts easily) c) the reflexive deletion (John washed himself ~ John washed) d) the reciprocal transformation (John married Mary ~ John and Mary married) e) the dative alternation (Mary gave the book to John ~ gave John the book) f) the locative alternation (John loaded the truck with hay ~ the hay onto the truck) g) the conative alternation (John cut the bread ~ cut at the bread) h) the object deletion alternation (John ate the bread ~ John ate). i) the preposition dropping alternation (John climbed up the hill ~ climbed the hill). Given that Levin’s list includes a number of other more lexically restricted alternations, it is clear that alternation types need to be generalized before they can be applied cross-linguistically. We will distinguish between the following general types of case alternations, which also find equivalents among voice-alternations to be considered in the next section: a) subject-demoting/deleting case alternations This type will include subject-demoting alternations frequently discussed under the heading of differential subject marking (e.g., genitive of negation in Russian, or ergative/oblique alternation related to volitionality in some ergative languages), but also subject-deleting alternations, as in the case of S/P labile verbs (like break; see Levin’s “inchoative-causative alternation”).

34

Andrej Malchukov and the Leipzig Valency Classes Project team

b) object-demoting/deleting case alternations Object-demoting alternations refer, in particular, to varieties of differential object marking (e.g., asymmetric alternations such as the ACC/NOM alternation in Hindi, but also symmetric ones, such as the ACC/PART alternation in Finnish). Object-deleting alternations will refer, in particular, to A/S-labile verbs (like eat; see Levin’s “object deletion alternation”). c) object rearranging case alternations These include, in particular, varieties of dative and locative alternations (cf. (e, f) above), also found in other languages. So the first question to be addressed is: What are the major varieties of case alternations in your language (subject-demoting/deleting; object-demoting/deleting; object rearranging)?

Of course, there may be further varieties of argument alternations, not listed above (for example, object incorporation). These varieties, as well as the alternations listed above, are relevant to the project to the extent they are sensitive to verb classification (e.g., some varieties of differential object-marking apply to any transitive verb, which does not yield an interesting clustering of verb types). More generally for our purposes most relevant are those alternations which are fairly productive (not restricted to a few lexical items), but – most importantly – are sensitive to lexical classes. That is, we are interested in alternations which are distinctive for the verbal lexicon (as sampled here) rather than in those which apply across the board or apply to just few items. After delimiting in this way the set of most relevant case-alternations in your language, the question is to be addressed is: To which Vs in your language do these alternations pertain?

For example, if your language features labile verbs, which of the verbs from the list are labile: S/P labile (cf. the causative-inchoative alternation), and S/A labile (cf. the unspecified object-deleting transformation)?

4 Diathetic alternations and valency changing operations As mentioned above, languages with richer morphology use diathetic alternations for many argument alternations left uncoded in English. Thus in Even, a Tungusic language, the “middle alternation” is signaled by the mediopassive marker, the “inchoative-causative alternation” is signaled by the causative marker (in competi-

Leipzig Questionnaire on valency classes

35

tion with the mediopassive), while equivalents of English verbs allowing for a “reciprocal alternation” commonly involve a lexicalized sociative marker (e.g. bakalda [find-SOC] ‘meet’). It is convenient to use the same (or similar) taxonomy for the domain of verbcoded diathetic alternations, as we adopted for (uncoded) case alternations. Also in this case we will distinguish between the following types: a) Subject demoting/deleting (anticausatives, etc.) b) Subject-Object rearranging (passives) c) Object demoting/deleting (antipassives) d) Object-Object rearranging (applicatives) In addition we address valency increasing alternations: e) Subject addition (causatives) f) Object addition (applicatives) It should be noted that in some cases the distinction between these subvarieties may be problematic (especially between subvarieties of valency rearranging vs. valency increasing applicatives). It is also convenient to treat Subject-Object rearranging operations (passives) together with subject demoting/deleting (anticausatives; reflexives, etc.), as they frequently employ the same markers. From this perspective, the major distinction will be between valency reducing vs. valency increasing diathetic alternations. Obviously, the set of valency/voice markers varies across languages; some of these languages distinguish between several such markers (e.g., anticausatives vs. reflexives), while other languages use the same polyfunctional marker. For such cases it is important to state both availability and the meaning of particular markers for certain Vs.

4.1 Valency-reducing operations Valency reducing operations come in several subtypes, as illustrated below: a) Subject demoting/deleting voice alternations – anticausative (cf. a) in § 3 above) – middle (cf. b) in § 3 above) – reflexive (cf. c) in § 3 above) – reciprocal (cf. d) in § 3 above) b) Object demoting/deleting voice alternations – antipassive (cf. g), h) in § 3 above) c) Subject-Object rearranging voice alternations – passive (differs from anticausative in that A may be expressed, or is implied)

36

Andrej Malchukov and the Leipzig Valency Classes Project team

Does your language have the voice alternations listed above? If not, how are these functions expressed (this information should be given in § 3 if the functions are expressed through argument alternations). If so, what Vs do these markers apply to?

What are the functions of these markers when applied to different Vs?

For example, in Russian the “reflexive” suffix -sja can be used in a reflexive function with some verbs (myt’-sja ‘wash’), while with other verbs it has anticausative (slomat’-sja ‘break’), passive (stroit’-sja ‘be built’), or reciprocal (vstrečat’-sja ‘meet’) function.

4.2 Valency-increasing and valency-rearranging operations The most important valency-increasing operations are causatives and applicatives; the latter however may rearrange rather than increase the valency. In some languages, the same marker is used both in causative and applicative functions; also for such cases it is important to determine which Vs select for which function. If your language has causative markers, which Vs can they apply to? Does the meaning of the causative marker differ with the V involved?

Note that some languages have several causative markers, for example, for building intransitive vs. transitive causatives. These can be used to test for transitivity of less prototypical transitive verbs. If your language has applicative markers, which Vs can they apply to? Does it have several applicative markers used with different Vs? Does the meaning of the applicative marker differ with the V involved?

There may be several subtypes of applicatives, depending on which object is promoted (for example, in Hoocąk (Siouan), there are 4 different applicative markers, including the benefactive applicative, the instrumental applicative and two types of locative applicatives). On the other hand, the general applicative in Salish has been claimed to have different meanings depending on the verb’s class. Applicatives may be used to render many of the alternations listed in § 3, including the dative (cf. e) in § 3 above), locative (cf. f)), but also preposition dropping (cf. i)). Other languages may use directional markers to code some of these alternations (cf. Russian na-gruzit’ seno na telegu [PREF-load hay on cart] ‘to load the hay on the cart’ vs. za-gruzit’ telegu senom [PREF-load cart hay.INSTR] ‘to load the cart with hay’; German laden vs. be-laden).

Leipzig Questionnaire on valency classes

37

4.3 Other valency/voice categories Of course, it is impossible to foresee (let alone, list here) all language-particular voice categories. While choosing to address certain voice alternations, one should again be reminded that voice constructions will be relevant to the project to the extent that they interact in an interesting way with the verb lexicon (in particular, are neither restricted to few lexical items, nor apply across the board to all verbs). Does your language have other voice categories? Which Vs do they apply to?

For example, some (Austronesian) languages show a variety of “voice” (or “focus”) forms (“actor focus”, “goal focus”, etc), used for ‘promotion’ of different objects to the subject position; for these languages it will be relevant which Vs allow for which voice constructions. On the other hand, head-marking languages of the “hierarchical type” show a direct-inverse alternation triggered by the relative prominence of the A and P arguments. In that case it is relevant to study the use of direct-inverse alternations with different groups of two and three argument verbs (in the latter case, it is also relevant which of the object arguments takes part in the alternation; e.g., Theme or Recipient of a ditransitive verb). But also for the domain of monotransitives some languages may show further differentiation; e.g., some languages (like Tlapanec) may have different inverse forms for different subtypes of 2-argument verbs.

II Advanced Questionnaire 5 Further properties of individual verbs 5.1 Morphological issues: complexity Indicate which of the Vs are morphologically complex?

In § 1 the contributors were prompted to select for the basic (nonderived) equivalent of verbs on the list. In some cases, however, this is impossible, as in the case where all ditransitives including GIVE are derived (e.g. applicative, as in Tzotzil). Therefore it is important to provide information about morphological complexity of Vs. This question is relevant insofar as morphological make-up may determine availability of a certain valency pattern. For example, in Malayalam, only derived ditransitives (causatives of transitives), take a double object construction, while basic (underived) ditransitives take a dative construction.

38

Andrej Malchukov and the Leipzig Valency Classes Project team

5.2 Semantic issues: polysemy Do Vs have other meanings?

This question is relevant insofar as the valency pattern may be motivated through one of the meanings of the polyfunctional item. For example, in some languages, which use the same verbs for both ‘hit’ and ‘throw’, this verb follows the allative pattern, as expected for caused motion verbs.

5.3 Semantic issues: etymology Do you know etymology of Vs?

The motivation for this question is the same as for the previous one: a verb may inherit the valency pattern from its original meaning.

6 Further properties of lexical classes 6.1 Lexical issues: open and closed valency classes Which other verbs belong to the same valency pattern as individual Vs? Is it an open or a closed class? For an open class, please specify which verbs belong to this class (in terms of relevant semantic or formal features). For a closed class, please list other verbs in this class.

For example, if some of the Vs are labile (see § 3 above), please give the list of other (S/P and S/A) labile verbs. Do other verbs, semantically similar to a V, participate in the same alternations as this V?

Do other verbs, semantically similar to a V, show the same diathetic alternations as this V?

References Apresjan, Jurij D. 1969. Èksperemental’noe issledovanie russkogo glagola. Moskva: Nauka. Haspelmath, Martin. 2005. Ditransitive Constructions: The Verb ‘Give’. In Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures, 426–429. Oxford: Oxford University Press. Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chicago Press.

Leipzig Questionnaire on valency classes

Malchukov, Andrej, Martin Haspelmath & Bernard Comrie. 2010. Ditransitive constructions: A typological overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions, 1–65. Berlin: Mouton de Gruyter. Sauerland, Ulrich. 1994. German diathesis and verb morphology. In Douglas A. Jones (ed.), Working Papers and Projects on Verb Class Alternations in Bangla, German, English, and Korean, 37–92. MIT AI Memo 1517.

39

Martin Haspelmath and Iren Hartmann

3 Comparing verbal valency across languages 1 Introduction That different verbs may have different valencies even when they are semantically similar has long been well known (at least since Tesnière 1959), and it is the reason why dictionaries contain (or should contain) valency information for each verb. For example, we need to specify that English wait takes a for-complement, while expect takes a direct object. That “the same” verbs across languages also often differ in valency is not quite so well known, though typological-comparative and contrastive works in linguistics have often discussed valency mismatches of the type illustrated in (1)–(2). (1) a. English I1 miss you2. b. French Vous2 me1 manquez. ‘I miss you.’ (Literally something like ‘You are missing to me.’) (Tesnière 1959: § 123.2) (2) a. English She1 filled the container2 with water3 . b. Chintang (Tibeto-Burman, Nepal) gagri-be2 phatt-e. Huĩsa-ŋa1 cuwa-O̸3 she-erg water-nom container-loc fill-pst ‘She filled the container with water.’ (Literally ‘She filled the water into the container.’) (Bickel et al. 2010: 387) Tesnière called such cross-linguistic mismatches “metataxis” (Tesnière 1959; cf. Koch 1994), and it is such differences between languages that we are primarily interested in here. But what does it mean for the valencies of two verbs to fail to match? How can we compare valencies across languages? These are not trivial questions, but answers to them are a prerequisite for any comparative or contrastive research on valency. This chapter will explain how we approached this problem, and what decisions we took for the Valency Patterns Leipzig (ValPaL) database (Hartmann et al. 2013). The basic principle is that we decided to match verbs across languages on the basis of verb meaning, and to match arguments across languages on the basis of

42

Martin Haspelmath and Iren Hartmann

individual argument meaning, i.e. argument microroles (see § 5, and also Hartmann et al. 2014).

2 Valency: coding frames and role frames The valency of a verb is the range of syntactic properties of other elements of the clause that depend on the particular choice of verb, i.e. that are verb-specific. These other elements of the clause are called arguments. The most salient argument properties are the coding properties, i.e. flagging (case or adpositional marking) and indexing (i.e. bound person marking associated with the verb).1 Examples of argument-coding elements are Nominative and Accusative case and the preposition of in (3) from English, and Dative and Nominative case as well as 3 rd person singular Subject indexing in (4) from German. (3) English [They]nom accused [her]acc [of plagiarism]. (4) German [Den Kindern] gefällt [der Schneemann]. the.pl.dat child.pl.dat please.3sg the.sg.nom snowman.sg.nom ‘The children like the snowman.’ Everything else in the clause is independent of the verb: A clause can contain locational and temporal setting adverbials (e.g. 5), or it can contain manner adverbials or illocutionary adverbials (e.g. 6), regardless of the kind of verb that is chosen. Such verb-independent elements are called adjuncts. (5) Last week they accused her of plagiarism in her school. (6) In all frankness, they accused her with verve. A clause can contain subordinators or diverse particles, it can exhibit special word order, and it can occur in different tenses (all illustrated by (7) from German), independently of its verb. (7) weil der Schneemann dem Kind doch gefiel because the snowman.sg.nom the child.sg.dat after.all please.pst.3sg ‘since the child liked the snowman after all’

1 See Haspelmath (2013) for the term “person indexing” and its relation to traditional terms like “agreement” and “bound pronoun”.

Comparing verbal valency across languages

43

Thus, there are many syntactic properties of clauses that do not depend on the verb, but those that do, viz. the verb’s valency properties, are generally regarded as core features of syntax. Comparing valency properties across languages is thus a core concern of comparative syntactic research. We can represent the valency of the English verb accuse and the German verb gefallen as in (8a–b). These representations are called coding frames in this work. (8)

two exemplary coding frames a. English accuse: b. German gefallen:

Here, Arg1, Arg2 and Arg3 are variables for distinct arguments that are primarily expressed by nominals; -nom, -acc and -dat indicate the case features of the arguments, of+ stands for a preposition that flags an argument, and V.subj[ ] refers to the person information on the verb that matches the person feature of one of the arguments, i.e. the argument is indexed on the verb (in English, accuse indexes, or “agrees with”, its Nominative argument, and German gefallen also indexes its Nominative argument). For a complete description of a verb’s behaviour, its coding frame needs to be linked to a (possibly partial) representation of the verb’s meaning in such a way that the argument variables are paired with variables for its semantic participants. We call such a semantic representation a role frame. The role frames of accuse and gefallen are shown in (9)–(10) with the linking to the valency frames: (9)

English accuse a. role frame

‘X accuses Y of Z.’ (the accuser) (the accusee) (the wrongdoing) | | | b. coding frame

(10) German gefallen a. role frame ‘X likes Y.’ (the liker) (the likee) | | b. coding frame To a significant extent, a verb’s coding frame can be predicted from its role frame (cf. Levin & Rappaport Hovav 2005). For example, in English a participant with an agent role (e.g. accuser, breaker, thrower) is always linked to the Nominative argument in the coding frame. But not infrequently across languages, there are quite a few unpredictable linkings which simply have to be learned by speakers and listed in dictionaries. It is these unpredictable, or less predictable, coding frames that are

44

Martin Haspelmath and Iren Hartmann

of greatest interest for the comparative study of valency, because they also tend to show the greatest cross-linguistic variability. In addition to coding properties, verbs may also determine other properties of their arguments, such as the ability to be the antecedent of a reflexive pronoun, to be the target of omission under coreference, or the ability to be passivized. These properties are generally called behavioural properties. Arguments that are coded in the same way may nevertheless show different behaviours. For example, German wiegen ‘have a weight of (a measurement)’ takes an Accusative argument that cannot be passivized (cf. 11a–b), while wiegen ‘determine the weight of (an object)’ takes an Accusative argument that can be passivized (cf. 12a–b). (11) a. Der Sand wiegt einen Zentner. the sand weighs one hundredweight ‘The sand has the weight of one hundredweight.’ (passive) b. *Ein Zentner wurde von dem Sand gewogen. one hundredweight was by the sand weighed (12) a. Frau Müller wiegt den Sand. Ms. Müller weighs the sand ‘Ms. Müller determines the weight of the sand.’ (passive) b. Der Sand wird von Frau Müller gewogen. the sand is by Ms. Müller weighed ‘The weight of the sand is determined by Ms. Müller.’ Thus, one can say that the two homonymous verbs wiegen have two different valency frames, perhaps notated as and , respectively. In other words, we might use syntactic-function labels like subject, object and extent in valency frames, thus taking into account not only coding properties of arguments, but also behavioural properties. However, in our comparative study of valency, we have not been able to take these other properties into account systematically, so we generally use the label coding frame rather than valency frame. (Word order is intermediate between coding properties and behavioural properties; see § 11 below on word order.)

3 Notes on terminology Before we go on to compare languages, we offer a few notes comparing linguistic traditions, especially with respect to terminology. The term “valency” was intro-

Comparing verbal valency across languages

45

duced by Tesnière (1959) and has been used in English at least since Svartvik (1966), but other terms have been widely used as well.2 The terms in (13) all have basically the same meaning. A terminological distinction between the more abstract “valency” and the more concrete “valency frame” is sometimes made, but is often redundant. (13)

a. b. c. d. e.

complementation subcategorization argument structure government model clause blueprint

e.g. Quirk et al. (1985: 1069–71) Chomsky (1965) e.g. Goldberg (1995) Mel’čuk (1974) (Russian model’ upravlenija) Grebe (1959) (German Satzbauplan)

Those nominals in a clause that are determined by (i.e. depend on the choice of) the verb are now generally called “arguments”, while the verb-independent elements are called “adjuncts”. Again, this distinction is widely recognized, but in the past the terminology varied quite a bit: (14)

a. b. c. d.

argument complement actant argument

adjunct modifier, adjunct circumstant satellite

(here) e.g. Vater (1978) Tesnière (1959) Dik (1997: 86–90)

What we call the “role frame”, i.e. the semantic representation of the verb meaning with the participant variables, has been variously called “logical structure” (e.g. Van Valin & LaPolla 1997), “lexical-conceptual structure” (e.g. Jackendoff 1990), or “semantic valency”. Throughout this chapter, we will work with the distinction between descriptive categories of particular languages and the comparative concepts of cross-linguistic research (Haspelmath 2010). In the practice of linguists, verbal valency is primarily used as a descriptive concept, needed to characterize the behaviour of particular verbs in particular languages. Thus, we need to develop a number of comparative concepts that allow us to compare valency patterns across languages. The most important concepts are the semantic concepts comparison meaning (meanings that are used to compare verbs across languages, e.g. ‘miss’ and ‘fill’ in (1)–(2)) and microrole (meanings that are used to compare arguments with similar roles across languages, e.g. ‘liker’ and ‘likee’ in (10)), and concepts for formal coding elements, in particular flags (cases and adpositions) and indexes (see § 6 below). We have also tried to work with a comparative concept of argument, but as we will see in the next section, this is more difficult to apply consistently across languages than the comparison meanings, the microroles, and the coding elements.

2 Note also that the alternative form “valence” is also common, especially in American English (e.g. Abraham 1978; Langacker 1988).

46

Martin Haspelmath and Iren Hartmann

4 Delimiting valency: arguments and adjuncts A lot of research on valency has been concerned with the practical task of compiling valency dictionaries, i.e. dictionaries of verbs that give a valency frame for every verb. This forces researchers to delimit the notion of valency very clearly, and to distinguish sharply between arguments (included in the valency frame of the verb) and adjuncts. But sometimes it is not quite clear whether a nominal is specific to the verb or not. Consider the bracketed phrases in (15). (15) a. b. c. d. e.

She put the book [on the table]. He lives [in Wisconsin]. She is sitting [on the sofa]. They cut the meat [with the knife]. He broke the window [with his fist].

These are locational and instrumental phrases which could be taken either as arguments or as setting or manner adverbials and thus as adjuncts. Is there a way to tell whether they are arguments or adjuncts? One way in which the distinction between arguments and adjuncts has often been framed is by considering the verb’s meaning. If a participant is entailed by the meaning of a verb, this participant is an argument, otherwise it is an adjunct (e.g. Van Valin 2005; Bickel 2011). According to this criterion, with the knife in (15d) would be an argument, because cut means ‘sever with a sharp instrument’, while with his fist in (15e) would not be an argument, because break only means ‘do something so that something becomes broken’. However, not all entailed participants can be overtly expressed, not all entailed participants would be considered arguments, and not all elements that are generally considered arguments are entailed participants. First, not all entailed participants can be overtly expressed. For example, the English verb tell allows the expression of both the addressee and the content, but lie does not allow the expression of the content: (16) a. She told her address to her friend. b. She lied (*her address) to her friend. Second, not all entailed participants would normally be considered arguments. For example, the verb break means ‘do something1 so that something2 becomes broken’. The first entailed participant, the action that causes the breaking event, can be expressed in a by-phrase (He broke the window by hitting it with his fist), but such a by-phrase would not normally be regarded as an argument. More generally, all verbs that denote a spatiotemporal event entail a time and a location, but these are the most typical adjuncts, not arguments.

Comparing verbal valency across languages

47

Third, some arguments are not entailed by the verb’s meaning, e.g. the beneficiary me in (17). (17) Roland baked me a cake. As has been widely discussed, bake does not entail a beneficiary (one can bake a cake without having someone specific in mind), but the beneficiary is generally regarded as an argument. Thus, because there is no close match between entailed participants and what are normally considered arguments, we do not adopt the entailment-based definition of argument. We think that the notion that best captures the intuition that lies behind the argument-adjunct distinction is the notion of verb-specificity. Elements that are verb-specific are arguments, and elements that are not verb-specific are adjuncts. Perhaps the clearest case of verb-specificity is coding-specificity, i.e. where a verb determines idiosyncratic coding of its arguments, so that the coding is not predictable on the basis of the verb’s meaning. This can best be seen in minimal pairs of verbs with very similar or identical meaning but different argument coding. Thus, the semantic role of the boy or der Junge is not very different in the (a) and (b) examples in (18)–(19), but only the (b) examples require a dative preposition or case. This is something that needs to be learned in addition to the meaning of the verb, and it is thus no accident that systematic valency research began in the context of language teaching (Helbig & Schenkel 1969).3 (18) English a. I showed the boy the solution. b. I demonstrated the solution to the boy. (19) German a. Ich unterstützte de-n Junge-n. I.nom support.pst.1sg the-acc boy-acc ‘I supported the boy.’ b. Ich half de-m Junge-n. I.nom helped the-dat boy-dat ‘I helped the boy.’ Another example of this type was given in the very first paragraph of this paper (English wait for someone vs. expect someone).

3 There was of course valency research before Helbig & Schenkel (1969), but they published the first valency dictionary, which was perhaps also the first book specifically on valency patterns in a language (incidentally, this work was carried out at Leipzig University and published in Leipzig).

48

Martin Haspelmath and Iren Hartmann

Another fairly clear case of verb-specificity is obligatoriness. In many languages, some kinds of nominals must occur when a given verb is chosen. This is true not only of subjects (which are fairly trivial cases of arguments and therefore sometimes even ignored by valency researchers), but also of many objects, and sometimes even of oblique arguments. For example, the object in (20a) cannot be omitted (contrasting with 20b), and the prepositional auf-argument in (21a) cannot be omitted (contrasting with 21b). (20) English a. The dragon devoured the princess. (*The dragon devoured.) b. The customer ate (the fish). (21) German a. Sie hat mich auf den Fehler hingewiesen. she has me on the mistake pointed ‘She pointed me to the mistake.’ a′. (*Sie hat mich hingewiesen.) b. Sie hat (auf Godot) gewartet. she has on Godot waited ‘She waited for Godot.’ But there are many arguments that exhibit neither coding specificity nor obligatoriness. In fact, in many languages no argument is ever obligatory, because all arguments can be omitted when they can be reconstructed from the context.4 And in many cases, the absence of an argument may be odd for pragmatic reasons, not for any grammatical reasons: (22) a. ??He lives. b. ?She is sitting. Hearing about someone living or sitting is not informative, so the location is usually specified to get a pragmatically felicitous utterance. Thus, obligatoriness is not as useful for delimiting valency as is often thought, not even for those languages where it does play a role.

4 However, often one can make a clear distinction between an anaphoric and an existential interpretation of argument omission, and when argument absence implies an anaphoric interpretation, this could be taken as evidence of verb-specificity and argumenthood. Thus, in She found out the omitted argument has to be definite and anaphoric (= She found out about it), whereas in She ate the omitted argument has to be indefinite and non-anaphoric (= She ate something).

Comparing verbal valency across languages

49

The crucial criterion for argumenthood is whether a nominal is limited in its cooccurrence options to a restricted and semantically arbitrary set of verbs (i.e. whether it is verb-specific), or whether it can occur with any verb, or at least with a large and semantically coherent class of verbs (i.e. whether it is verb-free). We can call this specificity of occurrence. Thus, while the boldfaced arguments in (23) are not coded in a (highly) verb-specific way and are not obligatory, they are still considered arguments because they cannot occur freely with any verb, as shown in (24). (23) a. He called his brother. b. She gave the shopkeeper too much money. c. He ran to the house. (24) a. *She laughed her brother. b. *He spent the shopkeeper too much money. c. *She sang to the house. One widely cited test for (non-)argumenthood that reflects specificity of occurrence is the ‘happen’ test: If a sentence has a paraphrase in which a phrase from the original sentence is removed and occurs in an anaphoric ‘happen’ (or ‘do so’) clause, then this phrase is an adjunct: (25) a. She called her brother in the morning. b. She called her brother, and this happened in the morning. (26) a. They accused her of plagiarism in her school. b. They accused her of plagiarism, and this happened in her school. (OR: and they did so in her school) This test shows clearly that temporal and locational setting adverbials are adjuncts, while obligatory and coding-specific nominals are arguments:5 (27) a. She called her brother. b. *She called, and this happened her brother. (28) a. They were waiting for Godot. b. *They were waiting, and this happened for Godot.6

5 In the ValPaL manual, we asked the contributors to determine argumenthood by means of a test of this kind. But several participants told us that no such test is available in their language (e.g. Seifart 2013 on Bora). Thus, this test is not readily available as a cross-linguistically applicable way of determining arguments (see also Haspelmath 2014). 6 This sentence is acceptable with an irrelevant sense (‘this happened for the sake of Godot’).

50

Martin Haspelmath and Iren Hartmann

The test can be applied to many languages, because anaphoric verbal expressions like ‘do so’ or ‘this happened’ are widely found. However, with (stationary)7 locational phrases and instrument nominals, the result of the test is not so clear: (29) a. She was sitting on the sofa. b. ?She was sitting, and this happened on the sofa. (30) a. They cut the meat with the knife. b. ?They cut the meat, and this happened with the knife. Quite generally, locational phrases and instrument nominals are hard to classify uniquely as arguments or adjuncts. The concept of valency and the argument/adjunct distinction is simply not very useful for these kinds of phrases.8 In addition to coding specificity and obligatoriness, quite a few other languageparticular criteria for distinguishing between arguments and adjuncts have been cited for different languages in the literature (cf. Haspelmath 2014). However, since we are interested in cross-linguistic comparison, language-particular argument criteria are not of interest to us. Thus, for quite a few cases we did not have a unique way of distinguishing between arguments and adjuncts, and the ValPaL database is therefore not consistent in this regard. If the verb ‘break’ is said to have three arguments in one language and two arguments in another language, this does not mean that there is a typological difference between the two languages. This may just be due to an arbitrary or at least idiosyncratic difference between the decisions taken by the authors. As a result, the number of arguments is not a kind of information that should be taken as important for cross-linguistic comparison, especially when the difference concerns locational phrases and instrument nominals.

5 Comparison meanings: verbs and microroles For cross-linguistic comparison of valencies, or more specifically coding frames, we need to be able to compare verbs (the valency-bearers), participant roles, and coding elements (flags and indexes, see § 6 below). For example, a comparison as in (2a–b), repeated here from above, presupposes that we know that both English

7 By contrast, directional locational phrases are clearly arguments: He ran to the house / *He ran, and this happened to the house. 8 Note also that the ‘happen’ test cannot be applied to subjects in English; and the criterion of verb-specificity might be taken to indicate that English subjects are not arguments, because they occur with all verbs. (However, nobody has suggested that subjects are not arguments, in English or other languages, so this has no practical consequences.)

Comparing verbal valency across languages

51

fill and Chintang phatt- mean ‘fill’, that the verbs in addition to an agent role have a substance role (in this example, the water) and a location role (in this example, the container), and that the substance role is coded by a preposition (with) in English, and with Nominative case in Chintang. (31) a. English She1 filled the container2 with water3 . b. Chintang (Tibeto-Burman, Nepal) gagri-be2 phatt-e. Huĩsa-ŋa1 cuwa-O̸3 she-erg water-nom container-loc fill-pst ‘She filled the container with water.’ (Literally ‘She filled the water into the container.’) (Bickel et al. 2010: 387) Thus, our ValPaL database contains entries such as those in (32a–b). The first line contains the verb form (its citation form), the second line contains the coding frame, and the third line contains the microroles, i.e. the participant roles relevant for this particular verb. (32) a. English fill

(1=filler, 2=filled container, 3=filling material) b. Chintang phatt (1=filler, 3= filled container, 2= filling material) In other words, the verb entries in our database contain the kinds of information that we saw above in (9)–(10), though in a slightly abbreviated notation, with microrole labels instead of full role frames with index numbers (thus, “1=filler, 2= location, 3=substance” is equivalent to “Arg1 fills Arg2 with Arg3”). The 80 pre-defined comparison meanings that we used to gather comparable verb forms for the database are identified by their English counterpart (which is also their label) and by a typical context. For example, the meaning ‘cover’ is identified by the label COVER as well as the typical context “The woman covered the boy with a blanket.”. We did not try to provide a more detailed semantic description, as we did not expect this to lead to greater comparability of verbs across languages.9

9 We are aware that the use of English labels to elicit counterpart verbs, as well the use of English as a general metalanguage of our project, may well have biased the set of comparison meanings toward the kinds of meanings that tend to have simple expression in English and related languages. Unfortunately, we felt that there was nothing we could do about this, other than avoiding verb

52

Martin Haspelmath and Iren Hartmann

Languages differ not only in their forms, but also in the kinds of meanings that they have words for, so the verbs that are given as counterparts to our 80 comparison meanings are not always perfect matches. For this reason, we do not call them “equivalents”, but “counterparts”. We asked the contributors to find the semantically closest verb in their language. Moreover, we said that the verb should have a “basic” flavour, i.e. verbs that are used very rarely should be avoided if a more common verb with similar meaning is available. Sometimes a basic verb may have a somewhat different meaning from the English label that serves as comparison meaning; in such cases, we said that deviation from the comparison meaning was less important than basicness. The reason for this was that we felt that there was no need to insist on exact matches, because this was not achievable anyway in many situations. And it must be kept in mind that the purpose of our database is not to provide exact translations from each language to each other language, but to compare languages with respect to their valency patterns.10 The relationship between the pre-defined comparison meanings and counterpart verbs can be many to many, so that if there were two basic verbs in the language that corresponded to a given meaning, both could be included. For example, in Sliammon, there are two verbs corresponding to EAT, ʔiɬtən and məkʷt, with two different coding frames (Watanabe 2013). Conversely, when a single verb corresponded to two different meanings, it was not necessary to enter it twice into the database. For example, the Mandinka verb bori is a counterpart both of RUN and of ROLL (Creissels 2013). Homonymous verbs are distinguished by a number, e.g. Nǁng nǁaa (1) (‘dress’) and nǁaa (2) (‘live’).11 The argument variables in the coding frames are represented by integers (1, 2, 3, …) whose order does not have any significance (though normally the linearly first argument gets variable 1, and so on). Each argument variable is linked to a microrole, so that we know what role the argument plays in the verb meaning. With typical transitive verbs, as in (33), the microroles are not so important, because they could normally be replaced by “agent” and “patient”, and with singleargument verbs, as in (34), they are completely redundant.

meanings for which we were aware from the beginning that there is something peculiar about their English counterparts. 10 Every translation distorts the original text to some slight extent, but translations are still eminently useful. We do not expect our cross-linguistic comparison to be any better (or worse) than the average professional translation. 11 Since a verb must have a unique coding frame in our database (this was a decision we took to simplify the database), if a verb had two slightly different meanings but different coding frames, it had to be broken up into two different verb entries that are treated as homonymous. For example, Yaqui chaae has the coding frame when it renders ‘scream’, but when it means ‘shout at’ (Estrada-Fernández et al. 2013). ‘Scream’ and ‘shout’ would be sufficiently similar to count as one meaning, and no doubt most linguists would say that we are dealing with the same verb here, but since there are two possible coding frames, two different ValPaL database entries are required.

Comparing verbal valency across languages

(33) a. EAT b. WASH c. HELP

eater washer helper

(34) a. COUGH b. SINK

cougher sunken entity

53

eaten food washed entity helpee

(Microroles sometimes have unusual-sounding labels such as helpee; we hope that readers and database users will quickly get used to them. We find transparent labels much more practical than opaque abbreviations or numbers.) The microroles are important for those verb meanings where cross-linguistic metataxis is not uncommon. We saw some examples in (1)–(2/31) above, and here is another contrasting pair of verbs that are counterparts of the same meaning from two languages: (35) Sri Lanka Malay a. verb: b. coding frame:

(Nordhoff 2013) mintha- ‘ask for’

(1=asker, 2=requested thing, 3=askee)

c. example: see baapa=ka car=yang su-mintha 1sg father=loc car=acc pst-beg ‘I asked my father for the car.’ (Lit. ‘I asked the car at my father.’) (36) Jaminjung a. verb: b. coding frame:

(Schultze-Berndt 2013) yanggi ganarrany ‘ask for’

(1=asker, 2=askee, 3=requested thing)

c. example: gurrany yanggi ya-wun-karra=yinyag mangarra-wu neg ask irr-du.A:1.P-put=1du.excl.P plant.food-dat ‘Don’t ask us two for food, you two!’ (Lit. ‘Don’t ask us to food.’) The microroles ‘askee’ and ‘requested thing’ do not clearly map onto the general roles ‘patient’, ‘source’ and/or ‘recipient’, but in our approach this does not matter: The coding frames in (35b) and (36b) are perfectly clear and comparable on the basis of the microroles. Intuitively, the Jaminjung sentence is more like English (using a dative case or the requested thing), while the Sri Lanka Malay sentence is literally ‘I asked the car at my father’, which is rather different from English, and more like Russian poprosit’ (ja poprosila mašinu u otca [I asked the.car at father]). So there is a metataxis relationship between (35) and (36).

54

Martin Haspelmath and Iren Hartmann

6 Coding elements We saw above (§ 2) that valency has both a coding and a behavioural component, but here we limit ourselves to coding frames. And while behavioural properties of arguments are quite heterogeneous, there are just two kinds of coding elements, which are quite straightforward to represent schematically: flags (adpositions or case markers) on arguments, and indexes on the verb that match arguments. Let us consider three additional examples, which illustrate the conventions that we use: (37) Japanese (standard) a. verb: kowasu ‘break’ b. coding frame:

(Kishimoto & Kageyama 2013)

c. example: Ken-ga (hanmaa-de) mado-o kowasi-ta. Ken-nom hammer-ins window-acc break-past ‘Ken broke the window (with a hammer).’ (37) Hoocąk a. verb: b. coding frame:

(Hartmann 2013) hožu ‘load’

c. example: wiiraruti=ra kšee hožu trailer=def apple put.in ‘I loaded apples on the trailer.’ (39) Icelandic a. verb: b. coding frame:

(Barðdal 2013) stela ‘steal’

c. example: Þjófur-inn stal peningum frá gömlu konu-nni. thief-the.nom stole money.dat from old.dat woman-the.dat ‘The thief stole money from the old lady.’ The following conventions are used to represent coding elements (flags and indexes) in coding frames in ValPaL: (i) Cases are represented by (abbreviations of) their category labels, such as nom (for nominative), dat (for dative), following the argument variable and linked to it by a hyphen.

Comparing verbal valency across languages

55

(ii) Adpositions are represented by their form, linked to the argument variable by a plus (+) sign, preceding it (for prepositions) or following it (for postpositions). (iii) Index-sets are represented by their category labels, linked to the verb by a period (e.g. V.agr in Icelandic). The index label is immediately followed by brackets which contain the argument variable number, so “V.agr[1]” in Icelandic means that the agreement index-set corresponds to the argument that also bears nominative case (the stealer in the case of the verb stela ‘steal’), and “und[2].act[1].V” in Hoocąk means that arguments 1 and 2 are indexed by the Actor and Undergoer sets, respectively (the loader and the loaded theme in the case of hožu ‘load’). (iv) Optional arguments may be enclosed in brackets. (We did not enforce this, because optionality is a difficult concept, and we were not able to provide this information consistently.)

7 Kinds of arguments In the most common case, verbal arguments are nominals (noun phrases, adpositional phrases) or person indexes on verbs. Such arguments are represented in our coding frames by free-standing integers (possibly accompanied by adpositions or with case labels attached to them) or by integers in brackets following the indexset label which is attached to the verb variable. But there are certain other kinds of arguments, in particular locational arguments, clausal arguments, and utterance arguments. A locational argument need not have any particular categorial form. It is most often expressed by an adpositional phrase, but the nature of the adposition is determined by the spatial meaning, not by the verbal valency. In many languages, locational adverbs may also be expressed by spatial adverbs which are neither adpositional phrases nor noun phrases. For example, a locational argument occurs with Italian sedere ‘sit’, illustrated in (40a–c). (40) Italian a. Mario siede in seconda fila. Mario sits in second row ‘Mario is sitting in the second row.’

(Cennamo & Fabrizio 2013)

b. Gli anzian-i sede-va-no intorno a-l fuoco. the old.man-pl sit-ipfv-3pl around to-the fire ‘The old men were sitting around the fire.’ c. Mario sied-e lì. Mario sits there ‘Mario is sitting there.’

56

Martin Haspelmath and Iren Hartmann

In ValPaL’s coding frames, locational arguments are notated with LOC, followed by the argument number. Thus, the coding frame of Italian sedere is . (A locational argument also occurs in the coding frame of Hoocąk hožu ‘load’, see (38b) above.) Clausal arguments are complement-clause arguments such as that-clauses and infinitival clauses in English (e.g. I hope that you’ll be there; I hope to be there). These were excluded from ValPaL, as the typology of complement clauses is an entirely separate domain of study. The only comparison meanings that frequently require a clausal argument are ‘know’ and ‘want’, as illustrated by (41b). However, we asked our contributors for contexts like (41a), where the ‘wanted thing’ argument is a nominal rather than a clause. Thus, Eastern Armenian uzel ‘want’ has the coding frame in ValPaL. (41) Eastern Armenian (Khurshudian & Daniel 2013) a. Jes uz-um em ajs aʁǯka-n. I.nom want-cvb.ipfv aux.1sg this girl.dat-def ‘I want this girl.’ b. Jes uz-um em gn-al. I.nom want-cvb.ipfv aux.1sg go-inf ‘I want to go.’ There is another comparison meaning that takes a special kind of argument: ‘say’ usually takes what we call an utterance argument (and marginally ‘tell’ and ‘ask’ can have such an argument as well). Like locational and clausal arguments, utterance arguments cannot be treated like nominal arguments because they are not coded by the usual flags (cases, adpositions) or indexes. Most commonly, they are simply juxtaposed (as in 42a–b), but occasionally there are special quotative markers, as in Japanese (see 42c). (42) a. Jakarta Indonesian Nadia omong kita mendingan pergi. Nadia say 1pl better go ‘Nadia said it would be best if we go.’ (or: ‘Nadia said: “We better go”.’) b. Ket Qarʲ da’ŋa bara: “eddi”. dem.m 3sg.m.dat 3.m.sg.say alive-1sg ‘He said to him: “I’m alive”.’ c. Japanese (standard) Ken-ga Mari-ni “hai” to it-ta. Ken-nom Mari-dat yes quot say-pst ‘Ken said “yes” to Mari.’

(Conners & Gil 2013)

(Vajda & Kryukova 2013)

(Kishimoto & Kageyama 2013)

Comparing verbal valency across languages

57

Utterance arguments are notated with UTT in ValPaL, followed by the argument number (thus, Jakarta Indonesian omong has the coding frame ).

8 Prominence-based splits Coding frames contain a unique coding element, i.e. a unique case or adposition associated with a free-standing argument number, or a unique index-set on the verb that indexes an argument. But sometimes languages exhibit prominencebased splits, i.e. depending on the argument’s inherent or contextual prominence features, different flags are employed. A well-known case of this is differential object marking in Spanish, where human direct objects have the preposition a (e.g. veo a Juan ‘I see Juan’), while nonhuman objects lack this preposition (e.g. veo la casa ‘I see the house’). In such cases, one would not say that the verb ver ‘see’ is associated with two different coding patterns, because the coding contrast does not depend on the verb, but on the nature of the argument. But an even better-known case is English, where a Nominative-Accusative distinction is made only with personal pronouns (I/me, he/him, she/her, we/us, they/ them, who/whom), i.e. with nominals that are higher in prominence. All other nominals make no distinction. When a distinction is only partial and depends on the kind of nominal, then the usual solution adopted by linguists is to encode the distinction in the syntactic rule and to make a separate (morphological) statement about the lack of the distinction in certain kinds of nominals. Thus, in Russian we would say that there is a general Nominative-Accusative distinction, even though it is syncretized in many inanimate nouns in the singular (e.g. nom/acc kniga/knigu ‘book’, but stol/stol ‘table’, mesto/mesto ‘place’), and in all inanimate nouns in the plural (e.g. nom/acc knigi/knigi ‘books’). Likewise, we decided to say that English has a general Nominative-Accusative distinction, even though it is syncretized in all nominals that are not personal pronouns (and in the personal pronouns you and it). Thus, the English verb see has the coding frame V.subj[1] > 2-acc> in ValPaL. Another case of a prominence-based split is Eastern Armenian, where a direct object is consistently in the Dative case when it is human, but in the Nominative case when it is inanimate (Khurshudian & Daniel 2015). In the glosses in the Eastern Armenian examples, the case labels dat and nom are used, e.g. (43) Eastern Armenian (Khurshudian & Daniel 2013) a. Maɾd-ə tesa-v tʁa-ji-n. man[nom]-def see.aor-aor.3sg boy-dat-def ‘The man saw the boy.’ b. Jes cʰamakʰ em tesn-um. I.nom land[nom] aux.1sg see-cvb ‘I see the land.’

58

Martin Haspelmath and Iren Hartmann

However, since the Dative and the Nominative occur in complementary distribution with direct objects, they are treated as a single case (“Dative-Nominative”) for the purposes of valency information, and the coding frame of tesnel ‘see’ in the database is .

9 Verb forms Following our guidelines, the verbs in our database have to be conventional expressions in the language, but they do not have to be basic or simple verbs, and they do not even have to be “verbs” in a very strict sense. The requirement that the entries are conventional (or fixed) expressions means that we did not want the contributors to provide ad-hoc paraphrases. For example, in Bezhta the verbal meaning ‘peel’ is rendered by ‘take (off)’ plus ‘skin’ (Comrie & Khalilova 2013). (44) Kibba k’atu-wa-s beš b-aɣo-yo. girl.erg potato-obl-gen1 skin(iii).abs iii-take-pst ‘The girl peeled the potato.’ (Lit. ‘The girl took (off) the skin of the potato.’) Another example is the meaning ‘sing’, which is rendered by doŋkilóo láa [song say] in Mandinka. This is a fully regular combination of a verb (‘say’) and its object (‘song’). Other objects would be equally possible (e.g. ‘tell a story’). Thus, the database has “no counterpart” for SING in Mandinka and PEEL in Bezhta. Since it is very difficult to distinguish between fixed expressions and freely created expressions, our database may not be homogeneous in this regard. For example, for ‘rain’, we have bi-ga o-da [rain-nom come-decl] in Korean, but in Bezhta, the very similar wodo guu-s [rain come-prs] was not counted as a conventional expression. This may not correspond to a real difference in the fixedness of the two expressions. The counterparts of our comparison meanings are not always really verbs in the languages, but we wanted to avoid the issue of the verb-adjective distinction and thus basically ignored the language-particular status of the counterparts. We were exclusively interested in the valency properties of expressions corresponding to meanings such as ‘be afraid’, ‘like’, ‘feel cold’, ‘be sad’, ‘be dry’, not in their word-class assignment. It is apparently the case that all counterparts of dynamic comparison meanings are actually verbs in all languages, but counterparts of meanings such as ‘be afraid’ and ‘be sad’ (and especially ‘be dry’) are very often called “adjectives” in language descriptions. In such cases, the entries often include copulas, but these copulas should be regarded as irrelevant to our underly-

Comparing verbal valency across languages

59

ing question of valency (we did not try to be consistent in including or excluding copulas).12 While fully regular ad-hoc paraphrases are not included, as just noted, complex verbal expressions are allowed in the database, and were in fact used in many cases (more than a quarter). They may be quite diverse formally: They may consist of a verb plus an additional locative particle (e.g. English sit down), or a verb plus a reflexive pronoun (e.g. German sich hinsetzen ‘sit down’), or a verb plus an incorporated noun (e.g. Bezhta komak bowal ‘help’, lit. ‘do help’), or two verbs (e.g. Yoruba mú ... wá ‘bring’, lit. ‘take ... come’). They may also include valency-changing affixes, especially causative affixes, e.g. Mandinka niki-ndi ‘teach’, which derives from nikiŋ ‘learn’. It is important to be aware that by including complex verbal expressions in the database, we are on the one hand maximizing the amount of data that we can include for each language, but on the other hand we are introducing a possible bias. The comparison meanings that we started out with almost all have simple counterparts in English (and closely related European languages), but not uncommonly, they do not have simple counterparts in other languages. Thus, the English verb set (Goddard 2013) consists of simple verbs, while the verb sets of other languages often include derived verbs such as causatives or other complex verbal expressions. Since complex verbal expressions may have different valency profiles than simple verbs, we may not get an unbiased picture of the general valency character of the language. We chose to include complex verbal expressions in order to maximize the amount of data in the database, and because it is not clear that complex verbal expressions are less fundamental to a language’s character than simple, monomorphemic verbs. The question of how best to compare verbal syntax across languages in a systematic way is still quite open, in our view, so we did not want to limit the available data in advance. We did try to annotate all verbs in the database for simple vs. complex status,13 so that users can filter out complex verbal expressions if they want. With complex verbal expressions that include a nominal element, the question may arise what the relation of this nominal element is to the verb’s valency. Consider the examples in (45)–(47).

12 One of our comparison meanings was ‘be a hunter’. Including this was motivated by the question whether the subject of the predicate ‘is a hunter’ is coded in the same way as the subject of other stative predicates such as ‘be sad’ or ‘be afraid’. We recognize that including such a meaning in a database on verbs is somewhat confusing, and as a result some contributors did not include a counterpart. 13 More precisely, a complex verb in the database is a verbal expression that includes an element that is relevant for the verb’s valency, such as a causative marker, an applicative morpheme or an incorporated noun.

60

Martin Haspelmath and Iren Hartmann

(45) Bezhta komak bowal [help do] ‘help’ Kibba iyo-l komak b-oo-yo. girl.erg mother-lat help(iii) iii-do-pst ‘The girl helped her mother.’ (Lit. ‘The girl did help to the mother.’) (Comrie & Khalilova 2013) (46) Yucatec Maya ch’a’ sahakil [take fear] ‘be afraid’ Le wíinik-o’ t-u ch'a'-ah sahak-il ti’ le báalam-o’. dem man-d2 pfv-sbj.3 take-cmpl fear loc dem jaguar-d2 ‘The main became afraid of the jaguar.’ (Lehmann 2013) (47) Icelandic finna lykt [find smell] ‘smell’ Björn-inn fann lykt-ina af strák-num. bear-the.nom found smell-the.acc of boy-the.dat ‘The bear smelled the boy.’ (Lit. ‘The bear found the smell from the boy.’) (Barðdal 2013) In these cases, one might at first want to say that we are dealing with transitive verbs, and that the incorporated nouns (‘help’, ‘fear’, ‘smell’) are objects of these verbs. This would not be incorrect, and indeed at some level this needs to be said, because otherwise the case-marking (ergative marking of the helper in (45), accusative marking of the incorporated noun in (47)) and the agreement (gender III agreement of the verb with ‘help’ in (45)) could not be explained. Certainly the Bezhta ‘do’ verb, the Yucatec Maya ‘take’ verb and the Icelandic ‘find’ verb are transitive, and this explains some facets of the complex verbal expressions. But from the comparative perspective, it is these complex verbal expressions as a whole which are the counterparts of the comparison meanings, and thus their valencies are as in (48a–c). (48) a. Bezhta komak bowal [help do] ‘help’

(1=helper, 2=helpee) b. Yucatec Maya ch’a’ sahakil ‘be afraid’

(1=fearer, 2=fear stimulus) c. Icelandic finna lykt ‘smell’

(1=smeller, 2=smelled entity)

Comparing verbal valency across languages

61

These “composed valencies” may sometimes look a bit odd, especially for Bezhta: A verb with an ergative but no absolutive argument is not normally possible, and most verbs normally agree with one argument in gender. In purely language-specific descriptions, composed valencies are often left aside, but strictly speaking, they are necessary for language descriptions, too. That the action of helping is standardly expressed by ‘do help’ in Bezhta is not fully predictable (alternatively, one could say ‘give help’, ‘extend help’, etc., or there could be a completely unrelated simple verb), and that the smelled entity is coded by the preposition af in Icelandic is not fully predictable (cf. also English take part, where it is not predictable that the activity is coded with the preposition in: take part in something). Thus, the general principle of the ValPaL database is that only arguments that can be varied freely are taken into account as variables in the coding frames. Elements which are fixed parts of complex verbal expressions are ignored. In the coding frame of Bezhta komak bowal, the variable V stands for the complex verbal expression. The internal structure of counterpart verbs (or verbal expressions) is not reflected in the database.

10 Argument mismatches: apotaxis In the simplest case, which is very frequent in the ValPaL database, the arguments of the counterpart verbs have roles that correspond to the microroles that we anticipated, and our anticipated microroles have corresponding arguments. But this is not always the case. What we anticipated may admittedly have been based on English and similar European languages. But languages may differ not only in the way in which arguments expressing the same roles are coded (this is metataxis, cf. § 1 above). They may also differ in the roles that can be expressed as arguments (this can be called “apotaxis”). Let us illustrate this with the verb meaning ‘dig’, which involves the creation of a hole (i) in the ground (ii) by a digger (iii), possibly with the goal of digging up a thing (iv) that was hidden in the ground. (In addition, digging of course requires an instrument, minimally a body part, but since the coding of instruments exhibits little variability, this is left aside here.) In German, the digger and the ground can be expressed simultaneously with the hole or with the hidden thing: (49) German a. Sie gruben ein tiefes Loch in die Erde. they dug a deep hole in the.acc earth ‘They dug a deep hole in the earth.’ b. Sie gruben in der Erde nach dem Schatz. they dug in the.dat earth after the treasure ‘They dug for the treasure in the earth.’

62

Martin Haspelmath and Iren Hartmann

But the hole and the hidden thing cannot be expressed simultaneously (*Sie gruben ein Loch nach dem Schatz ‘They dug a hole for the treasure’). In other languages in our database, our contributors gave us verbs that take primarily the ground as the argument of digging (e.g. Bezhta in 50, Hoocąk in 51), or primarily the hidden thing (e.g. Bora in 52). (50) Bezhta Öždi mäče b-äx-čä. boy.erg ground(iii).abs iii-dig-prs ‘The boy digs the ground.’ (51) Hoocąk Mąą=ra ha-k’e. earth=def 1e.a-dig ‘I’m digging the ground.’ (52) Bora Mújcuri ó tsehdí. barbasco.root(acc) 1sg dig ‘I dig for barbasco root.’

(Comrie & Khalilova 2013)

(Hartmann 2013)

(Seifart 2013)

Another manipulation verb meaning that shows variability is ‘peel’. In Russian, both the peel and the peeled object can be expressed as arguments (as seen in 53), while in German, only the peeled object can be an argument (at least with the verb schälen, 54). In Yucatec Maya, finally, only the peel is an argument (as seen in 55). (53) Russian (Malchukov & Jahraus 2013) Povarënok očisti-l kartošk-u ot kožur-y. kitchen.boy.nom peel.pfv-pst.m.sg potatoes-acc from skin-gen ‘The kitchen boy peeled the potatoes.’ (Lit. ‘The kitchen boy cleaned the potatoes of their skin.’) (54) German Die Großmutter schälte die Kartoffeln. the grandmother.nom peeled the potatoes.acc ‘Grandmother peeled the potatoes.’ (55) Yucatec Maya (Lehmann 2013) Le xibpal-o’ t-u lak-ah u sóol le che’-o’. dem boy-d2 pfv-sbj.3 detach-cmpl poss.3 shell dem wood-d2 ‘The boy removed the bark from the stick.’ (Lit. ‘The boy removed the stick’s shell.’)

Comparing verbal valency across languages

63

These differences between languages must be kept in mind when interpreting the cross-linguistic data of the ValPaL database. While most verbs have arguments that correspond largely to arguments in other languages, there are sometimes apotactic mismatches of the type just seen. Another example is the comparison meaning ‘shave’. A verb with this meaning can take the shaved person as its non-agent argument (e.g. in Chatino, in 56), or the body part (e.g. in Ainu, in 57), or the hair/the beard (e.g. in Balinese, in 58), or it can be intransitive, because the affected entity is understood from the context (e.g. in Xârâcùù in 59, and also in English). (56) Zenzontepec Chatino Yū=wá nka=téēʔ=yu j=yū. 3sg.m=dem cpl.caus-get.shaved=3sg.m obl=3sg.m ‘He shaved (lit. himself).’ (57) Ainu E=nan-uhu memke. 2sg.a=face-poss shave ‘Shave (lit. your face).’ (58) Balinese Anak=e muani ento nguris jenggot=ne. person=def male that shave beard=3.poss ‘The man shaved (lit. his beard).’ (59) Xârâcùù Nâ xii. he shave ‘He shaved.’

(Campbell 2013)

(Bugaeva 2013)

(Shibatani & Artawa 2013)

(Moyse-Faurie 2013)

Similar cases of apotaxis are also found with other verbs involving the human body, such as ‘hurt’, where either the affected person can be an argument of the verb (as in English I feel pain), or the affected body part (as in English My head is hurting). In general, apotactic variability in valency patterns can be attributed to metonymic shift (e.g. Waltereit 1998), but there is a lot of interesting cross-linguistic variation here that deserves further study (see also Michaelis & APiCS Consortium 2013 on diverse ways of expressing having a headache in pidgin and creole languages).

64

Martin Haspelmath and Iren Hartmann

11 Constituent order Unlike argument flagging and argument indexing, constituent order is not a coding element that is generally available. The primary way in which constituent order may serve to code semantic roles in some languages is by requiring agents to precede the verb and patients to follow the verb, as in typical SVO languages like English or Yoruba. Since constituent order is salient in some languages, we allowed the contributors to provide this information, but since it is often difficult to say whether a particular order is required for coding a relation, we made this an optional feature. In (60) we give a few examples of coding frames that include constituent order information. (60) English a. be afraid ‘fear’ Emai b. dia ‘live’ Mandarin Chinese c. zhuāng ‘load’ Mandinka d. ma ‘touch’

V.subj[1] > of+2> V > vbi+2> (LOC2 >) V > 3> 2 > V > 3+la>

Mandarin Chinese and Mandinka are atypical in that they have fairly rigid constituent order, but have some arguments that must occur in preverbal position, while others must occur in postverbal position: (61) Mandarin Chinese Gōngrén-men wǎng kǎchē-shàng zhuāng gàncǎo. worker-pl prep truck-loc load hay ‘The workers loaded hay onto the truck.’ (62) Mandinka Kambaan-óo ye saá maa fál-oo la. boy-def pfv.pos snake.def touch stick-def obl ‘The boy touched the snake with a stick.’

(Zhang Guohua 2013)

(Creissels 2013)

While constituent order is represented in the coding frames of these languages in ValPaL, order is not regarded as a coding element.

12 Alternations In many languages, verbs may systematically be associated with different valencies under different circumstances. Such situations are called valency alternations.

Comparing verbal valency across languages

65

Three well-known examples from English are the Passive alternation, the Dative alternation and the Causal-noncausal alternation (also known as “ambitransitive” or “labile” verbs). (63) Passive (V → be V’-en)



a. The bear caught a fish. b. A fish was caught by the bear. (64) Dative alternation (uncoded) →

a. She gave the money to her brother. b. She gave her brother the money. (65) Causal-noncausal alternation (uncoded)14



a. He opened the door. b. The door opened. Like valency frames, valency alternations generally do not affect all verbs equally and thus subclassify the verbal lexicon in a language. Alternations often introduce many additional complexities and thus could not be covered exhaustively in ValPaL. The general guideline was that not more than ten alternations had to be entered if the language had more. Also, alternations which more or less apply across the board (as in some languages causatives do) did not have to be included either, as they would not be very useful in distinguishing verb classes. Alternations are sometimes subdivided into voices (like passive, middle and antipassive) and valency-changing operations (like causative, applicative, desubjective), but this distinction cannot be made consistently across languages. It is sometimes correlated with inflectional vs. derivational status, but this distinction likewise cannot be made consistently across languages. Another frequently made association is between voice and information-structural function, but other alternations may also have informational-structural effects (e.g. Peterson 2007: Chapter 4). Thus, we do not distinguish between voices and other valency-changing alternations. Alternations can be grouped into coded alternations, where the verb undergoes a change (as in the English Passive alternation), and uncoded alternations, where the form of the verb is the same with the two alternate valency frames (as in 65 for English). In addition, alternations may preserve the semantic roles, or

14 This alternation has often been called “inchoative-causative“ alternation. See Haspelmath et al. (2014) for the more general and more transparent term “causal-noncausal”.

66

Martin Haspelmath and Iren Hartmann

they may change the available semantic roles (by removing a role as in (65), or by adding a role as in (66) below), as long as the same basic meaning of the verb is preserved. When an alternation was unproductive and occurred only with a small number of verbs, we were less interested in the alternation. (When the valency-frame variation affects only a single verb, it is not really an alternation, because alternations require some systematicity.) In (63)–(65) above, the alternations are presented as directed, with a basic alternant on the left-hand side of the arrow and a derived alternant on the righthand side. In coded alternations, it is generally easy to distinguish the basic from the derived alternant, because in most cases only the derived alternant involves some overt valency-changing morphology. Two more standard cases of this kind from other languages are given in (66)–(67). Mapudungun has an applicative suffix -l, and Chintang forms reciprocal verbs by reduplication and an affix -ka-. (66) Mapudungun (Zúñiga & Lienlaf 2013) → a. Ti pichi malen ye-tu-y kiñe rayen colegio mew. art little girl carry-tel-ind one flower school loc ‘The girl carried a flower to school.’ b. Ti pichi malen ye-l-fi kiñe rayen ñi kimeltuchefe. art little girl carry-appl-3.obj one flower 3.poss teacher ‘The girl carried a flower to the teacher.’ (67) Chintang (Schikowski et al. 2013) → a. Cha-ŋa puchak lauri-ŋa ten-o-s-e. child-erg snake stick-erg hit-[3sA.]3[s]P-prf-ind.pst ‘The child has hit the snake with a stick.’ b. Teı̃-ka-teı̃ lus-i-nɨŋ-kha. beat-recp-beat aux-1p[S]-neg-bgr ‘Let’s not beat each other!’ But when the alternation is uncoded, as in the Dative alternation and the Causalnoncausal alternation in English, it is often difficult or impossible to determine which of the alternants is basic. But given the setup of our database, where alternations are associated with basic verbs rather than abstract verb roots, this decision must be taken, be it in some arbitrary way.15 15 The alternative to this would have been a database with a list of verb roots (or stems), plus a list of coding frames that can be associated with the verb roots in a many-to-many fashion. Such a database would not have contained a data type “alternation” at all. We considered this alternative,

Comparing verbal valency across languages

67

Another issue that sometimes arises with alternations is that the same alternation occurs twice in our database, but in different directions. For example, Mandinka has a Causative alternation, which turns intransitive verbs into transitive verbs, as in (68). (68) Mandinka a. Saatéw-o janí-ta. village-def burn-pfv.pos ‘The village burned / was burned.’

(Creissels 2013)

b. Jáw-oo-lu yé saatéw-o jani. enemy-def-pl pfv.pos village-def burn ‘The enemies burned the village.’ It also has a Noncausative alternation, which turns transitive verbs into intransitive verbs, as in (69). (69) Mandinka (Creissels 2013) a. Kambaan-óo ye palantéer-oo teyi ber-óo la. boy-def pfv.pos window-def break stone-def obl ‘The boy broke the window with the stone.’ b. Kíl-oo teyí-ta. egg-def break-pfv.pos ‘The egg broke.’ These two alternations are of course really the same alternation in Mandinka. The reason they are treated as two alternations for the purposes of the database is that the relevant comparison meanings happened to be ‘burn (intr.)’ and ‘break (tr.)’ and alternations had to be unidirectional for the sake of the database

13 Conclusion Let us summarize briefly our discussion and our choices for the Valency Patterns Leipzig database (Hartmann et al. 2013). As in other areas of morphosyntax, comparison of verbal valency requires semantic comparative concepts as a basis of comparison, in particular verbal comparison meanings and microroles (§ 5). Distinguishing consistently between arguments and adjuncts across languages is diffi-

but decided that it would have been more difficult to extract the information that interested us from such a database.

68

Martin Haspelmath and Iren Hartmann

cult, but it is not really necessary to capture the most important aspects of valency variation, because the intermediate cases (especially locational and instrumental arguments/adjuncts) show relatively little variation across languages (§ 4). To be comparable across languages, coding frames must primarily contain information about argument flags (cases and adpositions associated directly with the argument nominals) and about argument indexes (person forms mostly associated with the verb) (§ 6). It is primarily nominal arguments that are of interest for valency comparison (§ 7), and prominence-based splits are not encoded in the valency frame (§ 8). Counterpart verb forms need not be verbs in the strict sense, but can be adjectives and complex verbal expressions, but they must be fixed conventionalized expressions. Incorporated nouns that are part of such complex expressions do not count for the valency of the verbal expression (§ 9). Cases of apotaxis, i.e. the availability of different roles for argument expression in verbs with similar meanings in different languages, are a non-negligible difficulty for comparing valencies that must be kept in mind (§ 10). Alternations have to be regarded as directed in a database which asks for each verb whether it undergoes a certain alternation or not (§ 12).

Acknowledgements We are grateful to Bernard Comrie and Andrej Malchukov for helpful comments on this paper.

Special abbreviations ACT AOR BGR CPL D1/2/3 POS PREP UND TEL

actor inflection aorist background completive aspect proximal/distal/anaphoric/demonstrative positive preposition undergoer inflection telic

References Abraham, Werner (ed.). 1978. Valence, Semantic Case, and Grammatical Relations. Amsterdam: Benjamins. Barðdal, Jóhanna. 2013. Icelandic Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary

Comparing verbal valency across languages

69

Anthropology. (Available online at http://valpal.info/languages/icelandic, Accessed on 2014-07-25) Bickel, Balthasar. 2011. Grammatical relations typology. In Jae Jung Song (ed.), The Oxford Handbook of Linguistic Typology, 399–444. Oxford: Oxford University Press. Bickel, Balthasar, Manoj Rai, Netra P. Paudyal, Goma Banjade, Toya N. Bhatta, Martin Gaenszle, Elena Lieven, Ichchha Purna Rai, Novel Kishore Rai & Sabine Stoll. 2010. The syntax of threeargument verbs in Chintang and Belhare (Southeastern Kiranti). In Andrej Mulchokov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions: A Comparative Handbook, 382–408. Berlin: Mouton de Gruyter. Bugaeva, Anna. 2013. Ainu Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/ainu, Accessed on 2014-07-25) Campbell, Eric. 2013. Zenzontepec Chatino Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/zenzontepecchatino, Accessed on 2014-07-25) Cennamo, Michela & Claudia Fabrizio. 2013. Italian Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/italian, Accessed on 2014-07-25) Chomsky, Noam A. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Comrie, Bernard & Zaira Khalilova. 2013. Bezhta Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/bezhta, Accessed on 2014-07-25) Conners, Thomas J. & David Gil. 2013. Jakarta Indonesian Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/ jakarta-indonesian, Accessed on 2014-07-25) Creissels, Denis. 2013. Mandinka Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/mandinka, Accessed on 2014-07-25) Dik, Simon C. 1997. The Theory of Functional Grammar. 2 vols. Berlin: Mouton de Gruyter. Estrada-Fernández, Zarina, Mercedes Tubino Blanco & Jesús Francisco Villalpando Quiñonez. 2013. Yaqui Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/yaqui, Accessed on 2014-07-25) Goddard, Cliff. 2013. English Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/english, Accessed on 2014-07-25) Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press. Grebe, Paul. 1959. Grammatik der deutschen Gegenwartssprache. (Duden). Mannheim: Bibliographisches Institut. Hartmann, Iren, Martin Haspelmath & Bradley Taylor (eds.). 2013. Valency Patterns Leipzig. Leipzig: MPI for Evolutionary Anthropology. (Available online at http://valpal.info) Hartmann, Iren, Martin Haspelmath & Michael Cysouw. 2014. Identifying semantic role clusters and alignment types via microrole coexpression tendencies. Studies in Language 38(3). 463– 484.

70

Martin Haspelmath and Iren Hartmann

Hartmann, Iren. 2013. Hoocąk Valency Patterns In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/hoocak, Accessed on 2014-07-25) Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. Language 86(3). 663–687. Haspelmath, Martin. 2013. Argument indexing: a conceptual framework for the syntax of bound person forms. In Dik Bakker & Martin Haspelmath (eds.), Languages Across Boundaries: Studies in Memory of Anna Siewierska, 197–226. Berlin: De Gruyter Mouton. Haspelmath, Martin. 2014. Arguments and adjuncts as language-particular syntactic categories and as comparative concepts. Linguistic Discovery 12(2). 3–11. Haspelmath, Martin, Andreea Calude, Michael Spagnol, Heiko Narrog & Elif Bamyacı. 2014. Coding causal-noncausal verb alternations: A form-frequency correspondence explanation. Journal of Linguistics 50(3). 587–625. Helbig, Gerhard & Wolfgang Schenkel. 1969. Wörterbuch zur Valenz und Distribution deutscher Verben. Leipzig: Bibliographisches Institut. Jackendoff, Ray. 1990. Semantic Structures. Cambridge, MA: MIT Press. Khurshudian, Victoria & Michael Daniel. 2013. Eastern Armenian Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/ languages/eastern-armenian, Accessed on 2014-07-25) Khurshudian, Victoria & Michael Daniel. 2015. Eastern Armenian. In Andrej Malchukov & Bernard Comrie (eds.). Valency Classes in the World’s Languages. Berlin: De Gruyter Mouton. Kishimoto, Hideki & Taro Kageyama. 2013. Japanese (standard) Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/ languages/japanese-standard, Accessed on 2014-07-25) Koch, Peter. 1994. Verbvalenz und Metataxe im Sprachvergleich. In Werner Thielemann & Klaus Welke (eds.), Valenztheorie – Werden und Wirkung, 109–124. Münster: Nodus. Langacker, Ronald W. 1988. The nature of grammatical valence. In Brygida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 91–125. Amsterdam: Benjamins. Lehmann, Christian. 2013. Yucatec Maya Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/yucatec-maya, Accessed on 2014-07-25) Levin, Beth & Malka Rappaport Hovav. 2005. Argument Realization. Cambridge: Cambridge University Press. Malchukov, Andrej & Alexander Jahraus. 2013. Russian Valency Patterns In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/russian, Accessed on 2014-07-25) Mel’čuk, Igor’ Aleksandrovič. 1974. Opyt teorii lingvističeskix modelej “smysl-tekst”: semantika, sintaksis. Moskva: Nauka. Michaelis, Susanne Maria & the APiCS Consortium. 2013. Experiencer constructions: headache. In Susanne Maria Michaelis, Philippe Maurer, Martin Haspelmath & Magnus Huber (eds.), The Atlas of Pidgin and Creole Language Structures, 260–263. Oxford: Oxford University Press. Moyse-Faurie, Claire. 2013. Xârâcùù Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/xaracuu, Accessed on 2014-07-25)

Comparing verbal valency across languages

71

Nordhoff, Sebastian. 2013. Sri Lanka Malay Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/sri-lankamalay, Accessed on 2014-07-25) Peterson, David A. 2007. Applicative Constructions. Oxford: Oxford University Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Schikowski, Robert, Balthasar Bickel & Netra Prasād Paudyāl. 2013. Chintang Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/ languages/chintang, Accessed on 2014-07-25) Schultze-Berndt, Eva. 2013. Jaminjung Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/jaminjung, Accessed on 2014-07-25) Seifart, Frank. 2013. Bora Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/bora, Accessed on 2014-07-25) Shibatani, Masayoshi & Ketut Artawa. 2013. Balinese Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/balinese, Accessed on 2014-07-25) Svartvik, Jan. 1966. On Voice in the English Verb. The Hague: Mouton. Tesnière, Lucien. 1959. Éléments de syntaxe structurale. Paris: Klincksieck. Vajda, Edward J. & Elena Kryukova. 2013. Ket Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/ket, Accessed on 2014-07-25) Van Valin, Robert D., Jr. & Randy J. LaPolla. 1997. Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press. Van Valin, Robert D., Jr. 2005. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge University Press. Vater, Heinz. 1978. On the possibility of distinguishing complements and adjuncts. In Werner Abraham. 1978. Valence, Semantic Case, and Grammatical Relations, 21–45. Amsterdam: Benjamins. Waltereit, Richard. 1998. Metonymie und Grammatik: Kontiguitätsphänomene in der fanzösischen Satzsemantik. Tübingen: Niemeyer. Watanabe, Honore. 2013. Sliammon Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/sliammon, Accessed on 2014-07-25) Zhang, Guohua. 2013. Mandarin Chinese Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/mandarin, Accessed on 2014-07-25) Zúñiga, Fernando & Leonel Lienlaf. 2013. Mapudungun Valency Patterns. In Iren Hartmann, Martin Haspelmath & Bradley Taylor (eds.), Valency Patterns Leipzig. Leipzig: Max Planck Institute for Evolutionary Anthropology. (Available online at http://valpal.info/languages/ mapudungun, Accessed on 2014-07-25)

Andrej Malchukov

4 Valency classes and alternations: parameters of variation 1 Introductory remarks As explained in Comrie et al. (this volume), the goal of the Leipzig Valency Classes Project is to study cross-linguistic universals and variation with respect to valency classes, as identified through distributional properties, in particular, through coding frames and alternations. The Leipzig Valency Classes Project follows up on some in-depth studies of European languages such as Levin (1993) on English and an earlier study by Apresjan (1969) on Russian. These studies have demonstrated that it is possible to arrive at a semantic classification of verbs through the study of their syntactic behavior, in particular, their potential to take part in alternations (such as the English dative alternation, locative alternation and passive alternation discussed by Levin). These results are highly interesting, but they raise the question of the extent to which valency classes arrived at on syntactic grounds are language-particular or universal. Clearly, some aspects of valency classes should be language-particular, otherwise valency dictionaries would not be necessary, yet other aspects are likely to show regularities, mirroring regularities in argument structure (role frames). The Leipzig Valency Classes Project is the first systematic large-scale attempt to explore cross-linguistic regularities in this domain. We establish valency classes on the basis of both the basic pattern (basic coding frame) and alternations–verb-coded (like passive alternation) as well as uncoded. In this respect the present project follows Apresjan (1969), which studies both coding frames and alternations, more closely than Levin (1993), which focuses on alternations exclusively. The study of coding frames is further informed by the typological studies of transitivity, such as the work by Tsunoda (1981, 1985) and subsequent work. Recently, the approaches like Levin’s that claim that the verb meaning can be recovered from its syntactic distribution have been challenged with the advent of Construction Grammar (Goldberg 1995). Construction Grammar differs from Levin’s work and related approaches inasmuch as it does not assume that the verbal semantics is (fully) recoverable from syntactic distribution, since syntactic context can also lead to reinterpretation (alternative “construal” in the terms of Croft (2012)) of verbal meaning. Yet, generally, these approaches are compatible with the proviso that for Construction Grammarians syntactic distribution defines both the core meaning of a verb and its possible extensions (“construals”) determined by the syntactic context. Clearly, the verb’s “core” meaning largely determines its possible construals, manifested through different constructions. In the present volume, Goddard’s paper stressing the importance of “lexico-semantic frames” for verb meaning is most in line with the Construction Grammar approaches.

74

Andrej Malchukov

In this chapter I will provide an overview of valency classes as identified through coding frames (section 2) and alternations (section 3). The final section (section 4) addresses the phenomenon of ambitransitivity and voice ambivalence. Generally, the aim of the chapter is to introduce the framework for the comparative study of the valency classes and to summarize some of the findings in other chapters of the volume dealing with individual languages. While this chapter focuses on variation in the domain of valency classes, the chapter by Wichmann (this volume) summarizes some of the quantitative results leading to the formulation of verb hierarchies based on patterns of alternations.

2 Valency classes: coding frames 2.1 Introduction Let us consider first the question of universality of valency classes as far as coding is concerned. On the one hand, there are clearly certain cross-linguistic regularities in coding of verbs within semantic classes reflecting similarities in argument structure (i.e. semantic roles licensed by the verbal lexeme). Thus a case marker called “allative” is used for marking Goals in very different languages, but may be extended to semantic roles which are similar, such as Recipients and Addressees. Also the transitive patterns will be used in language after language with verbs like KILL and BREAK describing canonical transitive actions with a typical agent and patient argument 1, but can be extended to other types of bivalent verbs. Such regularities are expected on iconicity assumptions, which predict iconic marking of an argument, but also allow that other semantically similar (micro)roles are coded in the same way. On the other hand, valency classes cannot be universal, since after all the coding frames (as manifested in case and agreement) are language-particular. It is easy to show that the availability of certain coding frames will have an impact on valency classes, insofar as the number of valency classes is partially dependent on the “resources” available (in our case, different coding frames). Thus, the Northern Tungusic languages Even and Evenki feature 12–14 cases, most of which are also used to code arguments, while in Eastern Tungusic languages (such as Nanai) their number reduces to 8, and in Southern Tungusic Manchu it is further reduced to 5 cases, of which only 4 can be used to code arguments (Malchukov & Nedjalkov, this

1 The latter statement is circular if a transitive pattern is defined on such verbs as BREAK and KILL; but it is not circular if a transitive pattern is described as a major or default pattern with bivalent verbs, as suggested by Lazard (1994) and others (see Haspelmath, this volume; and Blasi, this volume, for further discussion).

Valency classes and alternations: parameters of variation

75

volume). This has an obvious impact on diversification of valency classes, which correspondingly reduce in number as one moves from Northern Tungusic to Southern Tungusic. Thus, in Manchu, the Dative is used as a default oblique case for the second argument of bivalent intransitives with different argument structures, which would be coded distinctly in Northern Tungusic. For example, the NOM-DAT pattern in Manchu is found with verbs of motion (like ‘come’), static location (like ‘sit’), and emotion (like ‘fear’), which would take three distinct patterns in Even (with an allative, locative, and instrumental object, respectively). Still more instructive in that respect are languages like Vafsi (an Iranian language; Stilo 2009) with a general oblique case which can be used on both obliques and direct objects (when the latter are animate). Thus, diversification/neutralization of valency classes depends on the set of coding frames available (coding resources), although this dependency is not deterministic (one should take into account that some case systems may be elaborate for spatial relations, which are of less relevance for argument coding, and also that a reduced case system may be compensated for by adpositional use at least for some argument types).2 Perhaps the most important source of variation in coding patterns across languages concerns different extensions of the transitive frame for different verb types. This variation will be discussed in more detail below in connection with Tsunoda’s transitivity hierarchy. At this point, suffice it to say that this extension varies broadly across languages, which also has an immediate impact on differentiation of valency classes. To a certain extent this variation also depends on the argument-coding type: thus, the inventory of ‘flags’ (dependent role marking by case and adposition) tends to be richer than inventory of ‘index-sets’ (head-marking through agreement/cross-referencing). Note that minimal case systems are rather exceptional (see Arkadiev 2009 on two-term case systems), while elaborate indexing systems are equally rare (see Kratochvíl 2011 on Abui, which features no less than seven index-sets). As a result, head-marking languages tend to have an extension of a transitive frame across different types of bivalent verbs: thus in Hoocąk (Hartmann, this volume) effectively all types of bivalent verbs are transitive (featuring subject-object agreement) apart from the verbs of motion and static location, which take spatial arguments. Also “no-marking” languages of the isolating type seem to be more liberal in extending the transitive frame. Thus, Mandarin Chinese (Lu et al., this volume) can accommodate into the postverbal position many more arguments, including locative ones, than is usual for languages relying on dependent marking (see also (1)−(2) from Indonesian below). The same is true of Nǁng (Ernszt et al., this volume), where the vast majority of two-argument verbs are transitive.

2 See Say (2014) for further discussion of this point and for confirming evidence from European languages.

76

Andrej Malchukov

Yet, while such examples clearly show that valency classes are subject to crosslinguistic variation, this does not mean that this variation (in particular variation conditioned by availability of resources) is unconstrained. In what follows variation in the number of valency classes will be discussed in terms of the two competing motivations of Iconicity and Markedness. As will be shown below, both factors can be taken into account within the semantic map approach, which allows both for capturing universal regularities in coding and also for representing cross-linguistic variation.

2.2 Variation in valency classes: valency and transitivity: a first look The most basic distinction in valency classes is the transitive/intransitive distinction. Recent typological research has generally confirmed the universality of this distinction, while certain other subclasses have been shown to be more subject to cross-linguistic variation. Thus, many languages also have zero valent verbs as distinct from intransitives, but this pattern is not universal (for example, Atoyebi, this volume, notes that verbs with no arguments do not exist in Yorùbá). Similarly, some languages may feature a dedicated syntactic frame for ditransitive verbs (taking Agent, Theme and Recipient arguments), a double object construction, yet the latter pattern is not universal (Malchukov et al. 2010). The transitive-intransitive distinction is cross-linguistically most robust: thus, even Sri Lanka Malay (SLM), which shows little differentiation in verb classes (Nordhoff, this volume), still distinguishes between transitive and intransitive verbs insofar as only the latter may take an accusative object. As in other languages with extensive pro-drop (use of zero anaphora), a transitive verb in SLM is better defined as a verb which can be used in the transitive frame, rather than a verb that must be used in the transitive frame (cf. Schikowski et al., this volume, on Chintang). Yet some languages seem to challenge the universality of the transitive-intransitive distinction as well. Thus, in Jakarta Indonesian (Conners et al., this volume; cf. Gil 2009 and passim on Riau Indonesian), verbs of different semantic classes are surprisingly liberal in allowing different valency frames. Particularly surprising is that seemingly any monovalent verb may appear in the transitive frame with a bare postverbal object. (1)

Indonesian (a) Beli rumah. buy house ‘(He) bought a house.’ (b) Tidur rumah. sleep house ‘(He) is sleeping in a house.’

(Conners et al., this volume)

Valency classes and alternations: parameters of variation

77

Yet even Indonesian allows us to distinguish between the two classes if one takes into account alternative case frames for different verb types. Indeed, those verbs that deviate from the transitivity prototype (as defined by Hopper & Thompson 1980 and Tsunoda 1981) allow for an alternative pattern involving various prepositions. For example, an alternative pattern for verbs of location like ‘sit’ is a prepositional one: (2)

Indonesian (Conners et al., this volume) Mulaintar malem, tidur di rumah Ayu. beginning_later night sleep in house Ayu ‘Starting from tonight, you sleep at Ayu’s house.’

Within the approach adopted here this variation between unmarked bivalent (“transitive”) pattern and the prepositional one can be described as an (unmarked) alternation. Depending on which pattern is taken as basic, it can be further described either as preposition dropping or as preposition insertion. (As noted by Conners et al. the patterns differ in their frequencies, so for different patterns different coding frames can be considered as basic). In both cases it is clear that even though Indonesian underdifferentiates the notional transitives conforming to the transitivity prototype from other types of bivalent verbs, their neutralization is resolved through alternations. Also, the transitive-intransitive opposition can be weakened in certain languages such as Chintang (see (50)–(51) below) and Central Alaskan Yupik (44)–(45), which feature “rampant ambitransitivity” (i.e. pervasive lability of verbs). However, rampant ambitransitivity does not eliminate the transitive/intransitive distinction as such, since it does not carry over from the lexical to the syntactic level: syntactically the transitive and intransitive patterns are clearly distinguished in these languages. Apart from the traditional distinction between transitives and intransitives (with further subgroups of impersonals, and syntactic ditransitives), some linguists have proposed a more elaborate classification. Thus, Dixon (Dixon 1994; Dixon & Aikhenvald 2000) proposes the following classification, where E stands for an extension (roughly, an oblique argument): a) intransitive (S) b) transitive (A + O) c) extended intransitive (S + E) d) extended transitive (A + O + E) e) ditransitive (A + O + O) An advantage of this classification is that it clearly distinguishes verbal valency and transitivity: thus, a class of bivalent intransitives (extended intransitives) is distinguished from (mono)transitives, and trivalent monotransitives (extended transitives) are distinguished from ditransitives. It also allows us to capture further variation between valency classes. Thus, one can characterize languages like Chati-

78

Andrej Malchukov

no (Campbell, this volume), which feature a generalized oblique marker used for both (prominent) Ps and obliques, as neutralization between extended intransitive and (mono-)transitive constructions (or maybe, as assimilation of the former to the latter). One can also characterize languages like Sri Lanka Malay (Nordhoff, this volume) as having assimilation of extended intransitives to the intransitive class (hence the distinction between oblique arguments and adjuncts is lost), and of extended transitives to (mono-)transitives (hence oblique/indirect arguments of trivalent verbs pattern with adjuncts). Yet it is clear that this classification is also incomplete. Thus, Kishimoto et al. (this volume) propose the following six basic valency patterns for Japanese. (3)

Japanese a. intransitive XNOM hasiru ‘run’ sinu ‘die’

(Kishimoto et al., this volume) b. double subject XNOM YNOM itai ‘ache’ sukida ‘be fond of’

d. semi-transitive XDAT YNOM hituyooda ‘necessary’ wakaru ‘understand’

c. semi-intransitive XNOM YLOC/ABL tuku ‘arrive’ aru/iru ‘be’ saru ‘leave’ umareru ‘be born’

e. transitive XNOM YACC tateru ‘build’ korosu ‘kill’

f. ditransitive XNOM ZDAT/ LOC /ABL YACC ataeru ‘give’ nusumu ‘steal’

In this classification, ‘semi-intransitives’ roughly correspond to Dixon’s extended intransitives (with an oblique object), but ‘double subject’ predicates and ‘semitransitives’ (with an inverted pattern) cannot be easily accommodated into Dixon’s classification. Indeed, while extended intransitives show similarities to non-canonical object constructions, the latter patterns are rather similar to the non-canonical subject constructions and are occasionally analyzed as such (Onishi 2001; but see Shibatani 2001, who argues for a special status of these “double subject” constructions, distinct from transitive and intransitive patterns). These latter patterns are illustrated below: (4)

Japanese Otoko-ni(-wa) kuma-ga mie-ta. man-dat(-top) bear-nom see-pst ‘The man saw the bear.’

(Kishimoto et al., this volume)

(5)

Japanese Watasi-{ga/wa} onaka-ga itakat-ta. 1.sg-{nom/top} stomach-nom painful-prs ‘I had a pain in my stomach.’

(Kishimoto et al., this volume)

Valency classes and alternations: parameters of variation

79

Other deviant patterns, straddling between transitives and intransitives, are found elsewhere. Thus, apart from extended intransitives (with an oblique object) Bezhta features a special “affective” pattern with an oblique subject: (6)

Bezhta kibba-l quy tuq-iyo girl.obl-lat noise hear-pst ‘The girl heard the noise.’

(Comrie et al., this volume)

In addition, Bezhta also features a class of unergative verbs, which exceptionally mark a single argument by the ergative case (rather than the absolutive case found elsewhere with intransitive verbs): (7)

Bezhta öždi lalaƛo-yo boy.obl(erg) shout-prs ‘The boy shouts.’

(Comrie et al., this volume)

While it is possible to view such cases as instances of noncanonical marking of subjects (and objects), this characterization does not do justice to the intermediate status of both constructions with respect to transitivity. Meanwhile, the transitiveintransitive distinction is better viewed as a cline rather than a dichotomy (see Haspelmath (2001) for a similar view, and Haspelmath (this volume), for empirical evidence for this claim).3 Yorùbá (Atoyebi, this volume) provides another example of a construction whose characterization as transitive or intransitive is contentious. Like Bezhta, Yorùbá also shows a split among (notionally) monovalent verbs into an unergative and an unaccusative class. However, in Yorùbá, the unergative class is much broader (including ko̩rin SING, wúkó̩ COUGH, s̩éj̩ ú BLINK, sáré RUN, ré̩rìín LAUGH, s̩eré PLAY, fé̩ràn LIKE, dáná COOK and sunkún CRY) and further shows conflation with the transitive pattern insofar as agentive verbs combine with an inherent (non-referential) object: (8)

Yorùbá Bó̩lá ko̩ orin Bola sing song ‘Bola sang (a song).’

(Atoyebi, this volume)

3 Also the frequency data, whenever available, suggest that a transitive/intransitive distinction is better viewed as a cline since different verbs vary vastly with respect to the transitive/intransitive preference; see, in particular, Ernszt et al. (this volume) on Nǁng, and Schikowski et al. (this volume) on Chintang.

80

Andrej Malchukov

Yet another type that cannot be easily accommodated into Dixon’s typology is an object-experiencer pattern, exemplified here from Nen (Evans, this volume) and Ket (Vajda, this volume) (see also (38) below from CAY Eskimo). The pattern is called ‘transimpersonal’ (Malchukov 2008), as it features a non-referential (or dummy) subject. (9)

Nen zän ak-äm y-ram-t-e dog.abs thirst-erg 3sgU:α-make-nd:ipfv-3sga ‘The dog is thirsty.’

(10) Ket qoʁot da-taŋaja-q-in-di-t hunger inan.sbj-stretch-inside-pst-1sg.obj-pst-carry ‘I got hungry.’ (literally, ‘Hunger stretched me.’)

(Evans, this volume)

(Vajda, this volume)

From a diachronic perspective, transimpersonal constructions of the type illustrated in (9)–(10) can give rise to the “unaccusative” So pattern in a split intransitive system, yet their characterization as transitive or intransitive is equally problematic. This brief overview reveals that there are more distinct types of semitransitive verbs that have to be distinguished (more distinctions can be made in the domain of trivalent verbs as well; Malchukov et al. (2010)). Simultaneously, it reveals that there are certain cross-linguistic regularities in this domain insofar as minor patterns are restricted to certain semantic verb classes (cf. Lazard 1998). Thus, patterns with non-referential objects are frequently found with verbs of bodily reaction and excretion, which Lazard dubs ‘antiimpersonal’. Inverted patterns with non-canonical subjects are frequently found with emotion predicates (as already demonstrated for Bezhta and Japanese), and transimpersonal predicates with verbs of sensation (Evans 2004; Malchukov 2008). This brings us to the approaches which try to constrain and predict syntactic frames for particular semantic verb classes.

2.3 Valency and transitivity: transitivity hierarchies and semantic maps The first major attempt to predict syntactic patterns (valency frames) for verb classes is Tsunoda’s Transitivity Hierarchy. In the early 1980s, Tsunoda (1981) developed a functional approach to transitivity, which is in many ways similar to Hopper & Thompson’s (1980) influential account, but has a broader scope. While Hopper & Thompson (1980) studied the impact of clause-level factors (such as aspect or mood) or NP-related factors (such as A-animacy and Object individuation) on the encoding of a clause as transitive or intransitive, Tsunoda also took verb semantics

Valency classes and alternations: parameters of variation

81

into consideration. As is well-known, he proposed the following hierarchy starting from canonical transitives (“effective action verbs”) to less canonical ones: Effective action >> Perception >> Pursuit >>Knowledge >>Feeling >> Relation Fig. 1: Tsunoda’s Transitivity Hierarchy (Tsunoda, this volume, and passim).

The Transitivity Hierarchy (called Hierarchy of Two-Place Predicates, HTPP, in Tsunoda, this volume) states that a transitive pattern (nominative-accusative or ergative-absolutive) extends on the hierarchy from left to right and languages may show different cut-off points on the hierarchy. This hierarchy, which first appeared in Tsunoda (1981), has since undergone significant developments, but the general layout has remained the same (see Tsunoda, this volume, for further discussion). Thus, the typology predicts that if emotion verbs pattern transitively, then verbs of perception and pursuit will do so as well. Of course, such predictions would always be of a statistical nature, and may fail for individual lexical items; therefore they are formulated in terms of semantic classes rather than individual lexemes. But in this respect this work is no different from other work in lexical typology such as Dixon’s (1977) study of adjectives, which is predictive at the level of semantic classes (e.g., ‘if (some) terms of human propensity are coded as adjectives, (some) color terms will be as well’), rather than for individual lexical items, which might show idiosyncrasies reflecting details of historical development. The same is true for the Transitivity Hierarchy, as will be discussed later. Tsunoda’s Transitivity Hierarchy has attracted much attention in the typological literature since it was the first major attempt to predict the syntactic behavior of verbs on a semantic basis; yet, some questions have been raised as well. Lazard (1984) questioned the ordering of certain verb types on the hierarchy, while Lehmann (1991) noted that it conflates several dimensions. Partly in response to this criticism, Malchukov (2005) suggested decomposing the hierarchy into a two-dimensional map in order to disentangle two dimensions. One subhierarchy spanning verbs of effective action to verbs of motion represents a dimension of decreased patienthood 4 on the part of the object argument (the agent-patient frame is replaced by the agent-goal frame), while the second subhierarchy represents an additional dimension of decreased agentivity (a change from agent-patient to experiencer-stimulus).

4 Cf. the Affectedness Hierarchy introduced by Beavers (2011), which, however, suggests a still more regimented process with more intermediate verb types. Thus, on Beavers’ hierarchy, CUTverbs with a patient undergoing non-quantized change are intermediate between BREAK-verbs (with a change quantized) and HIT-verbs where a theme argument has a mere potential for change.

82

effective action

Andrej Malchukov

contact

pursuit

(motion)

perception cognition

emotion

(sensation)

Fig. 2: Two-dimensional Transitivity Hierarchy (semantic map).

Note that unlike the original hierarchy (Fig. 1), on the hierarchy in Fig. 2 verbs of pursuit are not ordered with respect to different varieties of experiential verbs as they correspond to different subhierachies. Indeed, the data from ValPaL, as reported in Figure 1 in Haspelmath (this volume), is consistent with this conclusion. One reason for decomposing a semantic map in my earlier work was to capture the fact that the departures from transitivity may be different along the two subhierarchies. In particular, only the second subroute (from canonical transitives to verbs of emotion and sensation) shows the inverted pattern with the first argument represented as an oblique. This is clearly due to the role structure: experiencer arguments are iconically coded by the dative case. In this way, semantic maps are well suited to display patterns of construction splits, for example, resulting from different extensions of the transitive pattern on the map. Thus, while many languages extend the transitive pattern from notional transitives to verbs of contact like HIT, some use a NOM-OBL or NOM-DAT pattern instead (see the chapters on Armenian and Sri Lanka Malay). Another use of these maps is to display alternations, which appear as areas of overlap on a semantic map. Thus, the following map shows the domain of the conative alternation in English (hit vs. hit at) centered on impact/contact verbs (the same is true of Yaqui), and the alternation between a transitive and the dative-nominative pattern with some experiential predicates in Japanese (cf. miru ~ mieru for LOOK/SEE, and kiniiru for LIKE). In principle also other (uncoded) alternations can be represented in this way. For example, a semantic map based on transitivity extensions can partially capture effects of the inchoative-causative scale suggested by Nedjalkov & Sil’nitskij (1969), Haspelmath (1993) and subsequent work. On this map, too, BREAK would occur closer to the transitive pole, and BURN would be found closer to the intransitive pole, and overlap of the range of the transitive and the intransitive frames would indicate ambitransitivity for individual items. (For example, while in English both verbs are ambitransitive, in Italian, only BURN is ambitransitive, while BREAK needs reflexive marking in anticausative use). The effects of the inchoative-causative hierarchy are captured on the assumption that if a verb is transitive it can be anticausativized by an A-deleting operation (if available in that language), while if it is intran-

Valency classes and alternations: parameters of variation

contact

pursuit

(motion)

perception cognition

emotion

(sensation)

83

effective action

Fig. 3: Unmarked alternations on the two dimensional transitivity scale. (Legend: solid line indicates a transitive extension, broken line an intransitive extension; areas defined by black lines indicate English conative alternations; those defined by gray lines indicate an alternation in Japanese)

sitive it may be causativized by an A-adding operation (if available). On a broad interpretation of anticausative operations as involving A-deletion, and a causative operation as including both morphological and periphrastic causatives, effects of the inchoative-causative scale follow. Thus, on our approach, the inchoative causative hierarchy can be more directly derived from the transitivity scale, than from the scales for A-adding or A-deleting operations discussed in section 3, and in more detail in the chapter by Wichmann (this volume). This is because, on our approach, verb meanings have in-built “biases”, so that BREAK is included in the transitive (bivalent) use, while BURN is included in its intransitive (monovalent) use. Similar maps (multidimensional hierarchies) can be proposed for the domain of monovalent and trivalent verbs (events). In many languages monovalent verbs behave uniformly, selecting for the same (intransitive) coding frame, but some show subdivision, as in split-intransitive languages, where agentive (“unergative”) and non-agentive (“unaccusative”) verb classes select for different coding frames (see Hartmann, this volume, on Hoocąk ). Furthermore, some languages show assimilation of one or the other of these classes to the transitive pattern. Thus, in Yorùbá, agentive verbs are assimilated to transitives (see (8) above), while in Nen and Ket patientive verbs (transimpersonals) are assimilated to transitives (see (9) and (10), respectively). This assimilation is only partial though: in both cases the minor class is assimilated, while the majority intransitive class shows a coding frame distinct from transitives. Moreover, these types of notionally monotransitive verbs may alternate between transitive and intransitive patterns in other languages. Thus, in Nǁng, unlike Yorùbá, typical unergatives (like RUN) may appear either in an intransitive or a transitive pattern (with a cognate object). And transimpersonal constructions in CAY Eskimo exemplified in (38) below may be rendered through an intransitive pattern with the same verb. Further discussion of the con-

84

Andrej Malchukov

struction splits and role clustering for monovalent predicates from a quantitativetypological perspective can be found in Bickel et al. (2014), and from a more general analytic perspective in Croft (2012: 256–258). Different groups of trivalent verbs also show construction preferences, which can be captured through the use of semantic maps (see Comrie et al., this volume). As discussed in Malchukov et al. (2010), core ditransitives (verbs like give, featuring Agent, Theme and Recipient) show preference for a double object construction (T= P=R), caused motion verbs (like put and throw) show preference for an indirective construction (T=P ≠ R), while verbs of impact (like hit and cover) show preference for the secondary object construction with an (instrumental) oblique (T ≠ P=R). LOAD-verbs provide another “bridge” between caused motion verbs and verbs of impact (see the semantic map of the ditransitive domain in Comrie et al., this volume). Not surprisingly, ditransitives and LOAD-verbs often allow for alternations, aligning themselves either with caused motion verbs or verbs of contact, with which they show similarities in the role structure (in Bickel’s terms, all of these classes feature Agent, Theme and Goal arguments; see Schikowski et al., this volume). This is familiar from English featuring dative alternation and a locative alternation, but is also attested elsewhere (see, e.g., Kász, this volume, on Modern Standard Arabic). Thus, in Balinese (Shibatani & Artawa, this volume), caused motion verbs predictably feature an “indirective” (allative) pattern, and verbs of impact like HIT feature a secundative pattern with the instrumental preposition. This intrinsic alignment pattern can be changed through the use of applicative morphology (Goal-applicative in -in in case of caused motion verbs and Instrumental-applicative in -ang in case of verbs of impact). Neither ditransitive verbs, nor LOAD-verbs (and ‘verbs of fixing’) show such intrinsic alignment. Ditransitives can appear in a double object or indirective construction without extra marking, while LOAD-verbs cannot appear without applicative morphology at all (are “precategorial” in Shibatani & Artawa’s terms). Instead, LOAD-verbs appear in an equipollent applicative alternation taking Goal applicative in the double object pattern or Instrumental applicative in the indirective pattern (see Shibatani & Artawa, this volume). Similarly to the use of maps in the other domains, alternations in the domain of trivalent verbs can be represented as areas of overlaps on a semantic map (see Comrie et al., this volume, for presentation of the semantic map of ditransitive verbs, and Malchukov et al. 2010 for further discussion). A simplified representation of the distribution of different ditransitive constructions in Balinese on the ditransitive semantic map is shown in Figure 4 below. (The representation is simplified, as it is confined to a subset of verb meanings and constructions considered; further, other uses of the -ang and -in markers are ignored; see Shibatani & Artawa, this volume, for details). The map shows an (unmarked) ditransitive alternation between an indirective and double object construction for (certain) semantic ditransitives (like baang ‘give (1)’), as well as an overlap between instrumental and goal applicatives for “precategorial roots” like buat ‘load’ and (certain) other ‘verbs of fixing’ resulting in an equipollent applicative alternation.

Valency classes and alternations: parameters of variation

External possession construction

Malefactive Source construction

85

Internal Possessor construction

BREAK him X

Patient Beneficiary construction SELL

BUILD him a house

Theme-Recipient ient construction GIVE G

SEND

Patient Instrumental construction

Theme-Goal Go io construction

(with HIT verbs) SPRAY/LOAD

(PUT,PULL) (PUT, ULL

Fig. 4: Marked and unmarked ditransitive constructions and alternations in Balinese. Legend: Basic indirective construction Basic secundative construction Basic neutral construction (Double Object Construction) Instrumental Applicative constructions (-ang-applicative) ---------Goal Applicative construction ( -in-applicative) –··–··–

2.4 Typological variation in coding frames: towards a predictive approach Building on Tsunoda’s work (Tsunoda 1981, 1985), as well as related typological and optimality-theoretic work, I proposed (Malchukov 2005) a predictive typology of argument marking for certain verb types. In essence the claim is that semantic maps/hierarchies are able to represent both universal and language-particular properties of valency classes. Two general factors which have the greatest impact on the choice of the frame are Iconicity or Faithfulness (choice of the most semantically fitting frame) and Markedness (choice of the transitive frame as a major default pattern; cf. Lazard 1998). Competition between the two factors is already suffi-

86

Andrej Malchukov

cient to model different extensions of the transitive frame across semantic maps. Thus, ranking Markedness over Iconicity can be used to represent (a range of) extension of the transitive frame. Note that Iconicity reflecting the semantic role structure would also make specific predictions concerning other patterns, not just of transitive vs. intransitive patterns. As already mentioned in connection with the two-dimensional map in Fig. 2, one of the reasons for decomposing Tsunoda’s onedimensional hierarchy was an attempt to represent the fact that different departures from transitivity result in different preferential frames. Thus, only the second subhierarchy on the map in Figure 2 is associated with the inverted Obl-S frame (which may border on noncanonical subject marking), while the first subhierarchy (including verbs of pursuit, among others) rather features an S-Obl frame (which may develop into noncanonical object marking). This approach is illustrated below for verbs of pursuit, i.e. verbs meaning ‘look for’, ‘follow’. In my earlier work (Malchukov 2005) I proposed that cross-linguistically recurrent case frames can be explained by three general functional constraints: – FaithRole: faithful encoding of semantic roles (de Hoop & Malchukov 2008; cf. the work of Kiparsky (1998) and Butt (2006), where both argument roles and cases are defined in terms of common features) – Transitive Default: assimilation to the majority pattern of bivalent verbs (Lazard 1998; Primus 1999) – Unmarked Case Constraint (proposed by Tsunoda (1981), following earlier work by Shibatani (1977)). The last constraint requires the use of the unmarked case (nominative in a nominative-accusative language or absolutive in an ergative-absolutive language) in all constructions (thus penalizing the use of impersonal structures). From an Optimality-Theoretic perspective, the Unmarked Case Constraint (UCC) as well as the TransDef may be regarded as markedness constraints5, and contrasted with the faithfulness constraint FaithRole as a reflection of iconicity. First, consider predictions of this approach with respect to coding of pursuit verbs in accusative languages. Starting from the assumption that the argument frame including Agent and Goal (non-attained object, unlike Patient) would be faithfully represented as NOM-OBL, the two optimal case frames will be NOM-OBL (if Faithfulness decides the competition), or NOM-ACC (if Transitive Default determines the competition). Schematically (>> stands for a “stronger” dominating factor/constraint): – NOM-OBL, if FaithRole >> {UCC, TransDef} – NOM-ACC, if {TransDef, UCC} >> FaithRole

5 In some OT approaches, like Woolford’s (2001), both markedness constraints are derived from a single Case Markedness Hierarchy: NOM, ABS >> ACC, ERG >> Oblique.

Valency classes and alternations: parameters of variation

87

Both patterns are indeed common: thus, the NOM-OBL pattern is typical for pursuit verbs in English (and is also found in Armenian, Russian, Even and Ket), while the transitive frame is typical for Japanese (and also found in Evenki, Yaqui and Bora). For ergative languages, the competition is somewhat more complex, since the faithful ERG-OBL pattern (Agent faithfully coded by the ergative, Goal argument coded by an oblique) will violate the unmarked case constraint. The alternative patterns result from different ranking of constraints: ERG-OBL emerges if faithfulness is the strongest constraint; ERG-ABS emerges if Transitive Default determines the competition; ABS-OBL wins if Unmarked Case Constraint dominates Faithfulness and Faithfulness dominates Transitive Default). Schematically: – ERG-OBL, if FaithRole >> {UCC, TransDef}; – ERG-ABS, if TransDef >> {UCC, FaithRole}; – ABS-OBL, if UCC >> FaithRole>> TransDef Indeed, all patterns here are attested: ERG-OBL is found in some Australian languages, such as Djaru (Tsunoda 1981), the transitive ERG-ABS pattern is attested in Central Alaskan Yupik, Jaminjung and Nen, among others, while the ABS-OBL pattern is found in some Caucasian languages, including Bezhta. In a similar way preferred coding patterns can be predicted for other verb types (see Malchukov (2005) for further discussion and references).

2.5 Other factors Yet, as also noted in this study, apart from the general factors of Markedness and Iconicity, there are other interfering factors that influence the resulting coding pattern. Some of these factors are also relevant for interpreting the results of the Leipzig Valency Classes Project. In particular the following factors may interfere with the results of the competition (cf. Malchukov 2005): a) complexity b) word class of the predicate c) verb polysemy d) inheritance The first factor pertains to cases when verb meaning in particular languages is expressed periphrastically (Haspelmath speaks of such cases as non-canonical counterparts of a verb meaning). Thus, the meaning of BLINK is rendered as ‘close eye’ in Bezhta in (11) and LOVE is rendered as (u)-som tukt- ‘liver aches’ in Chintang (Schikowski et al., this volume). (11) Bezhta öždi-s häy-dä pacpaƛo-s boy.obl-gen1 eye-pl close-prs ‘The boy’s eyes are blinking.’

88

Andrej Malchukov

When a verb meaning is expressed periphrastically by a verb-noun combination, the participant corresponding to a role variable of the lexeme in question (‘blinker’ in (11)) may be expressed as an oblique or a possessor (as is also the case in (11)) rather than a core argument. Even though such argument “excorporation” (a process opposite to argument incorporation/conflation) often produces non-canonical marking of arguments, the impact of this factor is limited, since lexical operations such as incorporation or excorporation have obvious semantic consequences. Indeed, if an argument is incorporated, the meaning of the verbal lexeme is narrower than V (cf. use of ‘hammer’ instead of ‘hit’); and if an argument is excorporated, its meaning is more general than V (cf. ‘blink’ vs. ‘close (eye)’). In both cases, the meaning of the verb changes substantially, which suggests that such cases should be disregarded if a study takes a predefined list of meanings as its starting point. The second factor pertains to word class of the predicate; in particular, it states that non-verbal predicates would disprefer a transitive frame. This would account for the fact that many sensation and emotion predicates pattern intransitively in languages such as Japanese. In typological studies this factor can be taken into account either by annotating the meanings for the verb class, or simply by excluding non-verbal expression of the meanings from consideration. It should be noted that although this factor is very important for many languages, it can be violated due to another factor: thus, in Japanese some emotion predicates basically of adjectival origin, such as suki ‘like’, have been realigned with a transitive frame: thus markedness (transitive default) is taking the upper hand. For the present project this pattern is again of marginal importance since the Leipzig Valency Classes Project is focused on verbal predicates. The last two factors are related and can be treated together. Verb polysemy is relevant for cases where the verb has several meanings that differ in argument structure but share a single frame that is expected only for one of its meanings. One example discussed in Malchukov (2005) is the use of the verb LOOK in Lezgian in an inverted DAT-ABS pattern. Note that this pattern occasionally occurs with inactive perception verbs (e.g., in Japanese mieru has a DAT-NOM pattern), but is virtually unattested for active perception verbs, which have an agent rather than experiencer as a first argument. Now, in Lezgian the same item can be used in both senses LOOK and SEE, which can account for a deviant pattern here. (12) Lezgian Gila kwe-z [za wuč-da-t’a] aku! now you-dat [I.erg do.what-fut-cond] see.ipfv ‘Now look (lit. see) what I am going to do!’

(Haspelmath 1993: 283)

A similar case can be found in Sri Lanka Malay (Nordhoff, this volume), where diyath LOOK takes a DAT-ACC pattern; again, the same verb has the meaning of SEE elsewhere.

Valency classes and alternations: parameters of variation

89

While polysemy can account for such cases and, theoretically, can be controlled for, in other cases the polysemy is not attested synchronically any more; rather the verb meaning has shifted but the corresponding verb pattern is retained. For example, Goldberg (1995) explains the double object pattern with verbs like deny by showing that the verb belonged to the semantic class of ditransitives. Such cases of inheritance are a frequent source of idiosyncratic patterns of argument coding. There are other factors that may influence case-marking. Thus, transitivity (and its possible extensions on the hierarchy) is influenced by a number of further factors, as discussed by Hopper & Thompson (1980) and Tsunoda (1981). However, many of these factors are variable clause-level properties (aspect, tense, mood, negation), which are more relevant for transitivity alternations rather than transitivity splits considered here. The role of aspect (or rather actionality) will be discussed below in connection with the distinction between result-verbs and mannerverbs proposed by Levin (this volume, and passim). Similarly, animacy (and other inherent features of an argument) may have an influence on the coding patterns, as is well known from the literature on differential case marking. Yet, the impact of this factor on coding frames is also of secondary importance, since it is not verb-sensitive. The situation can be different in headmarking languages, though, as indexing is more sensitive to animacy, and in some languages animacy distinctions can be lexicalized. The most striking case are languages such as Ojibwe, where both transitives and intransitives are split into animacy-based classes. (13) Ojibwe (Rhodes & Valentine, this volume) a. Ngii-miijin wiiyaas. ni - gii= miij- i -n(aa) wiiyaas 1sbj- pst= eat- inan.obj - n meat ‘I ate the meat.’ b. Ngii-mwaa moos. ni -gii= am -w -aa moosw 1sbj- pst= eat tr.an 3an.obj moose ‘I ate moose.’ Interestingly this distinction does not merely affect coding but extends to alternations as well. Thus, only animate transitives allow for reflexive and, less expectedly, for passivization as well.

90

Andrej Malchukov

3 Valency alternations 3.1 Introductory In this section I discuss general trends in verb classification as defined in terms of alternations rather than coding frames. In other words, the question to be addressed is to what extent the verb classes established by Levin (1993) for English or by Apresjan (1969) for Russian are universal. Here again, one does not expect strict universality, given that, as in the case of coding frames, the set of alternations is after all also language-specific. Thus, in English, Levin (1993) mentions some alternations which hardly have parallels in other unrelated languages, such as the ‘way-alternation’ (He whistled his way through the crowd). In other languages we also find idiosyncratic alternations such as the ‘Prenasalization Alternation’ in Indonesian, or the ‘Quantity-Ratio Alternation’ in Mandarin Chinese, which do not find ready parallels in other languages. On the other hand, some of the other alternations described by Levin for English (such as the Inchoative-causative alternation, the Passive alternation, or the Dative alternation) more easily find parallels across languages. These alternations too show variation across languages, in particular whether they are verb-coded or not: thus in some languages the inchoative-causative alternation manifests itself in ambitransitivity (I broke the stick ~ The stick broke), but in other languages it is verb-coded, either by a causative marker (as in Turkish) or anticausative marker (as in Russian, see Nedjalkov & Sil’nitskij 1969; Haspelmath 1993; Nichols et al. 2004). Similarly, an applicative marker may mark a dative alternation introducing a recipient or beneficiary, while it is unmarked in English. And even a passive, which is most typically verb-coded across languages, may remain uncoded in some languages such as Mandinka (Creissels, this volume). In short, many alternations can be compared across languages in syntactic terms: thus a causative adds an agent subject, a passive (or anticausative) deletes an agent subject, while an applicative (or a corresponding uncoded alternation) adds an object. Importantly, in the present context, once alternations are classified in this way, we can ask the further question whether these alternations are sensitive to verb types and whether cross-linguistic regularities can be established in this domain in the form of alternation hierarchies. This is also the question addressed in this chapter below, as well as in the chapter by Wichmann (this volume), which addresses specifically alternation hierarchies. Before we embark on this discussion we will have to delimit the domain of alternations as understood in the Leipzig Valency Classes Project.

3.2 Valency alternations: definition and related notions In the Database Manual, a valency alternation is defined as follows (see also Haspelmath & Hartmann, this volume):

Valency classes and alternations: parameters of variation

91

For the purposes of our project, a valency alternation is defined as a set of two different coding frames that are productively (or at least regularly) associated with both members of a set of verb pairs sharing the same verb stem.

The definition above is inclusive insofar as it takes into account both uncoded alternations (like the English Dative alternation), but also verb-coded alternations (like the English Passive alternation). Further, the definition is inclusive in yet another respect: apart from voice proper (remapping of semantic roles to syntactic arguments), it also includes other valency operations which change the number of roles (such as anticausative or causative). First, the distinction across languages is problematic, given that the same form may perform both functions (e.g., the Russian reflexive marker -sja has both a reflexive-passive and an anticausative function). More generally, as observed by Dixon & Aikhenvald (2000), it is difficult to classify valency-changing operations into those having exclusively syntactic effects (voice proper, like passives) versus those having exclusively semantic effects (like causatives). Rather, all valency-changing operations have syntactic, semantic and also pragmatic functions. Finally, in the discussion of the unmarked argument alternation in English and elsewhere, the question of meaning preservation is not an issue, as, for example, the inchoative-causative alternation clearly affects the number of arguments.

3.3 Delimiting valency alternations While our definition of alternation is fairly inclusive in a number of ways, it is restricted in other ways. Some of these restrictions are definitional, and follow from the definition of alternation adopted in the Leipzig Valency Classes Project, while the others are rather of a practical nature and pertain to the scope of the project. First, from our definition it follows that we are concerned with alternations involving arguments, not adjuncts, as the latter are not syntactically constrained by verb types (even though they may show semantic restrictions). For this reason, I will have less to say about valency pattern extension by an oblique NP, whose status as an argument or an adjunct is unclear. Daniel & Khurshudian (this volume) discuss such extensions in Armenian in great detail, but also treat them separately from alternations, as the status of the oblique NP is unclear. The same remarks apply to Emai, where numerous “insertion” alternations posited (Schaefer & Egbokhare, this volume), especially those which rely on serial verbs, are arguably more reminiscent of augmentation of the coding frame through an adjunct: 6

6 The issue of the argument/adjunct distinction in Emai is more complicated, though, since combinations with certain oblique arguments are restricted in a seemingly idiosyncratic way. Thus, for example, KILL in Emai (expectedly) allows for combination with an instrumental NP (introduced by a “precedence” SVC), but less expectedly disallows combination with a benefactive NP (intro-

92

Andrej Malchukov

(14) Emai (Schaefer & Egbokhare, this volume) òjè kpáyé àlèkè híán ólí úì. Oje replace Aleke cut the rope ‘Oje cut the rope instead of/in lieu of Aleke.’ While ultimately combinability with oblique arguments and adjuncts also contributes to verb classification, such more inclusive classification is not systematically attempted within the Leipzig Valency Classes Project. Yet, the chapters on Armenian, Nen, and especially Emai can provide a good illustration of this more inclusive approach to syntactic classification. Generally, however, oblique-adding operations contribute less to our general typology as they show less verb sensitivity. For example, in Sri Lanka Malay, valency extension with oblique NPs works indiscriminately with verbs of different semantics and does not produce any valency classification (Nordhoff, this volume). For this reason, valency extension by an oblique NP will be disregarded apart from cases where adding or removing an oblique is marked on the verb. The latter option is illustrated below by a non-promotional benefactive in Nǁng, which introduces a Beneficiary in the dative case: (15) Nǁng hng kxʼuu-a ǀʼhuun-a nǁaen 3pl make-ben Boer.pl-dat blanket.pl ‘They make blankets for the Boers.’

(Ernszt et al., this volume)

‘Relative roots’ in Ojibwe also belong to this type insofar as they introduce an oblique argument, which is obligatory in this construction (with some verbs it can appear as a “secondary” object, and with a few verbs exceptionally as a primary object): (16) Ojibwe Oodenaang ngii-zhaa. oodenaw -ing ni- gii= iN- -yaa town loc 1sbj pst to go.an.intr ‘I went to town.’)

(Rhodes & Valentine, this volume)

Second, the study is confined to the forms of the same lexeme, which would exclude cases of verbal periphrasis of the type influence ~ exert influence ~ undergo influence, which have been considered as alternations by Mel’čuk (1974) and Apresjan (1974) and described in terms of lexical functions. As noted above, various kinds of “excorporation” operations are disregarded as they make the cross-lin-

duced by a “succedence” SVC). On the whole, however, it seems fair to say that roles introduced through serial verbs are more reminiscent of adjuncts.

Valency classes and alternations: parameters of variation

93

guistic comparison of verb meanings problematic. In some cases complex expression of predicates cannot be ignored since some languages like Jaminjung (Schultze-Berndt, this volume) and many other Australian languages have a small class of verbs, with most predicates being expressed periphrastically. Such cases do not involve excorporation though, as neither part of a complex predicate counts as an argument. Here the light verbs may generally be regarded as valency operators, while the lexical verb can be taken to convey verbal meaning. Thus, this pattern is not unlike the patterns of analytic ambitransitive constructions of the type attested in Hindi and Chintang (when combination of the infinitive with the ‘be/go’-auxiliary triggers an intransitive reading and combination with the ‘make/put’-auxiliary a transitive reading), with the difference that Jaminjung lacks a basic (non-periphrastic) pattern (cf. Lehmann, this volume, for a similar view). (17) Jaminjung a. Ah, bag ga-jga-ny=ni bottle. ah break 3sg.s-go-pst=sfoc bottle ‘Ah, the bottle broke.’

(Schultze-Berndt, this volume)

b. Guyuwarn bag nga-rra-ji. bone break 1sg.s-put-refl ‘I broke a bone.’ At a more grammaticalized stage, these valency operators yield equipollent transitivity alternations. This is exemplified by the equipollent anticausative-causative alternation in Ket (as in (18)), signalled by the final part of a discontinuous stem. (18) Ket (Vajda, this volume) a. ām hiɣdɯl da-toʁojiŋ-q-aj-it mother child fem.sbj-dry-inside-prs-inan.obj-take ‘The mother starts drying off the child (once).’ b. da-toʁojiŋ-q-is-a-tn fem.sbj-dry-inside -prs-3sbj-go ‘She starts (spontaneously) drying off (once).’ Third, we confine our study to alternations involving referential rather than clausal arguments. For this reason, alternations involving multiclausal units are also left out. This pertains to analogs of such English alternations as the tough-alternation (It is tough to please John ~ John is tough to please) or raising alternations (It seems that John is clever ~ John seems to be clever). For the same reason, alternations related to complex (multiclausal) constructions such as Direct Quotation and Knowledge Complements in English are beyond the scope of the present study. Fourth, by definition an alternative expression should manifest itself in a different coding. For this reason word order variation that does not have further morphosyntactic repercussions is disregarded. Note that in this approach “locative in-

94

Andrej Malchukov

version” in English (The man came vs. There came a man) also does not count as an alternation, as coding remains the same as in the original construction (i.e. in the standard varieties the verb agrees with a postverbal subject). In this respect English contrasts with Mandinka, where the locative inversion (called “subjectoblique alternation” by Creissels) has an obvious impact on coding (both flagging and indexing). (19) Mandinka (Creissels, this volume) a. Musu-kéebáa fula tú-tá saatéw-o to. woman-old two remain-cmp.pos.intr village-def loc ‘Two old women remained in the village.’ b. A tú-tá jěe musu-kéebáa fula (la). 3sg remain-cmp.pos.intr there woman-old two obl ‘There remained two old women.’ It should be admitted, however, that some alternations may be problematic in this respect; even for English it has been shown that postverbal subjects fail on some subjecthood diagnostics (Givón 1997). Still more instructive in that respect is the NOM-DAT ~ DAT-NOM alternation in Icelandic, which arguably involves grammatical function inversion rather than mere word order variation: (20) Icelandic a. Mér hefur alltaf fallid tetta vel. me.dat has always fallen this.nom well ‘I have always liked this.’

(Barðdal, this volume)

b. Tetta hefur alltaf fallid mér vel. this.nom has always fallen me.dat well ‘This has always pleased me.’ This pattern can be described as an alternation, given that the first argument invariably displays syntactic subject properties (Barðdal, this volume), whether it is NOMmarked as in (20a) or DAT-marked as in (20b). Another complex case is found in Mandarin Chinese, where the status of a construction like (21c) is controversial. On the one hand, it can be regarded as an S=P ambitransitive alternation with a transitive counterpart in (21a). On the other hand, it can alternatively be analyzed as a result of P topicalization and A deletion (i.e. derived from (21b) rather than (21a)). (21) Mandarin a. (The basic construction) wǒ dǎ-suì-le bēizi. I hit-break-pfv cup ‘I broke the cup.’

(Lu et al., this volume)

Valency classes and alternations: parameters of variation

95

b. (Object topicalization) bēizi wǒ dǎ-suì-le. cup I hit-break-pfv ‘The cup, I broke [it].’ c. (Ambitransitive Alternation) bēizi dǎ-suì-le. cup hit-break-pfv ‘The cup broke / the cup was broken.’ Interestingly, Lu et al. (this volume) are able to show that a preverbal P in (21c) is a subject rather than a topic; so in (21a)–(21c) we are dealing with S/P ambitransitivity after all. Still, the ambitransitive alternation illustrated in (21) is remarkable in that it does not show verb-sensitivity. It is likely that this “rampant” ambitransitivity in Chinese is related to the lack of morphology. The connection between the presence of valency-changing morphology and restricted use of ambitransitivity has been proposed earlier (Haspelmath 1993), and seems to be supported by the data from (some) isolating languages. Fifth, we will largely disregard variation in coding frames (conventionally known as differential case marking) that depends on the properties of the argument itself rather than the verb. While the phenomenon of differential case marking (in particular, differential subject marking, DSM, and differential object marking, DOM) is not uniform (Malchukov & de Swart 2009), generally these issues are of limited importance for valency classification. Thus, a DOM driven by definiteness/specificity as in Turkish (or in Chatino; Campbell, this volume), will be found regularly with any transitive verb and does not produce any valency classes. Similarly, animacy-based DOM as in Hindi or Armenian (Daniel & Khurshudian, this volume) is of limited importance, as most verbs allow both animate and inanimate arguments. In the present context, of interest are only those cases where such differential case marking or differential agreement further interacts with the verb type (briefly discussed in Malchukov 2005). Thus, in Newari, the availability of an ergative pattern on the one hand depends on tense and on the other hand on the lexical class of the verb, so that two-argument verbs deviating from the transitive prototype can take this pattern optionally. As an example of verb-type sensitive indexing, one can cite the Austronesian language Manam, where some experiencer verbs (‘like’, ‘know’, ‘be bad at’) use object indexing only when the object is prominent, while canonical transitives (like ‘break’) invariably index the object (Malchukov 2005).

96

Andrej Malchukov

3.4 Syntactic types of alternations In this section I will outline a general classification of valency alternation in order to spell out the assumptions behind our approach, but also to present an overview (or rather a preview) of the findings in this domain from individual chapters. The classification adopted is fairly conventional. In the typological literature it is common to use the following general classification of valency-changing operations (Haspelmath & Müller-Bardey 2004; cf. Lazard 1984; Dixon & Aikhenvald 2000): a. valency-increasing – subject-adding (causatives, etc.) – object-adding (applicatives, etc.) b. valency-decreasing – subject-removing/demoting (passives, anticausatives) – object-removing/deleting (antipassives, P-incorporation) c. valency-rearranging – subject-object rearranging (inverses) – object-object-rearranging (applicatives) d. argument-identifying (reflexives, reciprocals) Of course, such a classification remains very general. In particular, it glosses over the distinctions between demoting and deleting operations, which in some approaches (like Lehmann’s, this volume) correspond to two different valency levels. Furthermore, some other valency-operations do not fit easily into the proposed classification, and will be addressed at the end of the section. The following remarks discuss different types of valency alternations based on the data in the contributions to the volume, and also address those constructions that defy classification in one way or another (see Lehmann, this volume, for more illustration and discussion). a. valency-increasing – subject-adding (causatives, etc.) Apart from causatives other A-adding alternations belong to this class such as ‘directive’, ‘speculative’, or ‘reportative’ forms in Central Alaskan Yupik: (22) Central Alaskan Yupik (Miyaoka, this volume) Tuqu-c-i-yuk-i-unga taquka-mun qimugte-mek. die-a-antip-think-antip-ind.1sg bear-all.sg dog-abm.sg ‘I thought the bear killed (made-die) a/the dog.’ Some of these operations are more appropriately defined as A/S-adding, giving a possibility of S-adding operations on zero-valent verbs. An S-adding operation can be illustrated by the adversative passive built on impersonal meteoverbs in Even

Valency classes and alternations: parameters of variation

97

(cf. Malchukov (1993, 1995); Nedjalkov (1992) on adversative passives in Tungusic languages): (23) Even a. Udn-a-n rain-nf-3sg ‘It rains.’

(Malchukov & Nedjalkov, this volume)

b. Etiken uda-la-w-ra-n old man rain-advp-nf-3sg ‘The old man is caught by the rain.’ – object-adding (applicatives, etc.) Apart from productive applicatives, this class includes applicatives of a more lexical kind called ‘extraversive’ derivation by Lehmann (this volume): (24) Yucatec Maya a. k-u haan-al ipfv-sbj.3 eat-incmpl ‛he eats’

(Lehmann, this volume)

b. k-u haan-t-ik ipfv-sbj.3 eat-trr-incmpl ‛he eats it’ b. valency-decreasing – subject-removing/demoting (passives, anticausatives) This class includes both A-removing and A-demoting operations. The former include anticausatives, but also different kinds of “middles” (cf. This book sells well.), or resultatives (cf. The door is open.), which do not allow the expression of the A argument. As noted above, one reason for treating the forms together is that the same marker is used whether an agent is implied or not. This is exemplified by the “mediopassive” form in -b- ~ -p- from Even: (25) Even Tör anga-p-ta-n door open-med-nf-3sg ‘the door opened/was opened’

(Malchukov & Nedjalkov, this volume)

Subject-demoting operations include canonical passives, but also other cases, which are not always associated with passives, such as the potential in Bezhta (with an ergative changing into oblique ‘poss-essive’ case).

98

Andrej Malchukov

(26) Bezhta (Comrie et al., this volume) hini-la yakʼi-ʔ-is unti-urγel holco-qa gisa self.obl-gen2 heart.obl-in-abl disease-sorrow(iii) he.obl-poss out b-aγo-yɬ-aʔa-s iii-take-pot-neg-prs ‘ he was not able to discuss (lit. take out) his sorrow.’) – object-removing/deleting (antipassives) Again, object-removing operations will cover both antipassives proper, with the P demoted to an oblique, and “absolute antipassives” which disallow expression of the object altogether (called “introversive” derivation by Lehmann, this volume). The former type is illustrated for Warrungu (Tsunoda, this volume), the latter for Ainu (Bugaeva, this volume), where the antipassive (i- ‘thing/person’) is also referred to as a “generalized object” marker: (27) Warrungu [bama-O̸] gajarra-nggo ganyji-gali-n. [man-abs] possum-ins carry-antip-nf ‘[The man] brought a possum [for us].’

(Tsunoda, this volume)

(28) Ainu (Bugaeva, this volume) konto kotan kor utar ne yak-ka a=puma-kor-e pa wa then village have pl cop if-even ind.a=wage-have-caus pl and i-rura pa. antip-carry pl ‘Then I paid the villagers and they carried the things.’ Object-incorporation, as illustrated for Yaqui (Estrada-Fernández et al., this volume) is also regarded as a P-deleting operation. (29) Yaqui a. transitive kuta-m wood-pl ‘Cut wood!’

(Estrada-Fernández et al., this volume)

b. intransitive chuk-ta! kuta-chuk-te! cut-tr wood-cut-intr ‘Cut wood!’ (lit. ‘do some wood-cutting!’) Indeed, in some cases the borderline between incorporation and antipassive is not clear (cf. the term ‘object-incorporating antipassive” in Givón 2001). This is also

Valency classes and alternations: parameters of variation

99

true of Yaqui, where an incorporated generic object ‘thing’ functions as a kind of antipassive. (30) Yaqui a. Juan ji’i-bwa Juan thing-eat.prs ‘Juan is eating.’

(Estrada-Fernández et al., this volume)

c. valency-rearranging – subject-object reversal (inverses) While some authors (Mel’čuk 2006) include passives among instances of ‘‘permutation’’, I understand rearrangement more restrictively, as swapping of subject and object position. Some direct-inverse alternations arguably belong to this type, even though for some languages this analysis is controversial (see Appendix 2 of Rhodes & Valentine, this volume, for discussion of complexities in Ojibwe). However, for some languages a syntactic reversal analysis is accepted. Thus, in Mapudungun the inversion analysis seems to hold to the extent that the notion of subject is applicable: Zúñiga (this volume) describes the inverse pattern as one where the Undergoer rather than the Actor is the Primary Argument (roughly, subject). (31) Mapudungun a. (Iñche) pe-fi-n chi machi. 1sg see-3p-1sg.ind art shaman ‘I saw the shaman.’

(Zúñiga, this volume)

b. (Iñche) pe-e-n-mew chi machi. 1sg see-inv-1sg.ind-3a art shaman ‘The shaman saw me.’ Inverse systems generally show limited verb-sensitivity, beyond being restricted to transitive verbs, more specifically to a subset taking animate objects. Thus in Ojibwe, only verbs of the Transitive Animate (TA) class take an alternative inverse pattern. An exception in this respect is Tlapanec (Wichmann 2010), which, apart from the direct/inverse alternation with canonical transitives (manifested in ergative vs. absolutive indexing patterns), features another alternation (dubbed ‘pegative-dative’ by Wichmann) with less canonical transitives (like ‘fear’ and ‘meet’, as well as some ditransitives like ‘give’): (32) Tlapanec a. ni-hm-y-éʔ nenʤ-ó pfv-use-ag-3sg.g>1sg.dat aux-3sg.dat ‘I used him (or: he served me).’

(Wichmann 2010: 661)

100

Andrej Malchukov

b. ni-hm-y-ǔ ʔ nenʤ-òʔ pfv-use-an-1sg.peg>3sg.g aux-1sg.dat ‘She used me.’ Subject-object inversion may be an appropriate analysis for voice (“focus”) alternations in some Austronesian languages such as Tagalog (Bisang 2006), yet, in many languages an alternative analysis is possible. Thus, Shibatani & Artawa (this volume) describe focus alternation in Balinese in terms of remapping between roles and pragmatic (topic) rather than syntactic functions. The ‘symmetric alternation’ in Yorùbá may also be said to represent this pattern. Atoyebi (this volume) describes it as “subject swapping with the object without affecting the meaning of the construction”: (33) Yorùbá (Atoyebi, this volume) a. àárè̩ ń s̩e ìya arúgbó yìí sickness prog do mother old this ‘This old woman is ill.’ (Lit. ‘Sickness is doing the old woman.’) b. ìya arúgbó yìí ń s̩àárè̩ mother old this prog do.sickness ‘This old woman is ill.’ On the other hand, this pattern is also reminiscent of A-incorporation. Typologically, the ‘symmetric alternation’ can be seen as an example of a transimpersonal construction (33a) being reanalyzed into the patient-subject construction (33b) (or, in other words, an experiencer object construction is reanalyzed into the experiencer subject construction). – object-object-rearranging (applicatives) This class covers different kinds of applicative formation resulting in object rearrangement: (34) West Greenlandic a. Niisi aningaasa-nik tuni-vaa. Niisi money-ins.pl give-ind.3s->3s ‘He gave Niisi money.’

(Fortescue 1984: 88)

b. Aningaasa-t Niisi-mut tunni-up-pai. money-pl Niisi-all give-appl-ind.3s->3p ‘He gave money to Niisi.’ As will be discussed later in 4.5, applicatives are often “ambivalent”, that is, they may have not only a valency-increasing or valency-rearranging, but also a valencyreducing effect.

Valency classes and alternations: parameters of variation

101

d. argument-identifying This type is not always distinguished from valency-reducing, but I follow those authors who, like Xrakovskij (1981) and Mel’čuk (2006), treat coreference as an operation separate from reduction or rearrangement (see also Kulikov 2011). These forms include the familiar cases of reflexive and reciprocal marking (both can be expressed by the reflexive marker in Russian; see (52) below). For the sake of consistency, an opposite operation of antireflexive (and antireciprocal) marking should be also theoretically acknowledged. An example of an antireflexive form (these often are polysemous with causatives or applicatives), is provided below for Balinese, where WASH (pandus) is basically intransitive: (35) Balinese (Shibatani & Artawa, this volume) a. Anak=e luh ento mandus (di telaga=ne). person=def female that af.wash in pool=def ‘The girl washed (herself in the pool).’ b. Anak=e luh ento mandus-ang anak cenit ento. person=def female that af.wash-caus person small that ‘The girl washed the child.’ There are other types of alternation which do not fit into the proposed classification. For example, it does not have a natural place for different alternations involving possessors, frequently referred to as Possessor Raising or Possessor Ascension (Apresjan 1969 qualifies such cases as ‘valency split’). Such alternations are frequently restricted by other factors (such as inalienability) which are not of immediate concern for the verb classification. For this reason they will not be addressed unless they are marked on the verb. Within the adopted classification, ‘Possessorto-Subject Raising’ may count as a kind of A-adding operation, and ‘Possessor-toObject Raising’ as a kind of P-adding operation. Now, it is interesting to note that there is some typological evidence for such an analysis insofar as Possessor-toSubject Raising may involve A-adding morphology of the causative type, while Possessor-to-Object Raising may involve P-adding marker of the applicative type. This is the case in Even, where an “adversative passive” (a kind of involitive-permissive causative) is used to introduce a possessor into the A function (note the obligatory reflexive-possessive marking on the direct object in (36b)): (36) Even (Malchukov & Nedjalkov, this volume) a. Huličan bödel-en en-se-n fox feet-ref.pos.sg hurt-nf-3sg ‘The fox’s paws hurt.’ b. Huličan bödel-i ene-le-v-re-n fox feet-ref.pos.sg hurt-inch-advp-nf-3sg ‘The fox’s paws began to hurt. (lit. The fox let her paws hurt).’

102

Andrej Malchukov

Similarly, cases of possessor-to-object raising may count as instances of an objectadding operation; again, not surprisingly, some languages use applicative markers for this function. For example, Hartmann (this volume) notes that in Hoocąk, the Benefactive/Malefactive applicative can signal the possession of an undergoer: (37) Hoocąk (Hartmann, this volume) Ceewasnį wikįnįjara wamąįragicgisną? Ceewasnį_wikįnįja=ra wa-mąącgis=ną cheese=def obj.3pl-cut=pot ’Can you cut the cheeses for me?’ / ’Can you cut all of my cheeses?’ Impersonal constructions are another problematic type of construction. They are sometimes conflated with A-deleting operations, but sometimes treated differently. Illustrations from Central Alaskan Yupik, Russian and Icelandic follow. (38) Central Alaskan Yupik Nanvaq ciku-a. lake.abs.sg freeze-ind.3sg.3sg ‘The lake is frozen.’ (lit. ‘It has frozen the lake.’)

(Miyaoka, this volume)

(39) Russian Lodku uneslo tečeniem boat-acc carry-pst.3.n current.ins ‘The boat was carried (away) by a current.’ (40) Icelandic Bátinn rak að landi. boat-the.acc drove to land ‘The boat drifted to the shore.’

(Barðdal, this volume)

These constructions will be regarded as A-deleting here, although their status will differ across languages, and their interpretation will differ depending on the approach. Thus, in some approaches, impersonal constructions are rather seen as varieties of the basic constructions with a zero subject (this is the approach taken by Mel’čuk 2006, and also by Relational Grammarians, although on different grounds). The latter interpretation is most natural for those languages where impersonal constructions may alternatively be treated as a case of subject ellipsis, as in case of the “quasipassive” construction in Nen (Evans, this volume). On the other hand, in some languages, such constructions can also be analyzed as constructions with non-canonical subjects. This is the usual interpretation of such constructions in Icelandic, where the accusative argument shows syntactic subject properties (Barðdal, this volume). From a diachronic perspective, the controversial status of these constructions is due to the fact that they represent different stages

Valency classes and alternations: parameters of variation

103

of reanalysis of transimpersonal constructions into intransitive constructions with patientive subjects (Malchukov 2008). Clearly, more alternations will eventually have to be distinguished. Thus, we have not had much to say about different subvarieties of causatives and passives, or different varieties of applicatives discussed in the literature. The reason is that our goal is to try to establish semantic classes through the use of syntactic diagnostics, which is more in line with our general approach of trying to establish verb types on syntactic grounds. Another aspect which makes some of the alternations problematic is that they may fulfill different functions when applied to different verb types; this aspect will be addressed under the heading of “voice ambivalence” in 4.4.

3.5 Concluding remarks on coding hierarchies vs. alternation hierarchies Once subclasses of alternations are established, the question arises whether different alternation types show certain preferences with respect to certain verb meanings, and whether hierarchies can be established in this domain. This question, which has not been systematically explored so far in typology (with some notable exceptions such as the inchoative-causative alternation), is addressed in the chapter by Wichmann (this volume), which presents and discusses results of statistical evaluation of the database. In this section we will confine ourselves to general theoretical and methodological considerations. First, let us briefly compare valency classes established by coding frames and alternations. Tsunoda (this volume, and passim) makes the point that preferences for both coding frames and alternations are in effect constrained by the same hierarchy (namely, the transitivity hierarchy). By contrast, Wichmann (this volume; cf. Wichmann, forthcoming) notes that such hierarchies are different for different alternations, and generally the established voice hierarchies offer limited support for Tsunoda’s hierarchy (i.e., in many cases they cannot be described as extensions from the most transitive type (Effective Action Verbs) to other verb types on the Transitivity Hierarchy). Surprisingly, Wichmann (this volume, and forthcoming) also finds a (statistically significant) correlation between a number of hierarchies. Now, I would like to argue that the results are not as confusing as they might appear at first sight. First, it is important to note that ultimately transitivity might be behind the correlation detected by Wichmann: importantly, all the alternations which show correlation with respect to the verb class (anticausative, passive, reflexive, reciprocal) are detransitivizing. By contrast, causatives do not show a correlation with other operations, as they are valency-increasing rather than valencydescreasing. Actually one would rather expect an inverse correlation here, which is also confirmed in Wichmann’s study (this volume). Second, while Tsunoda’s hierarchy is sensitive to “semantic transitivity” (transitivity prototype), many alter-

104

Andrej Malchukov

nations may be sensitive rather to syntactic transitivity: that is, the application of antipassives would normally be restricted to transitives (but see (58) from Bezhta for an exception). Thus, while coding patterns (coding frames) may be determined by semantic transitivity directly, alternations may rather show dependency in an indirect way mediated through coding (in particular, through formal transitivity).7 Third, some alternations may be insensitive to transitivity distinctions, as they are determined by some other shared meaning components. For example, in Ainu (Bugaeva, this volume), the locative applicative (o-applicative) regularly applies to verbs specifying motion, whether they are transitive (i.e. a caused motion verb) or intransitive. Of course, some alternations may still be sensitive to semantic transitivity, as discussed by Tsunoda (this volume). For example, in Even (Malchukov & Nedjalkov, this volume) application of resultative passive is restricted to just a subset of transitive verbs that correspond to core transitives (“Effective Action Verbs” in Tsunoda’s terms). Clearly this restriction is determined semantically; yet, such effects may either support or be in conflict with the Transitivity Hierarchy, depending on the function of individual constructions. For this reason, not all alternations can be naturally seen as constrained by the Transitivity Hierarchy, rather their applicability is constrained by the function of the individual alternation. By way of illustration consider the Reflexive hierarchy from Wichmann (this volume) (the Guttman Scale is derived from the number of verbs with a matching behavior across languages; the verbs at the top of the scale show the strongest preference for occurrence in a reflexive alternation): (41) Hierarchy for Reflexives WASH, COVER, SHAVE, SHOW, CUT, SEE, HIDE, DRESS, GIVE, TOUCH > LOOK AT, HEAR, PUT, BEAT, HUG, SMELL, TIE, THROW, HIT, KILL, LIKE, FEAR, WIPE > KNOW, PUSH, ASK FOR, TEAR, NAME, HELP > SEARCH FOR, THINK, TEACH, TAKE, SAY, CARRY, TELL, BREAK, SEND > FRIGHTEN, TALK, LOAD > BUILD, STEAL > BRING, PEEL, COOK, FOLLOW, EAT > FILL, MEET, GRIND, SING, BURN, DIG, BE SAD, POUR, ROLL > SHOUT AT, BE DRY, SCREAM, LAUGH, RUN, PLAY, FEEL PAIN, LEAVE, GO > JUMP, SIT, BLINK, BOIL, BE A HUNTER > LIVE, RAIN, SINK, BE HUNGRY, DIE, FEEL COLD, CLIMB > SIT DOWN > COUGH. (Wichmann, this volume; cf. Wichmann forthcoming)

7 The impact of coding on alternations is clearest in languages like Modern Standard Arabic (Kász, this volume), where certain types of alternations are constrained by a morphological pattern (so that for stem I verbs a passive alternation is signaled by stem VII, while for stem II verbs it is signaled by stem V).

Valency classes and alternations: parameters of variation

105

Even a cursory look reveals that this scale follows a certain pattern: it starts from “natural reflexives” (grooming verbs) at the top of the scale, extending first to bivalent verbs with animate objects, then to other bivalent verbs. Expectedly, monovalent verbs, which resist reflexive marking for functional reasons, are at the bottom of the hierarchy. In other words, the hierarchy above shows that the reflexive voice is first grammaticalized with grooming verbs, which normally involve coreferentiality of agent and patient, and may be extended to other verb classes only at a later stage. Importantly, when the reflexive marker is extended beyond its natural domain of application (verbs taking animate A and P arguments) it may change its function (for example, to anticausative; see below). Alternatively, it may be restricted in its application to grooming verbs. Admittedly, the case of grooming verbs might not count as a counterexample to Tsunoda’s Transitivity Hierarchy, which does not include grooming verbs, but generally, it is widely agreed that grooming verbs do not count as canonical transitives, which require distinctness of A and P arguments (Kemmer 1993; Næss 2007). Similarly, antipassive marking need not be associated with the verbs conforming to the transitivity prototype as defined by Hopper & Thompson and Tsunoda (cf. Dixon & Aikhenvald 2000). Indeed, antipassive marking and similar P-demoting/deleting operations are preferentially found with verbs of impact like HIT rather than with “effective action verbs” like BREAK (recall the similar restriction in the conative alternation in English: cf. hit at vs. *break at). This pattern is actually predicted by Levin’s (this volume, and passim) generalization to the effect that manner-verbs (i.e., verbs specifying the manner in which an event is implemented) allow for object manipulation (object-omission or object rearrangement) more readily than result verbs (i.e., verbs specifying the result of the action).8 This generalization, illustrated by Levin (this volume) for a number of languages, is also confirmed by our data and results in a higher ranking of manner verbs on the Pdemoting/deleting hierarchy in (42). (42) Hierarchy for Object-demoting/deleting alternations EAT, WASH, GIVE > STEAL, TEACH, SHAVE, COOK > CUT, WIPE, SEARCH FOR, HIT > KILL, ASK FOR, TAKE, BEAT > SEE, THROW, HEAR, TOUCH, LOOK AT > GRIND, BREAK, FILL, HUG, COVER, POUR, THINK, LOAD > TELL, KNOW, TEAR, HELP, TIE, SHOW, CARRY > SING, DIG, DRESS > CLIMB, BUILD, FEAR > SMELL, PUSH, PUT, SEND, LEAVE > PEEL, BLINK, SAY, TALK, SHOUT AT, NAME, RUN > JUMP, HIDE, FRIGHTEN, LIKE, PLAY, FOLLOW, LIVE, BE DRY >

8 This view also resonates with the approach of Role and Reference Grammar (Van Valin 2005 et passim), which assumes that activity verbs are intrinsically intransitive (that is, do not select for an Undergoer argument). Note that all activity verbs belong to the manner rather than the result class, although the opposite does not hold (that is, not all manner verbs belong to the activity class).

106

Andrej Malchukov

BRING, ROLL, LAUGH, BURN, SCREAM, GO, SINK > MEET, DIE, COUGH, BOIL, BE A HUNTER > FEEL PAIN, SIT > BE SAD > SIT DOWN, BE HUNGRY > RAIN > FEEL COLD. (Wichmann, this volume; cf. Wichmann forthcoming) On balance, the typical manner verbs (like WASH, WIPE, HIT, COOK) appear higher on the hierarchy in (42) than typical result-verbs (such as FILL, BREAK, KILL and BUILD), although some exceptions are also found (note in particular the high status of EAT, arguably a result verb). Thus, Italian (Cennamo, this volume), generally allows P-omissions with activities (manner verbs), but such omission is restricted with “resultative” verbs (result verbs), especially those having an inanimate P. Also the unmarked conative, locative and “applicative” alternations resulting in O-demotion/rearrangement are restricted to activities (manner verbs). Furthermore, some languages like Ainu (Bugaeva, this volume) preferentially use antipassive (or else feature deponent antipassives) for less transitive verb types in the mid-range of the hierarchy. Bugaeva (this volume) notes that antipassive is found with grooming verbs, mental verbs (emotion, perception) and interaction verbs, but curiously, not with notional transitives (Tsunoda’s Effective Action Verbs). This finding can be also attributed to the restriction on the use of result verbs in object-deleting alternations, as familiar from other languages (Levin, this volume). Yet, as is also clear for the hierarchy, the manner/result dichotomy is hardly the only factor which can account for the whole hierarchy. Clearly, other factors are also at play, which would for example account for why EAT and not HIT is at the top of the object-deletion hierarchy For example, in Hoocąk (Hartmann, this volume), EAT is the only verb taking the antipassive marker. In both cases such restrictions can be explained with reference to the basic function of an alternation and its natural domain of application (derivable from its function). We will have more to say about this in § 4, dealing with the phenomenon of ambivalence of voice markers. The same holds for other alternations. For example, on the Causative Hierarchy (Wichmann, this volume), intransitive verbs generally rank higher than transitives; yet a few intransitives like RAIN and BE_HUNTER appear at the very bottom of the hierarchy. This is of course hardly surprising, given that ‘causing to rain’ is a highly unusual situation (apart maybe from shamanistic practices), while BE_HUNTER in many languages is a nominal rather than a verbal expression, and for this reason can resist causativization. The above observations suggest that predictions concerning preferential coding of certain verb types for certain alternations (their ranking on individual alternation hierarchies) is determined by the following two factors: – coding pattern of the verb (in particular, its transitivity value) – function of a particular alternation (which may be more compatible with notional transitives, but need not be)

Valency classes and alternations: parameters of variation

107

The ranking (relative importance) of the two factors will differ somewhat for different (voice) alternations; thus, the role of coding (transitivity) is clearly a decisive constraint on formation of antipassives, less so for passives (in view of demotional/ impersonal passives of intransitives), and still less for causatives (as discussed below, many causative forms across languages allow causatives of transitive verbs, even though some are more restrictive in this respect). There may be other factors interfering with the hierarchies, and introducing “noise” in them (conflicting rankings, which might not be easily detectable in a Guttman Scale, but are more evident on NeighborNet plots; see Wichmann, this volume). One of these factors may be referred as blocking, where a certain pattern is blocked due to the availability of another (generally less marked pattern). One case of blocking relates to cases where one meaning of a particular voice form blocks another interpretation; thus, in Russian, reflexive forms with reflexive or (lexicalized) anticausative meaning are often awkward or even impossible in the passive use (thus, myt’-sja ‘wash (oneself)’ or pugat’-sja ‘be frightened’ can hardly be used in a passive construction with an instrumental agent). A somewhat different case of blocking can be illustrated for Even. In Even (Malchukov & Nedjalkov, this volume) the mediopassive marker -b- with a basic anticausative function applies regularly to different classes of transitive verbs, but does not readily apply to verbs of destruction like BREAK. For these verbs the mediopassive form is marginal, because they take part in equipollent alternations which provide a regular way of expressing the anticausative function (cf. čelge-l- ‘break (pfv.tr)’ vs. čelge-r‘break/get broken (pfv.intr.)). Such instances of ‘morphological blocking’ may introduce “noise” in the maps, but generally their effect is restricted, since here (as elsewhere) the effect of blocking is not absolute, but rather may lead to a different interpretation of a competing pattern (see Kiparsky 1982). Thus, availability of unmarked anticausative (i.e. S=P ambitransitive) with some transitive verbs in Italian does not exclude the possibility of deriving a reflexive si-form with an anticausative function; rather the two forms show aspectual differentiation (Cennamo, this volume). Yet another case of blocking is attested in Ainu (Bugaeva, this volume), where certain verb meanings do not allow specific voice alternations (applicative, antipassive) while semantically related verbs do allow them. As noted by Bugaeva (this volume), this has a structural explanation: in particular, it pertains to deponent verbs which already include a corresponding valency marker (applicative, antipassive) and therefore do not allow a further application of the same operation which would have yielded doubling of the respective markers. This gives an appearance that certain verb meanings belong syntactically to the wrong valency class, but such discrepancies can be attributed to structural factors interfering with the functionally based constraints.

108

Andrej Malchukov

4 Voice ambivalence and ambitransitivity 4.1 Ambitransitivity and alternation hierarchies Above we considered preferences for certain voice alternations as manifested in preferential coding of certain verbs for certain voice markers (passive, reflexive, causative and the like). Another aspect of such hierarchies that has not been systematically investigated within the Leipzig Valency Classes Project, but which must be briefly discussed is a preference resulting in preferential interpretation rather than coding. Such preferences can be fruitfully explored for polyfunctional markers that display certain recurrent patterns of voice ambivalence, as discussed in § 4.3 below. Yet perhaps most clearly these preferences are revealed by unmarked alternations which have often been discussed in terms of ambitransitivity and lability (see § 4.2). These alternations reveal verb sensitivity in the clearest way, even though parallel effects can be found with polysemous voice markers as well.

4.2 Classes of ambitransitive verbs In the literature dealing with ambitransitivity (lability) it is a common observation that this alternation works differently for different classes of verbs, as exemplified by familiar examples from English: (43) English a. I break the stick a′. the stick breaks b. I ate the meal b′. I ate In these examples, the ‘break’ verb shows a pattern of S/P lability (ambitransitivity), while the ‘eat’ verb shows A/S lability. As shown in the typological literature (Haspelmath 1993; Kazenin 1994; Dixon & Aikhenvald 2000), this distinctive behavior is not accidental and has been reported for other languages and other types of transitivity alternations. The following examples illustrate this pattern for Central Alaskan Yupik, where the alternation manifests itself in the choice of the ergative vs. intransitive construction (with a concomitant choice of transitive vs. intransitive agreement on the verb): (44) Central Alaskan Yupik a. Angute-m kuvya-ni allg-aa. man-rel.sg net-abs.3rsg.sg tear-ind.3sg.3sg ‘The man tears/tore his (own) net.’

(Miyaoka, this volume)

Valency classes and alternations: parameters of variation

109

b. Kuvya-a alleg-tuq. net-abs.3sg.sg tear-ind.3sg (i) ‘His net tears/tore (by itself).’ (ii) ‘His net was torn (by someone).’

(45) Central Alaskan Yupik a. Angute-m neqa ner-aa. man-rel.sg fish.abs.sg eat-ind.3sg.3sg ‘The man is eating the fish.’

(Miyaoka, this volume)

b. Angun neq-mek ner’-uq. man.abs.sg fish-abm.sg eat-ind.3sg ‘The man is eating a fish.’ There is a consensus in the literature that this distinctive behavior of the two verb types is motivated semantically: in particular, the verbs appearing in an S/P alternation are Patient-oriented (specifying an effective change on a patient), while the verbs appearing in an S/A alternation do not imply a change in the patient (or, as in case of EAT-verbs, change in the agent is of more central interest than the change in the patient). Interestingly, the distinction between the two types of transitive verbs is mirrored by a corresponding distinction of two types of intransitive verbs, which are frequently characterized as “unaccusative” vs. “unergative” verbs or, in more neutral terms, as patientive vs. agentive intransitives (Perlmutter 1978; Van Valin 1990; Levin & Rappoport Hovav 1995; Wunderlich 2006). This is of course a welcome result, given that in the case of unmarked alternations the choice of one use (transitive or intransitive) as basic is often contentious (see Levin & Rappoport Hovav 1995, and Wunderlich 2006 for discussion).9 The distinction is overtly marked (through distinct coding frames) in languages with “split-intransitivity” (aka semantic alignment; Donohue & Wichmann 2008) such as Hoocąk (Hartmann, this volume); in other languages it is manifested syntactically, through alternations (see Cennamo, this volume, on Italian). Furthermore, as shown in the typological literature (Kazenin 1994; Dixon & Aikhenvald 2000), the same distinction between patient-oriented and agent-oriented verbs can also be detected for valency-decreasing and valency-increasing alter-

9 Wunderlich (2006) suggests some aspectual diagnostics for distinguishing between transitivebased ambitransitives (like break) and intransitive-based ambitransitives (like dry). In other languages, other language-particular diagnostics may be useful. Thus in Salish (Watanabe, this volume), apart from classes of canonical intransitive and transitive verbs, two classes may be treated as ambitransitive (classes B and C in Watanabe’s classification). Of these two classes, class C can be argued to be transitive-based, as it resists causativization, as do canonical transitives.

110

Andrej Malchukov

nations. As shown later, a detransitivizing operation may have an anticausative effect with BREAK-verbs, and an antipassive effect with EAT-verbs. Conversely, a valency-increasing operation may have a causative effect with unaccusative intransitives, and an applicative effect with unergative intransitives. In terms of Kazenin (1994), alternations involving patient-oriented verbs of the BREAK-type are Patientpreserving, while alternations of the agent-oriented verbs like EAT are Agent-preserving. The distinction between two types of verbs is similar, if not identical, to the distinction of result vs. manner verbs introduced by Levin & Rappoport Hovav (1995; see also Levin, this volume, for discussion and references). In their work result verbs are those verbs that predicate a scalar change of the (patient) argument; hence the requirement that a patient argument of which the scalar change is predicated is overtly expressed. On the other hand, manner verbs do not imply a scalar change on the part of the patient exclusively (or some, like EAT, may additionally imply a scalar change on the part of the agent, as well); instead they include a manner component predicated of an agent argument. For that reason result verbs disallow P-alternations (are Patient-preserving, in Kazenin’s terms) but may allow A-alternation (like inchoative-causative alternations). By contrast, manner verbs disallow A-alternations (are A-preserving, in Kazenin’s terms), but may allow for P-alternations (see Levin, this volume, for extensive arguments in support of this latter claim).10 The popular distinction between unaccusative vs. unergative verbs and their transitive counterparts is just one case of what Levin (this volume, and passim) calls verb sensitivity in argument alternations: the same alternation performs different functions when applied to different verb types. Thus, apart from S=O and S=A ambitransitives, it is appropriate to distinguish a third “anaphoric” type (A=O) of ambitransitivity, less discussed in typological literature (but see Letuchiy (2009) on the typology of labile verbs). The latter can be illustrated by the familiar English examples of reflexive and reciprocal lability. (46) English a. I wash my hands vs. a′. I wash b. I met my friend vs. b′. we met

10 As noted by Levin (this volume), this correlation is not perfect given that not all result verbs allow for an agent-alternation (inchoative-causative alternation); she cites put and create as examples of result verbs that do not show S/P ambitransitivity. However, in this as well as other cases, one should not discard the possibility of extra conditions governing ambitransitivity (such as allowing for inanimate causer) that restrict its applicability for particular languages, nor analogical extensions from certain verb types to other similar verbs (cf. the examples from Mandinka in (49)). This granted, it appears that the preferred pattern is that a) manner verbs allow for P-alternations and disallow for A-alternation (are A-preserving), while b) result verbs allow for A-alternations and disallow for P-alternation (are P-preserving).

Valency classes and alternations: parameters of variation

111

Again, this verb-sensitivity is functionally based and recurrent cross-linguistically: the reflexive type is preferentially found with grooming verbs as well as some verbs of emotion (Kemmer 1993), while reciprocal lability is preferentially found with “natural reciprocals” (symmetric predicates; Nedjalkov 2007). Reflexive and reciprocal lability is illustrated below for Xârâcùù (Moyse-Faurie, this volume): (47) Xârâcùù a. Nâ xii è. 1sg shave 3sg ‘I am shaving him.’

(Moyse-Faurie, this volume)

b. Nâ xii. 1sg shave ‘I am shaving.’ c. Ri tôôbùtù ri. 3pl assemble 3pl ‘They are assembling them.’ d. Ri tôôbùtù. 3pl assemble ‘They are assembling.’ The second qualification is that although ambitransitivity is verb-sensitive it may extend from its core domain to other domains. Thus, Mandinka features a productive S=O alternation manifested in the use of intransitive vs. transitive versions of TAM markers. Expectedly, this alternation is found with patient-oriented transitives like BREAK, and is correspondingly dubbed the anticausative alternation by Creissels (this volume): (48) Mandinka anticausative alternation a. Máŋk-óo jolón-tá baŋk-óo to. mango-def fall-pf.pos.intr ground-def loc ‘The mango fell on the ground.’

(Creissels, this volume)

b. Kew-ó ye mur-óo jolóŋ baŋk-óo to. man-def pf.pos.tr knife-def drop ground-def loc ‘The man dropped the knife on the ground.’ Yet, the same S=O alternation is also extended to other verb types, including those whose meaning implies an agentive causer and thus resists a spontaneous interpretation in the intransitive use. For this reason, Creissels (this volume) qualifies it as an ‘unmarked passive’ rather than an ambitransitive anticausative alternation; yet formally we are dealing with the same alternation:

112

Andrej Malchukov

(49) Mandinka “unmarked passive” a. Kew-ó ye wot-ôo dádaa. man-def pf.pos.tr car-def repair ‘The man has repaired the car.’

(Creissels, this volume)

b. Wot-ôo dádáa-ta. car-def repair-pf.pos.intr ‘The car has been repaired.’ The possibility of extension of a certain alternation type beyond its core domain seems to depend on the availability of a dedicated construction (a dedicated voice marker but also a special construction associated with this function). This is of course most evident with voice markers, which when productive may be used with any verb provided that a plausible context is available. Yet, the same point can be made for unmarked alternations as well. Thus, as noted above, Central Alaskan Yupik has one main detransitivizing construction, which is, as expected, verb-sensitive. Chintang, by contrast, features two different detransitivizing constructions. One is a regular alternation of transitive and intransitive patterns reminiscent of Central Alaskan Yupik: in both cases an ergative intransitive alternation is manifested in both case and agreement. This alternation is used for S=O ambitransitivity; the A argument is, as expected, deleted in (50b): (50) Chintang (Schikowski et al., this volume) a. Sa-ŋa u-lett-o-kha phuŋ? who-erg 3[p]a-plant-3[s]p-nmlz2 flower ‘Who planted the flower?’ b. Makkai-ce u-lett-a-ŋs-e. maize-ns 3[p]s-plant-pst-prf-ind.pst ‘The maize plants have been planted.’ There is, however, another detransitivizing construction in Chintang that is used for S=A detransitivization. This construction illustrated in (51b) is intransitive (as far as A-marking and verbal agreement is concerned) but the P is retained. This pattern, which is reminiscent of analytic incorporation or an antipassive construction, is used when the object is non-referential. (51) Chintang (Schikowski et al., this volume) a. Debi-ŋa seu kond-o-ko. Debi-erg apple look.for-3[s]p-ind.npst[.3sa] ‘Debi is looking for an apple.’

Valency classes and alternations: parameters of variation

113

b. Debi seu kon-no. Debi apple look.for-ind.npst[.3ss] ‘Debi is looking for apples.’ Both S=O ambitransitivity, and especially S=A ambitransitivity are fairly productive processes in Chintang that extend far beyond patientive verbs (like BREAK) or agentive verbs (like EAT), although certain residual preferences remain. Here the extension is apparently facilitated by availability of two different detransitivizing constructions, without the danger of ambiguity. If, on the other hand, a language features only one general detransitivizing construction, more verb-sensitivity is expected, as we observed for Central Alaskan Yupik.

4.3 Markedness effects in uncoded and verb-coded alternations As noted by Kazenin (1994) and Dixon & Aikhenvald (2000), the distinction between the two types of verbs called agent-oriented vs. patient-oriented above holds irrespective of the direction of alternation. That is, similar effects of verb-sensitivity can be detected with polyfunctional voice markers, both valency-decreasing and valency-increasing. A familiar example from Russian can serve as an illustration. As is well-known, Russian features a special reflexive voice marker (the postfix -sja/s’) which appears in different voice alternations. As expected, it shows verbsensitivity. With some verbs (grooming and emotion verbs) it has a reflexive meaning (52a), with natural reciprocals and some semantically related verbs it has reciprocal meaning (52b), with perfective verbs allowing for spontaneous interpretation it has anticausative meaning (52c), with many transitive verbs (in imperfective aspect) it allows for passive interpretation (52d), while with a few verbs (of negative impact) it has antipassive meaning (52e): (52) Russian a. On mo-et rebenk-a. he wash-prs.3sg child-acc ‘He washes the child.’ a′. On mo-et-sja. he wash-prs.3sg-refl ‘He washes.’ b. On vstreča-et drug-a. he meet-prs.3sg friend-acc ‘He meets a friend.’ b′. Oni vstrečaj-ut-sja. they meet-prs.3pl-refl ‘They meet.’

114

Andrej Malchukov

c. On sloma-l palk-u. he break-pst.ms stick-acc ‘He broke the stick.’ c′. Palka sloma-la-s.’ stick break-pst.fm-refl ‘The stick broke.’ d. Rabočie stroj-at dom. construction_worker.pl build-prs.3pl house.acc ‘The construction workers build a house.’ d′. Dom stro-it-sja raboči-mi. house build-prs.3sg-refl construction_worker-pl.ins ‘The house is (being) built by construction workers.’ e. Sobaka kusa-et mal’čik-a. dog bite-prs.3sg boy-acc ‘The dog bites a boy.’ e′. Sobaka kusa-et-sja. dog bite-prs.3sg-refl ‘The dog bites.’ Of course the pattern of reflexive marking in Russian is much more complicated and the realization of particular meanings is subject to further restrictions (see Geniušienė 1987; Timberlake 2004 for discussion of complexities), still the general tendency is clear and falls into the expected typological pattern. Only verbs of consumption stand out in that they do not readily allow for antipassive meaning; still such pairs can be found even if they are not perfect examples, as they involve different (directional/aspectual) prefixes (Timberlake’s 2004 ‘quantifying reflexives’): f. On s”-e-l sup. he pfv-eat-pst.m soup.acc ‘He ate soup.’ f′. On na-e-l-sja. he pfv-eat-pst.ms-refl ‘He ate his fill.’ To conclude, polyfunctional voice markers display the same patterns of verb sensitivity as found with unmarked alternations; furthermore for such cases it is often easier to determine the basic function. Thus, in Even the ‘medio-passive’ marker -b- (exemplified above) is used in anticausative or “decausative” function (i.e.

Valency classes and alternations: parameters of variation

115

agentless passive use as in (25)), but with a few verbs like CUT it can acquire reflexive meaning. In Yaqui, the -man Reflexive has a regular reflexive interpretation with grooming verbs, but an anticausative interpretation with verbs like OPEN. For Kiche, Campbell (2000) notes that the antipassive, which has a regular use, has an anticausative effect with a few verbs like BREAK. And in Bezhta, antipassive, which is also regular, unexpectedly has a reflexive interpretation with a few verbs like WASH (and is for that reason called ‘‘Antipassive-2’’ by Comrie et al., this volume). Such cases are treated here as reinterpretations of a certain marker as it spreads into another domain. Similar reinterpretations will be described below in terms of voice ambivalence. Similar effects will be observed for other formal types of alternations, both directed and non-directed. Thus, in Japanese the equipollent transitivity alternation (which is admittedly difficult to differentiate from the detransitivizing type) is preferentially used in an anticausative function (with verbs like BREAK), but has a passive (“decausative” in Kishimoto et al.’s terms) function when extended to verbs of creation like BUILD. Still more instructive in that respect is Modern Standard Arabic (Kász, this volume), which features several equipollent valency alternations manifested through stem alternations (indicated by roman numbers). Here again we can detect certain core functions behind individual stem alternations. For example, the I->VII-alternation is most productive in a passive meaning but may be extended to anticausative function with BREAK-verbs, and the II->V-alternation seems most productive in the anticausative/middle function, but can be extended to reflexive for verbs like COVER, which have been shown to prefer reflexive meaning elsewhere (recall that COVER is high on the reflexive hierarchy in Wichmann, this volume). Finally, the same verb sensitivity is also attested if we turn to valency increasing operations. This manifests itself in the patterns of causative-applicative polysemy as observed in the literature (Dixon & Aikhenvald 2000; Shibatani & Pardeshi 2002). An example from Balinese (Austronesian; Shibatani & Artawa, this volume) may serve as an illustration. In Balinese, the same -ang marker has both causative and applicative uses. When applied to intransitives (inactive/stative predicates) its meaning is causative: (53) Balinese a. Celeng=e mati. pig=def pf.dead ‘The pig is dead.’

(Shibatani & Artawa, this volume)

b. Anak=e ento nge-mati-ang celeng=e. person=def that af-dead-caus pig=def ‘The man killed the pig.’ When the -ang marker is applied to transitives the meaning is that of the instrumental applicative (with a theme/instrument promoted to the main object):

116

Andrej Malchukov

(54) Balinese (Shibatani & Artawa, this volume) a. Ia nyikut natah=e aji tungked. s/he af.measure yard=def with stick ‘S/he measured the yard with a stick.’ b. Ia nyikut-ang tungked ka natah=e. s/he af.measure-caus stick to yard=def ‘S/he used a stick to measure the yard.’ And, as noted earlier, with grooming verbs (like WASH) the same marker performs rather an “antireflexive” function (see (35) above). Such cases would be discussed in more detail in § 4.4 dealing with voice ambivalence, which studies behavior of voice markers when they are extended beyond their natural domain of application.

4.4 Voice ambivalence: an overview Above we have considered alternation hierarchies constrained by verb types. In 4.3 we showed that similar hierarchies are at work in the ambitransitive alternation, where we are dealing with preferences in interpretation rather than preferences in coding. Finally, it has been observed (and illustrated with reflexive-marking) that similar hierarchies are found for polysemous voice-markers. In this section we will address these issues in more detail – that is, study how voice markers shift their function depending on the verb type to which they apply. This verb-sensitivity has already been illustrated for the case of the polyfunctional -sja marker in Russian, which has reflexive, anticausative, passive, reciprocal and antipassive functions (see (52)). While such uses are mostly distinguishable in syntactic terms (i.e. in terms of ‘diathesis’, correspondence of the basic and derived frames), they all share the feature of detransitivization. In what follows we will see that voice ambivalence is pervasive and can manifest itself in different, sometimes opposite, effects when applied to different verb types. On the approach adopted certain voice markers have a natural domain of application, which is motivated by the semantics of a particular alternation. For reflexives the natural domain of application is the group of bivalent verbs (preferentially transitives) which take (or at least allow) animate A and P arguments. If a reflexive marker is monofunctional it may be restricted (grammaticalized) with just these verbs, and not extended beyond this domain (to bivalent intransitives, or transitive verbs with inanimate objects). These languages will yield straightforward support for the hierarchy. Yet, in case a marker is polyfunctional, it extends across the verbal lexicon, and is reinterpreted along the way. From this perspective polysemy arises from the extension of the voice marker through the lexicon, where a verb’s meaning may favor a certain reinterpretation most natural for this verb type (for example, anticausative interpretation with transitives taking inanimate objects). In this section we will briefly discuss patterns of ‘voice ambivalence’, i.e.,

Valency classes and alternations: parameters of variation

117

the different interpretation of voice markers dependent on valency type of the verb to which they are applied (see Malchukov, forthcoming, for further discussion). The phenomenon of voice ambivalence is clearly related to voice polysemy, which has been extensively studied in the typological literature in connection with individual voice categories (see, e.g., Xolodovich 1969; Comrie 1975, 1976; Shibatani (ed.) 1976; Dixon 2000; Shibatani & Pardeshi 2002 on causatives; Xrakovskij 1974; Siewierska 1984; Shibatani 1985 on passives; Geniušienė 1987; Klaiman 1991; Kemmer 1993 on reflexives and middles; Peterson 2007 on applicatives; Nedjalkov 2007 on reciprocals). We will not attempt to review this literature, but note at this point that our focus is narrower, since we a) focus in the first place on polysemy resulting in distinct syntactic behavior and b) focus on this polysemy resulting from application to different verb types. This approach, even though somewhat restrictive in scope, allows us to bring a new perspective to patterns of voice polysemy described in terms of extensions of voice forms (or broadly, constructions) to different verb types, which leads to their reinterpretation.

4.4.1 Valency-decreasing markers with intransitive verbs In a language which has a valency-decreasing (detransitivizing) marker such as passive, antipassive or (verbal) reflexive, this marker routinely applies to transitive verbs (maybe subject to certain restrictions). Yet languages differ widely, as concerns extension of this operation to intransitive verbs (see, e.g., Keenan 1985). In some languages, an operation like passive or antipassive may be extended to intransitives. For passives, for example, it yields impersonal passives as familiar from European languages such as German. In other languages, however, passives may be restricted to transitive verbs (as in English). Finally, in some other languages the valency marker is extended but changes its function. Thus, for example, in some languages passives develop a modal function in the context of intransitives. Geniušienė (1987) provides the following example of the “modal deagentive” construction from Russian involving the use of the reflexive-passive form: (55) Russian Mne ne spit-sja. me.dat not sleep.prs.sg-pass ‘I cannot sleep.’ Reanalysis of the passive marker to a modal marker is more advanced when the diathesis does not change, as is the case of potential uses of the passive in Japanese: (56) Japanese Boku wa nemur-are-nakat-ta. I top sleep-pass-neg-pst ‘I could not sleep.’

(Shibatani 1985: 822–823)

118

Andrej Malchukov

Kishimoto et al. (this volume) the ‘potential passive’ does not affect the valency with intransitive verbs, while with transitives the potential form is either valencypreserving, or valency-decreasing (in the latter case being closer to a conventional passive). Similar reanalyses are observed for other valency-decreasing markers. Thus, in Bora, the reflexive marker (which may also perform a passive function with transitives), may only be used in modal (“attempted action”) function with intransitives: (57) Bora a. Wajpi tsájtyé-meí-hi. man carry-refl-pred ‘The man carried himself.’

(Seifart, this volume)

b. Wajpi dsɨɨné-meí-hi man run-refl-pred ‘The man tried to run.’ Now, modal use without the change of diathesis is also possible for transitive verbs, but for the latter it is rather marginal, while for intransitives it is the only use of this marker. Another interesting case of ambivalence is illustrated by antipassive markers in Bezhta (Comrie et al., this volume). When used with transitive verbs the Bezhta antipassive behaves like a garden variety of antipassives: the ergative construction changes to intransitive with the P argument marked by instrumental case: (58) Bezhta a. Öždi bäbä m-üq-čä. boy.obl(erg) bread(iii) iii-eat-prs ‘The boy eats the bread.’

(Comrie et al., this volume)

b. Öžö bäbälä-d O̸ -üⁿq-dǟ-š. boy(I) bread.obl-ins I-eat-antip-prs ‘The boy is busy eating the bread.’ Now, apart from diathetic effect, the antipassive has an aspectual function, expressing a durative or iterated event (which is a typical correlate of antipassive meaning also in other languages; Tchekhoff 1987; Cooreman 1994). Interestingly, the Bezhta antipassive can also apply to intransitive verbs. In this case, it does not affect verbal valency, and its contribution is reduced to its aspectual value: (59) Bezhta a. Öžö O̸ -ogic’-iyo. boy(I) I-jump-pst ‘The boy jumped once.’

(Comrie et al., this volume)

Valency classes and alternations: parameters of variation

119

b. Öžö O̸ -ogiyac-ca. boy(I) I-jump.antip-prs ‘The boy jumps many times.’ This raises a legitimate question, namely whether ‘antipassive’ is an appropriate designation for the category in the latter use (see Comrie et al., for discussion).11 In the present context, it is important to recognize that Bezhta falls into the familiar pattern insofar as the “infelicitous” application of antipassive marker to intransitives is concerned: if the antipassive marker is extended to intransitives, its meaning shifts.

4.4.2 Valency increasing markers based on transitives While passives preferentially apply to transitive verbs, causatives are known (at least since Nedjalkov 1964) to have a preference for intransitives. Admittedly, some languages do not show such restrictions: a causative of transitive is regularly used to derive ditransitives from transitives (in some languages all derived ditransitives are causatives). However, other languages (such as Ket) do not allow causatives of transitives at all. Furthermore, Yucatec Maya (Lehmann, this volume) and Chatino (Campbell, this volume) restrict the use of causatives to inactive intransitives. Still other languages may allow a causative form of transitives, but its function is different (cf. Kittilä 2009). One possibility is that a causative marker develops an applicative meaning in the context of transitives (Shibatani 2002, Peterson 2007). This polysemy pattern, which has already been illustrated above (in (53)–(54)) for Balinese, seems actually to be more wide-spread cross-linguistically than reported, as it may be masked by differences in terminology. Consider a case of Central Alaskan Yupik (Miyaoka, this volume), which features a peculiar category of “adversative”. When the adversative is used with intransitives, the (negatively) affected participant becomes an A: (60) Central Alaskan Yupik a. Kicaq kit’-uq. anchor.abs.sg sink-ind.3sg ‘The anchor sank.’

(Miyaoka, this volume)

b. Kic-i-aqa kicaq. sink-adv-ind.1sg.3sg anchor.abs.sg ‘I had the anchor sunk (me negatively affected).’

11 In Jaminjung, we have a similar pattern, described as a “telicity alternation” by Schultze-Berndt (this volume): the use of the imperfective forms excludes ergative marking on the A and thus affects verbal valency with transitive, but not with intransitive verbs.

120

Andrej Malchukov

In this function it comes close to permissive causatives or adversative passives, as illustrated for Japanese (see (63) below) and Even (in (23) above). However, when derived from transitives, the affected participant appears as a P (see the 1st person object agreement with the experiencer in (61b)): (61) Central Alaskan Yupik a. Neqe-m neqcaq ner-aa. fish-rel.sg bait.abs.sg eat-ind.3sg.3sg ‘The fish ate the bait.’

(Miyaoka, this volume)

b. Ner-i-anga neqe-m neqca-mnek. eat-adv-ind.3sg.1sg fish-rel.sg bait-abm.1sg.sg ‘The fish ate my bait (on me).’ Thus, the adversative category has the function of the ‘adversative causative’ (or adversative passive) when derived from intransitives, but of ‘adversative (malefactive) applicative’ when derived from transitives. A similar pattern is found in Mapudungun, where the valency-increasing marker -l functions as a causative (Aadding) marker with monovalent verbs and as an applicative (O-adding) marker with both monovalent and bivalent verbs (Golluscio 2010; Zúñiga, this volume). More discussion of the causative-applicative polysemy is provided by Austin (1997), Shibatani & Pardeski (2002), and Peterson (2007). While the causative-passive polysemy can be accounted for in terms of a general transitivizing function, it is impossible to describe a causative-passive polysemy in this way. Yet, the causative-passive polysemy is cross-linguistically widespread (Nedjalkov & Sil’nitskij 1969; Haspelmath 1990). In particular, it is common in Altaic languages, and illustrated below for Khakas (a Siberian Turkic language): (62) Khakas a. Causative use Iže-zi a-ɣa süt iz-ir-š’e. mother-3sg he-dat milk drink-caus-prs ‘His mother makes him drink milk.’

(Letuchiy 2006: 424, 433)

b. Passive use Paba-m xyr-tyr-š’a parikmaxer-ɣa. father-1sg shave-caus-prs hairdresser-dat ‘My father is shaved by the hairdresser.’ Such cases are usually accounted for in diachronic terms as reinterpretation of the causative marker through an intermediate stage of reflexive-permissive use (see Nedjalkov & Sil’nitskij 1969; Haspelmath 1990; Kulikov 2011 for discussion and references). Yet, such polysemies can also be found also with ambiguous markers for which the causative meaning can hardly be regarded as basic, or argued to be

Valency classes and alternations: parameters of variation

121

historically primary. This is the case with adversative passives in Japanese and Tungusic (see (23b) from Even): (63) Japanese Ken-ga ame-ni hur-are-ta. Ken-nom rain-dat fall-pass-pst ‘Ken got rained on.’

(Kishimoto et al., this volume)

While the origin behind this polysemy may be complex, the markedness pattern concerning the dependency of causative or passive interpretation remains the same. The same marker is preferentially used in the passive sense when applied to transitives, and in a causative sense when applied to intransitives (see Malchukov & Nedjalkov, this volume, on Manchu). This kind of polysemy is unusual from the perspective of valency change, as the same marker is used in two opposite functions, valency increasing and valency decreasing. Another paradoxical ambivalence of this kind, which has attracted less attention, is the antipassive-applicative polysemy. Consider again Central Alaskan Yupik (CAY), which has not been discussed in this context so far (my analysis is based on the comprehensive description by Miyaoka (this volume) and Miyaoka (2012), although my interpretation is somewhat different). Recall that CAY features a special adversative category, which functions as a malefactive applicative with transitives (as illustrated in (60b) and (61b)). Now, the same marker can be also used as an antipassive as well: (64) a. Central Alaskan Yupik Angute-m kuvya-ni allg-aa. man-rel.sg net-abs.3rsg.sg tear-ind.3sg.3sg ‘The man tears/tore his (own) net.’

(Miyaoka, this volume)

b. Angun kuvya-minek allg-i-uq. man.abs.sg net-abm.3rsg.sg tear-antip-ind.3sg ‘The man tore his (own) net.’ Moreover, CAY also features a special applicative marker -ut- that can be used with both intransitives and transitives. When used with intransitives, it is expectedly used to introduce a new object, but when built on transitives it may perform an antipassive function. A similar polysemy pattern has been observed for other languages as well. Thus, in Sliammon Salish (Watanabe, this volume), the same suffix -ʔǝm is used as an ‘indirective’ marker and as a suffix of the ‘active-intransitive’ forms: (65) a. Sliammon Salish Hǝy-ʔǝm=č ʔǝ=k̓w=k̓waxwa. make-A.intr=1sg.ind.sbj obl= det= box ‘I will make a box.’

(Watanabe, this volume)

122

Andrej Malchukov

b. indirective Hǝy-ʔǝm-θi=tθǝm ʔǝ=k̓w=k̓waxwa. make-ind-ctr+2sg.obj=1sg.ind.sbj+fut obl= det= box ‘I will make a box for you.’ Watanabe (this volume) notes that the issue of whether to treat the two forms as distinct is unresolved in Salish linguistics. Given that ‘indirective’ is a variety of benefactive applicative and ‘active-intransitive’ is a variety of (demoting) antipassive (with the notional patient introduced by the oblique preposition), I would classify this as belonging to the cases of applicative-antipassive polysemy discussed earlier.

4.5 Voice ambivalence and semantic maps Voice ambivalence of the type discussed above presents a challenge to the analysis of voice markers in individual languages, as well as for typological studies aiming to restrict such polysemies. Note that such polysemies cannot be simply dismissed as a case of homophony when they are cross-linguistically recurrent. And analyses in terms of general meaning (e.g., imperfective meaning shared by all uses of Bezhta antipassive) are often not informative enough to derive the diathetic effect we are interested in. Our approach is most in line with a prototype approaches to definitions of linguistic categories, as commonly practiced in typology (cf. Shibatani’s (1985) analysis of passives; and Dixon & Aikhenvald’s (2000) description of other voice categories from a similar perspective). On the latter approach, the basic meaning of antipassives (“the antipassive prototype”) would include both voice related features (A promoted to S; P deleted/demoted), as well as aspect-related (Verb: imperfective) components. In case of intransitives, only the aspectual function is available since the antipassive rule applies vacuously (the base verb is intransitive, so neither A nor P needs to be manipulated). In what follows I assumed the definitions of the prototypes of passives, antipassives, causatives and applicatives as suggested in the typological literature (see Dixon & Aikhenvald 2000 for a summary). In that approach, the following polysemies can be readily explained in terms of shared syntactic components: – Causatives of transitives and intransitives are A-adding operations (therefore may be coded by the same marker) – Applicatives of transitives and intransitives are P-adding operations (therefore may be coded by the same marker) – Causatives and applicatives of intransitives are transitivizers (therefore may be coded by the same marker) – Passives and antipassives of intransitives are detransitivizers (therefore may be coded by the same marker)

Valency classes and alternations: parameters of variation

123

These polysemies are unremarkable, but in the present context it is important that they can be explained in terms of shared syntactic features. Yet, the same approach can be easily extended to explain “surprising” polysemies such as the causativepassive polysemy and applicative-antipassive polysemy, which either increase or decrease verbal valency: – Causatives of transitives and passives share the feature of Agent-to-oblique demotion (therefore may be coded by the same marker) – Applicatives of transitives and antipassives share the feature of P-demotion (therefore may be coded by the same marker) It is of crucial importance that a connection of passives to causatives is selective as it relates passives to causatives of transitives (they share the feature of A-tooblique demotion), but not to causatives of intransitives.12 Similarly, a connection of antipassive to applicative is selective in that it obviously relates antipassives to applicatives of transitives rather than intransitives (cf. the description of the applicative derivation for transitives and intransitives Dixon & Aikhenvald 2000: 14). In this way we can arrive at the following map of voice categories capturing selective similarities between individual categories. This map (adopted from Malchukov, fothcoming) can be conceived as a kind of semantic map, with a qualification that the term “syntactic map” may be more appropriate as it captures similarities in the abovementioned syntactic properties (these are indicated by connecting lines: see Fig. 5 on arrows). Of course, this map remains incomplete, as it only includes a subset of voice categories and further ignores polyfunctionality patterns beyond the voice categories (such as connections between antipassives and imperfectives discussed in connection with the Bezhta examples (58)–(59)). Also, it does not capture the frequent polysemy of reflexive, anticausative, passive and antipassive markers, which has been given a diachronic explanation in Haspelmath (2003). Extending this map to include further categories is a matter for future research, yet it is clear that an approach in terms of shared syntactic components can go a long way in explaining

12 Demotion of an A to an oblique is clearly a preferred option for causative of transitives in languages with dependent marking. This scenario covers both cases when a causee is coded as an indirect object (conforming to Comrie’s 1976 ‘Paradigm Case’ of causative formation), but also cases when a causee is introduced as an oblique (e.g., instrumental oblique; cf. Dixon 2000). It does not cover cases when a patient rather than the causee is demoted, which seems to be a minority pattern in dependent marking languages. Neither does it cover cases where a causee is marked as a primary object (by object agreement), as is common in head-marking languages. The causative formation in such languages does not share any features (i.e. A-to-Oblique demotion) with passives, and in fact, I am not aware of cases of passive-causative polysemy reported for such languages.

124

PASSIVE (intr)

Andrej Malchukov

CAUS (intr)

APPL (intr)

CAUS (tr.)

APPL (tr.)

PASSIVE (tr.)

ANTIPAS SIVE (tr.)

Fig. 5: Semantic map for core voice categories.

and constraining voice ambivalence.13 More discussion of voice ambivalence can be found in Malchukov (forthcoming). More importantly in the present context, the map above is intended to capture relations of (local) markedness: the relative unmarkedness (naturalness) of combining a voice category and a valency class is iconically represented by the size of the cells. In this way it is shown, for example, that causatives of intransitives represent a particularly natural (unmarked) combination (the causative prototype, in Dixon’s approach), while causatives of transitives are a more marked category (a prototype extension). The unmarked combinations function as “attractors” in a dynamic system conditioning a meaning shift on the part of the marked combinations, which, being less stable, will tend to be either absent or reinterpreted. The direction of reinterpretation is indicated by the arrows. These arrows also allow for a diachronic interpretation even though evidence for certain transitions is stronger than for some others. Yet, the markedness account is more general than the diachronic account postulating unidirectionality, as it also covers more complex scenarios, as when the categories develop from a third source (e.g., both passives and causatives developing from a source construction that specializes in the passive function in the context of transitives, and in the causative function in the context of intransitives). Moreover, while the dynamicized semantic maps representing grammaticalization channels (as for instance in Van der Auwera & Plungian 1998) are “context-free”, the transition network in Figure 5 is intended to capture the relations of (local) markedness. Thus, methodologically, the most novel feature of

13 Thus, the above mentioned reflexive-anticausative-passive polysemy can equally be captured in terms of shared syntactic components, if reflexive is defined as {A->S; P->S}, anticausative as {A>O̸; P->S}, and passive as {A->O̸ ~ Oblique; P->S}. Given these definitions, the predicted polysemy pattern connects the anticausative with the reflexive, on the one hand, and with the passive, on the other hand, in accordance with the actual attestations.

Valency classes and alternations: parameters of variation

125

the transition diagram above is that it tries to capture both the iconicity of linguistic signs (the underlying semantic map representation) and (local) markedness. Iconicity restricts possible transitions in a network (through categories sharing certain features), while local markedness determines the direction of a transition.

5 Conclusions As is clear from the discussion above, valency classes as manifested in coding and alternations are not universal, but do not show random variation either. For coding, the two major factors shown to be responsible for cross-linguistic variation are Iconicity (or Faithfulness) imposing marking most faithful to the role structure and Markedness imposing the use of a default patterns for certain verb types (e.g., transitive pattern for bivalent verbs). Interaction of these two factors can be captured through implicational hierarchies (such as Tsunoda’s Transitivity Hierarchy) or (multidimensional) semantic maps where verb types similar in argument structure are adjacent in the conceptual space. There are other factors (such as word class or verb polysemy) which can interfere with predictions of the function-based maps and can account for counterexamples. In any case, given variation in coding inventories (flags or indices), all generalizations in this domain will be of statistical rather than absolute nature. Valency classes as detected through alternations are basically sensitive to the same factors, although here the relation is more complicated in several respects. First, applicability of an alternation may be sensitive to functional factors other than semantic role (as in the case of animacy restricting application of the reflexive alternation); such factors play a restricted role in coding (at least as far as flagging is concerned). Second, applicability of certain alternations may be further restricted by coding: in particular, availability of an antipassive is better predictable on the basis of syntactic transitivity rather than “notional transitivity” in the sense of Hopper & Thompson and Tsunoda. Again, as in the case of coding, structural factors may interfere with the functional constraints and block the application of an alternation for certain verb meanings (due to availability of a competing form or a competing function). That said, classification of valency classes by alternations may also be shown to be governed by similar factors, and represented through interaction of Faithfulness and Markedness constraints. On this view, faithfulness constraints determine the core domain of application of a particular (voice) alternation, while markedness constraints permit extensions of the alternation beyond this domain. As shown above, when a verb-coded alternation is extended beyond its core domain it may undergo reinterpretation producing voice ambivalence. Thus, interpretation preferences detected in patterns of voice ambivalence in effect mirror coding preferences manifested by alternation hierarchies, and both can be captured by semantic maps (transition networks) such as shown in Figure 5.

126

Andrej Malchukov

Acknowledgements I am grateful to the other members of Leipzig Valency Project (Bernard Comrie, Martin Haspelmath, Iren Hartmann, Søren Wichmann) for useful feedback, as well as to the other contributors of this volume for inspiration. The usual disclaimers apply.

Abbreviations α, β, γ

ABM ADVP AF AN CMP FEM INAN INCMP INV LAT MED ND NF POS POT PRED TRR

different series of undergoer prefixes, varying according to TAM but with no straightforward semantics detectable until they are combined with TAM suffix series (Nen) ablative-modalis adversative passive actor focus animate completive aspect feminine class inanimate incomplete invers lative (motion to) mediopassive non-dual nonfuture positive potential predicative transitivizer

References Apresjan, Jurij D. 1969. Èksperemental’noe issledovanie russkogo glagola [Experimental study of the Russian verb]. Moskva: Nauka. Apresjan, Jurij D. 1974. Leksičeskaja semantika. Sinonimičeskie sredstva jazyka. [Lexical Semantics]. Moscow: Nauka. Arkadiev, Peter M. 2009. Poor (two-term) case systems: Limits of neutralization. In Andrej Malchukov & Andrew Spencer (eds.), Handbook of Case, 686–699. Oxford: Oxford University Press. Austin, Peter. 1997. Causatives and applicatives in Australian Aboriginal languages. In Kazuto Matsumura & Toru Hayashi (eds.), The Dative and Related Phenomena, 165–225. Tokyo: Hitsuji Shobo. Beavers, John. 2011. On affectedness. Natural Language & Linguistic Theory. 29. 335–370.

Valency classes and alternations: parameters of variation

127

Bickel, Balthasar, Taras Zakharko, Lennart Bierkandt & Alena Witzlack-Makarevich. 2014. Semantic role clustering: an empirical assessment of semantic role types in non-default case assignment. Studies in Language. 38(3). 485–511. Bisang, Walter. 2006. Widening the perspective: Argumenthood and syntax in Chinese, Japanese and Tagalog. In Daniel Hole, André Meinunger & Werner Abraham (eds.), Datives and Other Cases. Between Argument Structure and Event Structure, 331–381. Amsterdam: Benjamins Butt, Miriam. 2006. Theories of Case. Cambridge: Cambridge University Press. Campbell, Lyle. 2000. Valency-changing derivations in K’iche’. In R. M. W. Dixon & Alexandra Y. Aikhenvald (eds.), Changing Valency: Case Studies in Transitivity, 236–281. Cambridge: Cambridge University Press. Comrie, Bernard. 1975. Causatives and Universal Grammar. Transactions of the Philological Society. 1–32. Comrie, Bernard. 1976. The syntax of causative constructions: Cross-language similarities and divergences. In Masayoshi Shibatani (ed.), Syntax and Semantics 6: The Grammar of Causative Constructions, 261–312. New York: Academic Press. Cooreman, Ann. 1994. A functional typology of antipassives. In Barbara Fox & Paul Hopper (eds.), Voice: Form and Function (Typological Studies in Language 27). Amsterdam: John Benjamins. Croft, William. 2012. Verbs: Aspect and Causal Structure. Oxford: Oxford University Press. Dixon R. M. W. 1977. Where have all the adjectives gone? Studies in Language 1. 19–80. Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press. Dixon, R. M. W. 2000. A typology of causatives: form, syntax and meaning. In Robert M. W. Dixon & Alexandra Y. Aikhenvald (eds.), Changing Valency. Case Studies in Transitivity, 30–83. Cambridge: Cambridge University Press. Dixon, Robert M. W. & Alexandra Y. Aikhenvald (eds.). 2000. Changing Valency. Case Studies in Transitivity. Cambridge: Cambridge University Press. Donohue, Mark & Søren Wichmann (eds.). 2008. Typology of Languages with Semantic Alignment. Oxford: Oxford University Press. Evans, Nicholas. 2004. Experiencer objects in Iwaidjan languages. In Peri Bhaskararao & Karumuri Venkata Subbarao (eds.), Non-Nominative Subjects, vol. 1, 169–193. Amsterdam: Benjamins. Fortescue, Michael D. 1984. West Greenlandic. London: Croom Helm. Geniušienė, Emma. 1987. The Typology of Reflexives. Berlin: Mouton de Gruyter. Gil, David. 2009. How much grammar does it take to sail a boat. In Geoffrey Sampson, David Gil & Peter Trudgill (eds.), Language Complexity as an Evolving Variable, 19–33. Oxford: Oxford University Press. Givón, Talmy. 1997. Grammatical relations: an introduction. In Talmy Givón (ed.), Grammatical Relations: a Functionalist Perspective, 1–85. Amsterdam: John Benjamins. Givón, Talmy. 2001. Syntax. A Functional-Typological Introduction, vol. 1 & 2. Amsterdam: John Benjamins. Goldberg, Adele E. 1995. Constructions: A Construction Grammar Approach to Argument Structure. Chicago: The University of Chicago Press. Golluscio, Lucia. 2010. Ditransitives in Mapudungun. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies In Ditransitive Constructions: A Comparative Handbook, 710– 757. Berlin: Mouton De Gruyter. Haspelmath, Martin. 1990. The grammaticalization of passive morphology. Studies in Language 14. 25–72. Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternation. In Bernard Comrie & Maria Polinsky (eds.), Causatives and Transitivity, 87–120. Amsterdam: John Benjamins. Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages. In Alexandra Aikhenvald, Robert W. M. Dixon & Masayuki Onishi (eds.) Non-Canonical Marking

128

Andrej Malchukov

of Subjects and Objects (Typological studies in language, 46), 53–85. Amsterdam: Benjamins. Haspelmath, Martin. 2003. The geometry of grammatical meaning: Semantic maps and crosslinguistic comparison. In Michael Tomasello (ed.), The New Psychology of Language, vol. 2, 211–243. New York: Lawrence Erlbaum Associates Publishers. Haspelmath, Martin & Thomas Müller-Bardey. 2004. Valency change. In Geert Booij, Christian Lehmann & Joachim Mugdan (eds.), Morphology: A Handbook on Inflection and Word Formation, vol. 2, 1130–1145. Berlin: Mouton de Gruyter. de Hoop, Helen & Andrej Malchukov. 2008. Case-marking strategies. Linguistic Inquiry 39. 565– 587. Hopper, Paul J. & Sandra A. Thompson. 1980. Transitivity in grammar and discourse. Language 56. 251–299. Kazenin, Konstantin I. 1994. On the lexical distribution of agent-preserving and object-preserving transitivity alternations. Nordic Journal of Linguistics 17. 141–154. Keenan, Edward L. 1985. Passive in the world's languages. In Timothy Shopen (ed.), Language Typology and Syntactic Description, vol. 1: Clause Structure, 243–281. Cambridge: Cambridge University Press. Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam: John Benjamins. Kiparsky, Paul, 1982. Lexical morphology and phonology. In The Linguistics Society of Korea (ed.), Linguistics in the Morning Calm. Selected Papers from SICOL-1981, 3–91. Seoul: Hanshin Publishing Co. Kiparsky, Paul. 1998. Partitive case and aspect. In Miriam Butt & Wilhelm Geuder (eds.), The Projection of Arguments: Lexical and Compositional Factors, 265–307. CSLI: Stanford. Kittilä, Seppo. 2009. Causative morphemes as non-valency increasing devices. Folia Linguistica 43(1). 67–94. Klaiman, Mimi H. 1991. Grammatical Voice. Cambridge: Cambridge University Press. Kratochvíl, Frantisek. 2011. Transitivity in Abui. Studies in Language 35(3). 588–635. Kulikov, Leonid. 2011. Voice typology. In Jae Jung Song (ed.), The Oxford Handbook of Linguistic Typology, 368–398. Oxford: Oxford University Press. Lazard, Gilbert. 1998. Actancy. Berlin: Mouton de Gruyter. Lehmann, Christian. 1991. Predicate classes and PARTICIPATION. In Hansjakob Seiler & Waldfried Premper (eds.), Partizipation: das sprachliche Erfassen von Sachverhalten, 183–239. Tübingen: G. Narr. Lehmann, Christian. 2006. Participant roles, thematic roles and syntactic relations. In Tasaku Tsunoda & Taro Kageyama (eds.), Voice and Grammatical Relations. In Honor of Masayoshi Shibatani, 167–190. Amsterdam: Benjamins. Letuchiy, Alexander. 2006. Case marking, possession and syntactic hierarchies in Khakas causative constructions in comparison with other Turkic languages. In Leonid Kulikov, Andrej Malchukov & Peter de Swart (eds.), Case, Valency, and Transitivity, 417–441. Amsterdam: John Benjamins. Letuchiy, Alexander. 2009. Towards a typology of labile verbs: Lability vs. derivation. In Alexandre Arkhipov & Patience Epps (eds.), New Challenges in Typology: Transcending the Borders and Refining the Distinctions. 223–244. Berlin & New York: Mouton de Gruyter. Levin, Beth. 1993. English Verb Classes and Alternation. A Preliminary Investigation. Chicago: The University of Chicago Press. Levin, Beth & Malka Rappaport Hovav. 1995. Unaccusativity: At the Syntax-Lexical Semantics Interface (Linguistic inquiry monographs 26). Cambridge, Mass.: MIT Press. Malchukov, Andrej. 1993. The syntax and semantics of adversative constructions in even. Gengo Kenkyu 103. 1–36. Malchukov, Andrej. 1995. Even. Munich: LINCOM.

Valency classes and alternations: parameters of variation

129

Malchukov, Andrej. 2005. Case pattern splits, verb types, and construction competition. In Mengistu Amberber & Helen de Hoop (eds.), Competition and Variation in Natural Languages: the Case for Case, 73–117. Amsterdam: Elsevier. Malchukov, Andrej. 2008. Split intransitives, experiencer objects and ‘transimpersonal’ constructions: (re-) establishing the connection. In Mark Donohue & Søren Wichmann (eds.), Typology of languages with semantic alignment, 76–101. Oxford: Oxford University Press. Malchukov, Andrej. Forthcoming. Ambivalent voice: markedness effects in valency change. In Taro Kageyama & Wesley M. Jacobsen (eds.), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: Mouton de Gruyter. Malchukov, Andrej & Peter de Swart. 2009. Differential case marking and actancy variation. In Andrej Malchukov & Andrew Spencer (eds.), Handbook of Case, 339–356. Oxford: Oxford University Press. Malchukov, Andrej, Martin Haspelmath & Bernard Comrie (eds.). 2010. Ditransitive constructions: a typological overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions, 1–65. Berlin: Mouton de Gruyter. Mel’čuk, Igor´. 1974. Opyt teorii lingvisitičeskix modelej tipa Smysl ⇔ Tekst. [Theory of linguistic models]. Moskva: Nauka. Mel’čuk, Igor. 2006. Aspects of the Theory of Morphology. Berlin: Mouton de Gruyter. Miyaoka, Osahito. 2012. A grammar of Central Alaskan Yupik (CAY). Berlin: Walter de Gruyter. Næss, Åshild. 2007. Prototypical Transitivity. Amsterdam: Benjamins. Nedjalkov, Igor V. 1992. Zalog, vid, vremja v tunguso-man’čžurskix jazykax [Voice, aspect, tense in Tungusic languages]. Habilschrift. St. Petersburg. Nedjalkov, Vladimir P. 1964. O svjazi kauzativnosti i passivnosti. [On the relation between causatives and passives]. In J. M. Skrebnev (ed.), Voprosy obščego i romano-germanskogo jazykoznanija [Problems of Romance and Germanic linguistics], 301–310. Ufa: Baškirskoe knižnoe izdatel’stvo. Nedjalkov, Vladimir P. (ed.). 2007. Reciprocal Constructions, 5 vols. Amsterdam: John Benjamins. Nedjalkov, Vladimir P. & Georgij G. Sil’nitskij. 1969. Tipologija morfologičeskogo i leksičeskogo kauzativov [Typology of morphological and lexical causatives]. In Alexandr A. Xolodovič (ed.), Tipologija kauzativnyx konstrukcij. Morfologičeskij kauzativ, 5–50. Leningrad: Nauka. Nichols, Johanna, David A. Peterson & Jonathan Barnes. 2004. Transitivizing and detransitivizing languages. Linguistic Typology 8. 149–211. Onishi, Masayuki. 2001. Non-canonically marked subjects and objects: Parameters and properties. In Alexandra Aikhenvald, Robert W. M. Dixon & Masayuki Onishi (eds.). NonCanonical Marking of Subjects and Objects (Typological studies in language, 46) 1–53. Amsterdam: Benjamins Perlmutter, David M. 1978. Impersonal passives and the unaccusative hypothesis. In Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society. 157–189. Peterson, David. 2007. Applicative Constructions. Oxford: Oxford University Press. Primus, Beatrice. 1999. Cases and Thematic Roles. Tübingen: Niemeyer. Say, Sergej. 2014. Bivalent Verb Classes in the Languages of Europe: A Quantitative Typological Study. Language Dynamics and Change 4. 116–166. Shibatani, Masayoshi (ed.). 1976. The Grammar of Causative Constructions. (Syntax and Semantics 6). New York: Academic Press. Shibatani, Masayoshi. 1977. Grammatical relations and surface cases. Language 53. 789–809. Shibatani, Masayoshi. 1985. Passives and related constructions. Language 61(4). 821–848. Shibatani, Masayoshi. 2001. Non-canonical constructions in Japanese. In Alexandra Aikhenvald, Robert W. M. Dixon & Masayuki Onishi (eds.), 307–355. Amsterdam: Benjamins. Shibatani, Masayoshi (ed.). 2002. The Grammar of Causation and Interpersonal Manipulation. Amsterdam: John Benjamins.

130

Andrej Malchukov

Shibatani, Masayoshi & Prashant Pardeshi. 2002. The causative continuum. In Masayoshi Shibatani (ed.), The Grammar of Causation and Interpersonal Manipulation, 85–126. Amsterdam: John Benjamins. Siewierska, Anna. 1984. The Passive: A Comparative Linguistic Analysis. London: Routledge. Stilo, Donald. 2009. Case in Iranian: from reduction and loss to innovation and renewal. In Andrej Malchukov & Andrew Spencer (eds.), Handbook of Case, 700–715. Oxford: Oxford University Press. Tchekhoff, Claude. 1987. Antipassif: Aspect imperfectif et autonomie du sujet. Bulletin de la Société de Linguistique de Paris 82. 43–67. Timberlake, Alan. 2004. A Reference Grammar of Russian. Cambridge: Cambridge University Press. Tsunoda, Tasaku. 1981. Split case-marking in verb types and tense/aspect/mood. Linguistics 19. 389–438. Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21(2). 385–396. Van Valin, Robert D. 1990. Semantic Parameters of Split Intransitivity. Language 66(2). 221–260. Van Valin, Robert D. 2005. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge University Press Van der Auwera, Johan & Vladimir A. Plungian. 1998. Modality’s semantic map. Linguistic Typology 2(1). 125–139. Wichmann, Søren. 2010. Ditransitive constructions in Tlapanec. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions. A Comparative Handbook, 651–678. Berlin & New York: W. de Gruyter. Wichmann, Søren. Forthcoming. Quantitative tests of implicational verb hierarchies. In Taro Kageyama & Wesley M. Jacobsen (eds.), Transitivity and Valency Alternations: Studies on Japanese and Beyond. Berlin: Mouton de Gruyter. Woolford, Ellen. 2001. Case patterns. In Géraldine Legendre, Jane Grimshaw & Sten Vikner (eds.), Optimality-Theoretic Syntax, 509–545. Cambridge, Mass.: MIT Press. Wunderlich, Dieter. 2006. Towards a structural typology of verb classes. In Dieter Wunderlich (ed.), Advances in the Theory of the Lexicon, 57–166. Berlin: Mouton de Gruyter. Xolodovič, Alexander A. (ed.) 1969. Tipologija kauzativnyx konstrukcij. Morfologičeskij kauzativ. [Typology of causative constructions: Morphological causative]. Leningrad: Nauka. Xrakovskij, Viktor S. 1974. Passivnye konstrukcii. [Passive constructions]. In Viktor S. Xrakovskij (ed.) Tipologija passivnyx konstrukcij [Typology of passive constructions], 5–45. Leningrad: Nauka. Xrakovskij, Viktor. 1981. Diateza i referentnost’ [Diathesis and Referentiality]. In Viktor Xrakovskij (ed.), Zalogovye konstrukcii v raznostrukturnyx jazykax, 5–38. Leningrad: Nauka.

Martin Haspelmath

5 Transitivity prominence 1 Differences in the degree to which transitive encoding used It is often taken for granted that languages have a large number of transitive verbs, or even that the typical two-argument verb is transitive. And indeed, this paper will provide further evidence for this view. But we know that languages differ in the extent to which they make use of transitive encoding – in other words, in their degree of transitivity prominence. For example Hawkins (1986) highlighted a number of systematic contrasts between English and German, among them the much stronger tendency for English to employ transitive verbs. Thus, verbs like ‘help’ and ‘follow’ are encoded transitively in English, but not in German, where the helpee and followee arguments are in the Dative (rather than the Accusative) case. (1) a. English b. German

Henom helped heracc. Ernom half ihrdat.

(2) a. English b. German

Theynom followed themacc. Sienom folgten ihnendat.

It is unclear, however, what is the typical situation in the world’s languages. Is German more typical or is English more typical? Or are both quite atypical, in line with other features where European languages turn out to be rather unusual (cf. Cysouw 2011)? It is sometimes thought that English is particularly transitivityprominent (e.g. Bossong 1998: 271), but the following two examples, from a language of southern Africa and a language of Amazonia, illustrate the possibility that non-European languages may also exhibit a strong proclivity for transitive encoding. In (3), a verb of directed motion takes its goal argument as if it were a patient, and in (4), an utterance verb takes the addressee as if it were a patient, i.e. both show transitive encoding.1 (3)

N‖ng ha ǁʼaa tya ǃuu 3sg go.to that world ‘He goes to that country.’

(Ernszt et al. 2013)

1 It should be noted that by transitivity, I mean transitive encoding, i.e. a formal concept. Transitivity is often understood as a semantic concept of some kind, but semantics is quite irrelevant to my understanding of transitivity. See § 3 for my definition of the term.

132 (4)

Martin Haspelmath

Bora wajpi ihjyúcunú ováhtsa-ke man shout boy-acc ‘The man shouted at the boy.’

(Seifart 2013)

Typological studies such as Tsunoda (1985) and Malchukov (2005) have tried to formulate generalizations concerning the kinds of verb meanings that tend to be coded non-transitively in different languages. We know from these studies that verbs of emotion (‘fear’, ‘like’), verbs of cognition (‘know’), and verbs of pursuit (‘follow’, ‘look for’) are less likely to be encoded transitively across languages, but so far languages have not often been ranked in terms of their transitivity prominence. Linguists have made a number of individual observations, like Hawkins’s study alluded to above, and as also illustrated by the following quote from Lazard (2002: 153–154) (see also Dahl 1990: 7). “in most languages the major two-actant construction [= the transitive construction] is not limited to the expression of prototypical actions [= ‘break’-type actions], and not even to actions as such. ... it may even be used in describing a property or a location, as in English This room sleeps four persons, or French L’école jouxte la mairie (‘The school is beside the town hall’). In this respect, there are great differences from one language to the next. In English and French the extension of the major two-actant construction to the expression of processes which are not prototypical actions is particularly large: this may be a characteristic typological feature of Western European languages. The extension seems to be larger in English than in French. In Russian, it is certainly smaller.” (Comments in square brackets added by me)

Another interesting study is Müller-Gotama (1994), who examined a dozen languages and classified them by their degree of “transparency” with respect to a set of criteria, one of them being “subject range” and “object range” (i.e. the range of semantic roles that can occupy these syntactic positions). English is said to have wider subject and object range (i.e. high transitivity prominence), like Chinese and Indonesian (two other right-branching languages), while left-branching languages like Korean and Malayalam have narrow subject and object ranges. As far as I am aware, the only other typological work that uses quantification is Bossong’s (1998) study of experiential verbs, which is restricted to European languages.2 (Bossong’s paper will be discussed further below in § 4.1.) This article thus provides the first quantification of the degree to which languages world-wide tend to use transitive encoding in their verbal lexicon. For such a quantitative assessment, we need systematic data from around the world (the Valency Patterns Leipzig database, § 2.1), we need a sample of verbs (§ 2.2), and we need a rigorous definition of transitivity as a comparative concept (§ 3). The results (§ 4) are somewhat surprising: We find that the European languages in our sample are not particularly transitivity-prominent. Languages from other 2 After this work was completed, I became aware of Say’s (2014) thorough quantitative study (restricted to European languages).

Transitivity prominence

133

parts of the world use transitive encoding to an even higher degree. I briefly discuss how the myth of high transitivity prominence in Western European languages could have arisen but also emphasize that measuring the degree of transitivity prominence in such a way that it corresponds to our intuitions is very difficult.

2 The language sample and the verb sample 2.1 Cross-linguistic data: The ValPaL database The data for this paper come from the Valency Patterns Leipzig (ValPaL) database, which contains data from 36 languages world-wide (Hartmann et al. 2013). We brought together a consortium of 36 author teams (experts in their respective lan-

Tab. 1: The 36 languages of ValPaL. language

family

macro-area

Mandinka N‖ng Yoruba, Emai Modern Standard Arabic

Mande Tuu Benue-Congo Afro-Asiatic

Africa

Eastern Armenian, German, English, Icelandic, Italian, Russian Bezhta Chintang, Mandarin Chinese Ket Ainu Even, Evenki Korean Standard Japanese, Mitsukaido Japanese, Hokkaido Japanese

Indo-European

Eurasia

Sri Lanka Malay, Jakarta Indonesian, Xârâcùù, Balinese Nen

Austronesian

Jaminjung

Mirndi

Australia

Sliammon Ojibwe Hoocąk Yaqui Zenzontepec Chatino Yucatec Maya

Salishan Algonquian Siouan Uto-Aztecan Otomanguean Mayan

North America

Bora Mapudungun

Boran Mapudungun

South America

Nakh-Daghestanian Tibeto-Burman Yeniseian Ainu Tungusic Korean Japanese

Papunesia

Morehead-Wasur

134

Martin Haspelmath

guages; see Appendix A) to provide a dataset of 80 verbs with detailed valency information. The individual datasets are comparable because they consist of counterparts to the same set of 80 basic verb meanings. The aggregated database was published online (valpal.info). The database groups the verbs into valency frames, or more specifically coding frames, because it only takes into account two types of argument-coding devices: Flags (case-markers and adpositions) and index-sets (sets of person-number cross-referencing markers). In addition, there is information about alternations undergone by the verbs, as well as a large number of glossed examples. Assembling this database from the author teams was very time-consuming, so it was impossible to gather data from more languages. This would have been desirable, because our 35 languages do not give a satisfactory picture of world-wide diversity, also because they are not evenly distributed. The languages are arranged by macro-area (broadly following Glottolog) and family in Table 1 (see Appendix 1 for the names of the contributors).

2.2 The verb sample Sampling verb meanings in such a way that they are representative of the verb meanings of languages presents a different kind of challenge. While we have a good idea of what set of languages would be representative but lack data, in the case of verb meanings we do not know how to even address the issue of representativeness, other than by intuition. Thus, we had to be content with a set of verb meanings that intutively seemed to satisfy the following criteria: – they are diverse in terms of the kinds of meanings and the number of associated arguments – they are reasonably common in language use (in traditional societies) – they can be expected to have counterparts in all or most languages The 80 verb meanings are listed in Table 2. Each verb meaning is associated with a set of verb-specific micro-roles which (potentially) correspond to arguments of verbs. For example, the verb meaning cover (as in ‘the mother covered the child with a blanket’) has three microroles: the coverer (in our example, the mother), the covered thing (the child) and the cover (the blanket). These micro-roles are linked to language-specific arguments in the coding frames via reference numbers. For example, in Bora (see (5)), the coverer (1) is expressed as the nominative argument of the verb wátájcó ‘cover’, the covered thing (2) is expressed as the accusative argument and the cover (3) is expressed as the allative argument. This is expressed by the coding frame . There are two other Bora verbs in (5b–c).

Transitivity prominence

135

Tab. 2: The 80 verb meanings of ValPaL. EAT HUG LOOK AT SEE SMELL FEAR FRIGHTEN LIKE KNOW THINK SEARCH FOR WASH DRESS (tr) SHAVE HELP

(5) Bora a. cover

FOLLOW MEET TALK ASK FOR SHOUT AT TELL SAY NAME BUILD BREAK KILL BEAT HIT TOUCH CUT

HIDE SHOW GIVE SEND CARRY THROW TIE PUT POUR COVER FILL LOAD BLINK COUGH CLIMB

JUMP SING GO LEAVE LIVE LAUGH SCREAM FEEL PAIN FEEL COLD DIE PLAY BE SAD BE HUNGRY ROLL (intr) JUMP

SINK (intr) BURN (intr) BE DRY RAIN BE A HUNTER GRIND WIPE DIG PUSH BRING STEAL TEACH HEAR COOK BOIL (intr)

(Seifart 2013) wátájcó coverer covered thing cover

b. follow

úraavyé follower followee c. roll (intr) víyiivye rolling entity rolling location

1 NP-nom 2 NP-acc 3 NP-all

1 NP-nom 2 NP-acc

1 NP-nom 2 NP-all

With the Icelandic verb horfa ‘look at’, the looker argument (1) is coded by nominative case and subject indexing on the verb, and the looked at entity (2) is expressed by a prepositional phrase with the preposition á (see 6a)). This verb thus has the coding frame . Another Icelandic verb is given in (6b). (6) Icelandic a. look at

b. like

(Barðdal 2013) horfa looker looked at entity

1 NP-nom & V.subj 2 á+NP

líka liker liked entity

1 NP-dat 2 NP-nom & V.subj

136

Martin Haspelmath

So far we have seen how verbs are selected and how their valency properties are recorded in the database by coding frames. But since the coding information (flags such as case-markers and adpositions, index-sets such as subject cross-referencing) is language-specific, this does not allow us to compare valency information across languages yet. How I define transitivity cross-linguistically will be explained in the next section.

3 How to define transitivity cross-linguistically In order to measure the degree of transitivity prominence across languages, we need a rigorous way of defining transitivity in such a way that for each coding frame, we can decide unambiguously whether it is transitive or not. A scalar notion of transitivity (e.g. Hopper & Thompson 1980) is thus not suitable for my purposes. My definition of transitivity follows Lazard (2002) and Haspelmath (2011). I start out from the typical transitive verb ‘break’ and define transitive encoding as the encoding that is used by this verb, as summarized in (7). (7) definition of transitivity A verb is considered transitive if it contains an A and a P argument. A and P are defined as the arguments of a verb with at least two arguments that are coded like the ‘breaker’ and the ‘broken thing’ micro-roles of the ‘break’ verb. Let us consider two languages in order to illustrate how this works. In Hoocąk, the breaker micro-role of the ‘break’ verb gišiš is encoded by the actor index-set on the verb, and the broken thing is encoded by the undergoer index-set, as summarized by the coding frame in (8). There are no case-markers or adpositions, so Hoocąk is typologically rather different from languages such as Bora and Icelandic. Still, we can compare the different types of languages by looking at the coding of an exemplary verb like the ‘break’ verb. All the verbs that have two arguments (A and P) that use the same encoding are regarded as transitive verbs. Thus, the verb horoǧoc ‘look at’ is counted as a verb with transitive encoding in Hoocąk. (8) Hoocąk a. break

b. look at

(Hartmann 2013) gišiš breaker broken thing

1 act.V 2 und.V

horoǧoc looker looked at entity

< 1 2 und[2].act[1].V> 1 act.V 2 und.V

Transitivity prominence

137

Another example is Even (see (9)), which has a lot of case-marking, like Bora and Icelandic. The ‘break’ verb čelgel- requires the breaker to be coded with nominative case and verbal subject indexing, and the broken thing with accusative case. In addition, the verb can have an instrumental argument, so that it has the coding frame .3 Now let us consider bele- ‘help’ (9b). This verb has two arguments, but the second argument is coded with dative case, so it is not a P argument, and hence the verb does not count as transitive. (9) Even a. break

b. help

(Malchukov 2013) 2-acc 3-instr V.subj[1]> 1-nom & V.subj 2-acc 3-instr

čelgelbreaker broken thing breaking instrument

Marco ha visto i ragazzi.) Mark has.prs.3sg seen.pp.m.sg the.m.pl boy.m.pl ‘Mark has seen them.’ Thus, verb agreement is always with the A/S argument in simple tenses. In compound tenses there occurs split agreement when P is realized by a pronoun: the finite verb always agrees with A, whilst the past participle agrees with the pronominal P (ergative orientation). In compound tenses the past participle agrees with S if the verb is unaccusative, reverting instead to the unmarked masculine singular if the verb is unergative (active alignment). Although typically a dependent-marking language, Italian, like other pro-drop languages, shows patterns of head marking, as exemplified in (8), where the arguments (A, P and R) are coded on the verb, through agreement for A, and the dative and accusative clitics for R and P, respectively (gli ‘to him’-DAT and li ‘they’-ACC), with the independent NPs optionally occurring, but extranuclear: (8) Glieli diede ieri (Lucio i libri he.dat.cl.it.m.pl.acc.cl give.pst.3sg yesterday (Luke the.m.pl books.m.pl a Marco). to Mark) ‘He gave them to him yesterday.’ (lit. to him them (he) gave yesterday)

2.3 Argument coding Arguments in Italian are coded by means of verb agreement, – that singles out A/ S from P – and adpositions, which occur with non-core arguments (e.g., recipient/ beneficiary phrases) and adjuncts (e.g., goal phrases). The adpositions (i.e. prepositions) which most typically occur to mark arguments are the following: a ‘to’, su

422

Michela Cennamo

‘on’, di ‘of’, in ‘in’, da ‘from’, con ‘with’. Marginally, word order too plays a role, differentiating A from P when they cannot be identified on the basis of agreement, as in (9). There are, in fact, no cases in Italian, apart from a residual case system in clitic pronouns, illustrated in (10)–(11) for clitics (Vincent 1988: 291; Salvi & Vanelli 2004: 196–199; among others and § 2.1): Tab. 2 1sg

2sg

3sg

1pl

2pl

3pl

ACC

mi

ti

lo (M)/la (F), si (REFL)

ci

vi

li (M)/le (F) si (REFL)

DAT

mi

ti

gli (M)/le (F) si (REFL)

ci

vi

gli (loro) si (REFL)

(9)

a. Maria (A) vide Marco (P). Mary see.pst.3sg Mark ‘Mary saw Mark.’ b. Marco (A) vide Maria (P). Mark see.pst.3sg Mary ‘Mark saw Mary.’

(10) a. Maria mi /ti /lo /la /si /ci /vi Mary i.acc /you.acc /he.acc /she.acc /refl /we.acc /you.pl.acc /li /le vide. /they.m.acc /they.f.acc see.pst.3sg ‘Mary saw me/you/him/her/herself/us/you/them.’ (11) Gli studenti mi /ti /gli /le /ci the.m.pl student.m.pl i.dat /you.dat he.dat /she.dat /we.dat /vi /si comprarono un libro. you.dat.pl refl buy.pst.3pl a.m book.m.sg ‘The students bought me/you/him/her/us/you.pl/themselves a book.’

2.4 Voice and valency rearranging ~ valency changing patterns and markers Italian has a rich and very productive system of valency rearranging and valency changing strategies. Among the valency rearranging constructions, (i) the oblique subject alternation, involving an oblique argument surfacing as subject (§ 4.2) and (ii) the body-part possessor ascension (§ 4.3) are most widely used, unlike (iii) the conative (where the unattained P is marked adpositionally, headed by the

Valency patterns in Italian

423

preposition di) (§ 6.1), (iv) the locative (§ 6.2), and (v) the applicative alternations (§ 6.3), which are marginal instead. The valency changing patterns comprise (i) object omission (§ 4.1), (ii) the factitive construction (formed with the verbs fare ‘do/make’, lasciare ‘let’ + infinitive) (§ 5.1), (iii) the reflexive/reciprocal/middle pattern, marked by the reflexive morpheme si, signalling coreference between the A and P arguments (§ 5.2), (iv) anticausatives, subdividing into three classes, according to the presence, absence or optionality of an overt marker of reduced transitivity, the reflexive morpheme si (§ 5.3), (v) the reflexive passive, formed from the third person singular/plural reflexive morpheme si plus a verb in the active voice and a Nominal agreeing with it (§ 5.4), (vi) impersonal reflexives, marking defocusing of A/S and P, signalled by the reflexive morpheme si (§ 5.5), (vii) impersonal of (lexicalized) reflexives, with the defocused A/S signalled by the 1 st person plural oblique clitic pronoun ci (§ 5.6), (viii) the analytic passive (formed with the auxiliaries essere ‘be’, venire ‘come’, andare ‘go’ + the past participle of the lexical verb) (§ 5.7).

3 Valency patterns 3.1 Identification of valency classes Two useful syntactic tests for differentiating non-core arguments from adjuncts in Italian are: (i) preposing/free permutation, that is only possible with adjuncts in unmarked focus, as shown in (12b) vs. (13b, d), and (ii) variation in the number and type of adposition(s), possible with adjuncts, but not with arguments, as illustrated in (12a) vs. (13a). The locative phrase per Roma in (13a) is, in fact, an argument, with the preposition selected by the verb, unlike a Roma in (12a), that is an adjunct, and as such it can be replaced by another locative phrase (Salvi 1988: 32– 35; Mereu 2010; Cennamo and Lenci 2011): (12) a. Ieri ho incontrato Marco a Roma / in treno. yesterday have.prs.1sg meet.pp.m.sg Mark at Rome / in train ‘Yesterday I saw Mark in Rome/on the train.’ b. Ieri a Roma ho incontrato Marco. / A Roma ieri ho incontrato Marco.

(13) a. (Ieri) sono partito per Roma / *a Roma. yesterday be.prs.1sg leave.pp.m.sg for Rome / *in Roma ‘Yesterday I left for Rome.’ b. *(Ieri) per Roma sono partito.

424

Michela Cennamo

c. Carlo parlò di politica tutta la sera. Charles talk.pst.3sg of politics.f.sg all.f.sg the.f.sg evening.f.sg ‘Charles talked about politics for the whole evening.’ d. *di politica parlò Carlo tutta la sera The possibility of occurring as a core argument (subject) in a related transitive pattern, as in (14), helps one to detect the argument status of an instrumental adpositional phrase (Van Valin 2001: 94), that behaves as an adjunct with respect to the preposing test (14c) (see however Cennamo & Lenci 2011 for the non-full reliability of preposing/free permutation as a test for adjunthood/argumenthood in Italian): (14) a. Marco ruppe la finestra con un sasso. Mark break.pst.3sg the.f.sg window.f.sg with a.m rock.m.sg ‘Mark broke the window with a rock.’ b. Un sasso ruppe la finestra. a.m rock.m.sg break.pst.3sg the.f.sg window.f.sg ‘A rock broke the window.’ c. Con un sasso Marco ruppe la finestra. with a.m. pebble.m.sg Mark break.pst.3sg the.f.sg window.f.sg ‘Mark broke the window with a pebble.’ (lit. ‘With a pebble Mark broke the window.’) d. Marco mangiò la pasta con le mani. Mark eat.pst.3sg the.f.sg pasta. f.sg with the.f.pl hands.f.pl ‘Mark ate the pasta with his hands.’ e. *Le mani mangiarono la pasta. the.f.pl hands.f.pl eat.pst.3pl the.f.sg pasta. f.sg *‘The hands ate the pasta.’ f. Con le mani Marco mangiò la pasta, con with the.f.pl hand.f.pl Mark eat.pst.3sg the.f.sg pasta. f.sg with la forchetta la carne. the.f.sg fork the.f.sg meat.f.sg ‘Mark ate pasta with his hands, meat with the fork.’ Thus the ungrammaticality of (14e) as opposed to (14b), shows that only instrumental phrases that are the semantic argument of a verb can occur as subject.

Valency patterns in Italian

425

3.2 Avalent verbs (coding frame: ((S) > V.subj[3SG/S]) This class includes different subtypes of predicates and constructions denoting metereological phenomena (e.g., piovere ‘rain’, grandinare ‘hail’, nevicare ‘snow’, tuonare ‘thunder’, lampeggiare ‘lighten’, etc.) (see (69) in the Appendix) and astronomical events (albeggiare ‘dawn’, imbrunire ‘grow dark’, far giorno ‘daybreak’, far notte ‘get dark’, etc.) (Salvi 1988: 56, 70; La Fauci 2009: 60–61; Cennamo 2010 and further references therein): (15) Piove a_dirotto. rain.prs.3sg heavily ‘It is pouring with rain.’ The states of affairs described by these verbs can be conceptualized as processes, that is as dynamic situations lacking a final point, and therefore compatible with for X time adverbials, specifying the duration of an event, as in (16a), or they can be conceptualized as events with a telic endpoint, as shown by their cooccurrence with in X time adverbials, specifying the completion of the event, focusing on its final point, as in (16b): (16) a. Piovve per tutto il giorno. rain.pst.3sg for all.m.sg the.m.sg day.m.sg ‘It rained for the whole day.’ b. Piovve in cinque minuti. rain.pst.3sg in five minute.m.pl ‘It rained within five minutes.’ The different conceptualization of the process as telic or atelic corresponds to a different auxiliary choice in compound tenses, respectively BE and HAVE/BE, as in (16c–d): (16) c. Mi è/*ha piovuto sulla testa i.dat be.prs.3sg/*have.prs.3sg rain.pp.m.sg onto_the.f.sg head.f.sg ‘It has rained on my head.’ d. È/ha piovuto per giorni. be.prs.3sg/have.prs.3sg rain.pp.m.sg for day.m.pl ‘It has rained for days.’ Weather verbs such as lampeggiare ‘lighten’, tuonare ‘thunder’, respectively a verb of light emission and of sound, however, tend to select the auxiliary HAVE in com-

426

Michela Cennamo

pound tenses, regardless of the punctuality/durativity of the event, as shown in (17): (17) a. Oggi ha lampeggiato tutto il giorno today have.prs.3sg lighten.pp.m.sg all.m.sg the.m.sg day.m.sg /*? è lampeggiato tutto il giorno. /*? be.prs.3sg flash.pp.m.sg all.m.sg the.m.sg day.m.sg ‘Today it lightened all day.’ b. Questa mattina alle cinque ha lampeggiato. this.f.sg morning.f.sg at.the five have.prs.3sg lighten.pp.m.sg /*?è lampeggiato. /*?be.prs.3sg flash.pp.m.sg ‘This morning it lightened at five.’ Several weather verbs allow also an intransitive use (18a), sometimes with a partial change of meaning (18b–e), with inanimate, definite or indefinite subjects (18a–c, e) less frequently with animate ones, as in (18d), most typically in post-verbal position if the subject is indefinite, as in (18c) (but see (18e), with an indefinite preverbal subject), and in pre/post-verbal position, if definite, as in (18a) (La Fauci 2009: 61; Cennamo 2010). For instance, the verb lampeggiare ‘lighten’ takes up the meaning of ‘emitting intermittent light’ in (18b) and piovere ‘rain’, becomes a telic change of location, with the meaning ‘come down, fall’, selecting the auxiliary BE in compound tenses, as shown in (18c–d): (18) a. Il cielo ha lampeggiato minacciosamente. the.m.sg sky.m.sg have.prs.3sg lighten.pp.m.sg threateningly ‘The sky lightened threateningly.’ b. Il semaforo lampeggiava. the.m.sg traffic_light.m.sg flash.impf.3sg ‘The traffic-light flashed.’ c. Piovvero granate/sono piovute granate da rain.pst.3pl granade.f.pl/be.prs.3pl rain.pp.f.pl granade.f.pl from tutte le parti. all.f.pl the.f.pl areas.f.pl ‘There were (lit. rained) granades from everywhere.’ d. Mario è piovuto a casa mia alle tre Mario be.prs.3sg rain.pp.m.sg at house.f.sg my.f.sg at.the.f.pl three di notte. of night ‘Mario unexpectedely turned up at three o clock at night.’

Valency patterns in Italian

427

e. Vetri grandinavano dal cielo. piece_of_glass.m.pl hail.impf.3pl from.the.m.sg sky.m.sg ‘There hailed down pieces of glass from the sky.’ Weather verbs denoting light emission such as lampeggiare ‘lighten’ also allow a transitive use under the meaning ‘flash’, with both an adpositional argument ([±animate], headed by the preposition a), as in (19a), and a canonical object (19b): (19) a. Ho lampeggiato alla macchina di fronte /a Marco. have.prs.1sg flash.pp.m.sg to.the.f.sg car.f.sg in front.of /to Mark ‘I flashed at the car in front of me/at Mark.’ b. Ho lampeggiato i fari. have.prs.1sg flash.pp.m.sg the.m.pl light.m.pl ‘I flashed the lights.’ There also occur complex predicates consisting of the verb fare ‘do, make’ in the third person singular form, and a nominal (e.g., notte ‘night’, buio ‘dark’), in fixed patterns such as far giorno ‘daybreak’, far buio ‘get dark’, in alternation with the pattern with the reflexive morpheme si, as in (20): (20) a. Ha fatto giorno. have.prs.3sg make.pp.m.sg daylight ‘It is already daylight.’ b. Si è fatto giorno. refl be.prs.3sg make.pp.m.sg daylight ‘It is already daylight.’ Whereas in the pattern with the reflexive morpheme si there occurs the auxiliary BE in compound tenses, as in (20b), in the structure without si there occurs the auxiliary HAVE, as in (20a) (see also § 5.2).

3.3 Monovalent verbs (coding frame: V.subj[S]) This class comprises verbs of different aspectual and lexico-semantic properties, whose interplay determines their variable morphosyntactic behaviour in relation to indexing properties such as past participle agreement and auxiliary selection in compound tenses (i.e. perfective tenses), discussed in § 2.2 with reference to the existence of semantically determined patterns of active alignment in Italian and their manifestations.

428

Michela Cennamo

In particular, monovalent verbs in Italian subdivide into two subclasses, socalled unaccusatives/class SP verbs, and unergatives/class SA verbs, characterized by the presence/lack of past participle agreement with the S argument, the subject, and the choice of auxiliary, respectively BE/HAVE, as shown in (21): (21) a. I bambini hanno giocato sul prato. the.m.pl child.m.pl have.prs.3pl play.pp.m.sg on.the.m.sg lawn.m.sg ‘The children played in the lawn.’ b. I ragazzi sono partiti ieri mattina alle the.m.pl boy.m.pl be.prs.3pl leave.pp.m.pl yesterday morning at.the sei. six ‘The boys left yesterday morning at six o’clock.’ Semantically, unaccusatives/class SP verbs are characterized by an Undergoer (patient/theme) subject, as with inherently telic change of state and motion verbs, i.e. achievements/accomplishments such as nascere, ‘be born’, morire ‘die’, scoppiare ‘burst’, esplodere ‘blow’, partire ‘leave’, andare ‘go’ and verbs denoting the existence and continuation of a state/position (e.g., bastare ‘suffice’, esistere ‘exist’, restare ‘remain’). Unergatives/class SA verbs on the other hand, have an Actor/ agent subject and denote a dynamic situation lacking a final point, as with activities (both motional and non-motional): camminare ‘walk’, giocare ‘play’, lavorare ‘work’. The distinction, however, is not clear-cut, with several examples of mismatches between the syntactic and semantic planes, whereby for instance nonagentive indefinite change verbs (i.e. verbs which do not specify a telic endpoint (Sorace 2000: 864)) such as arrossire ‘blush’, and internally caused verb of change of state, select either auxiliary, as illustrated in (21c) and agentive motion verbs such as correre ‘run’, saltare ‘jump’, select BE when the direction of motion is overtly expressed, i.e. when they are telicized, as in (21d). (21) c. Maria ha arrossito /è arrossita. Mary have.prs.3sg blush.pp.m.sg /be.prs.3sg blush.pp.f.sg ‘Mary blushed.’ d. Maria ha corso nel parco /è Mary have.prs.3sg run.pp.m.sg in.the.m.sg park.m.sg /be.prs.3sg corsa a casa. run.pp.f.sg to home.f.sg ‘Mary ran home.’ The imperfect correspondence between the semantics of predicates and their syntax suggests that the distinction is a gradient, the choice between the different

Valency patterns in Italian

429

indexing strategies reflecting a number of different dimensions, namely telicity, control and affectedness of S, animacy, as well as the abstract/concrete nature of the situation described by the verb. The interplay among these parameters allows one to locate monovalent verbs along a lexico-aspectual continuum, illustrated in Figure 1, characterized by verbs denoting a telic dynamic change and an Undergoer subject at one pole and atelic (non-motional) activity and an Actor subject at the opposite pole. They realize the core of the categories of unaccusativity/unergativity, consistently showing BE selection and past participle agreement, and HAVE selection plus lack of past participle agreement, respectively (morire ‘die’, andare ‘go’, partire ‘leave’, cadere ‘fall’ and giocare ‘play’, ridere ‘laugh’, urlare ‘scream’, abitare ‘live’ in the Appendix): Change of location (It. arrivare ‘arrive’) Unaccusatives (consistent selection of BE) Change of Condition (It. nascere ‘be born’ (def.) Indefinite Change of Condition marcire ‘rot’, arrossire ‘blush’ (indef.) Continuation of a Pre-existing Condition (It. rimanere ‘remain’, durare ‘last’) Existence of a Condition (It. esistere ‘exist’) Uncontrolled Process Bodily function (it. tossire ‘cough’) Emission (of substance/light/smell) (it. squillare ‘ring’, rimbombare ‘resound/roar’, profumare ‘smell’) Weather verbs (It. piovere ‘rain’, nevicare ‘snow’) Controlled Process (motional) (It. camminare ‘walk’, nuotare ‘swim’) Controlled Process (non-motional) Controlled, affecting (It. abdicare ‘abdicate’, cedere ‘yield’) Controlled, unaffecting (It. lavorare ‘work’, giocare ‘play’) Unergatives (consistent selection of HAVE) Fig. 1: The Auxiliary Selection Hierarchy (ASH) (adapted from Sorace 2000, 2004).3

Variation in auxiliary selection occurs as one moves away from the core of the categories and is maximal in the middle of the hierarchy, i.e. at the stative centre, where telicity is irrelevant and the subject has no/low agentivity and control. The degree of variation in auxiliary selection therefore reflects the position of the verb along the hierarchy: it increases as one moves away from the core of the categories, i.e. with the decrease in the aspectual specification of the situation expressed by the verb and in the degree of agentivity and control of the subject, being also sensitive to animacy, as shown in (22a), where the verb cedere ‘yield’ alternates HAVE/ BE if the subject is inanimate, whilst it selects only HAVE if the subject is animate.

3 We follow a slightly modified version of the ASH, consistent however with its parameters and implicational paths, since synchronic and diachronic evidence from some southern Italian varieties as well as diachronic data from other domains in Late Latin (e.g., pleonastic reflexives and accusative subjects) point to the core of unaccusatives as realised by telic changes of state (Cennamo 1999, 2001, 2002, 2008, 2009).

430

Michela Cennamo

This accounts for auxiliary alternation with stative verbs (giacere ‘lie’, scarseggiare ‘be short of’, vivere ‘live’), continuation of states (sopravvivere ‘survive’), and with verbs denoting uncontrolled process (activity) (attecchire ‘catch on’, squillare ‘ring’) (22b), including weather verbs (piovere ‘rain’, nevicare ‘snow’) (see discussion in § 2.1). Variation also occurs with indefinite change, i.e. verbs which do not encode a telic endpoint (Sorace 2000: 864) (e.g, marcire ‘rot’, arrossire ‘blush’) (23c) and manner of motion verbs (volare ‘fly’, atterrare ‘land’, saltare ‘jump’, correre ‘run’) (21d):

(22) a. Marco ha ceduto /il pavimento Marco have.prs.3sg give-in.pp.m.sg /the.m.sg floor.m.sg è ~ ha ceduto. be~have.prs.3sg yield.pp.m.sg ‘Mark has given in/the floor has yielded.’ b. I telefoni hanno squillato /sono the.m.pl phone.m.pl have.prs.3sg ring.pp.m.sg /be.prs.3pl squillati. ring.pp.m.pl ‘The phones have rung.’ Some monovalent verbs, both unergative and unaccusative (dormire ‘sleep’, piangere ‘cry’, and scendere ‘go down’, salire ‘go up’ in the Appendix, respectively), allow also transitive uses, although only with a restricted range of Ps, as shown in (23a–c) and in some cases only with so-called cognate objects, expressing semantic features of the verb, as in (23d–e) and often based on the verbal lexeme/root itself (23f) (Salvi 1988: 60; Lo Duca 2000: 220–221). The transitive use of unaccusatives denoting telic and directed motion, such as scendere ‘go down’, uscire ‘go out’, salire ‘climb, go up’ (23g) and continuation of a pre-existing condition, such as rimanere ‘remain’ in Neapolitan (23h), has however strong diatopic connotations, being characteristic of southern varieties (Serianni 1989: 380; Jezek 2003: 88; Cennamo 2011b).

(23) a. Marco non ha ancora lavorato quella pratica. Mark not have.prs.3sg yet work.pp.m.sg that.f.sg file.f.sg ‘Mark has not worked at that file yet.’ b. Marco lavora il legno. Mark work.prs.3sg the.m.sg wood.m.sg ‘Mario works wood.’

Valency patterns in Italian

431

c. Marco (si) lavorerà il capoufficio /lavorerà il Mark rfl work.fut.3sg the.m.sg boss.m.sg /work.fut.3sg the.m.sg capoufficio ai fianchi. boss.m.sg at.the.m.pl hip.m.pl ‘Mark will (try to) convince his boss.’ d. Pianse lacrime amare. cry.pst.3sg tear.f.pl bitter.f.pl ‘He cried bitter tears.’ e. Dormì il sonno del giusto. sleep.pst.3sg the.m.sg sleep.m.sg of.the.m.sg just_man.m.sg ‘He slept the sleep of the just.’ f. Marco ha vissuto una vita difficile. Mark have.prs.ind.3sg live.pp.m.sg a.f life.f.sg tough.f.sg ‘Mark lived a tough life.’ g. Devo scendere /salire la spesa. must.prs.1sg go_down.inf /go.up.inf the.f.sg shopping.f.sg ‘I must take the shopping down/up.’ h. Ho rimasto le chiavi a casa. have.prs.1sg remain.pp.m.sg the.f.pl key.f.pl at home.f.sg ‘I left my keys at home.’

3.4 Bivalent verbs (coding frame: A > V.subj [A] (con+I) Verbs sharing this coding frame belong to different subclasses, varying both in their aspectual characteristics and in the inherent (e.g., animacy) and relational properties of the A and P arguments (e.g., control and affectedness). They comprise achievements (e.g., spezzare ‘crack’), different types of accomplishments (e.g., aprire ‘open’, affondare ‘sink’, guarire ‘heal’, including degree achievements/gradual completion verbs like aumentare ‘increase’, i.e., verbs denoting the gradual approximation to a telos that may not be reached (Bertinetto & Squartini 1995), activities (e.g., colpire ‘hit’, lavare ‘wash’) and states (e.g., amare ‘love’, vedere ‘see’) (Cennamo 2003, 2011c and references therein). The various subtypes of bivalent verbs are identified by means of three syntactic tests in Italian: passivization, anticausativization and optionality of the object. Resultatives ((causative) achievements and some accomplishments) – i.e. bivalent verbs denoting a non-reversible definite change of state on an inanimate, referential, definite, rhematic P argument, carried out by a highly agentive definite, animate, thematic A, with a high degree of control over the situation described by the

432

Michela Cennamo

verb/predicate –, allow both passivization and anticausativization, but do not allow omission of the object, as shown in (24a–c): (24) a. La finestra fu rotta da Marco. the.f.sg window.f.sg be.pst.3sg break.pp.f.sg by Mark ‘The window was broken by Mark.’ a′. Marco ruppe la finestra. Mark break.pst.3sg the.f.sg window.f.sg ‘Mark broke the window.’ b. La finestra si ruppe. the.f.sg window.f.sg refl break.pst.3sg ‘The window broke’ c. *Marco ruppe.4 Mark break.pst.3sg ‘Mark broke.’ Other bivalent verbs, for instance accomplishments with an animate object such as uccidere ‘kill’, nutrire ‘nourish’, allow, instead the omission of the object, with an habitual interpretation, as in (25a) or to refer to a general property/ability, as in (25b): (25) a. Ha ucciso (più volte), ecco perché è in have.prs.3sg kill.pp.m.sg (more times) that_is why be.prs.3sg in carcere. jail.m.sg ‘He has killed (several times), that is why he is in jail.’ b. Il latte nutre. the.m.sg milk.m.sg nourish.prs.3sg ‘Milk is nourishing.’ (lit. ‘nourishes’) Interestingly, these verbs do not occur in the anticausative alternation: the corresponding intransitive pattern with the original object as subject can only have a

4 This sentence is only acceptable in the informal style in some contexts with a different meaning of the verb rompere ‘break’, that is no longer an achievement, but an activity, meaning ‘bore, wear out’: (i) Marco ha rotto/ rompe. Mark have.prs.3sg break.pp.m.sg/ break.prs.3sg ‘Mark has worn me out/wears people out.’ (lit. has broken me/breaks).

Valency patterns in Italian

433

reflexive/middle5 interpretation, depending on the verb, as shown in (26a–b). It can never denote the spontaneous manifestation of an eventuality, characteristic of anticausatives, unlike resultatives (achievements/accomplishments) with an inanimate P like rompere ‘break’, spezzare ‘crack’, aprire ‘open’ (see further discussion in § 5): (26) a. reflexive La giovane si è uccisa. the.f.sg young-woman.f.sg refl be.prs.3sg kill.pp.f.sg ‘The young woman committed suicide.’ b. reflexive/middle I giovani si nutrivano di bacche. the.m.pl young_man.m.pl refl nourish.impf.3pl of berry.f.pl ‘The young men used to feed themselves on berries.’ Bivalent resultative verbs with an animate object and a highly thematically specified A (i.e. agentive) such as assassinate ‘murder’ do not allow instead object omission and a corresponding intransitive pattern, with either A or the original object, P, surfacing as subject, as illustrated in (27a–b): (27) a. *Marco assassinò ( T (a+R) (L) (I) (di+X)) Trivalent verbs have three nuclear arguments, A, the agent, surfacing as subject, P, the theme object, and the goal/beneficiary of the eventuality described by the verb, realized either by animate/animate-like NPs (e.g., collective nouns denoting people, institutions) – encoded adpositionally (headed by the preposition a ‘to’, di ‘of’)/as a dative clitic pronoun – or as a locative phrase, according to the verb. Various subtypes of trivalent verbs (and related coding frames) can be identified, among which ditransitives, involving verbs of physical transfer, denoting situations in which an agent ‘transfers’ the possession of an object to a goal/beneficiary (e.g., dare ‘give’, inviare ‘send’, vendere ‘sell’, prestare ‘lend’, comprare ‘buy’, restituire ‘return’), as in (36a), including also verbs of mental transfer (e.g., mostrare ‘show’, dire ‘tell’, spiegare ‘explain’, offrire ‘offer’, promettere ‘promise’), exemplified in (36b) (Salvi 1988; Cennamo 2011c): (36) a. Marco diede il libro a Giovanni. Mark give.pst.3sg the.m.sg book.m.sg to John ‘Mark gave the book to John.’ a′. Marco gli diede il libro. Mark he.dat.cl give.pst.3sg the.m.sg book.m.sg ‘Mark gave him the book.’ b. Marco promise un regalo ai ragazzi Mark promise.pst.3sg a.m present.m.sg to.the.m.pl boy.m.pl ‘Mark promised a present to the boys.’ b′. Marco offrì la sua collezione alla Mark offer.prs.3sg the.f.sg his.f.sg collection.f.sg to.the.f.sg Biblioteca. library.f.sg ‘Mark offers his collection of books to the library.’

Valency patterns in Italian

439

With a number of verbs (e.g., mettere ‘put’, accusare ‘accuse’, chiamare ‘call’) the non-core nuclear argument can be realised as a locative phrase, as in (37a), a different adpositional argument, as in (37b), or as a canonical object (i.e. a direct object), as in (37c): (37) a. Marco mise il libro sul tavolo. Marco put.prs.3sg the.m.sg book.m.sg on.the.m.sg table.m.sg ‘Mark put the book on the table.’ b. Marco accusò Luca di plagio. Mark accuse.pst.3sg Luke of plagiarism.m.sg ‘Mark accused Luke of plagiarism.’ c. Marco ha chiamato sua figlia Sara. Mark have.prs.3sg call.pp.m.sg his.f.sg daughter.f.sg Sarah ‘Mark called his daughter Sarah.’ Also in the benefactive construction, exemplified by verbs like portare ‘carry’, disegnare ‘draw’, the beneficiary is coded like the recipient/goal argument of ditransitives, realized either adpositionally, headed by a ‘to’/per ‘for’, if a full noun, or by the dative if a pronominal clitic, as shown in (38) (Salvi 1988; Malchukov et al. 2010; Zuñiga & Kittilä 2010 for a crosslinguistic perspective): (38) a. Marco portò la spesa a Giovanna /per Mark carry.pst.3sg the.f.sg shopping.f.sg to Jane /for Giovanna. Jane ‘Mark carried the shopping to Jane.’ b. Marco disegnò un cuore a Giovanna /per Giovanna. Mark draw.pst.3sg a.M heart.m.sg to Jane /for Jane ‘Mark drew a heart to Jane.’ c. Marco le portò la spesa /le Mark she.dat.cl carry.pst.3sg the.f.sg shopping.f.sg /she.dat.cl disegnò un cuore (le=to her). draw.pst.3sg a.m heart.m.sg ‘Mark carried her the shopping.’ If the beneficiary is not an argument of the verb it can only be encoded as a per ‘for’-phrase, as in (38d). Both the beneficiary argument and adjunct, however, can be indexed by a dative pronominal clitic, as in (38c, e):

440

Michela Cennamo

(38) d. Marco ha studiato /analizzato i Mark have.prs.3sg study.pp.m.sg /analyse.pp.m.sg the.m.pl documenti per Lucio (*a Lucio). document.m.pl for Lucio (*to Lucio) ‘Mark studied/analysed the documents for Lucio.’ e. Marco gli ha studiato /analizzato i Mark he.dat.cl have.prs.3sg study.pp.m.sg /analyse.pp.m.sg the documenti. (gli = per Lucio) document.pl ‘Mark studied/analysed the documents for him.’

4 Uncoded argument alternations Italian has a very pervasive pattern of P deletion, a valency decreasing alternation, and a number of argument rearranging strategies, with varying degrees of productivity, reflecting the aspectual characteristics of verbs and the inherent and relational properties of core and non-core arguments. The most productive uncoded alternations are object omission (§ 4.1), the oblique subject alternation (§ 4.2) and the body-part possessor ascension alternation (§ 4.3). Other, less productive alternations are discussed in § 6.

4.1 Object Omission Several verbs in Italian allow an intransitive use, with optionality of the P argument (insegnare ‘teach’, sentire ‘hear’, fare ‘do, make’, cuocere ‘cook’, pulire ‘clean’, macinare ‘grind’, rubare ‘steal’, guardare ‘look at’, vedere ‘see’, odorare ‘smell’, temere ‘fear’, pensare ‘think’, cercare ‘look for’, lavare ‘wash’, chiedere ‘ask’, chiamare ‘call’, costruire ‘build’, uccidere ‘kill’, toccare ‘touch’, tagliare ‘cut’, sbucciare ‘peel’, dare ‘give’, versare ‘pour’, riempire ‘fill’, caricare ‘load’ in the Appendix). This category comprises three main subtypes, identifiable through the interplay of the inherent and structural aspects of verb meaning with the degree of thematic specification of the subject (i.e. agentivity/control), the inherent characteristics of P (e.g., animacy), the degree of semantic implication (i.e. ‘lexical solidarity’) between the verb and the P argument (Coseriu 1971; Jezek 2003: 101 for Italian), as well as the linguistic and extra-linguistic context. As a general rule the object is optional with verbs denoting states and dynamic situations lacking an inherent final/terminal point, as with activity verbs and active accomplishments (activity verbs allowing an accomplishment use (Van Valin 2005: 32–33)), or accomplishments with animate objects, in iterative uses, whereby

Valency patterns in Italian

441

the focus is on the event itself rather than on its impingement on the P argument (Levin 1993: 33; Lo Duca 2000; Cennamo 2003, 2011c, 2012b; Jezek 2003: 94–104; Siller-Runggaldier 2003; among others). In type (i) the omitted P argument can be [±animate], [±referential], indefinite or reconstructable from the context (so-called unspecified/indefinite object/strong optionality (Allerton 1980: 68–69; Levin 1993: 33)). This group includes states (e.g., vedere ‘see, amare ‘love’), activity verbs (e.g., cucire ‘sew’, leggere ‘read’, studiare ‘study’, cucinare ‘cook’), as well as active accomplishments like leggere ‘read’, scrivere, write’, mangiare ‘eat’, dipingere ‘paint’, cucinare ‘cook’, etc, i.e. verbs of consumption and creation, and allow the omission of the object both in imperfective and perfective contexts, as illustrated in (39): (39) a. Marco mangiò e poi uscì. Mark eat.pst.3sg and then go.pst.3sg ‘Mark ate and then went out.’ b. Marco spazzò e lavò a fondo prima di Mark sweep.pst.3sg and wash.pst.3sg thoroughly before to partire. leave.inf ‘Mark swept and washed the house thoroughly before leaving.’ Type (ii), “generalized null objects” (Lo Duca 2000: 228–232), comprises other activity verbs, such as visitare ‘visit’ and different types of accomplishments (e.g., psych verbs such as affascinare ‘enchant’, active accomplishments like ritrarre ‘draw/paint and indefinite change/gradual completion verbs/degree achievements like corrodere ‘corrode’). The unexpressed P is [±human][±generic] (most typically) [+plural] and is either an Experiencer, as with psych-verbs (e.g., abbrutire ‘abase’, angosciare ‘grieve’, annoiare ‘bore’) or a Patient (e.g., corrodere ‘corrode’, stancare ‘wear out’, graffiare ‘scratch’, mordere ‘bite’). This group only allows the intransitive variant in atelic and imperfective contexts, most typically habitual/attitudinal (Bertinetto & Lenci 2012: 860), as shown in (40) (Lo Duca 2000: 229; Jezek 2003; Cennamo 2011c): (40) a. Giovanna affascina (*ha affascinato). Jane enchant.prs.3sg have.prs.3sg enchant.pp.m.sg ‘Jane is charming (lit. enchants).’ b. l’eccessivo lavoro abbrutisce /logora the.excessive.m.sg work.m.sg abase.prs.3sg /wear_out.prs.3sg (*ha abbrutito/ha logorato). (*have.prs.3sg abase/wear_out.pp.m.sg) ‘Excessive work abades/wears people out.’

442

Michela Cennamo

c. l’acido /l’invidia corrode (*ha corroso) the.acid /the.envy corrode.prs.3sg (*have.prs.3sg corrode.pp.m.sg) ‘acid/envy corrodes’ With some activity, hurt verbs taking an animate P (e.g., mordere ‘bite’, graffiare ‘scratch’, pizzicare ‘sting’), however, the intransitive use is also possible in perfective contexts, to refer to a generic animate P, as in (40d) (Cennamo 2011c): (40) d. Quel cane ha morso, per questo that.m.sg dog.m.sg have.prs.3sg bite.pp.m.sg owing.to this ha la museruola. have.prs.3sg the muzzle.f.sg ‘That dog has beaten people, this is why it wears a muzzle.’ The A or P nature of the unexpressed argument/optional argument with some verbs is signalled by past participle agreement in predicative structures, as shown in (41). If the predicative element, the past participle, refers to the A argument, it must agree with it, as in (41a), where the masculine singular ending -o in vestito ‘dressed’, refers to A, the subject, il pittore ‘the painter’. If the past participle is in the masculine plural ending, it refers to the unexpressed P argument, as in (41b) (Rizzi 1986; Lo Duca 2000: 229–230): (41) a. Il pittore ritrae /ritrasse vestito di the.m.sg painter.m.sg draw.prs.3sg /draw.pst.3sg dress.pp.m.sg of bianco. white ‘The painter drew (the painting wearing) a white dress.’ (lit. the painter draws/drew dressed.sg of white) b. Il pittore ritrae /ritrasse vestiti di the.m.sg painter.m.sg draw.prs.3sg /draw.pst.3sg dress.pp.m.pl of bianco. white ‘The painter draws/drew people wearing a white dress.’ (lit. the painter draws/drew dressed.pl of white) c. I pittori dipingono/ dipinsero vestiti di the.mpl painter.m.pl draw.prs.3pl draw.pst.3pl dess.pp.m:pl of bianco. white ‘The painters draw/drew people wearing a white dress.’ (lit. the painters draw/drew dressed.pl of white) (Lo Duca 2000: 229)

Valency patterns in Italian

443

Thus, with a masculine plural subject, as in (41c), the pattern is ambiguous, since the past participle may refer to either the A argument (pittori ‘painters’) or to the unexpressed generic P (Bernard Comrie, p.c.). Type (iii), “definite null objects” (Lo Duca 2000: 232) includes verbs that only allow the intransitive variant if P is recoverable from the linguistic context (anaphoric null object), as in (42a) or from discourse (deictic null object), as in (42b), where the unexpressed P may refer to the Speech Act Participants (speaker and/or hearer) (Lo Duca 2000: 233–234; Jezek 2003: 100): (42) a. Ho ascoltato la proposta e ho have.prs.1sg listen.pp.m.sg the.f.sg proposal.f.sg and have.prs.1sg rifiutato. refuse.pp.m.sg ‘I listened to the proposal and I turned it down.’ b. Marco stanca /ha stancato. Mark tire.prs.3sg /have.prs.3sg tire.pp.m.sg ‘Mark wears me/us out/has worn me/us out.’ Although with activity verbs the intransitive variant with an unexpressed inanimate P usually denotes “general attitudes, abilities, dispositions” (Lo Duca 2000: 227; Jezek 2003: 97), as in (41), the intransitive pattern with some activity verbs does not refer to a generic activity, but to an event, whose semantic feature(s) is/ are encoded and made explicit by the object (lexical solidarity (Coseriu 1971; Jezek 2003: 99–100)/(low degree of) Individuation of O (Olsen and Resnik 1997)). It may concern single elements, e.g., tavola ‘table’ in sparecchiare/apparecchiare ‘clear the table/lay the table’ (43a), personale ‘staff’ in (43b) or a (narrow) range of objects, such as vehicles (e.g., macchina ‘car’, moto ‘motor-cycle’) for parcheggiare ‘park’ (43c) (Jezek 2003: 99): (43) a. Ho dimenticato di sparecchiare /apparecchiare. have.prs.1sg forget.pp.m.sg to clear.inf /lay_the_table.inf ‘I forgot to clear/lay the table.’ b. Non assumono più in quella azienda. not employ.prs.3pl no_longer in that.f.sg firm.f.sg ‘They no longer employ people in that firm.’ c. Marco ha parcheggiato lontano. Mark have.prs.3sg park.pp.m.sg far_away ‘Mark parked far away.’ Also the linguistic and situational context as well as the nature of A (the subject) (e.g., animacy) play an important role in determining the interpretation of the un-

444

Michela Cennamo

expressed O. With activity verbs allowing a resultative use such as bere ‘drink’ the [±animate] nature of the subject and the discourse context allow one to understand whether the pattern refers to a habitual activity, the drinking of alcohol (44a), or whether it refers to a specific type of liquid (44b–c): (44) a. Secondo me hai bevuto (sc. alcohol). according.to me have.prs.2sg pp.m.sg ‘I think you are drunk (lit. you have drunk).’ b. Hai bevuto? (sc. acqua, coke …).’ have.prs.2sg drink.pp.m.sg (sc. water, coke …) ‘Have you drunk? (sc. water, coke …).’ Posso mettere la bottiglia in frigorifero? Can.prs.1sg put.inf the.f.sg bottle.f.sg into fridge.m.sg ‘Can I put the bottle back into the fridge?’ c. La mia macchina beve (sc. benzina). the.f.sg my.f.sg car.f.sg drink.prs. 3sg (sc. petrol.f.sg) ‘My car consumes a lot (lit. drinks).’ (Jezek 2003: 100) Interestingly, the omission of the object is only possible in literal uses, as in (45a) (Lo Duca 2000: 233; Jezek 2003: 100). In figurative uses O must be expressed (45b) vs. (45c): (45) a. Marco frenò /ha frenato improvvisamente Mark brake.pst.3sg /have.prs.3sg brake.pp.m.sg suddenly (sc. la macchina). (sc. the.f.sg car.f.sg) ‘Mark suddenly braked.’ b. Marco ha frenato il loro entusiamo. Mark have.prs. 3sg restrain.pp.m.sg the.m.sg their enthusiasm.m.sg ‘Mark restrained their enthusiasm.’ c. *Marco ha frenato. Marco have.prs.3sg restrain.pp.m.sg ‘Mark has restrained.’ Not only activity verbs, but also accomplishments taking an animate P, such as uccidere, ammazzare ‘kill’ allow its omission in order to express the event itself, as in (46a):

445

Valency patterns in Italian

(46) a. Ha ucciso (più volte), ecco perché have.prs.3sg kill.pp.m.sg (more times) that.is why è in carcere. be.prs.ind.3sg in jail.m.sg ‘He has killed (several times), that is why he is in jail.’

= (25a)

The possibility of omitting the P argument with these verbs reflects the degree of thematic specification of the A argument, that is low for uccidere ‘kill’, but high for assassinare ‘murder’, whose subject is highly agentive. This accounts for the non omissibility of P with this verb, as shown in (46b) (see also discussion in § 3.3): (46) b. *Marco ha assassinato ripetutamente, ecco perché Mark have.prs.3sg murder.pp.m.sg repeatedly that._is why è in carcere. be.prs.3sg in jail ‘*Mark murdered several times, that is why he is in jail.’ To sum up, the optionality of the object in Italian involves states, activity verbs and different types of accomplishments, with focus on the event itself rather than on its impingement on the P argument (Levin 1993: 33; Lo Duca 2000; Cennamo 2003, 2011c, 2012b; Jezek 2003: 94–104; among others). It is banned, instead, with achievements. The aspectual characteristics of verbs interact, in turn, with other features such as the elements of meaning lexicalized in the verb (e.g., the type of result encoded), animacy, definiteness, referentiality as well as with the linguistic and extra-linguistic context, confirming a general tendency also observable crosslinguistically (Lazard 1984; Bossong 1991, 1998; Aissen 2003; Cennamo 2003 and further references therein).

4.2 The oblique subject alternation This alternation involves the upgrading of the oblique (i.e. adpositional) argument of a bivalent verb to subject function, with the original A argument unexpressed, as illustrated in (47). It occurs with verbs of different aspectual classes (achievements, accomplishments, activities) and comprises four main subtypes: (i) the Time subject alternation, occurring with verbs such as vedere ‘see’, trovare ‘find’, segnare ‘mark’, causare ‘cause’ (47a) as in English (Levin 1993: 79), (ii) the Instrument subject alternation, including natural forces (47b), instantiated by verbs like seccare, asciugare ‘dry’, riscaldare ‘heat’, variare ‘vary’, riprodurre ‘reproduce’, and intermediary instruments, as in (47c), occurring with verbs like rompere ‘break’, caricare ‘load’, scavare ‘dig’, sotterrare ‘bury’, (iii) the Instrument subject alternation with object omission, a pattern confined to modal or negative polarity contexts, where the predicate refers to the activity itself as carried out by an Instru-

446

Michela Cennamo

ment, as in (47c), (iv) the Locatum subject alternation, where the subject is realized by an originally oblique argument, the entity whose location is described by the verb (acqua ‘water’ in (47d) (e.g., riempire ‘fill’, legare, ‘tie’ coprire ‘cover’, inondare ‘flood’, circondare ‘surround’, infettare ‘infect’, ornare ‘decorate’) (Levin 1993: 79–82 for English; Lo Duca 2000 for Italian): (47) a. Il 2010 ha segnato un nuovo corso the.m.sg 2010 have.prs.3G mark.pp.m.sg a.m new.m.sg course.m.sg nell’azienda.’ in_the_firm.f.sg ‘2010 marked a new era in the firm.’ a′. Nel 2010 il nuovo direttore ha in.the.m.sg 2010 the new.m.sg director.m.sg have.prs.3sg segnato un nuovo corso nell’azienda. mark.pp.m.sg a.m new.m.sg course.m.sg in_the_firm.f.sg ‘In 2010 the new director marked a new era in the firm.’ b. Il sasso ruppe il vetro. the.m.sg pebble.m.sg break.pst.3sg the.m.sg window_pane.m.sg ‘The pebble broke the windowpane.’ b′. Marco ruppe il vetro con il sasso. Mark break.pst.3sg the.m.sg glass.m.sg with the.m.sg pebble.m.sg ‘Mark broke the glass with the pebble.’ c. Questa penna scrive bene. this.f.sg pen.f.sg write.prs.3sg well ‘This pen writes well.’ c′. Generalmente scrivo con questa penna. generally write.prs.ind.1sg with this.f.sg pen.f.sg ‘Usually I write with this pen.’ d. L’acqua riempì il secchio. the_water.f.sg fill.pst.3sg the.m.sg pail.m.sg ‘Water filled the pail.’ d′. Marco riempì il secchio di acqua /con Mark fill.pst.3sg the.m.sg pail.m.sg of water.f.sg /with l’acqua. the_water.f.sg ‘Mark filled the pail with water.’

Valency patterns in Italian

447

4.3 The body-part possessor ascension alternation This alternation involves a possessor object and a possessed body part (Levin 1993: 71–72), which may be encoded either as an object noun phrase (e.g., la spalla ‘the shoulder’ in (48a), Marco in (47b)), or may be expressed as a prepositional phrase (e.g., sulla spalla ‘on his shoulder’ in (48b), di Marco ‘Mark’s shoulder’, in (48a), with the possessor expressed as a genitive within the NP). This pattern is fairly common, occurring mainly with accomplishments lacking an inherent endpoint (e.g., legare ‘tie’), and activities (e.g., toccare ‘touch’, colpire ‘hit’) as shown in (48): (48) a. L’uomo colpì la spalla the.man.m.sg hit.pst.3sg the.f.sg shoulder.f.sg di Marco. of Mark ‘The man beat Mark’s shoulder.’

(possessed body part)

b. L’uomo colpì Marco sulla /alla spalla. the.man.m.sg hit.pst.3sg Mark on.the.f.sg /at.the.f.sg shoulder.f.sg ‘The man beat Mark on his shoulder.’ The choice of the preposition varies according to the body part involved, as in (48b) vs. (48d): (48) c. L’uomo colpì l’occhio di Marco. the.man.m.sg hit.pst.3sg the.eye.m.sg of Mark ‘The man poked Mark’s eye.’ d. L’uomo colpì Marco nell’occhio. the.man.m.sg hit.pst.3sg Mark in.the.eye.m.sg ‘The man poked Mark in the eye.’ Interestingly, in the corresponding patterns with a pronominalized possessor, there occur the dative and accusative forms respectively, as shown in (48e–f): (48) e. L’uomo gli (sc. Marco) colpì l’occhio. the.man.m.sg he.dat.cl (sc. Mark) hit.pst.3sg the-eye.m.sg ‘The man poked his eye.’ f. L’uomo lo (sc. Marco) colpì the.man.m.sg he.acc.cl (sc. Mark) hit.pst.3sg nell’occhio. in.the.eye.m.sg ‘The man poked him in the eye.’

(< (48c))

(< (48d))

448

Michela Cennamo

4.4 Possessor-attribute alternation Also the (in)transitive variant of a pattern involving a possessor object and a possessed attribute is extensively used. The attribute may be encoded either as an object noun phrase, as in (49a), or may be expressed adpositionally, headed by the preposition per ‘for’, as in (49b). This pattern occurs with type (i) experiencer verbs, i.e. with bivalent states with an experiencer subject (A) and a theme object (P): temere ‘fear’, conoscere ‘know’, ammirare ‘admire’, apprezzare ‘appreciate’, stimare ‘esteem’, disprezzare ‘despise’, amare ‘love’, ricordare ‘remember, invidiare ‘envy’, detestare ‘detest’, … (see Levin 1993: 72–74 for analogous verbs in English): (49) a. Temo l’arroganza di Giovanni. fear.prs.1sg the.arrogance.f.sg of John ‘I fear John’s arrogance.’ b. Temo Giovanni per la sua arroganza. fear.prs. 1sg John for the.f.sg his.f.sg arrogance.f.sg ‘I fear John because of his arrogance.’

5 Coded alternations: voice and valency changes As pointed out in § 2.4, the system of voice alternations and valency changing strategies involves the following patterns and related markers: (i) the factitive construction (§ 5.1), (ii) the reflexive/reciprocal/middle construction (§ 5.2), (iii) anticausatives (§ 5.3), (iv) the reflexive passive (§ 5.4), (v) impersonal of (lexicalized) reflexives (§ 5.5), (vi) impersonal reflexives (§ 5.6), (vii) the analytic passive (§ 5.7). The discussion proceeds from valency increasing mechanisms and alternations to valency decreasing ones.

5.1 The factitive construction This valency increasing alternation is realized in Italian through the adjoining of two predicates, a factitive verb such as (the causative) fare ‘make’, and (the permissive) lasciare ‘let’, the governing or matrix verb, and a dependent infinitive. The two verbs form a complex predicate, functioning as one unit, with ensuing restructuring of their original argument structure. If the adjoined infinitive is monovalent, its original subject (cucchiaio ‘spoon’ in (50a)) becomes the object of the complex predicate far(e)/lasciar(e) cadere (postverbal if nominal, as in (50a), preverbal if pronominal, as in (50b)). If the adjoined infinitive is bivalent or trivalent, its origi-

Valency patterns in Italian

449

nal object and indirect objects (pacco ‘parcel’ in (50b), a Luca in (50c)) function, respectively, as the direct object and indirect object of the complex predicate far(e)/ lasciar(e) scartare ‘make/let unwrap’ in (50b), far(e)/lasciar(e) inviare ‘make/let send’ in (50c). The original subject (i.e. the causee) is expressed, instead, as either an indirect object (headed by the preposition a ‘to’) or as an agentive phrase (headed by da ‘by’), if a full nominal, as in (50b). The different encoding of the causee reflects a difference in control, low in the former form (i.e. the a-phrase), high in the latter (i.e. the da-phrase). If the causee is realized by a clitic pronoun, it goes in the dative and occurs before the complex predicate fare/lasciare + V, as in (50d): (50) a. Marco fece /lasciò cadere il cucchiaio. Mark make.pst.3sg /let.pst.3sg fall.inf the.m.sg spoon.m.sg ‘Mark made/let the spoon fall.’ a′. Marco lo fece /lasciò cadere. Mark it.acc.cl make.pst.3sg /let.pst.3sg fall.inf ‘Mark made/let it fall.’ b. Marco fece /lasciò scartare il pacco a Mark make.pst.3sg /let.pst.3sg unwrap.inf the.m.sg parcel.m.sg to Luca /da Luca. Luke /by Luke ‘Mark made/let Luke unwrap the parcel.’ c. Marco fece /lasciò inviare il libro a Luca Mark make.pst.3sg /let.pst.3sg send.inf the.m.sg book.m.sg to Luke da Anna. by Anna ‘Mark made/let the book to be sent to Luke by Anna.’ d. Marco gli fece /lasciò inviare il Mark he.dat.cl make.pst.3sg /let.pst.3sg send.inf the.m.sg libro da Anna. book.m.sg by Anna ‘Mark made/let Anna send him the book.’ With ditransitives, although the causee is generally encoded as a da-phrase if it is a nominal, as in (50c–d), marginally it can also be realized as an a-phrase, although with a different ordering of the causee and the recipient (50e), which cannot be adjacent, since they are both marked by the preposition a ‘to’, as shown in (50f):

450

Michela Cennamo

(50) e. Marco fece /lasciò inviare a(d) Anna (causee) Mark make.pst.3sg /let.pst.3sg send.inf to Anna il libro a Luca (recipient). the.m.sg book.m.sg to Luke ‘Mark made/let Anna send the book to Luke.’ f. Marco fece /lasciò inviare il libro *a(d) Mark make.pst.3sg /let.pst.3sg send.inf the.m.sg book.m.sg to Anna (causee) a Luca (recipient). Anna to Luke If the dependent infinitive is a stative verb, in the pattern with the permissive verb lasciare ‘let’, the causee can only be realised as an a-phrase, never as a da-phrase (50g): (50) g. Marco lasciò vedere il quadro a Luca /*da Luca. Mark let.pst.3sg see.inf the.m.sg painting.m.sg to Luke /*by Luke ‘Mark let Luke see the painting.’ When the dependent infinitive is a reflexive verb, in the factitive construction the verb occurs without the reflexive morpheme si, as in (51): (51) Marco fece/lasciò ribellare (*ribellare) Giovanni. Mark make.pst.3sg/let.pst.3sg rebel.inf (*rebel himself) John ‘Mark made/let John rebel.’ With bivalent (object) experiencer verbs (type ii), in the causative construction the causee can be overtly expressed, realised as a da-phrase, only if the verb denotes a change of state and the subject is thematically specified, i.e. agentive, as with spaventare in (52a), that can be contrasted with the ungrammaticality of (52b), with the stative verbs preoccupare ‘worry’ (type ii) and temere (type i), if the object is inanimate (52g) (see also § 3.4) (Bentley 2006: 103–105; Cennamo 2011a): (52) a. Marco fece spaventare i ragazzi da Luca /*a Luca. Mark make.pst.3sg frighten.inf the.m.pl boy.m.pl by Luke /to Luke ‘Mark made Luke frighten the boys.’ (lit. ‘made frighten the boys by Luke/*to Luke.’) b. Marco fece preoccupare i ragazzi *da /a Luca. Mark make.pst.3sg worry.inf the.m.pl boy.m.pl by /to Luke ‘*Mark made Luke worry the boys.’

Valency patterns in Italian

451

Interestingly, with class (ii) experiencer verbs the causee can only be introduced by the preposition da ‘by’, never by the preposition a ‘to’, as shown in (51a), in line with the aspectual and thematic constraints on causativization with these verbs. Conversely, with experiencer stative verbs of class (iii), inversion predicates such as piacere, where A, the experiencer, is in the dative, the causee can only be headed by the preposition a ‘to’, never by da ‘by’, as in (52c): (52) c. Marco fa piacere la matematica a tutti Mark make.prs. 3sg like.inf the.f.sg maths.f.sg to all.m.pl /*da tutti. /*by all.m.pl ‘Mark makes everybody like maths.’ With class (i) experiencer verbs (e.g., amare ‘love’, ammirare ‘admire’, odiare ‘hate’, etc.) the animacy of P appears to determine the coding of the causee, realised as a da-phrase if animate (52d, f) and as an a-phrase if inanimate, as in (52e, g): (52) d. Marco fa amare i suoi studenti da Mark make.prs.3sg love.inf the.m.pl his.m.pl student.m.pl by tutti /a tutti. all.m.pl /to all.m.pl ‘Mark makes everybody be fond of/love his students.’ e. Marco fa amare la matematica a tutti Mark make.prs.3sg love.inf the.f.sg maths.f.sg to all.m.pl /*da tutti. /*by all.m.pl ‘Mark makes everybody like maths.’ f. Marco fa temere il suo cane da tutti Mark make.prs.3sg fear.inf the his.m.pl dog by all.m.pl /*a tutti. /*to all.m.pl ‘Mark makes everybody fear his dog.’ (lit. ‘makes fear the his dog by/to all.’) g. Marco fa temere il suo comportamento a Mark make.prs.ind.3sg fear.inf the.m.sg his.m.sg behaviour.m.sg to tutti /*da tutti. all.m.pl /*by all.m.pl ‘Mark makes everybody fear his behaviour.’ (lit. ‘makes fear the his behaviour to everybody/by everybody.’)

452

Michela Cennamo

5.2 The reflexive voice A major and very pervasive valency decreasing marker is the reflexive morpheme si, the unstressed third person singular/plural reflexive pronoun, occurring with a variety of functions, signalling coreference between the A and P arguments of bivalent verbs, and between A and R with trivalent ones, as in its reflexive-middle uses (§ 5.2.1 and 5.2.5), a suppressed A, as in its anticausative (§ 5.3) and passive functions (§ 5.4), as well as a logically implied but unexpressed (A/S or P) argument with variable referential status or even lack of an argument, as in its impersonal uses ( § 5.5 and § 5.6). With all si-patterns the auxiliary selected is BE.

5.2.1 The reflexive-middle continuum Depending on the syntactic valency of the verb, the argumental/non-argumental status of si, the animacy and degree of control/affectedness of the A subject, different subclasses of reflexives can be identified: direct, indirect, reciprocal reflexive, middle, inherent, anticausative, passive. The distinction among different types of reflexive patterns, however, is to be seen as a gradient, with overlapping of categories at their periphery, in their non-canonical realizations. It is not always easy, in fact, to detect the function of the reflexive morpheme si, owing to the complex interplay between syntactic and semantic features, resulting also from different diachronic paths (Cennamo 1993, 1995, 1999, 2011c, 2014).

5.2.2 Direct and Indirect/Dative Reflexive In a direct reflexive pattern the verb is bivalent, A is animate, high in Potency and coreferent with P, which is signalled by the reflexive morpheme si, that is an argument of the verb, as shown by the applicability of the substitution test, whereby si can be replaced either by the tonic form se stesso ‘oneself’ or by a clitic complement pronoun, as in (53a): (53) a. Marco si protegge con la menzogna. Mark refl protect.prs.3sg with the.f.sg lie.f.sg ‘Mark protects himself by lying.’ a′. Marco protegge se stesso /lo protegge con Mark protect.prs.3sg refl himself /it.acc protect.prs.ind.3sg with la menzogna. the.fsg lie.fsg ‘Mark protects himself by lying.’ In compound tenses the pattern with the tonic form se stesso and its variants (feminine singular se stessa, masculine plural se stessi, feminime plural se stesse) take

Valency patterns in Italian

453

the auxiliary HAVE, unlike the structure with si, that selects BE, as shown in (53b–c): (53) b. Marco ha protetto se stesso con la Mark have.prs.3sg protect.pp.m.sg refl.tf self.m.sg with the.f.sg menzogna. lie.f.sg ‘Mark protected himself by lying.’ (lit. ‘Mark has protected himself with the lie.’) c. Marco si è protetto con la menzogna. Mark refl be.prs. 3sg protect.pp.m.sg with the.f.sg lie.f.sg ‘Mark protected himself by lying.’ (lit. ‘Mark has protected himself with the lie.’) In the indirect/dative reflexive the verb is trivalent and si is an argument of the verb, R, coreferent with the A subject, denoting either the goal/beneficiary of the verbal activity or Possession. In its canonical realization, si has the syntactic function of an indirect object, as in (54a). In its non-canonical realization, it is not an argument of the verb, but denotes the degree of involvement/participation of the subject in the verbal activity, as in (54b) (so-called Benefactive/Ethic Dative), frequently used in informal registers (Cennamo 2011b and references therein): (54) a. Marco si comprò una cravatta. Mark refl buy.pst.3sg a.f tie.f.sg ‘Mark bought himself a tie.’ b. Marco si lesse un libro. Mark refl read.pst.3sg a.m book.m.sg ‘Mark read a book.’ (lit. ‘read himself a book’)

5.2.3 Reciprocal Reflexive The reflexive also marks reciprocal situations, as with (di)transitives, with the two nuclear arguments A and P acting on each other and being both agent and patient of the verbal activity (benefactive in Indirect Reciprocal Reflexive), as shown in (55) (Cennamo 2011b and references therein): (55) Essi si abbracciarono. they refl hug.pst.3pl ‘They hugged each other.’

454

Michela Cennamo

5.2.4 Middle reflexive In its middle function the reflexive is not an argument of the verb (as shown by the non-applicability of the substitution test), but a marker of the degree of involvement of the non-agentive subject, lacking control over the verbal process. Most typically, the notion of control usually differentiates the reflexive and middle patterns, as in (56) (Jezek 2003: 123; Cennamo 2011b; among others): (56) a. involuntary action Luca si è bagnato mentre riparava Luke refl be.prs.3sg get_wet.pp.m.sg while repair.pst.3sg il lavandino. the.m.sg basin.m.sg ‘Luke got wet while repairing the basin.’ b. voluntary action Luca si è bagnato per rinfrescarsi. Luke refl be.prs.3sg get_wet.pp.m.sg for refresh.inf.refl ‘Luke got wet deliberately in order to get cooler.’

(middle)

(reflexive)

Interestingly, with object experiencer verbs denoting a state such as preoccupare ‘worry’ (i.e. type (ii)), the reflexive pattern can only have a middle interpretation, as in (56c), whilst telic object experiencer verbs like spaventare ‘frighten’, allow both a middle and a reflexive interpretation, depending on the agentivity of the subject (i.e. control), as shown in (56d): (56) c. middle interpretation I ragazzi si preoccupano. (*di proposito). the boy.m.pl refl worry.prs.3pl (of purpose) ‘The boys get worried on purpose.’ (lit. ‘worry on purpose’) d. reflexive interpretation Quei due si spaventano intenzionalmente ogni volta che this.m.pl two refl frighten.prs.3pl intentionally every time that ne hanno l’occasione. of.it have.prs.ind.3pl the.opportunity.f.sg ‘They deliberately frighten each other every time they have the opportunity for it.’ (Belletti & Rizzi 1988: 298)

Valency patterns in Italian

455

5.2.5 Inherent reflexives Some monovalent and bivalent (most typically denominal) stative (e.g., fidarsi ‘trust’, vergognarsi ‘shame’) and activity (e.g., congratularsi ‘congratulate’, ribellarsi ‘rebel’) verbs with an animate, often non-agentive subject, obligatorily take the reflexive morpheme, as in so-called inherent reflexives (Salvi & Vanelli 2004: 206), illustrated in (57) for a bivalent verb (Jezek 2003: 193 for a partial sample list): (57) Giovanna si è congratulata con il vincitore. Jane rfl be.prs.3sg congratulate.pp.f.sg with the.m.sg winner.m.sg ‘Jane congratulated the winner.’ (lit. ‘congratulated herself with the winner’)

5.2.6 Other alternations involving the reflexive With some verbs the presence of the reflexive does not fall under any of the above categories. For instance, with stative verbs of spatial configuration such as sedere ‘be seated’, the corresponding reflexive form sedersi ‘to sit down’, denotes a change of state, thus si acts as a telicizer, as shown in (58b) (Salvi & Vanelli 2004: 50; Cennamo 2011c): (58) a. Marco sedeva sempre lì. Mark sit.impf.3sg always there ‘Mark was always seated there.’ b. Marco si sedeva sempre lì. Mark refl sit.impf.3sg always there ‘Mark always sat down there.’ With non-canonical, oblique object experiencer verbs like dispiacere ‘be sorry’, dolere ‘regret’, interessare ‘interest’ (i.e. class iii), the alternation between the reflexive/non reflexive form functions as an applicative-like pattern (§ 6.3), with the reflexive registering on the verb the upgrading of an adpositional argument to subject function, as in (59b), with the original subject (P) encoded adpositionally (Cennamo 2011c): (59) a. Questo dispiace molto ai ragazzi. this displease.prs.3sg a_lot to.the.m.pl boy.m.pl ‘This greatly displeases the boys.’ (lit. ‘disappoints to the boys’) b. I ragazzi si dispiacciono molto di ciò. the boy.m.pl refl displease.prs.3pl a_lot of this ‘The boys are very disappointed for this.’ (lit. ‘displease themselves of this’)

456

Michela Cennamo

With other (class i) experiencer verbs (i.e. bivalent states) (e.g., dimenticare ‘forget’, ricordare ‘remember’) and other activity verbs (e.g., sbagliare ‘make a mistake’) the presence of the reflexive signals the lack of control of the subject over the verbal process, as shown in (59c–d):

(59) c. Marco ha dimenticato le chiavi da (+control) Mark have.prs.3sg forget.pp.m.sg the.f.pl key.f.pl at Giovanna per riveder-la. Jane for see_again.inf-she.acc ‘Mark has forgotten the keys at Jane’s in order to see her again.’ d. Marco si è dimenticato le chiavi da (–control) Mark refl be.prs.3sg forget.pp.m.sg the.f.pl key.f.pl at Giovanna *per riveder-la. Jane for see-again.inf-she.acc ‘Mark happened to forget his keys *in order to see her again.’

5.3 Coded and uncoded anticausatives This type of valency decreasing operation may be both coded, marked by the reflexive morpheme si, and uncoded, with formal identity between the intransitive and the corresponding transitive verb form. The process is presented as taking place spontaneously (Haspelmath 1987, 1993; among others), without an external causer (A), regarded as either suppressed both syntactically and semantically (Haspelmath 1987: 7) or as suppressed at argument structure (i.e. the lexical syntactic representation) but retained in the lexical semantic representation (Levin & Rappaport Hovav 1995: 20–21, 84; Koonz-Garboden 2009: 97). The pattern has different morphosyntactic realizations, characterized by the obligatory presence, absence or optionality of si, and by the selection of the auxiliaries BE, HAVE or HAVE/BE in compound tenses (Cennamo 1995, 2011c; Cennamo & Jezek 2011; among others). Three parameters determine the different distribution of si and of perfective auxiliaries: the aspectual nature of the verb/predicate, (inherent/relational) properties of the subject (animacy, control/affectedness), lexical (the meaning components encoded in the verb). According to the interplay among these factors and their morphosyntactic manifestations, three subtypes of anticausatives can be identified (Folli 2002; Jezek 2012; Manente 2008): Class 1: [+si], [+BE], comprises inherently telic predicates: achievements/accomplishments (e.g., rompersi, ‘break’, spezzarsi ‘crack’, spegnersi ‘go out’), as shown in (60a), although including also degree achievements/gradual completion verbs (e.g., svuotarsi ‘empty’), activities (e.g., esprimersi ‘express’) and states (e.g., basarsi ‘base’):

Valency patterns in Italian

457

(60) a. La sedia si è rotta. the.f.sg chair.f.sg refl be.prs.3sg break.pp.f.sg ‘The chair has broken.’ Class 2: [–si] [+BE], consists of predicates with different degrees of telicity, denoting a gradual change of state or the gradual approximation to a terminal point along a scale, which may or may not be attained, as with degree achievements/gradual completion verbs (e.g., aumentare ‘increase’, migliorare ‘improve’) (Centineo 1995; Sorace 2000: 864). This class comprises, however, also some inherently telic verbs such as guarire ‘heal’, affondare ‘sink’, cambiare ‘change’ and, marginally, activities (e.g., continuare ‘continue’): (60) b. I prezzi sono aumentati per mesi. the.m.pl price.m.pl be.prs.3pl raise.pp.m.pl for month.m.pl ‘Prices have gone up for months.’ Class 3: [±si], involves verbs describing a complex event consisting of a change process and an optional telos. The existence of a final state is possible but not necessary (Folli 2002; Schäfer 2008). It comprises accomplishments (e.g., fonder(si) ‘melt’, bruciar(si) ‘burn’, cuocer(si) ‘cook’, gelar(si) ‘freeze’) and degree achievements/gradual completion verbs (e.g., sbiadire ‘fade away’, ingiallire ‘discolour/ turn yellow’). With these verbs the focus is on the attainment of a final state in the pattern with si, as in (60c), and on the process if si is lacking, and ensuing telic/ atelic interpretation with related BE/HAVE selection in compound tenses, as in (60c, d) vs. (60e), respectively (Sorace 2000: 874–875): (60) c. Il bosco si è bruciato (in poco tempo). the.m.sg forest.m.sg refl be.prs.3sg burn.pp.m.sg in short time ‘The forest burnt in a short time.’ d. Il bosco è bruciato in poco tempo. the.m.sg forest.m.sg be.prs.3sg burn.pp.m.sg in short time ‘The forest burnt in a short while.’ e. Il bosco ha bruciato per ore. the.m.sg forest.m.sg have.prs.3sg burn.pp.m.sg for hour.f.pl ‘The forest burnt for hours.’ With some verbs, however (e.g., bruciare ‘burn’) BE is also acceptable in an atelic context (60f) and marginally, in the informal speech, HAVE can occur in a telic one (60g) (Manente 2008: 200, 212; Cennamo & Jezek 2011: 817) (59f): f. Il bosco è bruciato per ore. the.m.sg forest.m.sg be.prs. 3sg burn.pp.m.sg for hour.f.pl ‘The forest burnt for hours.’

458

Michela Cennamo

g. Il bosco ha bruciato in poche ore. the forest have.prs.3sg burn.pp.m.sg in few hour.f.pl ‘The forest burnt in a few hours.’ Si therefore functions both as a detransitivizer, a marker of the suppressed causer (Cennamo 1995; Bentley 2006: 134) and as the marker of the final (Folli 2002)/ result/target state (Manente 2008; Jezek 2008) in the lexical meaning of the verb. Thus, there appear to be three general semantic constraints on anticausativization in Italian: (i) Aspectual, reflecting the inherent/compositional temporal structure of the verb/predicate, (ii) Thematic, whereby only verbs with a thematically underspecified subject (i.e. agent, instrument, natural cause) can become the subject of a corresponding anticausative form, (iii) Inherent properties of the subject: only the inanimate object of a bivalent telic/atelic verb can become the subject of a corresponding anticausative form. As for the marking of this alternation, degree achievements/gradual completion verbs generally show labile behaviour (lacking an overt morphological marking, the reflexive morpheme si), with exceptions/mismatches reflecting residues of early stages where anticausativization was marked through the morpheme si with all aspectual classes of verbs, alternating with the labile pattern (Cennamo 2012a).

5.4 Reflexive Passive The third person singular/plural reflexive morpheme si also marks a P-oriented pattern, as in so-called reflexive passives, where the original P of a corresponding transitive verb occurs as subject and A is suppressed and cannot be overtly expressed. In the unmarked word order the [±animate] subject occurs in the preverbal position, it is definite, referential and conveys given information, as in (61) (Cennamo 1995: 85–86). In compound tenses the auxiliary selected is always BE essere, in all reflexive patterns. The pattern is possible with all bivalent and trivalent verbs (Bentley 2006; D’Alessandro 2007; Cennamo 2011b and further references therein). (61) I libri_gialli si leggono con piacere. the.m.pl thriller.m.pl refl read.prs.3pl with pleasure ‘One reads thrillers with pleasure.’ (lit. ‘thrillers are read with pleasure’)

5.5 Impersonal of (lexicalized) reflexives In the corresponding impersonal form of (direct/indirect/middle/inherent) reflexives, the defocused A/S argument (the subject), surfaces as ci, the 1 personal plural pronoun replacing impersonal si, whereby one finds the ci si V pattern instead of the sequence si si V, as in (62) (Cennamo 2011b, and references therein):

Valency patterns in Italian

459

(62) a. Ci si pente. imps refl repent.prs.3sg ‘One regrets.’ b. Ci si è pentiti-e. imps refl be.prs.3sg repent.pp.m.pl-f.pl ‘One has repented. / We repented.’ c. Ci si compra una penna. imps refl buy.prs.3sg a.f pen.f.sg ‘One buys oneself a pen.’ d. Ci si è comprati-e-a una penna. imps refl be.prs.3sg buy.pp.m.pl-f.pl-fsg a.f pen.f.sg ‘One has bought oneself a pen.’ In compound tenses there occurs split agreement: the finite verb is in the unmarked third person singular, whilst the past participle is in the plural (masculine or feminine, according to the context), as in (62b, d), following the general rule of Italian whereby in impersonal patterns the nominal predicate is in the plural (Salvi 1988; Cennamo 1993: 36), albeit optionally agreeing with the P argument (penna in (62d)).

5.6 Impersonal Reflexive Si can also signal an indefinite S/A, as in (63a–b) or P (in the so-called impersonal passive pattern, where A can be overtly expressed, surfacing as an adpositional phrase headed by da ‘by’) as in (63c): (63) a. Si parte /si lavora /si è bambini refl leave.prs.3sg /refl work.prs.3sg /refl be.prs.3sg child.m.pl /allegri. /happy.m.pl ‘One/they (indef.)/we leave/work; one is a child/is happy.’ b. Si vendono libri /si mangia /si mise un refl sell.prs. 3pl book.m.pl /refl eat.prs. 3sg /refl put.pst.3sg a.m libro sul tavolo. book.m.sg on.the.m.sg table.m.sg ‘One/they (indef.)/we sell books/one, they (indef.), we eat/one, they (indef.) we put a book on the table.’

460

Michela Cennamo

c. Si è pagati meno dallo Stato. refl be.prs.3sg pay.pp.m.pl less by.the.m.sg Government.m.sg ‘One/they (indef.)/we are paid less by the Government.’ With transitive verbs, in the unmarked word order the subject occurs in the postverbal position, and is most typically indefinite, non-referential and conveys new information, as in (63b) (Cennamo 1995: 85–86; Bentley 2006; D’Alessandro 2007; Cennamo 2011b and references therein). In most northern dialects, Florentine and in the corresponding regional varieties, although with varying degrees of acceptability (Salvi 1988: 102), reflecting the definiteness/indefiniteness of the nominal (D’Alessandro 2007), one finds, however, the non-agreeing form (63d–e)6 (Cennamo 1997: 153–54), that in standard Italian is only possible when si is cliticised onto the finite verb, in fixed phrases such as (63f): (63) (Example (63d) is Genoese) d. Se ‘leze fasil’mente i ‘libri. refl read.prs.3sg easily the.m.pl books.m.pl ‘One reads the books easily.’ e. Si vende libri /i libri. refl sell.prs.3sg book.m.pl /the.m.pl book.m.pl ‘One sells books/the books; books/the books are sold.’ e′. Li si vende. (li = the books) they.clm.acc refl sell.prs.3sg ‘One sells them; they are sold.’ f. Affittasi stanze (*affittansi) /vendesi rent.prs. 3sg.refl room.f.pl (*rent.prs.3pl /sell.prs. 3sg.refl appartamenti (*vendensi). apartment.m.pl sell.prs. 3pl ‘Rooms are for rent/apartments are on sale.’ When the nominal (object) is pronominalised, however, in standard Italian the non-agreeing pattern occurs, with the verb in the default third person singular inflection, as in (63e′). The logically implied understood participant can be generic or existential/indefinite, also optionally comprising the Speech Act Participants, as

6 In Florentine and other Tuscan, Umbrian and some Lazio varieties, however, (63e) has a different, first person plural interpretation, ‘we sell books/the books’, since si+3SG has replaced the 1PL ending (Cennamo 1997: 158, 2014: 82).

Valency patterns in Italian

461

in the first person plural, inclusive interpretation, according to the temporal reference of the clause and the aspectual nature of the predicate. With unaccusatives, in fact, the generic interpretation is impossible with specific time reference, unlike with unergatives, as shown in (64a), that only allow an inclusive interpretation: (64) a. Ieri si è partiti presto. yesterday refl be.prs.3sg leave.pp.m.pl early ‘We left early.’ (*One left early) b. Ieri si è camminato a lungo. yesterday refl be.prs.3sg walk.pp.m.sg at length ‘Yesterday they (indef.)/we walked for a long time.’ The inclusive interpretation of impersonal si, however, can be ‘suspended’ with unaccusatives if the predicate is temporally unbounded, e.g., in a hypothetical clause, as in (64c), that allows both an inclusive and a generic interpretation (Cinque 1988: 150; D’Alessandro 2007: 154 f.; Cennamo 2014: 76): (64) c. Ieri, se si fosse arrivati tardi yesterday if refl be.pst.subj.3pl arrive.pp.m.pl late all’appuntamento. at.the.appointment.m.sg ‘Yesterday, if we had arrived/they (indef.) had arrived late at the appointment.’ Unaccusatives and unergatives also trigger a different agreement pattern in compound tenses with impersonal si, with which the auxiliary BE is selected, with all verb classes. The past participle, in fact, is in the umarked masculine singular form with unergatives (and transitives), as in (65a–b), whereas it is in the plural (masculine/feminine) with unaccusative verbs/predicates, as in (65c–d) (thus also with equative structures and passives): (65) a. unergative Si è giocato. refl be.prs.3sg play.pp.m.sg ‘One/they (indef.)/we have played.’ b. transitive Si è letto /pagato. refl be.prs.3sg read.pp.m.sg pay.pp.m.sg ‘One/they (indef.)/we have read/paid.’

462

Michela Cennamo

c. unaccusative Si è partiti presto. refl be.prs.3sg leave.pp.m.pl early ‘We left early.’ d. passive/equative Si è letti /pagati /felici. refl be.prs.3sg read.pp.m.pl /pay.pp.m.pl /happy.m.pl ‘One/they (indef.)/we are read/paid/happy.’ Interestingly, whereas impersonal si is possible with all verb classes in finite tenses, in non-finite forms of the verbs its occurrence is lexically constrained. For instance, in the gerund it is impossible with unaccusatives (66a) (Salvi & Vanelli 2004: 76), whilst being only marginally acceptable with unergatives (66b): (66) a. *Essendosi partiti all’alba. be.ger.refl leave.pp.m.pl at.the.dawn ‘Having we left at dawn.’ b. ??Essendosi lavorato /dormito fino a tardi. be.ger.refl work.pp.m.sg /sleep.pp.m.sg till at late ‘Having one/they (indef.)/we worked/slept till late.’

5.7 (Analytic) passives Passives, i.e. patterns with a P subject and suppression of A, optionally surfacing as an adpositional phrase headed by da ‘by’/da parte di ‘on behalf of’ in Italian, are expressed by means of a marked verb morphology involving the auxiliaries essere BE, venire COME, andare GO + the past participle of the lexical verb, agreeing in gender and number with the subject, with different aspectual constraints. BE (essere) occurs in simple and compound tenses (i.e. in dynamic-eventive and resultative functions), with all verbs, as in (67a). COME (venire) occurs only in simple tenses, i.e. only with a dynamic-eventive function (67b); GO (andare) occurs mainly with deontic value (dynamic-eventive passive), as in (66b), and only marginally in non-deontic function, in a resultative construction, with a restricted number of telic verbs and inanimate subjects (e.g., distruggere ‘destroy’, perdere ‘lose’), as in (67d) (Bentley 2006: 346–349 and references therein): (67) a. Il tetto fu distrutto da un fulmine. the.m.sg roof.m.sg be.pst.3sg destroy.pp.m.sg by a.m lightening.m.sg ‘The roof was destroyed by a lightening.’

Valency patterns in Italian

463

b. Il tetto verrà riparato domani. the.m.sg roof.m.sg come.fut.3sg repair.pp.m.sg tomorrow ‘The roof will be repaired tomorrow.’ c. Il tetto va sostituito /riparato. the.m.sg roof.m.sg go.prs.3sg replace.pp.m.sg /repair.pp.m.sg ‘The roof ought to be replaced/repaired.’ (lit. ‘goes repaired’) d. Il tetto è andato distrutto the.m.sg roof.m.sg be.prs.3sg go.pp.m.sg destroy.pp.m.sg /*rotto. /*break.pp.m.sg ‘The roof has been destroyed.’ (lit. ‘has gone destroyed/*broken’) Passivization is possibile with achievements (e.g., strappare ‘tear’), accomplishments (e.g., bruciare ‘burn’, affondare ‘sink’, costruire ‘build’, uccidere ‘kill’), activities (e.g., dire ‘tell’, abbracciare ‘hug’), and states (e.g., vedere ‘see’, amare ‘love’), most typically with postverbal subjects if they are indefinite/generic, although preverbal subjects are equally possible with achievements and accomplishments (68a): (68) a. Ogni anno vengono /sono distrutti querceti every year come.prs.3pl /be.prs.3pl destroy.pp.m.pl oak-grove.m.pl e aranceti /querceti e aranceti and orange-grove.m.pl oak-grove.m.pl and orange-grove.m.pl vengono /sono distrutti … come.prs.3pl /be.prs.3pl destroy.pp.m.pl ‘Every year oak-groves and orange-groves are destroyed.’ In contrast, passivization of activity verbs with an indefinite/generic P is impossible, regardless of the pre-postverbal position of the subject (68b–c) vs. (68d) (Van Valin & La Polla 1997: 149, 657, note 4; Cennamo 2003: 53 for an analogous constraint in English). A different strategy is used, the reflexive passive, as shown in (68e), the corresponding patterns of the ungrammatical (68c): (68) b. *Carne fu mangiata /mele furono mangiate meat.f.sg be.pst.3sg eat.pp.f.sg apple.f.pl be.pst.3pl eat.pp.f.pl (da Anna). by Anna ‘Meat was eaten/apples were eaten (by Anna).’ c. *?fu mangiata carne/furono mangiate mele (da Anna) d. La carne fu mangiata /le mele the.f.sg meat.f.sg be.pst.3sg eat.pp.f.sg the.f.pl apple.f.pl

464

Michela Cennamo

furono mangiate da Anna. be.pst.3pl eat.pp.f.pl by Anna ‘The meat was eaten/the apples were eaten by Anna.’ e. Si mangiò carne /si mangiarono mele. refl eat.pst.3sg meat.f.sg /refl eat.pst.3pl apple.f.pl ‘Meat was eaten/apples were eaten.’ (Other possible interpretation: ‘One, they (indef.) ate meat/apples.’)

6 Other argument rearranging alternations There also occur a number of valency rearranging operations involving core (namely P) and non-core arguments, that are very frequent in other languages (e.g., English) but that appear to be confined to a limited number of activity verbs in Italian: the conative (§ 6.1), the locative (§ 6.2) and the applicative(-like) (§ 6.3) constructions.

6.1 The conative alternation The intransitive variant of a transitive pattern, with the prepositional encoding of the P argument reflecting the lack of attainment of the verbal process, and related low degree of affectedness of P, the so-called conative alternation (Guerssel et al. 1985; Levin 1993: 41–42; Cennamo 2003: 56, 2011c; among others), exemplified in (68a), is only marginally attested in Italian (see also discussion in § 3.4). Unlike in English and other languages, where this alternation occurs with verbs of impact/ concussion/contact/motion (Levin 1993: 42, Cennamo 2003 and further references therein), in Italian this pattern is attested mainly with verbs of saying (e.g., discutere ‘discuss’, raccontare ‘tell’, accennare ‘hint, mention’) and of mental process (e.g., pensare ‘think’). Also other activity verbs allow this alternation when P is inanimate (e.g., applaudire ‘applaud’, ammiccare ‘wink’, supplire ‘replace’), at times in figurative uses and only with abstract P arguments. In the last use marginally also accomplishments may be found, as in (69) (the issue, however, needs further investigation) (see also discussion in Lenci 2012: 8, 13): (69) a. conative coding Marco raccontò della propria esperienza. Mark narrate.pst.3sg of.the.f.sg own.f.sg experience ‘Mark talked about his experience.’ (lit. ‘told of his experience’)

Valency patterns in Italian

465

b. canonical coding Marco raccontò la propria esperienza. Mark narrate.pst.3sg the.f.sg own.f.sg experience ‘Mark narrated his experience.’ c. canonical coding Marco riparerà i danni causati Mark remedy. prs.fut.3sg the.m.pl damage.m.pl cause.pp.m.pl dal suo predecessore. by.the.m.sg his predecessor ‘Mark will remedy the damages caused by his predecessor.’ d. conative coding Marco riparerà ai danni causati Mark remedy. prs. fut.3sg to.the.m.pl damage.m.pl cause.pp.m.pl dal suo predecessore. by.the.m.sg his predecessor.m.sg ‘Mark will remedy the damages caused by his predecessor.’

6.2 The locative alternation Also the locative alternation has a low degree of productivity in Italian. In this type of alternation the canonical/non-canonical (i.e. adpositional) encoding of the P argument of a bivalent verb, headed by the preposition di ‘of’, reflects the holistic/ partitive interpretation of the location argument (Levin 1993: 50). When the location argument is expressed as a canonical object, as in (70b), it is associated with a holistic/affected interpretation; when the location argument is realized as an adpositional phrase, as in (70b), it is associated with a partitive interpretation (Levin 1993: 50–51; Iwata 2008 and references therein). (70) a. Caricai la macchina di libri. load.pst.1sg the.f.sg car.f.sg of book.m.pl ‘I loaded the car with books.’ b. Caricai i libri in macchina. load.pst.1sg the.m.pl book.m.pl in car.f.sg ‘I loaded the books onto the car.’

6.3 The applicative(-like) pattern Italian also seems to allow, although, marginally, an applicative-like alternation, a transitive pattern (71a) derived from an intransitive construction (71b) with an origi-

466

Michela Cennamo

nal adjunct ‘promoted’ to object status. Unlike in canonical applicative constructions, however, there occurs no change in the verb morphology. This pattern appears with some activity verbs with an emotional overtone (e.g., urlare, gridare ‘scream’, piangere ‘cry’, strillare ‘yell’, soffrire ‘suffer’: (71) a. Marco urlò il proprio dolore. Mark scream.pst.3sg the.m.sg own.m.sg sorrow/pain.m.sg ‘Mark screamed with pain.’ (lit. ‘Mark screamed his pain/sorrow.’) b. Marco urlò per il dolore. Mark scream.pst.3sg owing.to the.m.sg pain/sorrow.m.sg ‘Mark screamed with pain/under sorrow.’

7 Conclusion The analysis of the coding and behavioural properties of different valency classes of verbs in Italian reveals the existence of a number of regularities, reflecting the interplay of the structural aspect of a verb meaning (i.e. its event structure template) with the elements of meaning lexicalized in its root (e.g., state, target/result state, manner, location etc.) and the inherent/relational properties of their arguments (e.g., animacy and control/affectedness). As regards the coding properties of verbs, there appear to emerge the following general characteristics: Flagging (instantiated by adpositional marking, with a minimal case distinction for non-subject clitic and 1/2/3 SG tonic pronouns), although mainly used to mark non-core arguments and adjuncts, may be used to mark an atypical P, as shown by the adpositional marking of P with some activity/ state verbs, reflecting its animacy and low degree of affectedness, as with the verb pensare ‘think’. Indexing (realized by verb agreement and auxiliary selection) is the main coding strategy for core arguments, S, A and P: the verb agrees with A/S in simple tenses in the active voice, whilst agreeing with the P subject in the passive. In perfective contexts (i.e. in compound tenses) indexing is sensitive to aspectual and thematic notions, namely telicity and control/affectedness of S, as well as to animacy, thereby differentiating two subtypes of S. Consistent BE selection+past participle agreement occurs with achievements and accomplishments, i.e. verbs of telic change and a patient S (SP), whilst consistent selection of HAVE+ lack of agreement occur with activity verbs, i.e. dynamic atelic verbs with an agent S (SA). Variable auxiliary selection is displayed instead by stative verbs, whose flexible behaviour reflects the range of possible interpretations, resulting from their (compositional) aspectual reclassification (e.g., as achievements and accomplishments). The presence/lack of agreement also signals the SP/P status of the underlying unexpressed argument with impersonal si constructions, and the P status of the omitted

Valency patterns in Italian

467

argument of some divalent activity verbs allowing object omission in a predicative construction, such as visitare ‘visit’, ritrarre ‘paint’. Also for the behavioural coding of verbs, the distribution of syntactic alternations across verbs reflects aspectual, thematic notions as well as the nature of arguments (e.g., animacy). In particular, the omission of P occurs with activities, active accomplishments (i.e. verbs which do not lexicalize a final point), states, as well as with accomplishments with an animate P, such as ammazzare, uccidere ‘kill’, whilst being impossible with achievements and accomplishments with an inanimate P. The oblique subject alternation reflects, instead, the type and degree of telicity lexicalized in the verb, whilst the conative, locative and applicative alternations only apply to activity verbs. The degree of aspectual specification of a verb also plays a role in allowing the overt expression of the causee in causatives with experiencer verbs, that is impossible for stative object experiencer verbs (e.g., preoccupare ‘worry’). On the whole the Italian data confirm general trends observed cross-linguistically, such as the uniform coding and behavioural properties of achievements and accomplishments lexically encoding telicity (i.e. a result state) as opposed to the variability of verbs encoding different types and degrees of telicity, such as active accomplishments and degree achievements, with some properties confined to specific clusters of verbs, such as activities and states. Thus, as regards one pervasive manifestation of indexing, auxiliary selection, the thematic and aspectual parameters determining it in Italian – telic dynamic change ~ patienthood and atelic, non-motional activity ~ agentivity with gradient effects, along Sorace’s 2000, 2004 ASH – also mould it in several European languages (Dutch, French, German) (Sorace 2000; Keller & Sorace 2003; Diedrichsen 2013) as well as in other typologically distant languages such as Chinese (Liu 2007). As for the most widely used alternations, non-reflexive anticausatives tend to occur with degree achievements and marginally activities and states, confirming the preference for P-lability for verbs with lower semantic transitivity across languages (Letuchiy 2009). Also the semantic constraints on the omission of P are not specific to Italian but widespread cross-linguistically, although varying in their morphological realization (e.g., A-lability in Italian and English vs. the presence of the reflexive for indefinite object deletion and a habitual action in Russian and for indefinite object deletion in Dyirbal or the antipassive in Tzutujil (a Mayan language) for accomplishments under a habitual meaning) (Levin 1993: 33; Comrie 1985: 328; Malchukov, this volume; Van Valin & La Polla 1997: 302; Cennamo 2003; among others). In contrast, the lexical restrictions on the conative alternation appear to be peculiar to Italian, where this alternation involves verbs of saying and mental process rather than verbs of impact/concussion/contact/motion, as in English (Levin 1993: 41–42) and other languages (e.g., Warlpiri and Hungarian) (Simpson 1991: 327–333; Moravcsik 1978; Cennamo 2003: 86 and further references therein). The Italian data also point to the usefulness of an approach integrating the insights from typological research on transitivity and valency with the event struc-

468

Michela Cennamo

ture perspective on verb meaning and on the role played by the elements of meaning lexicalized in the root in determining argument realization.

Appendix The table below summarizes the various valency patterns of Italian, with related coding frames and some uncoded and coded alternations. The verbs are listed according to their number of arguments, from zerovalent to trivalent ones.

COUGH

RUN

JUMP

SING

SCREAM

DIE

BE SAD

BE HUNGRY

 47

 49

 52

 53

 58

 61

 63

 64

avere fame

essere triste

morire

urlare

cantare

saltare

correre

tossire

sbattere le palpebre

7 This entry is not part of the database.

BLINK

lampeggiare

‘lighten, flash’7

 46

piovere

RAIN

 69

Verb form

Meaning label

#

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

(S) >V.subj[3SG/S]

Coding frame schema

BE

BE

HAVE

HAVE

BE/ HAVE

BE/ HAVE

HAVE

HAVE

BE/ HAVE

BE/ HAVE

AUX























Object Omission























Oblique Subject























Anticausative























Body part possessor









+













Impersonal of Reflexives

+

+

+

+

+

+

+

+

+





Impersonal Reflexive









m







m





Reflexive Passive









indir.













Reflexive

Valency patterns in Italian

469

cadere

scendere

BE DRY

APPEAR

BE ILL

CRY

FALL

‘go down’8

‘go up’

SIT

SIT DOWN

GO

LIVE

FEEL PAIN

FEEL PAIN

FEEL COLD

 68

 81

 82

 83

 84





 50

 51

 54

 56

 59

 59

 60

Coding frame schema

1 > V.subj[1] (> LOC2)

1 > V.subj[1] (> LOC2)

1 > V.subj[1] (> LOC2)

1 > V.subj[1] (> LOC2)

1 > V.subj[1] (> LOC2)

1 > V.subj[1] (> LOC2)

1 > V.subj[1] (> LOC2)

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

1 > V.subj[1]

8 This and the following entry are not part of the database.

sentire freddo

avere dolore

sentire dolore

abitare

andare

sedersi

sedere

salire

piangere

essere malato

apparire

essere arido

essere affamato 1 > V.subj[1]

BE HUNGRY

 64

Verb form

Meaning label

#

HAVE

HAVE

HAVE

HAVE

BE

BE

BE

BE

BE

BE

HAVE

BE

BE

BE

BE

AUX































Object Omission































Oblique Subject































Anticausative































Body part possessor











+



















Impersonal of Reflexives

+

+

+

+

+



+

+

+

+

+

+

+



+

Impersonal Reflexive





















m









Reflexive Passive































Reflexive

470 Michela Cennamo

FEEL COLD

PLAY

LAUGH

LIKE

EAT

HUG

LOOK AT

SEE

SMELL

FEAR

FRIGHTEN

KNOW

SEARCH FOR

 60

 62

 57

  8

  1

  2

  3

  4

  5

  6

  7

  9

 11

cercare

conoscere

spaventare

temere

odorare

vedere

guardare

abbracciare

mangiare

piacere

ridere

giocare

avere freddo

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

a+1 > V.subj[2] > 2

1 V.subj[1] (di+2)

1 V.subj[1] (con+2)

1 > V.subj[1] (> LOC2)

HAVE

HAVE

HAVE

HAVE

BE/HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

+



m





+

m



+













+













+



























































+

+

+

+

+

+

+

+

+

+







+

+

+

+

+

+

+

+

+



+

+

+

+

+

+

+

+

+

+

+

+





m



dir., indir; (dir.) recipr.

dir.; (dir.) recipr.

mid.; (dir.) recipr.

m; (dir.) recipr.

dir.; (dir.) recipr.

dir./indir.; (dir.) recipr.

dir./ indir.; (dir.) recipr.

m; (dir.) recipr.

indir.

indir.; (dir.) recipr.







Valency patterns in Italian

471

Meaning label

WASH

DRESS

HELP

FOLLOW

MEET

SHOUT AT

PEEL

CLIMB

LEAVE

DIG

#

 12

 13

 15

 16

 17

 20

 33

 48

 55

 73

dissotterrare

lasciare

scalare

sbucciare

sgridare

incontrare

seguire

aiutare

vestire

lavare

Verb form

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

Coding frame schema

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

AUX





+









+

+

+

Object Omission

+





+











+

Oblique Subject





















Anticausative





















Body part possessor

+

+



+

+

+



+

+

+

Impersonal of Reflexives

+

+

+

+

+

+

+

+

m

+

Impersonal Reflexive

+

+

+

+

+

+

+

+

+

+

Reflexive Passive

indir.

–; (dir.) recipr.



indir./mid.; (indir.) recipr.

–; (dir.) recipr.

mid.; (dir.) recipr.

–; (dir.) recipr.

dir.; (dir.) recipr.

dir.; (dir.) recipr.

dir./ indir.; (dir.) recipr.

Reflexive

472 Michela Cennamo

HEAR

HEAR

COOK

WANT

BOIL (tr)

ROLL (tr)

TALK

LEAVE

CARRY

THROW

PUT

POUR

 78

 78

 79

 87

126

128

 18

 55

 38

 39

 41

 42

versare

mettere

lanciare

trasportare

partire

parlare

rotolare

bollire

volere

cuocere

udire

sentire

1 > V.subj[1] > 2 (> LOC3)

1 > V.subj[1] > 2 (> LOC3)

1 > V.subj[1] > 2 (> LOC3)

1 > V.subj[1] > 2 (> LOC3)

1 V.subj[1] (da+2) (per+3)

1 > V.subj[1] > (a+2) (di+3)

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

1 > V.subj[1] > 2

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

BE/HAVE

?

HAVE

HAVE

HAVE

HAVE

















+





+



















+









cod.











unc.

unc.





















+ cod./ + unc.





+

+

+

+



+

+

+

m

+



+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+





+

+

+

+

+

+

indir.; (indir.) recipr.

dir./indir.; (indir.) recipr.

dir.; dir./ indir. recipr.

recipr.



indir.

dir./mid.

indir.

– (dir.) recipr.

indir.; (ind.) recipr.



indir.; (dir.) recipr.

Valency patterns in Italian

473

Meaning label

PUSH

SHAVE

ASK FOR

TELL

SAY

HIDE

SHOW

GIVE

SEND

#

 74

 14

 19

 21

 22

 34

 35

 36

 37

inviare

dare

mostrare

nascondere

dire

raccontare

chiedere

radere

spingere

Verb form

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (> LOC3)

Coding frame schema

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE

AUX









+



+

+

+

Object Omission







+







+



Oblique Subject





+ cod.

+ cod.











Anticausative



















Body part possessor

+

+

+

+

+

+

+

+

+

Impersonal of Reflexives

+

+

+

+

+

+

+

+

+

Impersonal Reflexive

+

+

+

+

+

+

+

+

+

Reflexive Passive

–; indir

dir./indir.; indir

dir./mid.; (indir.) recipr.

dir./indir.; (dir./ indir) recipr.

dir./indir.; (indir.) recipr.

dir.; (indir.) recipr.

–; (indir.) recipr.

dir./ indir.; (indir.) recipr.

dir./indir.; (dir.) recipr.

Reflexive

474 Michela Cennamo

BRING

TEACH

BREAK

KILL

BEAT

HIT

TOUCH

CUT

COVER

FILL

 75

 77

 25

 26

 27

 28

 29

 30

 43

 44

riempire

coprire

tagliare

toccare

urtare

colpire

uccidere

rompere

insegnare

portare

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (a+3)

1 > V.subj[1] > 2 (a+3)

HAVE

HAVE

HAVE

HAVE

?

HAVE

HAVE

HAVE

HAVE

HAVE

m





m



+

+



+



+

+



+

+

+

+



+

+ cod.

+ cod.

+ cod.









+ cod.











+

+

+









+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

dir./indir.; (indir.) recipr.

dir./indir./; (dir.) recipr.

indir./mid.; (dir./indir.) recipr.

dir./indir./ mid.; (dir.) recipr.

–; (dir.) recipr.

dir./indir./ mid.; (dir.) recipr.

dir.

mid.; (dir./ indir) recipr.



indir.; (indir.) recipr.

Valency patterns in Italian

475

ASSASSI-NATE assassinare

BURN (tr)

SINK (tr)

TAKE

TEAR

WIPE

GET

CLEAN

 91

116

125

 31

 32

 72

 86

 90

pulire

ricevere

pulire

strappare

prendere

affondare

bruciare

macinare

GRIND

 71

Verb form

Meaning label

#

1 > V.subj[1] > 2 (da+3)

1 > V.subj[1] > 2 (da+3)

1 > V.subj[1] > 2 (da+3)

1 > V.subj[1] > 2 (da+3)

1 > V.subj[1] > 2 (da+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

1 > V.subj[1] > 2 (con+3)

Coding frame schema

HAVE

HAVE

HAVE

HAVE

HAVE

?

HAVE

HAVE

HAVE

AUX

+



+











+

Object Omission

+



+





+

+



+

Oblique Subject







+ cod.



+ uncod.

+ cod./ unc.





Anticausative

+



+







+





Body part possessor

+

m

+

+

+



+



+

Impersonal of Reflexives

+

+

+

+

+

+

+

+

+

Impersonal Reflexive

+

+

+

+

+

m

+



+

Reflexive Passive

dir./ indir.; (m) recipr.



dir./indir.; (m) recipr.

indir./mid.; (indir.) recipr.

indir.; (dir./indir.) recipr.



mid.



indir.

Reflexive

476 Michela Cennamo

MAKE

LOAD

NAME

TIE

STEAL

BE A HUNTER

 85

 45

 23

 40

 76

 70

no verbal counterpart

rubare

legare

chiamare

caricare

fare

costruire

N/A

1 > V.subj[1] > 2 (a+3) (da+4)

1 > V.subj[1] > 2 (a+3) (con+4)

1 > V.subj[1] > 2 > 3

1 > V.subj[1] > 2 (su+3)

1 > V.subj[1] > 2 (di+3)

1 > V.subj[1] > 2 (di+3)

HAVE

HAVE

HAVE

HAVE

HAVE

HAVE



+





+



+

































+











+

+

m

+

+

+

+

+

+

+

+

+

+

Legend: + = occurs regularly; m = occurs marginally; − = occurs never; _ = no data; dir. = direct; indir. = indirect; mid.= middle.

BUILD

 24



+

+

+

+

+

+



indir.; (indir.) recipr.

dir./indir.; (dir.) recipr.

dir.; (dir.) recipr.

indir.

indir.; (indir.) recipr.

indir.

Valency patterns in Italian

477

478

Michela Cennamo

Abbreviations ACC cod CL DAT dir EXP F FUT IMPF IMPS indir INF M

Accusative coded clitic dative direct experiencer feminine future (tense) imperfect (tense) impersonal indirect infinitive Masculine

mid PP PL PRS PST REFL SG SV(O) TF unc 1 2 3

middle past participle plural present (tense) past (tense) reflexive singular Subject Verb (Object) tonic form uncoded first person second person third person

References Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural Language and Linguistic Theory. 21(3). 435–483. Allerton, David J. 1980. Valency and the English Verb. London: Academic Press. Arad, Maya. 1998. Psych-notes. UCL Working Papers in Linguistics 10. 203–222. Aranovich, Raúl (ed.). 2007. Split Auxiliary Systems. Amsterdam: Benjamins. Belletti, Adriana & Luigi Rizzi 1988. Psych verbs and theta theory. Natural Language and Linguistic Theory 6. 291–352. Bentley, Delia. 2006. Split Intransitivity in Italian. Berlin: Mouton de Gruyter. Bentley, Delia. 2008. The interplay of focus structure and syntax: evidence from two sister languages. In Robert D. Van Valin, Jr. (ed.), Investigations of the Syntax-SemanticsPragmatics Interface, 263–284. Amsterdam: Benjamins. Bertinetto, Pier Marco & Mario Squartini. 1995. An attempt at defining the class of gradual completion verbs. In Piermarco Bertinetto, Valentina Bianchi, James Higginbotham & Mario Squartini (eds.), Temporal Reference, Aspect, and Actionality, 11–28. Turin: Rosenberg and Sellier. Bertinetto, Pier Marco & Alessandro Lenci. 2012. Habituality, pluractionality and imperfectivity. In Robert Binnick (ed.), The Oxford Handbook of Tense and Aspect, 852–880. Oxford: Oxford University Press. Bossong, Georg. 1991. Differential object marking in Romance and beyond. In Dieter Wanner & David Kibbee (eds.), New Analyses in Romance Linguistics. Selected Papers from the XVIII Synposium on Romance Languages, 143–170. Amsterdam: John Benjamins. Bossong, Georg. 1998. Le marquage différentiel de l’objet dans les langues d’Europe. In Jack Feuillet (ed.), Actance et Valence dans les Langues de l’Europe, 259–294. Berlin: Mouton de Gruyter. Cennamo, Michela. 1993. The Reanalysis of Reflexives: a Diachronic Perspective. Naples: Liguori. Cennamo, Michela. 1995. Transitivity and VS order in Italian reflexives. In Yaron Matras & HansJürgen Sasse (eds.), Verb-Subject Order and Theticity in European Languages. STUF 48(1/2). 84–105. Cennamo, Michela. 1997. Passive and impersonal constructions. In Martin Maiden & Mair Parry (eds.), Dialects of Italy, 145–161. London: Routledge.

Valency patterns in Italian

479

Cennamo, Michela. 1999. Late Latin pleonastic reflexives and the Unaccusative hypothesis. Transactions of the Philological Society 97(1). 103–150. Cennamo, Michela. 2001. L’Inaccusatività in alcune varietà campane. In Federico Albano Leoni, Eleonora Stenta Krosbakken, Rosanna Sornicola & Carolina Stromboli (eds.), Dati Empirici e Teorie Linguistiche. Atti del XXXIII Congresso Internazionale della Società di Linguistica Italiana, 427–453. Rome: Bulzoni. Cennamo, Michela. 2002. La selezione degli ausiliari perfettivi in napoletano antico: fenomeno sintattico o sintattico-semantico? Archivio Glottologico Italiano 87(2). 175–222. Cennamo, Michela. 2003. (In)transitivity and object marking: some current issues. In Giuliana Fiorentino (ed.), Romance Objects, 49–104. Berlin: Mouton de Gruyter. Cennamo, Michela. 2008. The rise and development of analytic perfects in Italo-Romance. In Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers, 115–142. Amsterdam: Benjamins. Cennamo, Michela. 2009. Argument structure and alignment variations and changes in Late Latin. In Jóhanna Barðdal & Shobana Chelliah (eds.), The Role of Semantics and Pragmatics in the Development of Case, 307–346. Amsterdam: Benjamins. Cennamo, Michela. 2010. Impersonali, verbi. In Raffaele Simone (ed.), Enciclopedia dell’italiano, vol I, 637–639. Rome: Istituto dell’Enciclopedia Italiana G. Treccani. Cennamo, Michela. 2011a. Psicologici, verbi. In Raffaele Simone (ed.), Enciclopedia dell’italiano, vol II, 258–260. Rome: Istituto dell’Enciclopedia Italiana G. Treccani. Cennamo, Michela. 2011b. Riflessivi, verbi. In Raffaele Simone (ed.), Enciclopedia dell’italiano, vol II, 320–322. Rome: Istituto dell’Enciclopedia Italiana G. Treccani. Cennamo, Michela. 2011c. Transitivi e intransitivi, verbi. In Raffaele Simone (ed.), Enciclopedia dell’italiano, vol II, 535–539. Rome: Istituto dell’Enciclopedia Italiana G. Treccani. Cennamo, Michela. 2012a. Aspectual constraints on the (anti)causative alternation in Old Italian. In Jóhanna Barðdal, Michela Cennamo & Elly van Gelderen (eds.), Thematic issue on Argument Realization and Change. Transactions of the Philological Society 110(3). 394–421. Cennamo, Michela. 2012b. Unexpressed objects and the semantics of predicates in Italian. Workshop on Contrastive Studies of Verbal Valency in European Languages, 45 th Annual Meeting of the Societas Linguistica Europaea, Stockholm, 31 August–1 September 2012. Cennamo, Michela. 2014. Passive and impersonal reflexives in the Italian dialects. Synchronic and diachronic aspects. In Paola Benincà, Adam Ledgeway & Nigel Vincent (eds.), Diachrony and Dialects, 71–95. Oxford: Oxford University Press. Cennamo, Michela & Antonella Sorace. 2007. Auxiliary selection and split intransitivity in Paduan. In Raúl Aranovich (ed.), Split Auxiliary Systems, 65–99. Amsterdam: Benjamins. Cennamo, Michela & Elisabetta Jezek. 2011. The anticausative alternation in Italian. In Giovanna Massariello & Silvana Dal Masi (eds.), Le Interfacce, 809–823. Rome: Bulzoni. Cennamo, Michela & Alessandro Lenci. 2011. Gradience in subcategorization? Locative phrases with Italian verbs of motion. Workshop on Explorations in Syntactic Government and Subcategorization, Cambridge, 31 August–3 September 2011. To appear in Studia Linguistica. Centineo, Giulia. 1995. The distribution of si in Italian transitive/inchoative pairs. In Mandy Simmons & Theresa Galloway (ed.), Proceedings from Semantic and Linguistic Theory V., 54– 71. Ithaca: Cornell University Press. Cinque, Guglielmo. 1988. On si constructions and the theory of arb. Linguistic Inquiry 19(4). 521– 581. Comrie, Bernard. 1985. Causative verb formation and other verb-deriving morphology. In Timothy Shopen (ed.), Language Typology and Syntactic Description III. Grammatical Categories and the Lexicon, 309–348. Cambridge: Cambridge University Press. Comrie, Bernard. 1989. Language Universals and Linguistic Typology. Second edition. London: Blackwell.

480

Michela Cennamo

Coseriu, Eugenio. 1971. Solidarietà lessicali. In Eugenio Coseriu (ed.), Teoria del Linguaggio e Linguistica Generale. Sette Studi, 303–316. Bari: Laterza. Diedrichsen, Elke. 2013. Auxiliary selection in German. Constructional gradience with perfect formation. In Elly van Gelderen, Michela Cennamo & Jóhanna Barðdal (eds.), Argument Structure in Flux: the Naples-Capri Papers, 404–434. Amsterdam: John Benjamins. Dixon, Robert M.W. 1994. Ergativity. Cambridge: Cambridge University Press. D’Alessandro, Roberta. 2007. Impersonal si Constructions. Agreement and Interpretation. Berlin: Mouton de Gruyter. Fiorentino, Giuliana. 2003 (ed). Romance Objects. Berlin: Mouton de Gruyter. Fiorentino, Giuliana. 2003. Prepositional objects in Neapolitan. In Giuliana Fiorentino (ed), Romance Objects, 117–152. Berlin: Mouton de Gruyter. Fiorentino, Giuliana. 2010. Accusativo preposizionale. In Raffaele Simone (ed.), Enciclopedia dell’italiano, vol I, 17–19. Rome: Istituto dell’Enciclopedia Italiana G. Treccani. Folli, Raffaella. 2002. Constructing telicity in English and Italian. PhD Thesis. University di Oxford. Guerssel, Mohamed, Hale, Kenneth, Laughren, Mary, Levin, Beth & Eagle, Josie White. 1985. A crosslinguistic study of transitivity alternations. In William H. Eilfort, Paul D. Kroeber & Karen L. Peterson (eds.) Papers from the Parasession on Causatives and Agentivity, 48–63. Chicago Linguistic Society 21.2. Chicago, IL: University of Chicago. Haspelmath, Martin. 1987. Transitivity alternations of the anticausative type. Arbeitspapier n. 5. Institut für Srachwissenschaft, Köln. 1–51. Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Bernard Comrie & Maria Polinsky (eds.), Causatives and Transitivity, 87–120. Amsterdam: Benjamins. Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typology. Linguistic Typology 15(3). 535–689. Iacobini, Claudio. 2006. Italian as a fusional language. In K. Brown (ed.), Encyclopedia of Language and Linguistics, 2 nd Edition, vol 6, 64–69. Oxford: Elsevier. Iwata, Seizi. 2008. Locative Alternation. A Lexical-Constructional Approach. Amsterdam: Benjamins. Keller, Frank & Antonella Sorace. 2003. Gradient auxiliary selection and impersonal passivization in German: an experimental investigation. Journal of Linguistics 39(1). 57–108. Koontz-Garboden, Andrew. 2009. Anticausativization. Natural Language and Linguistic Theory 27(1). 77–138. Jezek, Elisabetta. 2003. Classi di Verbi tra Semantica e Sintassi. Pisa: ETS. Jezek, Elisabetta, 2012. Inaccusativité, structure événementielle et verbes pronominaux en Italien. In Christine Bracquenier & Louis Begioni (eds.), L’Aspect dans les Langues Naturelles, 159– 178. Rennes: Presses Universitaires des Rennes. La Fauci, Nunzio 2009. Compendio di Sintassi Italiana. Bologna: Il Mulino. Lazard, Gilbert. 1984. Actance variations and categories of the object. In Frans Plank (ed.), Objects. Towards a Theory of Grammatical Relations, 269–292. London: Academic Press. Lenci, Alessandro. 2012. Argument alternations in Italian verbs: a computational study. In Valentina Bambini, Irene Ricci, Pier Marco Bertinetto & Collaborators (eds.), Linguaggio e Cervello − Semantica/Language and the Brain − Semantics. Atti del XLII Congresso Internazionale di Studi della Società di Linguistica Italiana (Pisa, SNS, 2008), Roma: Bulzoni. Volume 2. 1–26 (CD ROM, II.B.1). Lepschy, Anna Laura & Giulio Lepschy. 2006. Italian. In K. Brown (ed.), Encyclopedia of Language and Linguistics, 2 nd edition, vol 6, 545–549. Oxford: Elsevier. Letuchiy, Alexander. 2009. Towards a typology of labile verbs: Lability vs. derivation. In Alexandre Arkhipov & Patience Epps (eds.), New Challenges in Typology. Transcending the Borders and Refining the Distinctions, 223–244. Berlin: Mouton de Gruyter. Levin, Beth. 1993. English Verb Classes and Alternations. Chicago: University of Chicago Press.

Valency patterns in Italian

481

Liu, Feng-hsi. 2007. Auxiliary selection in Chinese. In Raúl Aranovich (ed.), Split Auxiliary Systems, 181–205. Amsterdam: Benjamins. Lo Duca, Maria Giuseppa. 2000. Proprietà valenziali e criteri di descrizione lessicografica: un caso di alternanza argomentale. In R. Simone (ed.), Thematic issue on Classi di Parole e Conoscenza Lessicale. Studi Italiani di Linguistica Teorica e Applicata 2. 219–242. Maiden, Martin. 1995. A Linguistic History of Italian. London: Longman. Maiden, Martin & Cecilia Robustelli. 2007. A Reference Grammar of Modern Italian, 2 nd edition. London: Hodder Arnold. Malchukov, Andrej, Haspelmath, Martin & Bernard Comrie. 2010. Ditransitive constructions: a typological overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions. A Comparative Handbook, 1–64. Berlin: Mouton de Gruyter. Manente, Mara. 2008. L’aspect, les auxilaires ‘être’ et ‘avoir’ et l’hypothèse inaccusative dans une perspective comparative Français/Italien. PhD Thesis. University of Venice. Mereu, Lunella. 2010. Argument versus Adjunct PPS. Where does the difference lie? MS. Rom: University of Roma3. Mithun, Marianne & Wallace Chafe. 1999. What are S, A and O? Studies in Language 23. 569–596. Moravcsik, Edith. 1978. On the case-marking of objects. In Joseph Greenberg et al. (eds.), Universals of Human Language, vol. 4, 249–289. Stanford: Stanford University Press. Olsen, Mari B. & Philippe Resnik, 1997. Implicit object constructions and the (in)transitivity continuum. In Kora Singer, Randall Eggert & Gregory Anderson (eds.), Proceedings of the Chicago Linguistics Society 33. Papers from the Main Session, 327–336. Chicago: Chicago Linguistics Society. Rizzi, Luigi. 1986. Null objects in Italian and the theory of pro. Linguistic Inquiry 17(3). 501–557. Salvi, Giampaolo. 1988. La frase semplice. In Lorenzo Renzi (ed.), Grande Grammatica Italiana di Consultazione, vol. 1, 29–113. Bologna: Il Mulino. Salvi, Giampaolo & Laura Vanelli. 2004. Nuova Gammatica Italiana. Bologna: Il Mulino. Schäfer, Florian. 2008. The Syntax of (Anti-)Causatives. External Arguments in Change-of-State Contexts. Amsterdam: Benjamins. Serianni, Luca. 1989. Grammatica Italiana. Torino: UTET. Siller-Runggaldier, Heidi. 2003. Changes of valence and their effect on objects. In Giuliana Fiorentino (ed.), Romance Objects, 187–216. Berlin: Mouton de Gruyter. Simpson, Jane. 1991. Warlpiri Morphosyntax. A Lexicalist Approach. Dordrecht: Kluwer. Sorace, Antonella. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76. 859–890. Sorace, Antonella. 2004. Gradience at the lexicon-syntax interface: evidence from auxiliary selection and implications for unaccusativity. In Artemis Alexiadou, Elena Anagnostopolou & Martin Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface, 243–268. Oxford: Oxford University Press. Thornton, Anna 2005. Morfologia. Roma: Carocci. Van Valin, Robert D. Jr. 2001. An Introduction to Syntax. Cambridge: Cambridge University Press. Van Valin, Robert D. Jr. 2005. Exploring the Syntax-Semantics Interface. Cambridge: Cambridge University Press. Van Valin, Robert D. Jr. & Randy J. La Polla. 1997. Syntax: Strucure, Meaning and Function. Cambridge: Cambridge University Press. Vincent, Nigel. 1988. Italian. In Martin Harris & N. Vincent (eds.), The Romance Languages, 279– 313. London: Routledge. Voigt, Rainer. 2008. Italian language in Ethiopia and Eritrea. In Siegbert Uhlig (ed.), Encyclopaedia Aethiopica, vol. 3, 222–224. Wiesbaden: Harrassowitz. Zúñiga, Fernando & Seppo Kittilä (eds.). 2010. Benefactives and Malefactives. Typological perspectives and case studies. Amsterdam: Benjamins.

Michael Daniel and Victoria Khurshudian

14 Valency classes in Eastern Armenian

1

1 Introduction Eastern Armenian is an Indo-European language which, together with its sister language Western Armenian and their various dialects and its earlier stage, Classical Armenian, forms the Armenian branch of the Indo-European family. Eastern Armenian is the official language of the Republic of Armenia and is spoken by over three million people living there, not to count the Diasporas mainly in the countries of the former Soviet Union and Iran. Within the Republic, the language is subject to strong dialectal variation. The standard language is closest to the dialect of the Yerevan area (the so-called Araratian dialect), but even between these two variants the differences are quite noticeable. Our data come from the Eastern Armenian National Corpus (www.eanc.net), an open electronic resource including over 100 million word tokens and covering Eastern Armenian from the moment of its standardization (early 19th century) to the present. Naturalistic examples, however, are often lengthy and contain a lot of irrelevant data. For the sake of simplicity many examples cited below are constructed, based on one of the authors’ native knowledge. For most of these constructed examples, parallel naturalistic examples may be found in the online valency database (valpal.info) to which this volume is a sister project. When naturalistic, examples are marked as EANC, additionally indicating whether they come from translated fiction, newspaper or original fiction (the name of the author is given in the latter case). Examples are given both in the Armenian script and in a transliteration which is close to the traditional Latin transliteration of Armenian (HübschmannMeillet) but is slightly modified to better match IPA. The paper is structured as follows. Section 2 provides a general overview of Eastern Armenian morphosyntax. Section 3 is a discussion of transitivity issues and the two transitivity changing derivations, the mediopassive and the causative, the only marked valency alternations existing in Eastern Armenian. The question of whether Eastern Armenian is primarily a transitivizing or detransitivizing language in terms of Nichols et al. (2004) is briefly addressed. Section 4 covers unmarked alternations, including reciprocal, object omission, contentive-locative and the most unusual of all, the proprietive (ablative-genitive) alternation, showing

1 We thank Timofey Arkhangelskiy, Martin Haspelmath, Hrach Martirosyan and the anonymous reviewer, whose comments on its draft greatly improved this paper. The study was implemented in the framework of the Basic Research Program at the National Research University Higher School of Economics (HSE).

484

Michael Daniel and Victoria Khurshudian

alternative construals of the same inanimate participant as the Source or (retrospective) Possessor. Section 5 introduces the notion of extended valencies: semantic roles that represent identical participants and are marked in the same way with all verbs but that are optional with some verbs while obligatory with other verbs. Using this notion helps to establish connections between verb classes that are different only in the degree to which the participant is integrated into the respective situation (cf. Beneficiary with ‘build (for)’ vs. ‘send (to)’ vs. ‘give to’). Finally, Section 6, building on the data on case frames (e.g., dative verbs or transitive verbs) and alternations available for specific verbs (first of all availability of marked alternations), groups them into classes of common morphosyntactic behavior to produce a valency-based classification of Eastern Armenian verbs. Section 7 is a brief summary of the paper.

2 Grammatical background Armenian nominal morphology is close to the agglutinative prototype, with however some traces of flexion inherited from the Classical Armenian period. Verbal morphology is relatively rich; but many forms are periphrastic, combining dependent forms (termed “converbs” below, though some of them are not used outside periphrasis) with the auxiliary (copula). Armenian is a left-branching language rich in postpositions. The clausal word order is flexible and determined by the communicative structure. Thus, focus is expressed by moving the auxiliary clitic from its neutral post-converb position and hosting it to the constituent to be focused (Comrie 1984). Armenian is a language with accusative alignment and a very clear differential object marking (DOM) pattern: dative is used for human direct objects (hDO), nominative for inanimate direct objects (iDO). (1) inanimate DO: nominative marking

Ռուբին սիրում էր միայն ժողովրդական երգեր … Rubi-n siɾ-um ēɾ miajn žoʁovɾdakan jeɾg-eɾ … rubi-def love-cvb.ipfv aux.pst.3sg only folk song-pl ‘Rubi only loved folk songs.’ (translated, eanc)

(2) human DO: dative marking

Դու սիրել ես մեկին, նա` ուրիշին։

Du siɾ-el es mek-i-n, na uɾiš-i-n. you.sg love-cvb.pfv aux.2sg one-dat-def (s)he other-dat-def ‘You loved with a [girl], she loved with someone else.’ (eanc)

Valency classes in Eastern Armenian

485

Non-human animate direct objects (aDO) behave inconsistently (are ambiguous): they take either nominative or dative marking with no apparent change in meaning. To the best of our knowledge, this topic has not been studied in detail; for an overview see (Dum Tragut 2010). Note that our use of glosses and case terminology, in (1) and (2) as well as below, is morphological rather than function oriented. We use the terms nominative and dative in the sense of case marking rather than syntactic function, and we do not use the term accusative at all, because there is no dedicated case marking for DO in Eastern Armenian. This goes against some traditional approaches to Armenian which do isolate the accusative case on a functional basis, may be partly due to the fact that Classical Armenian had a dedicated accusative proclitic, which developed into a pronominal marker of accusative in Western Armenian. The list of the cases traditionally isolated by Armenian grammars, together with their main functions, is as follows (below, syntactic relations are given as abbreviations, and conventional semantic roles with a first capital): Tab. 1: Case forms and their functions. nominative

S, iDO, Goal, address, citation

Direct?

genitive dative

attributive noun hDO, Recipient, Goal, superessive

Oblique?

instrumental locative

Instrument, Comitative Place (plus metaphorical extensions such as some temporal meanings etc.) Source (plus metaphorical extensions such as the starting point in time etc.)

ablative

Table 1 above highlights one further problematic point of the traditional Armenian case inventory. From the morphological point of view, distinguishing two different cases, the dative and the genitive, is partly misleading. In nouns, the genitive is identical to the dative. Separate morphology exists in first and second person pronouns. However, pronominal declension is cross-linguistically often not fully homologous to nominal declension, both in terms of the material used and the categories displayed. In addition to the analogy with pronominal declension, the traditional classification is also based on the morphological distribution. In Eastern (but not in Western) Armenian, adnominal genitive never combines with the suffixed definite article, while dative most often occurs with it. The definite article, which also seems to be a default option for the S/A functions, reflecting the known link between subjecthood and topicality (see e.g., Maslova & Bernini 2006), thus comes close to forming a part of the case marking system. But the definite article is, on the other hand, in complementary distribution with first and second person possessive suffixes, which do not have similar limitations determined by morphosyntactic context.

486

Michael Daniel and Victoria Khurshudian

Thus, morphologically, Armenian nouns only distinguish two core cases. Direct is the case of S/A (subject) and iDO; oblique is the case of the noun in the attributive position, hDO and Recipient; and both have peripheral spatial uses. Nonetheless, when discussing flagging below, we will use the traditional distinction between the genitive and the dative (but discarding the accusative label, adopted in some traditional grammars that are more oriented towards the syntactic categories). For the direct object marking the label DO will be used to cover both dative and nominative marking for hDO, aDO and iDO when these do not need to be differentiated (as when discussing transitivity issues in general). Cf. the following table; the labels used in this paper are boldfaced:

Tab. 2: ‘Core case’ – related morphological and syntactic categories of Eastern Armenian noun phrases. pure morphology

Direct

Oblique

traditional labels 1 (used in this paper) traditional labels 2 (morphology strongly contaminated by function) syntactic semantic functions

Nom

Dat

Gen Gen

generalized syntactic functions (used in this paper)

Nom

Acc

Dat

S

iDO

hDO

S

DO

Rec

Poss

IO

Poss

3 Transitivity and morphologically marked valency alternations In Classical Armenian, transitivity strongly correlated with the thematic vowel of the verb. Thematic -e- occurred in both transitive (the largest class) and intransitive verbs but changed to ‑i- in the mediopassive. Minor classes in -a-, -u- and ‑o- (single verb gom ‘be, exist’) were intransitive, although some forms of these verbs were labile (or, in paradigmatic terms, ambiguous between passive and active). In modern Eastern Armenian all themes except ‑e- and ‑a- are virtually lost. With very few exceptions (like kaɾdal ‘read’ or tal ‘give’), ‑a‑ thematic verbs are intransitive. On the other hand, ‑e‑ thematic verbs include both transitive and intransitive verbs (containing the descendants of both the transitive and intransitive ‑e- classes of Classical Armenian). Intransitive theme ‑i‑ is only preserved in Western Armenian where it is used with some primary intransitive verbs, derived decausatives and all mediopassives (see Donabedian 1997). Thematic verbs in ‑o‑ and ‑u- are lost almost completely in both languages.

Valency classes in Eastern Armenian

487

Eastern Armenian has two valency derivations marked on the verb: mediopassive and causative (Kozinceva 1974, 1981). The causative marking is inherited from the Classical Armenian where, according to Kortlandt (1999), the causative suffix -ujcʰan- (-ucʰan-) results from a reanalysis of the TAM morphology of some verbs with causative semantics, including ‘show’, with subsequent spread to other verbs as a regular causativization device. In the modern Armenian languages, this sequence has been contracted, resulting in ‑cʰn- in Eastern Armenian. Probably, the fact that the suffix is historically complex is reflected in that its -n- is lost in mediopassives derived from causatives: cf. jeral ‘boil (intr)’ → jera-cʰn-el ‘boil (tr)’ → jeracʰ-v-el ‘boil (intr)’. Almost all causatives belong to the -e- group, thus being marked for transitivity; they are mostly derived from intransitive verbs. In addition to that, a ‘give’causative pattern (infinitive + tal ‘give’, apparently gravitating towards distant causative meaning) exists, but is considerably less frequent than the morphological causative in the corpus. (3) ‘learn’: morphological causativization

Լռի՛ր, անզգա՜մ աղջիկ, թե չէ քեզ կսովորեցնեմ, թե ում հետ գործ ունես։

Lr-iɾ, anzgam aʁǯik, tʰe čʰē kʰez be.silent-imp.2sg shameless girl if cond-be.3sg you.sg.dat k-sovoɾ-ecʰn-em, tʰe um het goɾc un-es. cond-learn-caus-prs.1sg that who.gen with work have-prs.2sg ‘Silence, you brazen girl, lest I teach you who you are dealing with.’ (Bagrat Ajvazian, eanc) (4) ‘give’: periphrastic causative

Ուզենա մի օրում հազար ոչխար մորթել կտա։

Uzen-a mi ōɾ-um hazaɾ vočʰχaɾ moɾtʰ-el want-sbjv.prs.3sg one day-loc thousand sheep slaughter-inf k-ta. cond-give.prs.3sg ‘At his wish, he could have had a thousand sheep slaughtered in one day.’ (Aksel Bakunc, eanc) The valency decreasing thematic alternation present in Classical Armenian has been lost in Eastern Armenian and weakened in Western Armenian. Instead, a mediopassive marker ‑v- of unclear origin has been introduced in both modern languages. Verbs in -al which, in Modern Eastern Armenian form a closed group, form mediopassives – as well as some other forms – based on the aorist stem; cf. (59) below. This marker has a wide range of meanings (subsumed under the label Mediopassive), as shown in the following examples:

488

Michael Daniel and Victoria Khurshudian

(5) mediopassive: ‘passive’

Ծրագրի համաձայն՝ գազամուղ կկառուցվի թե՛ հայկական, թե՛ իրանական տարածքներում։

Cɾagɾ-i hamaʒajn gazamuʁ k-karucʰ-v-i tʰe hajkakan, project-gen according pipeline cond-build-med-prs.3sg that Armenian tʰe iɾanakan taɾackʰ-neɾ-um. that Iranian territory-pl-loc ‘According to the project, a pipeline will be constructed on both Armenian and Iranian territory.’ (newspapers, eanc) (6) mediopassive: ‘passive’ … այդ նավթամուղը կառուցվել է 1990-ականների վերջերին։ ajd navtʰamuʁ-ə karucʰ-v-el ē 1990-akan-neɾ-i that pipeline-def build-med-cvb.pfv aux.3sg 1990-adjvz-pl-gen veɾǯ-eɾ-i-n. end-pl-dat-def ‘… that pipeline was built at the end of the 1990’s.’ (newspapers, eanc) (7) mediopassive: ‘reflexive’

Արամյանը կուզենար ինքն էլ գնալ նայել, բայց ծուլանում էր հագնվել։

Aɾamjan-ə k-uzen-aɾ inkʰ-n ēl gn-al naj-el, bajcʰ Aramian-def cond-want-pst.3sg self-def also go-inf look-inf but culan-um ēɾ hagn-v-el. feel.lazy-cvb.ipfv aux.pst.3sg dress-med-inf ‘Aramian would also like to go and have a look, but he didn’t feel like getting dressed.’ (Žirajr Avetisian, eanc) (8) mediopassive: ‘reciprocal’

Եվ ահա այդքան տարբեր այդ բնավորությունները կապվել էին իրար։

Jev aha ajdkʰan taɾbeɾ ajd bnavoɾutʰjun-neɾ-ə kap-v-el and here so different that character-pl-def link-med-cvb.pfv ēin iɾaɾ. aux.pst.3pl each.other ‘And thus those characters so different became linked to each other.’ (Xarčil Hračyan, eanc)

Valency classes in Eastern Armenian

489

(9) mediopassive: ‘decausative’

Իսկ Լոռու մարզի Լուսաղբյուր գյուղի շրջակա դաշտերում այրվել է հեկտար խոտ։

3

Isk Lor-u maɾz-i Lusaʁbjuɾ gjuʁ-i šɾǯaka and Lori-gen province-gen Lusaghbur village-gen neighbouring dašt-eɾ-um ajɾ-v-el ē 3 hektaɾ χot. field-pl-loc burn-med-cvb.pfv aux.3sg 3 hectare grass ‘And on the fields near to Lusaghbyur village in the district of Lori three hectares of grass burnt away.’ (newspapers, eanc) Another relatively frequent meaning is a potential passive (‘easy to P’ – see Appendices). Both causative and mediopassive derivations are productive in Eastern Armenian. It is not straightforwardly clear whether Armenian should be classified as a transitivizing or detransitivizing language in terms of Nichols et al. (2004). While Nichols classifies Western Armenian as more transitivizing than not, Haspelmath (1993), using a different inventory of criteria, classifies Eastern Armenian as what may be interpreted as detransitivizing in Nichols’ terms. There are several arguments supporting classifying Eastern Armenian as more detransitivizing than transitivizing. First, in terms of text frequency, mediopassives are more than three times as frequent as morphological causatives in the EANC. Second, along the lines of Haspelmath (1993), one may compare primary intransitive and primary transitive meanings, i.e. intransitive and transitive meanings expressed by morphologically non-derived verbs. Out of those meanings which are considered as ambiguous between the two classes and often expressed by ambitransitive, or labile, verbs, only ‘boil’ is intransitive in Eastern Armenian, while such meanings as ‘burn’, ‘roll’, ‘cook’, ‘pour’, ‘tear’, ‘break’, and ‘open’ are transitive. Remarkably, the basic verb for ‘sink’, ‘drown’ (of a ship or a human) also belongs to primarily transitive verbs, while intransitive ‘sink’ is its mediopassive derivate: (10) ‘sink’: a primary transitive verb

Այդ օրը Նելսոն Ստեփանյանը իր ղեկավարած խմբով խորտակեց թշնամու երկու պահականավ։ Ajd ōɾ-ə Nelson Stepʰanjan-ə iɾ ʁekavaɾ-ac χmb-ov that day-def Nelson Stepanian-def self.gen direct-ptcp.res group-ins χoɾtak-ecʰ tʰšnam-u jeɾku pahakanav. sink-aor.3sg enemy-gen two patrol.ship ‘That day Nelson Stepanian and the crew under his command sank two enemy patrol ships.’ (Džori Balajan, eanc)

490

Michael Daniel and Victoria Khurshudian

(11) ‘sink’: a primary transitive verb (translated, EANC)

Ու՞մ է ախորժելի խորտակել հարազատ աղջկան։ Um ē aχoɾželi χoɾtak-el haɾazat aʁǯk-a-n. who-dat be.3sg tasty sink-inf own girl-dat-def ‘Who would feel happy about drowning his own daughter?’ (translated, eanc)

(12) ‘sink’: decausative mediopassive

Առնետները լքում են խորտակվող նավը, – բուռն հրճվանքով կանչեց Աղասին։ Arnet-neɾ-ə lkʰ-um en χoɾtak-v-oʁ nav-ə, - burn rat-pl-def flee-cvb.ipfv aux.3pl sink-med-ptcp.sbj ship-def violent hɾčvankʰ-ov kančʰ-ecʰ Aʁasi-n. exultation-ins exclaim-aor.3sg Aghasy-def ‘The rats are deserting the sinking ship, exclaimed Aghasy in passionate exultation.’ (Rafael Atajan, eanc)

Third, it is a known fact that morphological causatives tend to combine with intransitive verbs and to express contact causation (cf. Kulikov 2001). Both tendencies are not absolutely straightforward in Eastern Armenian. A few transitive verbs peripherally do form morphological causatives and thus often convey the meaning of distant rather than contact causation. A few causatives from intransitive verbs, on the other hand, denote distant causation, apparently because they have corresponding lexical, non-derived causatives (cf. 14–15). Armenian morphological causatives thus are not a fully prototypical representative of this cross-linguistic category for which Nichols et al.’s typology of (de)transitivizing languages was designed. (13) morphological causative of a transitive verb

Տանը Շուշանը Գեղամիկին մածուն էր ուտեցնում։ Tan-ə šušan-ə Geʁamik-i-n macun ēɾ house.dat-def Shushan-def Gegham-dat-def yogurt be.aux.pst.3sg ut-ecʰn-um. eat-caus-cvb.ipfv ‘At home Shushanik was feeding little Gegham with yogurt.’ (Sero Xanzadian, eanc)

(14) ‘die’: morphological causative conveys distant causation

… ուրիշներին մահացնելը պակաս հանցանք է, քան չմեռնելը։ Uɾiš-neɾ-i-n mah-acʰn-el-ə pakas hancʰankʰ ē kʰan other-pl-dat-def die-caus-inf-def less crime be.3sg than čʰ-mern-el-ə. neg-die-inf-def ‘… to cause others to die is a lesser crime than not to die.’ (translated, eanc)

Valency classes in Eastern Armenian

491

(15) ‘kill’: direct causation conveyed by a primary transitive

Հաջորդ օրը ես պարսատիկով մի ագռավ սպանեցի ու գցեցի Բողարի առաջ։ Haǯoɾd ōɾ-ə jes paɾsatik-ov mi agrav span-ecʰ-i u next day-def I catapult-ins one crow kill-aor-1sg and gcʰ-ecʰ-i Boʁaɾ-i araǯ. throw-aor-1sg Boghar-gen in.front.of ‘The next day I killed a crow with a catapult and threw it in front of Boghar [name of a dog].’ (Vaxtang Ananian, eanc)

Mediopassive and causative may combine in one wordform, as in (16). In this case, the mediopassive marker follows the causative marker. This suggests that, in terms of derivation ~ inflection scale, the causative is at least not less derivational than the mediopassive marker, supporting our claim that Eastern Armenian mediopassive, though a more recent acquisition, is a more grammatical device than the causative. (16) causative and mediopassive combined in one wordform

Հետ-պատերազմյան տասնամյակների ընթացքում այդ կապերը փաստորեն սառեցված էին։ Het-pateɾazmjan tasnamjak-neɾ-i əntʰacʰkʰ-um ajd kap-eɾ-ə pʰastoɾen post.war decade-pl-gen course-loc that tie-pl-def in.fact sar-ecʰ-v-ac ēin. freeze-caus-med-ptcp.res. aux.pst.3pl ‘In the post-war decades those connections have in fact stopped (lit. were frozen).’ (newspapers, eanc)

Mediopassive may be derived from an intransitive verb: (17) mediopassive from an intransitive verb

Եթե մենակ ապրվեր, աստված Եվային չէր ստեղծի

… Jetʰe menak apɾ-v-eɾ astvac Jeva-ji-n if alone live-med-sbjv.pst.3sg God Eve-dat-def čʰ-ēɾ steʁc-i … neg-be.aux.pst.3sg create-cvb.conneg ‘If [one could] live alone, God wouldn’t have created Eve …’ (Stepan Zorian, eanc)

Note also that there are some verbs with causative morphology but with no primary verb from which they may be derived (derivationally bound roots), such as tʰakʰ(-)cʰn-el ‘hide’, veɾ(-)cʰn-el ‘take’ (from which a mediopassive veɾ-cʰ-v-el ‘be

492

Michael Daniel and Victoria Khurshudian

taken’ is derived) and probably l(-)cʰn-el ‘fill’, from the list of the valency database meanings. Each of these three verbs suggest a possible non verbal source of derivation (see Appendices). However, all these deviations from typical mediopassive and causative behavior are relatively peripheral, the much more frequent situation being that mediopassive is formed from transitive verbs, and causative from intransitive verbs to convey contact causative meaning. Statistical data for the occurrence of individual verbs with mediopassive and/or causative marking is provided in Table 3 in the appendix.

4 Unmarked valency alternations In addition to causative and mediopassive derivations, some verbs show variation in argument marking that is not coded on the verb. Just as with marked alternations, they may be valency decreasing (e.g., reciprocal alternation), valency rearranging (contentive-locative, or ‘filling’-‘loading’ alternation) or ambiguous between the two (proprietive alternation). For the reasons explained in § 5, we consider most unmarked valency increasing alternations (e.g., introducing Beneficiary) separately under the name of extended valencies. All alternations, including extensions, are semantically driven. Availability of an alternation may be explained by the properties of the situation designated by the verb and its linguistic construal. Alternations are also different in that some of them are directional, distinguishing between the primary and secondary valencies (such as reciprocal and all extensions) or symmetrical (such as contentive-locative, proprietive and probably object omission), where it is hard or impossible to argue that one of the valencies is a primary one.

4.1 Reciprocal alternation Reciprocal alternation involves a change from a bivalent (transitive or other) construction to the intransitive construction with non-singular (plural or co-ordinated) subject (as English He’s friends with Alice → They, Alice and he, are friends). The alternation is directional in the sense that the primary schema cannot be predicted based on the derived one. It occurs with verbs belonging to different valency classes provided that the situation may be considered as symmetrical: the participants are simultaneously involved in the same activity towards each other, and this symmetrical involvement is typical enough for the activity to be covered by the same verbal lexeme (hence the relative cross-linguistic validity of this class of meanings). There is a certain tendency for the reciprocal alternation to combine with verbs introducing a comitative postpositional phrase – comitative verbs such as ‘marry’ or ‘converse’. This can be attributed to the fact that the semantic functional proper-

Valency classes in Eastern Armenian

493

ties of the comitative category itself, be it expressed by adpositional or case marking, presuppose participants that are ranked differently but show the same involvement in the situation, or express the same semantic role (see Arkhipov 2009 for a recent overview of comitatives). (18) primary valency (gen +with) – Չեմ կարող ամուսնանալ նրա

նա երբեք ապահարզան չի տա։

հետ, – ասացի ես, – տանը կին եմ թողել,

Čʰ-em kaɾ-oʁ amusnan-al nɾa het, asa-cʰi neg-aux.1sg be.able-ptcp.subj marry-inf (s)he.gen with say-aor.1sg jes, tan-ə kin em tʰoʁ-el, na jeɾbekʰ apahaɾzan I home.dat-def woman aux.1sg leave-cvb.pfv (s)he never divorce čʰ-i ta. neg-aux.3sg give.cvb.conneg ‘I cannot marry her, I said, I have a wife at home, and she’ll never give me a divorce.’ (translated, eanc) (19) unmarked reciprocal alternation

Նրանք ամուսնացան։

Nɾankʰ amusn-acʰ-an. they marry-aor-3pl ‘They got married.’ (constructed) (20) primary valency (dat)

Մարդը հանդիպեց աղջկան։ Maɾd-ə handip-ecʰ aʁǯk-a-n. man-def meet-aor.3sg girl-dat-def ‘The man met the girl.’ (constructed)

(21) unmarked reciprocal alternation Նրանք հանդիպեցին (իրար:) Nɾankʰ handip-ecʰ-in (iɾaɾ). they meet-aor.3pl (each.other) ‘They met (each other).’ (constructed) The reciprocal alternation is limited to a rather small set of verbs and is strictly lexical. With some verbs, reciprocal semantics is also conveyed with mediopassive forms (cf. 8). The reciprocal alternation combines with verbs with different valency patterns. It is not a property of any valency class but a property of individual verbs which cuts across various valency classes. Other unmarked alternations, although also

494

Michael Daniel and Victoria Khurshudian

semantically motivated, are associated with specific valency patterns and thus single out separate valency classes or subclasses.

4.2 Object omission Object omission occurs with generalized objects with verbs that often move their objects out of focus to highlight the process or event, e.g., ‘eat’, ‘read’ or ‘swallow’. Other verbs that easily lose their DO are the verbs strongly associated with one typical object such as ‘spit’ or ‘urinate’. Typically for accusative languages, these objects are omitted without affecting the verb or the marking on other arguments, including the subject. (22) object omission

Նա թքեց, շրջվեց և գնաց։ Na tʰkʰ-ecʰ, šɾǯ-v-ecʰ jev gn-acʰ. (s)he spit-aor.3sg turn-med-aor.3sg and go-aor.3sg ‘He spat, turned around and left.’ (constructed)

(23) focus on the object

Նա արյունոտ թուք թքեց։ Na aɾjunot tʰukʰ tʰkʰ-ecʰ. (s)he bloody saliva spit-aor.3sg ‘He spat blood-shot saliva.’ (constructed)

These verbs may be considered A-preserving labile (ambitransitive) verbs but, even more than with P-preserving labile verbs in ergative languages, it is sometimes difficult to distinguish between syntactically elliptical, implicit or absent objects.

4.3 Contentive-locative alternation Contentive-locative alternation is characteristic of few verbs such as verbs of filling or loading2 and reflects an alternative interpretation of a container as a Patient or a Place, with a respective interpretation of the filler/load as an Instrument or a Patient (or Theme). 2 The alternation corresponds to that available for ‘filling’ and ‘loading’ verbs in many languages including SAE (like the English load) and was termed locative alternation in this volume and in ValPaL. However, we prefer not to use just the term locative for this alternation because it gives an impression – false to our eyes – that one of these two constructions is primary and the other is derived (I loaded the truck with bricks → I loaded the bricks onto the truck). We see no grammatical evidence, at least for Armenian, to judge what of the two construction is primary.

Valency classes in Eastern Armenian

495

(24) container construed as Patient, filler as Instrument

Մարդը բաժակը լցրեց ջրով։

Maɾd-ə bažak-ə lcʰɾ-ecʰ ǯɾ-ov. man-def glass-def fill-aor.3sg water-ins ‘The man filled the glass with water.’ (constructed) (25) container construed as Place, filler as Theme Մարդը ջուր լցրեց բաժակը / բաժակի մեջ։ Maɾd-ə ǯuɾ lcʰɾ-ecʰ bažak-ə / bažak-i meǯ. man-def water fill-aor.3sg glass-def / glass-gen inside ‘The man poured water into the glass.’ (constructed) In terms of extensions (see below), these verbs in the two constructions are alternative extensions of the transitive pattern: S and DO (loadee) + spatial extension by Goal (locus) vs. S and DO (locus) + spatial extension by Instrument (loadee); see Appendix 1.

4.4 Proprietive (ablative-genitive) alternation Proprietive (ablative-genitive) alternation occurs with a few verbs that designate taking something off the surface of an object or taking away part of the object or someone’s possession. It may be viewed as an alternative construal of an object as a Possessor (in a part-whole relation) or a Source (cf. English I stole a book from him vs. I stole his book, where the possessive relation is that of legal possession). It is unclear whether either of the two valency patterns should be considered as primary. (26) ‘steal’: ablative Source/Possessor

Աղջիկը գիրքը գողացավ ուսուցչից։

Aʁǯik-ə giɾkʰ-ə goʁ-acʰ-av usucʰčʰ-icʰ. girl-def book-def steal-aor-3sg teacher-abl ‘The girl stole the book from the teacher.’ (constructed) (27) ‘steal’: internal genitive Source/Possessor

Աղջիկը գողացել է ուսուցչի գիրքը։

Aʁǯik-ə goʁacʰ-el ē usucʰčʰ-i giɾkʰ-ə. girl-def steal-cvb.pfv aux.3sg teacher-gen book-def ‘The girl stole the teacher’s book.’ (constructed)

496

Michael Daniel and Victoria Khurshudian

(28) ‘pluck’: ablative Source/Possessor

Ես ծառից խնձոր պոկեցի։

Jes car-icʰ χnʒoɾ pok-ecʰ-i. I tree-abl apple pluck-aor-1sg ‘I plucked an apple from a tree.’ (constructed) (29) ‘pluck’: internal genitive Source/Possessor

Ես ծառի խնձորը պոկեցի։

Jes car-i χnʒoɾ-ə pok-ecʰ-i. I tree-gen apple-def pluck-aor-1sg ‘I plucked the apple from the tree (lit. the tree’s apple).’ (constructed) Proprietive alternation shows a strong structural similarity to the widely attested external possessor construction (Payne & Barshi 1999) and to the much more rarely attested internal future (prospective) Possessor construction in e.g., the Northern Samoyedic languages of Uralic language family (see Khanina & Shluinsky in prep.). The proprietive alternation links the construction with an internal past-Possessor to the external past-Possessor construed as Source. From a SAE perspective, external possessor seems a construction derived from the internal one, while it would seem the opposite for the past or future Possessor construction. In Eastern Armenian, internal and external past-Possessors with verbs of removal however seem equally basic. Note that in all these cases, we cannot classify the alternation as valency changing or rearranging. The change in the number of verbal arguments results from moving a participant noun phrases between the position internal to one of the primary argument noun phrases and the position of a clause-level case. For the reciprocal alternation, object omission and extensions (see § 5), as well as of course for the two coded alternations, it is clear which of the valency patterns should be considered primary. For the contentive-locative and proprietive alternations this is an open issue. It would be useful to estimate text frequency of the two available options for each of the verbs; this is left for future research.

5 Extended valencies The main valency classes may be extended by additional arguments which are secondary as compared to primary, non-extended ones. Extensions are different from primary arguments in that the same (in terms of semantics and marking) extension may be optional, frequent or obligatory, depending on the verb. On the other hand, extensions, as opposed to adjuncts, may use flagging that is strongly associated with a set of verbs and cannot be used on adjuncts with a similar semantic role at all. Thus, in Armenian, animate Goal marking by a dative is optional

Valency classes in Eastern Armenian

497

with verbs like ‘bring’ and obligatory with ‘give’, but cannot be freely used as an adjunct with other verbs, where a special benefactive postposition hamar is used. Extensions thus introduce an additional slot associated with a specific marking and a specific semantic role, providing a fuller frame of the situation in terms of the participant set. The reason why valency classes formed by extensions are considered secondary, and why extensions should be distinguished from valency alternations, is that it is difficult to draw a line between those verbs with which an extension is optional and those that require it. It is in the latter case that verbs are traditionally considered as forming separate valency classes. Extensions often depend on the situation and contextual interpretation. Rather than licensed simply by the verb’s lexical class, they involve pragmatics (cf. Shibatani 1996 on benefactives). The important thing about extension is that the role is marked in the same way with all verbs, whether it is optional or obligatory. In other words, extensions may be represented on a scale of noun phrase occurring in different contexts with the same marking and the same or similar involvement in the situation but ranging syntactically from adjuncts through optional arguments to obligatory arguments. Tab. 3: ‘Extended’ valencies. adjunct

optional argument

obligatory argument

circumstantial

alternation-like

valency class-like

Thus, addition of a Beneficiary dative is an extension of the basic valency. Some verbal meanings may hardly ever allow it (‘burn’), some are compatible with it depending on the situation – whether it includes or not a Beneficiary participant – ‘cook’ or ‘build’; and some require it, like ‘give’. Obviously, this depends on how deeply the role of Beneficiary is integrated into the semantics of the verb, and it seems useless to try to cut this scale into separate valency classes – we will treat them as extensions ranging from optional to obligatory. Contrast dative contact verbs described below, which are a clear-cut valency class because licensing the contact dative is a lexical feature. Other examples of extensions are instrumental extension with an Instrument (36), possible with many transitive verbs but almost obligatory with verbs such as ‘cut’ even if the Instrument may be implicit (cf. Haspelmath 1993 on “instrumental verbs”); various spatial extensions (Place, Goal, or Source), possible with many verbs but obligatory even though sometimes implicit with some motion verbs; and ablative extension with a Source, possible with many verbs including motion but obligatory even though sometimes implicit with verbs such as ‘take’. Adding an optional argument may also be done by introducing a noun phrase flagged in a way which cannot be used as a free adjunct. Thus, the Cause/Stimulus with English be afraid is marked as of sth and not because of sth.

498

Michael Daniel and Victoria Khurshudian

5.1 Benefactive extension In Eastern Armenian, Benefactive extension occurs with verbs designating situations that allow one to introduce an indirectly affected human participant. The primary valency scheme is extended by a benefactive slot expressed by a dative noun phrase. The verbs are numerous and belong to different valency classes. Unlike the reciprocal alternation, the availability of the benefactive extension is not determined lexically but depends on the situation and may be more or less admissible with different speakers (cf. Shibatani 1996). Some verbs are more prone to it than others. In many languages, creation verbs easily introduce a dative Beneficiary participant. Ditransitive verbs are verbs for which the same extension is obligatory (Recipient). (30) ‘give’: obligatory Beneficiary (Recipient) with ditransitive verbs

Ես ձեզ տվեցի մեր Աստծո Աստվածաշունչը։

Jes ʒez tv-ecʰi meɾ astco astvacašunčʰ-ə. I you.pl.dat give-aor.1sg our lord-gen holy.script-def ‘I gave you our Lord’s Holy Script.’ (constructed) For the Beneficiaries that are less expected, a postpositional phrase with hamar ‘for’ may be used. (31) postpositional Beneficiary

Ես դա անում եմ քո մոր համար։ Jes da an-um em kʰo moɾ hamaɾ. I that do-cvb.ipfv aux.1sg you.sg-gen mother.gen for ‘I am doing that for your mother.’ (constructed)

5.2 Spatial extension Spatial extension introduces a slot expressed by the nominative for Goal (32), locative for Place, ablative for Source (33) or a postpositional phrase; more rarely a dative in Place (34) or Goal function (see Arkhangelskiy, Semionova in prep. on the locative domain in Eastern Armenian). This expansion slot, again, may be optional or obligatory, depending on the verb, ranging from clearly spatial adjuncts with almost any verb to almost obligatory spatial slots with motion verbs. (32) motion: spatial extension

Իսկ խանը չի էլ շտապում մտնել պալատ, բարձրանում է աշտարակներից մեկի սանդուղքներով վեր ու հասնելով վերջիններին՝ նայում ներքև։

Valency classes in Eastern Armenian

499

Isk χan-ə čʰ-i ēl štap-um mtn-el palat, but khan-def neg-aux.3sg even hurry-cvb.ipfv enter-inf palace baɾʒɾan-um ē aštaɾak-neɾ-icʰ mek-i sanduʁkʰ-neɾ-ov veɾ u rise-cvb.ipfv aux.3sg tower-pl-abl one-gen stairs-pl-ins up and hasn-el-ov veɾǯin-neɾ-i-n naj-um neɾkʰev. arrive-inf-ins last-pl-dat-def look-cvb.ipfv down ‘But the khan is not even in a hurry to enter into the palace, he goes up the stairs of one of the towers and, after reaching the last steps, looks down.’ (Gevorg Devrikian, eanc) (33) motion: spatial extension by Source

Գոմշասայլերը գինու տակառներ էին բերում քաղաքից։ Gomšasajl-eɾ-ə gin-u takar-neɾ ēin beɾ-um bull.cart-pl-def wine-gen barrel-pl aux.pst.3pl bring-cvb.ipfv kʰaʁakʰ-icʰ. town-abl ‘Bull carts were bringing wine barrels from the town.’ (Zarzand Darian, eanc)

(34) posture verb: spatial extension by Place

Թոմը նստեց ցանկապատին։

Tʰom-ə nst-ecʰ cʰankapat-i-n. Tom-def sit-aor.3sg fence-dat-def ‘Tom sat on the fence.’ (constructed)

5.3 Instrumental extension Instrumental extension introduces an Instrument coded by instrumental case. It is obligatory with verbs denoting processes typically carried out with instruments, such as ‘cut’ or ‘dig’, although these may be implicit when the object is typical for this process (knife, spade) or out of focus. (35) implicit instrument

Հացը կտրի՛ր։

Hacʰ-ə ktɾ-iɾ. bread-def cut-imp.2sg ‘Cut the bread.’ (constructed)

500

Michael Daniel and Victoria Khurshudian

(36) focused instrument

Սուր դանակով ամեն ինչ ավելի արագ կկտրես։ Suɾ danak-ov amen inčʰ aveli aɾag k-ktɾ-es. sharp knife-ins every thing more fast cond-cut-prs.2sg ‘You will cut everything faster with a sharp knife.’ (constructed)

6 Valency classes Below, we provide a classification of Eastern Armenian verbs into valency classes. A valency class is a group of verbs whose similar morphosyntactic behavior – which, for the purposes of the present volume, only includes argument flagging, marked valency derivations, unmarked valency alternations and extensions – is motivated by a similar linguistic construal of the real world situation and reflects participant sets that are perceived as similar. The latter proviso means that we may isolate different valency classes within a set of verbs that have identical morphosyntactic behavior based on a semantic analysis of these verbs, such as different valency classes or at least subclasses within a set of (morphosyntactically identical) dative verbs or ablative verbs (see below). For convenience, morphosyntactic groups are boldfaced, and valency classes are underlined. The main morphosyntactic distinction is that between intransitive and transitive verbs. This distinction is partly manifested in the verbal morphology: intransitive verbs use the theme ‑a‑ (with few exceptions). It also correlates with derivational properties: intransitive verbs rarely form mediopassives and transitive verbs do not frequently combine with the causative marker; and with behavioral features: transitive verbs take direct objects while intransitives do not. Transitivity is a macrofeature that divides Eastern Armenian verbs into major groups but is not sufficiently semantically motivated to delineate true valency classes: experiential verbs are aligned with transitive verbs, and the valency properties of contact verbs (see below) are in a way independent of transitivity. In a wider sense, transitivity may be considered as a category comprising more than two classes, as the cardinality of the valency, including zero-argument verbs, intransitive verbs with only one argument, dual intransitive verbs with two arguments but no direct object, semitransitive verbs which have only some of the transitivity properties listed above (e.g., take a direct object but do not form a mediopassive), transitive and then hypertransitive verbs. (To a certain extent, dual intransitive and hypertransitive verbs map onto Dixon and Aikhenvald’s (2000) extended intransitive and extended transitive verbs; but we have preferred to use different terms here to reserve the notion of extending valency to the kind of phenomena treated above in § 5.) These classes are discussed below in more detail. The reasons to introduce this scale instead of a binary opposition include inconsistent behavior of some dual intransitive and semitransitive verbs, which shows that the cardinality of the valency correlates with the conventional notion of formal transitivity.

Valency classes in Eastern Armenian

501

There is no morphosyntactic evidence for placing ‘give’ and other ditransitive verbs into a separate class; rather, they are transitive verbs with a benefactive extension ranging from optional (‘bring’) to obligatory (‘give’), in the same way as motion verbs are more or less obligatorily extended by spatial arguments. It is possible that Recipients and Beneficiaries with some verbs display special syntactic IO properties (such as e.g., reflexivization control), but in morphosyntactic terms ‘give’ and ‘bring’ behave in the same way. More generally, there are verbs that add further arguments, optional or obligatory, to the transitive pattern: ‘give’-verbs add a dative argument, ‘take’-verbs add an ablative argument etc. In terms of valency cardinality (number of arguments), they form a class of hypertransitive verbs. Figure 2 shows this classification together with major types of extensions that yield dual intransitive and hypertransitive verbs (dative and ablative verbs).

Tab. 4: Morphosyntactic classification of Eastern Armenian verbs. dative verbs

zeroargument

dative verbs ablative verbs true intransitive

dual intransitive

semitransitive

transitive

hyper-transitive

A more fine-grained classification that is semantically oriented and aimed at isolating true valency classes based on the similarity of participant sets, has to include, apart from flagging (case marking), also available valency derivations and unmarked alternations. Another important thing to note is that a verb may belong to different classes depending on the context: thus, ‘laugh’ may be a true intransitive or a dual intransitive, depending on the presence of Target (‘laugh’ vs. ‘laugh at’).

6.1. Weather verbs Weather verbs take no arguments. In Eastern Armenian, as in many other languages, verbs that take no arguments are meteorological verbs, including ‘rain’ (37), ‘snow’, ‘get dark’ etc. These verbs allow no mediopassive or causative derivation. Note the alternative intransitive construction for raining, as in (38), which is however structurally different and cannot qualify as an alternation. (37) weather: zero-valency

Մի անգամ ուժեղ անձրեւում էր, առանց հովանոցը փակելու բարձրացա երթուղային տաքսի։

Mi angam užeʁ anʒɾev-um ēɾ, arancʰ hovanocʰ-ə one time strong rain-cvb.ipfv aux.pst.3sg without umbrella-def

502

Michael Daniel and Victoria Khurshudian

pʰak-el-u baɾʒɾ-acʰa jeɾtʰuʁajin takʰsi. close-inf-gen rise-aor.1sg route taxi ‘One day it rained strongly, and (I) got into a minibus without folding my umbrella.’ (newspapers, eanc) (38) weather: alternative construction

Անձրև էր գալիս։

Anʒɾev ēɾ gal-is. rain aux.pst.3sg go-cvb.pfv ‘It rained; rain fell (lit. went).’ (constructed)

6.2 Intransitive verbs Intransitive verbs with one argument can typically be causativized but cannot form a mediopassive. They are not semantically homogeneous and thus may be divided into several different valency classes, including some motion and change of posture verbs (‘run’, ‘stand up’, ‘sit down’), internal states (‘ache’, ‘fear’, ‘be cold’), physiological processes (‘cough’, ‘die’, ‘sweat’) and sound production (‘laugh’, ‘scream’).

6.2.1 Motion verbs Motion verbs easily take or are more typical with a spatial extension of Source (ablative or postpositional phrase), Goal (postpositional phrase, nominative or dative; see Arkhangelskiy & Belyaev 2011; Arkhangelskiy & Semionova in prep.) or Place (locative or postpositional phrase). Spatial extension is not obligatory with any of them. Some motion verbs form morphological causatives. (39) ‘jump’ extended by Place (locative postpositional phrase)

Երեխան ցատկեց հատակի վրա։

Jeɾeχa-n cʰatk-ecʰ hatak-i vɾa. child-def jump-aor.3sg floor-gen on ‘The child jumped on the floor.’ (constructed) (40) ‘come’ extended by Goal (nominative in the lative sense)

Տղան եկավ գյուղ։

Tʁa-n jek-av gjuʁ. boy-def come-aor.3sg village ‘The boy came to the village.’ (constructed)

Valency classes in Eastern Armenian

503

(41) ‘go’: no extension (‘leave’)

Տղան գնաց։

Tʁa-n gn-acʰ. boy-def go-aor.3sg ‘The boy left.’ (constructed) (42) motion verb: causativized

Որպես պատիժ ուսուցիչը երեխաներին երկու կիլոմետր վազեցրեց։ Voɾpes patiž usucʰičʰ-ə jeɾeχa-neɾ-i-n jeɾku kilometɾ as punishment teacher-def child-pl-dat-def two kilometer vaz-ecʰɾ-ecʰ. run-caus-aor.3sg ‘As a punishment, the teacher made the children do a two-kilometer run.’ (constructed)

Most motion verbs, however, cannot form morphological causatives. A few of them have lexical causatives (ex. 43) and need not form morphological causatives (cf. however the discussion of examples 14–15 on the morphological causative of ‘die’, although a primary verb ‘kill’ also exists). More frequently than morphological or lexical causatives, intransitive motion verbs form periphrastic causatives using tal ‘give’ as a causativizer, which is a pattern more typical of distant than contact causativization; cf. (42) and (44). (43) ‘drop’: lexical causative for ‘fall’

Սիմոնյանը ձեռքից ցած գցեց բաժակը, փլվեց իր աթոռին ու հեծկլտաց։ Simonyan-ə ʒerkʰ-icʰ cʰac gcʰ-ecʰ bažak-ə Simonyan-def hand-abl down drop-aor.3sg glass-def pʰl-v-ecʰ iɾ atʰor-i-n u hecklt-acʰ. collapse-med-aor.3sg his chair-dat-def and sob-aor.3sg ‘Simonyan dropped the glass down from his hand, collapsed in his chair and sobbed.’ (Xarčik Hračian, eanc)

(44) ‘run’: periphrastic causative

Եվ այնպես վազել կտամ ես նրան, որ երբեք ոտքը չդիպչի քարին։ Jev ajnpes vaz-el k-ta-m jes nɾa-n, voɾ jeɾbekʰ and so run-inf cond-give-prs.1sg I (s)he.dat-def that never votkʰ-ə čʰ-dipčʰ-i kʰaɾ-i-n. foot-def neg-touch-sbjv.prs.3sg stone-dat-def ‘And I will make him run so that his feet will never be touching the ground.’ (translated, eanc)

504

Michael Daniel and Victoria Khurshudian

6.2.2 Sound production verbs Sound production verbs are one argument verbs, but may optionally include a postpositional Target. They may be causativized. (45)

Իշխանները ծիծաղեցին։ Išχan-neɾ-ə cicaʁ-ecʰ-in. prince-pl-def laugh-aor-3pl ‘The princes started to laugh.’ (Derenik Demirčian, eanc)

(46) sound production: causativised

Ծիծաղեցնելը ծիծաղեցնում եմ, բայց ոչ ոք չի ծիծաղում։ Cicaʁ-ecʰn-el-ə cicaʁ-ecʰn-um em, bajcʰ vočʰ vokʰ laugh-caus-inf-def laugh-caus-cvb.ipfv aux.1sg but no one čʰi cicaʁ-um. neg-aux.3sg laugh-cvb.ipfv ‘I do (try to) make them laugh, only no one is laughing.’ (translated, eanc)

(47) sound production: postpositional Target

Ծնողները գոռգոռացին ուսուցչի վրա։ Cnoʁ-neɾ-ə gorgor-acʰ-in usucʰčʰ-i vɾa. parent-pl-def yell-aor-3pl teacher-gen on ‘The parents had a fight with (yelled at) the teacher.’ (constructed)

(48) sound production: postpositional Target

Ծնողներիս վրա մի ծիծաղի՛ր։

Cnoʁ-neɾ-i-s vɾa mi cicaʁ-iɾ. parent-pl-gen-1poss on proh laugh-imp.2sg ‘Don’t laugh at my parents!’ (constructed)

6.2.3 Plain internal states Plain internal states ‘ache’ and ‘be cold’ do not have typical extensions. They may introduce additional participants such as Cause as a postpositional phrase or in the ablative, which is however never obligatory. (49) internal state with Cause

Ձյունն այնքան ճերմակ է, որ աչքերս ցավում են արևից։

Valency classes in Eastern Armenian

505

Ʒjun-n ajnkʰan čeɾmak ē, voɾ ačʰkʰ-eɾ-s cʰav-um snow-def so white be.3sg that eye-pl-1poss ache-cvb.ipfv en aɾev-icʰ. aux.3pl sun-abl ‘My eyes ache because of the sun, so white is the snow.’ (constructed) Verbs like ‘fear’ may license ablative Stimulus/Cause. (50) ‘fear’ with Stimulus/Cause

Մի՛ վախեցիր թշնամիներից, ընկերներից վախեցի՛ր։

Mi vaχecʰ-iɾ tʰšnami-neɾ-icʰ, ənkeɾ-neɾ-icʰ vaχecʰ-iɾ. proh fear-imp.2sg enemy-pl-abl friend-pl-abl fear-imp.2sg ‘Do not be afraid of the enemies, be afraid of the friends.’ (constructed)

6.2.4 Dative verbs A major and important intransitive class with two arguments are dative verbs. Intransitive dative verbs include contact verbs like ‘touch’ or ‘hit’ and interaction verbs such as ‘wait for’ or ‘help’. There is a discussion of dative intransitive verbs in (Blume 1998; see also Haspelmath 2001). Blume suggests that dative intransitive verbs typically designate complex events consisting of two sub-events and introduce two agentive-like participants. Her label for this class is interaction verbs, such as ‘listen’ (one of the participants is volitionally perceiving sound, while the other participant produces it), ‘follow’ (one of the participants is moving in the wake of another participant’s movement), or ‘help’ (one of the participants is doing something which facilitates what the other is trying to accomplish). These are of course very rough semantic definitions, but they do show that in each case there are two events, each with its own Agent (cf. however listening to a waterfall). Eastern Armenian shows that the interpretation of dative verbs may not always be straightforward. To start with, there are verbs of contact, such as ‘hit’ or ‘touch’, where the dative seems to inherit its allative Goal function from its purely spatial contexts (see Table 1). But more abstract verbs also introduce a dative second argument, such as ‘look at’, where this argument is by no means agentive. On the one hand, this verb is easily explained by a spatial metaphor; on the other, it is not unlike Blume’s verbs of communication / social gesture, such as ‘wave at’, that, for Blume, constitute a semantic subtype of interaction verbs. There is no a priori clear distinction between contact and interaction verbs, and the latter may be viewed as a gradual extension of the notion of target (a sub-type of Goal) to the domain of interpersonal interaction, from ‘hit’ / ‘touch’ to ‘look at’ / ‘shout at’ to ‘follow’ to ‘help’ to ‘meet’.

506

Michael Daniel and Victoria Khurshudian

(51) ‘touch’: a verb of contact

Սրի սառը շեղբը դիպավ նրա վզին։ Sɾ-i sarə šeʁb-ə dip-av nɾa vz-i-n. sword-gen cold edge-def touch-aor.3sg (s)he.gen neck-dat-def ‘The cold blade of the sword touched his neck.’ (constructed)

(52) ‘hit’: a verb of contact

Արսենը Պանդուխտի գլխին հարվածեց փայտով։ Aɾsen-ə Panduχt-i glχ-i-n haɾvac-ecʰ pʰajt-ov. Arsen-def Pandukht-gen head-dat-def hit-aor.3sg wood-ins ‘Arsen hit Panduxt on the head with a piece of wood.’ (Sahen Totikyan, eanc)

(53) ‘meet’: interaction verb

Երբեք նման ապուշի չեմ հանդիպել։

Jeɾbekʰ nman apuš-i čʰ-em handip-el. never such idiot-dat neg-aux.1sg meet-cvb.pfv ‘I have never met such an idiot.’ (constructed) (54) ‘help’: interaction verb

Բայց քանի որ ես հիվանդ չեմ և ուզում եմ գնալ, ինձ ոչ մի բժիշկ չի օգնի։ Bajcʰ kʰani voɾ jes hivand čʰ-em jev uz-um em but as.much that I sick neg-be.1sg and want-cvb.ipfv aux.1sg gn-al, inʒ vočʰ mi bžišk čʰ-i ōgn-i. go-inf I.dat not one doctor neg-aux.3sg help-cvb.conneg ‘But as I am not ill and want to leave, there is no doctor who can help me.’ (Vahagn Grigoryan, eanc)

Isolating dative interaction verbs in Armenian is superficially complicated by the fact that, as discussed in § 2, human DO take dative marking, and many interaction verbs – such as ‘help’, ‘meet’ or ‘follow’ – are typically used with human second participants, making them similar to transitive verbs. However, inanimate noun phrases with interaction verbs, though rare, still use the dative, not the nominative, as a transitive verbs would; cf. the following examples: (55) ‘follow’: interaction verb with an inanimate participant

Սիրունյանը հազիվ էր կարողանում հետևել կնոջ խոսքերին։ Siɾunjan-ə haziv eɾ kaɾoʁan-um hetev-el Sirunyan-def hardly aux.pst.3sg be.capable-cvb.ipfv follow-inf

Valency classes in Eastern Armenian

507

knoǯ χoskʰ-eɾ-i-n. woman.sg.gen speech-pl-dat-def ‘Sirunyan was hardly able to follow what the woman was saying.’ (Gagik Rštuni, eanc) (56) ‘meet’: interaction verb with an inanimate participant Նրանք ամենևին չէին սպասում, թե իրենց ծրագիրը

դիմադրության հանդիպել։

կարող էր որևէ

Nɾankʰ amenevin čʰ-ein spas-um, tʰe iɾencʰ they by.no.means neg-aux.pst.3pl wait-cvb.ipfv that self.pl.gen cɾagiɾ-ə kaɾoʁ eɾ voɾeve dimadɾutʰjan handip-el. plan capable be.pst.3sg any resistance.dat meet-inf ‘They never expected that their plan would possibly meet any opposition.’ (Ašot Šajbon, eanc) Some dative verbs may show inconsistent transitive properties, which makes this class more transitive than the previous one. Thus, ‘look at’ may be used transitively in sense of ‘watch’, ‘look after’. The dative verbs in Eastern Armenian are a morphosyntactic class independent of transitivity. In addition to the intransitive dative verbs discussed above, transitive non-derived and morphological causatives also may license a dative – see below.

6.3 Transitive verbs In Eastern Armenian, transitive verbs may be defined syntactically as verbs that take a direct object (a noun phrase flagged with dative or nominative, according to the differential object marking rules) and form a mediopassive.

6.3.1 Semitransitive verbs There are two types of semitransitive verbs whose behavior is not consistent with respect to these two criteria. The first comprises a few verbs that take direct objects but do not form a mediopassive. On the list of verbal meanings in the valency database, these are two experiential verbs ‘like’ (havanel) and ‘want’, ‘hide’ which is a defective causative (a morphological causative whose root is never used without the causative marker) and the verb ‘hunt’. At least for the experiential verbs this may be interpreted as a non-complete alignment of non-transitive meanings with the transitive pattern – or that, in other words, passivization prefers true Patients. On the other hand, many transitively aligned experiential verbs, such as ‘love’ or ‘remember’, do form a mediopassive and thus are true transitives from the formal point of view. Apparent semitransitivity of the defective causative

508

Michael Daniel and Victoria Khurshudian

‘hide’ may be explained morphologically, as mediopassives from causatives are dispreferred (they are frequent with three and marginally occur with eight more verbs; see the table in the Appendix 3) although not impossible (cf. 16 above). (57) ‘want’: direct object, but no mediopassive – Ես թույն եմ ուզում, – ասաց նա։ Jes tʰujn em uz-um, as-acʰ na. I poison aux.1sg want-cvb.ipfv say-aor.3sg (s)he ‘I want (to take) poison, he said.’ (translated, eanc) (58) ‘hide’: direct object, but no mediopassive Չէ որ լավ է ճանաչում մորը, նա ամեն ինչ

որևէ բան թաքցնել։

նկատում է, դժվար է նրանից

Čʰē voɾ lav ē čanačʰ-um moɾ-ə, na amen inčʰ no that well aux.3sg know-cvb.ipfv mother.dat-def (s)he every what nkat-um ē, džvaɾ ē nɾan-icʰ voɾevē ban notice-cvb.ipfv aux.3sg difficult be.3sg. (s)he-abl whatever thing tʰakʰ(-)cʰn-el. hide-caus-inf ‘Does he not know his mother well enough – she notices everything, it’s so hard to hide anything from her.’ (Zori Balajan, eanc) As noted above, verbs in -al, a closed class in modern Eastern Armenian, form mediopassives (as well as some other forms) based on the aorist stem. Functionally, they are regular causatives. (59) mediopassive based on the aorist stem

Ազատ և զգոն սովետական հայրենասերների օգնությամբ հիտլերյան այդ ապականությունը որսացվում և ոչնչացվում է։

Azat jev zgon sovetakan hayɾenaseɾ-neɾ-i ōgnutʰj-amb hitleɾjan ajd free and vigilant Soviet patriot-pl-gen help- ins Hitlerian that apakanutʰjun-ə voɾs-acʰ-v-um jev vočʰnčʰ-acʰ-v-um dirtiness-def hunt-aor-med-cvb.ipfv and destroy-caus-med-cvb.ipfv ē. aux.3sg ‘With the help of Soviet free and vigilant patriots the Hitler’s filth is hunted and destroyed.’ (newspapers, eanc) Second, there are several transitive verbs that are (more) often used without an object, because it is either generalized or implicit – what is described above as the object omission alternation – A-preserving ambitransitive verbs.

Valency classes in Eastern Armenian

509

(60) object omission: general object Ու երգեց սիրուց, երգեց կյանքից։ U jeɾg-ecʰ siɾ-ucʰ, jeɾg-ecʰ kjankʰ-icʰ. and sing-aor.3sg love-abl sing-aor.3sg life-abl ‘And he sang about love, and he sang about life.’ (Gurgen Mahari, eanc) (61) non-default object

Հիվանդը միզում է արյունով մեզ։ Hivand-ə miz-um ē aɾjun-ov mez. sick-def urinate-cvb.ipfv aux.3sg blood-ins urine ‘The patient urinated with blood (urinated urine with blood).’ (constructed)

(62) default object

Երեխան միզում է թփի տակ։ Jeɾeχa-n miz-um ē tʰpʰ-i tak. child-def urinate-cvb.ipfv aux.3sg bush-gen under ‘The child peed in the bushes.’ (constructed)

6.3.2 True transitive verbs The next morphosyntactic class includes all other, true transitive verbs that (most often) require a direct object and form a mediopassive. Possible direct objects in Eastern Armenian include true Patients whose state is subject to change, such as deformation (‘cut’, ‘break’), handling with heat, fire, water etc. (‘cook’, ‘burn’, ‘wash’), coming into existence or ceasing to exist (‘make’, ‘build’, ‘kill’), as well as Themes (‘put’, ‘throw’) and Stimuli of experiential verbs aligned with transitives (‘love’, ‘see’). They may be further classified according to the alternations they allow and to the availability of extensions to hypertransitive verbs (introducing a third argument) they allow or require. The transitive prototype includes verbs of caused change of state brought about in one of the participants (the Patient) by another (an Agent): ‘burn’, ‘beat’, ‘tear’, ‘cut’, ‘dig’, ‘wash’, ‘scrape’, ‘cook’, ‘hug’, ‘open’, ‘kill’, ‘sink’, ‘break’. Instrumental extension is typical of if not obligatory for some verbs in this class, because verbs such as ‘cut’ or ‘dig’ or ‘wash’ are always done with a special instrument or means or its substitute. If the instrument is out of focus, it remains implicit in most contexts, but if it has to be additionally characterized – cut with a sharp knife – or if it is a substitute – cut with an iron bar – it is an explicit noun phrase marked by instrumental case. Its connection to the default participant makes instrumental extension similar to the object omission alternation (cf. spit vs. spit blood-shot saliva and spit a fruit stone).

510

Michael Daniel and Victoria Khurshudian

(63) implicit instrument

Դու ինքդ քո գերեզմանը կփորես, իսկ հետո ես քեզ կթաղեմ։ Du inkʰ-d kʰo geɾezman-ə k-pʰoɾ-es, isk heto jes you.sg self-2poss you.sg-gen grave-def cond-dig-prs.2sg and then I kʰez k-tʰaʁ-em. you.sg.dat cond-bury-prs.1sg ‘You will dig your own grave, and then I’ll bury you.’ (constructed)

(64) explicit instrument

Նա սկսեց փորել հին բահով, որը մի րոպեից ջարդվեց։ Na sks-ecʰ pʰoɾ-el hin bah-ov, voɾ-ə mi ɾope-icʰ (s)he start-aor.3sg dig-inf old spade-ins that-def one minute-abl ǯaɾd-v-ecʰ. break-med-aor.3sg ‘He started digging with an old spade that broke in a minute.’ (constructed)

(65) focus on the instrument Մի անգամ տեսա, որ մեկը փայտով ծեծում է ոչխարին։ Mi angam tes-a, voɾ mek-ə pʰajt-ov cec-um ē one time see-aor.1sg that one-def stick-ins beat-cvb.ipfv aux.3sg vočʰχaɾ-i-n. sheep-dat-def ‘Once I saw someone beating a ram with a stick.’ (newspapers, eanc) (66) instrumental extension: non-default instrument Համբույրներով, քո աչերով ինձ այրեցիր։ Hambujɾ-neɾ-ov, kʰo ačʰ-eɾ-ov inʒ ajɾ-ecʰiɾ. kiss-pl-ins you.sg.gen eye-pl-ins I.dat burn-aor.2sg ‘(You) burnt me with your kisses, with your eyes.’ (internet forum, eanc)

6.3.3 Experiential transitives Quite a few experiential and perception verbs are fully aligned with the transitive pattern, taking Stimuli as DO and undergoing mediopassive derivation. By experiential verbs we do not mean exclusively verbs that denote sensations or perceptions, but a wider class including in addition to perception (e.g., ‘see’) also cognition (e.g., ‘know) and emotional or physiological states (e.g., ‘be glad’ or ‘be cold’) with a human participant usually classified as Experiencer. In Eastern Armenian, experiential transitives include ‘love’ (emotion, attitude), ‘hear’ (perception), ‘re-

Valency classes in Eastern Armenian

511

member’ and ‘know’ (cognition). The verb ‘see’, although it does form a mediopassive, does not belong here. It is a semitransitive verb, as its mediopassive has a reciprocal rather than passive interpretation: (67) ‘see’: transitive

Ես ցամաք եմ տեսնում։ Jes cʰamakʰ em tesn-um. I land aux.1sg see-cvb.ipfv ‘I see land.’ (constructed)

(68) ‘see’: reciprocal mediopassive

Անցյալ օրը մի կին եկավ ու խնդրեց ինձ հետ տեսնվել … Ancʰjal ōɾ-ə mi kin jek-av u χndɾ-ecʰ inʒ het past day-def one woman come-aor.3sg and ask-aor.3sg I.dat with tesn-v-el. see-med-inf ‘The other day a woman came and asked to meet me (to see each other).’ (newspapers, eanc)

The meaning ‘smell’, although formally a transitive construction, is based on a periphrastic construction ‘take the smell of’ (alternatively, zgal ‘feel’): (69)

Ես մարդկային ոգու հոտ եմ առնում։ Jes maɾdkajin vog-u hot em arn-um. I human soul-gen smell aux.1sg take-cvb.ipfv ‘I can smell the smell of a human being.’ (constructed)

The verb ‘look’ can be used transitively for intentional situations of watching: (70) ‘look’: transitive version Շատ կարեւոր է, որ մարդիկ սովորեն լավ ֆիլմեր նայել։ Šat kaɾevoɾ ē voɾ maɾd-ik sovoɾ-en lav film-eɾ very important be.3sg that man-pl learn-sbjv.prs.3pl good film-pl naj-el. look-inf ‘It is very important that people learn to watch good movies.’ (newspapers, eanc)

6.4 Caused motion verbs Caused motion verbs belong to the same valency class as intransitive motion verbs above except that they introduce an Agent. These are verbs like ‘put’, ‘push’, ‘roll’, ‘throw’, ‘bring’, ‘send’ and causatives of intransitive motion and change of posture verbs such as ‘cause to run’ or ‘cause to sit down’.

512

Michael Daniel and Victoria Khurshudian

(71) caused motion: primary

Գետակն անփութորեն գլորում էր իր ջրերը։

Getak-n anpʰutʰoɾen gloɾ-um ēɾ iɾ brook-def carelessly roll- cvb.ipfv aux.pst.3sg self.gen ǯɾ-eɾ-ə. water-pl-def ‘The small river was carelessly rolling its water.’ (Mkrtič Sargsyan, eanc) (72) caused motion: mediopassive

Հողաթմբի տակից ջրերը գլորվում էին խռովահույզ և աղմկում, վազում առաջ։ Hoʁatʰmb-i tak-icʰ ǯɾ-eɾ-ə gloɾ-v-um embankment-gen under-abl water-pl-def roll-med-cvb.ipfv ēin χrovahujz jev aʁmk-um vaz-um araǯ. be.aux.pst.3pl agitatedly and rush-cvb.ipfv run-cvb.ipfv forth ‘From under the embankment the water was quickly rolling, roaring and running forth.’ (Cerun Togomyan, eanc)

Just like intransitive motion verbs, caused motion verbs may be extended by a spatial slot. (73) caused motion: extended by Goal

Ամռանն անհնար է երեխաներին տուն մտցնել։ Amr-an-n anhnaɾ ē jeɾeχa-neɾ-i-n tun summer-dat-def impossible be.3sg child-pl-dat-def house mt-cʰn-el. enter-caus-inf ‘In summer it is impossible to make the children come home (enter the house).’ (constructed)

(74) caused motion: extended by Goal

Բակում ձիապանը բարձում էր իրերս, նանին պայուսակի մեջ դարսում էր գաթա և ճանապարհի պաշարը։ Bak-um ʒiapan-ə baɾʒ-um ēɾ iɾ-eɾ-s, nani-n yard-loc coach-def load-cvb.ipfv aux.pst.3sg thing-pl-1poss Mom-def pajusak-i meǯ daɾs-um ēɾ gatʰa jev čanapaɾh-i bag-gen inside put-cvb.ipfv aux.pst.3sg gata and road-gen pašaɾ-ə. stock-def ‘In the courtyard, the coach was loading my luggage, and Mom was putting gata [a traditional Armenian pastry] and food for the travel into a bag.’ (Aksel Bakunc, eanc)

Valency classes in Eastern Armenian

513

Unlike intransitive motion verbs, some caused motion verbs may represent property transfer events (‘give’-events), with an object passed from one human participant to another. In this case they are extended by the benefactive slot which introduces a special type of Goal of the motion: Recipient of the Theme, its prospective possessor, a human participant that is coded by dative. This extension is possible with ‘throw’, typical but optional with verbs like ‘bring’ or ‘send’, and obligatory with ‘give’ (which is often considered to form a separate valency class but is here viewed as a special case of caused motion verbs). (75) ‘throw’: optional Beneficiary slot (Recipient/Goal)

Նա ինձ նետեց բանալիները և վազեց-գնաց։

Na inʒ net-ecʰ banali-neɾ-ə jev vaz-ecʰ gn-acʰ. (s)he I.dat throw-aor.3sg key-pl-def and run-aor.3sg go-aor.3sg ‘He threw me the keys and ran away.’ (constructed) (76) ‘send’: extension by Beneficiary slot (Recipient/Goal)

Այնտեղից մեզ նամակներ ուղարկիր տուն։

Ajnteʁ-icʰ mez namak-neɾ uʁaɾk-iɾ tun. there-abl we.dat letter-pl send-imp.2sg home ‘Send us letters from there, back home.’ (constructed) (77) ‘give’: obligatory Beneficiary slot (Recipient/Goal)

Բայց իմ կինը ծաղիկը տվեց մարդուն։

Bajcʰ im kin-ə caʁik-ə tv-ecʰ maɾd-u-n. but my woman-def flower-def give-aor.3sg man-dat-def ‘But my wife gave the flower to the man.’ (constructed) Morphosyntactically, such animate Goals-Recipients are similar to inanimate spatial Goals that often use the dative. Truly spatial animate Goals, on the other hand, use a dedicated personal space postposition mot ‘at X’s place; to X’s place’, even if the participant is contextually also a Beneficiary: (78) intransitive motion: postpositional animate Goal

Կինս վերադարձել է ինձ մոտ։

Kin-s veɾadaɾʒ-el ē inʒ mot. woman-1poss return-cvb.pfv aux.3sg I.dat at ‘My wife came back to me.’ (constructed)

514

Michael Daniel and Victoria Khurshudian

Other caused motion verbs are used for transfer situation whose focus is on the Source rather than on the Goal: ‘receive’, ‘take from’, ‘steal’ etc. These verbs are extended by a Source coded by an ablative: (79) caused motion (→ transfer): extended by Source

Այդպիսով կհասցնենք ավելի շատ բան ստանալ կյանքից։ Ajdpis-ov k-hascʰn-enkʰ aveli šat ban stan-al kjankʰ-icʰ. thus-ins cond-manage-prs.1pl more many thing receive-inf life-abl ‘In that way we’ll manage to get much more from life.’ (translated, eanc)

(80) caused motion (→ transfer): extended by Source

Նա ինձնից ձին մի օրով վերցրեց։

Na inʒn-icʰ ʒi-n mi ōɾ-ov veɾcʰɾ-ecʰ. (s)he I-abl horse-def one day-ins take-aor.3sg ‘He borrowed my horse for one day.’ (constructed) The same valency is also applied to ‘ask for something’ and ‘ask a question’: (81)

Ինձնից ճամփան մի հարցրու։ Inʒn-icʰ čampʰa-n mi haɾcʰɾu. I-abl road-def proh ask.imp.2sg ‘Don’t ask me the way.’ (constructed)

6.4.1 Removal verbs The property that makes removal verbs a separate valency class is the proprietive alternation: the Source may be construed as a Possessor and coded by a genitive internal to the Patient noun phrases. Cf. English he took my pen vs. he took a pen from me. Unlike English, in Eastern Armenian this construal extends to two types of inanimate Possessors in part-whole relation to the Patient, e.g., with verbs like ‘pluck’ and further to situations like wiping off dust from a surface. For these two alternations, both variants are equally probable. (82) ‘pluck’: ablative

Ես ծառից խնձոր պոկեցի։

Jes car-icʰ χnʒoɾ pok-ecʰi. I tree-abl apple pluck-aor.1sg ‘I plucked an apple from a tree.’ (constructed)

Valency classes in Eastern Armenian

515

(83) ‘pluck’: internal genitive

Ես պոկեցի ճյուղի խնձորը։

Jes pok-ecʰ-i čjuʁ-i χnʒoɾ-ə. I pluck-aor-1sg branch-gen apple-def ‘I plucked the apple from the branch.’ (constructed) (84) ‘wipe’: internal genitive Source/Possessor

Վարդանն անզոր նստեց ծալովի աթոռի վրա և ձեռքով մաքրեց ճակատի քրտինքը։ Vaɾdan-n anzoɾ nst-ecʰ calovi atʰor-i vɾa jev ʒerkʰ-ov Vardan-def weak sit-aor.3sg pliable chair-gen on and hand-ins makʰɾ-ecʰ čakat-i kʰɾtinkʰ-ə. clean-aor.3sg forehead-gen sweat-def ‘Vardan limply sank to a folding chair and wiped sweat off his forehead with his hand.’ (Jakob Xačikyan, eanc)

(85) ‘wipe’: ablative Source/Possessor Արագաշարժ «մաքրիչները» հազիվ

մաքրել թաց ձյունը։

են հասցնում հողմապակու վրայից

Aɾagašaɾž «makʰɾičʰ-neɾ-ə» haziv en hascʰn-um quick wiper-pl-def hardly aux.3pl manage-cvb.ipfv hoʁmapak-u vɾaj-icʰ makʰɾ-el tʰacʰ ʒjun-ə. windshield-gen on-abl clean-inf wet snow-def ‘The quick wipers hardly were in time to take the snow away from the windshield.’ (Zori Balajan, eanc)

6.4.2 Filling verbs A small subgroup of caused motion verbs is formed by the filling verbs ‘load’ and ‘pour’, which are different from the other caused motion verbs in that they allow the contentive-locative alternation. This alternation is due to the fact that there is competition for the Patient’s slot between the container being filled and the Theme being transferred. (86)

Մայրը դույլը լցրեց ջրով։ Majɾ-ə dujl-ə lcʰɾ-ecʰ ǯɾ-ov. mother-def bucket-def pour-aor.3sg water-ins ‘Mother filled the bucket with water.’ (constructed)

516

Michael Daniel and Victoria Khurshudian

(87)

Մարդը ջուրը լցրեց բաժակը / բաժակ-ի մեջ։ Maɾd-ə ǯuɾ-ə lcʰɾ-ecʰ bažak-ə / bažak-i meǯ. man-def water-def pour-aor.3sg glass-def / glass-gen in ‘The man poured water into the glass.’ (constructed)

This puts verbs of filling between plain transitive verbs obligatorily extended with an instrumental slot (the container is the Patient, the filler is the means or the Instrument) and caused motion verbs (the filler is the Theme and the container is the Place).

6.4.3 Creation verbs While removal verbs use the ablative in its primary ‘free’ meaning as a Source, there are other verbs whose ablative is more lexically determined. Such is the optional use of the ablative with creation verbs (‘do’, ‘make’ or ‘build’ etc.) where the ablative designates material or components of the object coming into existence. (88) creation verb: extended by material Source

Պառավը նրա համար խոտից շրջազգեստ կարեց։ Parav-ə nɾa hamaɾ χot-icʰ šɾǯazgest kaɾ-ecʰ. old.woman-def (s)he.gen for grass-abl dress sew-aor.3sg ‘The old woman sewed her a dress out of grass.’ (constructed)

The second group of ablative verbs are a few verbs whose ablative optionally indicates the point where the Patient is affected (the locus of impact) such as ‘tie to’ or ‘take by’: (89) contact verb: locus of contact Սա՞ ինչ բան է, – նա բարկությամբ քաշեց լաթի ծայրից։ Sa inčʰ ban ē, – na baɾkutʰj-amb kʰaš-ecʰ latʰ-i this what thing be.3sg (s)he wrath-ins pull-aor.3sg clothes-gen cajɾ-icʰ. end-abl ‘What is this? – He furiously pulled at the tip of the clothes.’ (translated, eanc)

6.5 Transitive contact verbs Transitive contact verbs are transitive counterparts to intransitive dative verbs and designate caused motion events that result in physical contact between the Theme and the Goal. This class includes e.g., ‘throw’ as a non-derived verb:

Valency classes in Eastern Armenian

517

(90) ‘throw’: caused contact

Ես գնդակը նետեցի պատին, այն հետ թռավ և ջարդեց պատուհանը։

Jes gndak-ə net-ecʰ-i pat-i-n, ajn het tʰr-av jev I ball-def throw-aor-1sg wall-dat-def that back fly-aor.3sg and ǯaɾd-ecʰ patuhan-ə. break-aor.3sg window-def ‘I threw the ball at the wall, it bounced and broke the window.’ (constructed) The verb ‘attach’ is a morphological causative of ‘touch’ and is a transitive counterpart to intransitive ablative verbs licensing ablative ‘locus of contact’. (91) ‘attach’: ablative ‘locus of contact’ Այծը կապիր ծաոից, որ չփախչի։ Ajc-ə kap-iɾ car-icʰ, voɾ čʰ-pʰaχčʰ-i. goat-def connect-imp.2sg tree-abl that neg-run.away-sbjv.prs.3sg ‘Tie the goat to a peg so it cannot run away.’ (constructed)

6.6 Verbs of speech Verbs of speech take as a direct object noun phrases designating the content of the speech act. In most cases this slot is taken by a reported speech clause, which may become a formal S under passivization. They also introduce a third participant, Addressee, coded by dative. Example (93), however, is ambiguous between Addressee of speech and the target-Goal of dative verbs. (92) speech: dative Addressee Կհավատա՞ս, Գրիգորի Պանտելևիչ, հենց

քրտինք եկավ։

որ նա ինձ այդ ասաց, վրաս պաղ

K-havat-as, Gɾigoɾi Pantelevičʰ, hencʰ voɾ na inʒ ajd cond-believe-prs.2sg Grigorij Panteleevich just that (s)he I.dat that as-acʰ, vɾa-s paʁ kʰɾtinkʰ jek-av. say-aor.3sg on-1.poss cold sweat come-aor.3sg ‘Would you believe this, G.P., the moment he told me that – I was all covered with cold sweat.’ (translated, eanc) (93) … եւ ես քիչ էր մնում բացականչեի նրան … Jev jes kʰičʰ ēɾ mn-um bacʰakančʰe-i and I little aux.pst.3sg remain-cvb.ipfv shout-sbjv.pst.1sg

518

Michael Daniel and Victoria Khurshudian

nɾa-n (s)he.dat-def ‘… and I hardly kept myself from shouting to him … ’ (newspapers, eanc) (94) ‘say’: passive

Եվ ասվել էր՝ մի սպանի՛ր։ Jev as-v-el ēɾ mi span-iɾ. and say-med-cvb.pfv aux.pst.3sg proh kill-imp.2sg ‘And it was said: do not kill.’ (constructed)

The table summarizing the classification of Eastern Armenian verbs into morphosyntactic groups and then valency classes is found in the appendix.

7 Summary In Eastern Armenian, a transitive verb takes a direct object (dative/nominative marking according to the animacy split) and forms a mediopassive. Some verbs may be classified as semitransitive in the sense that they only have one of these two definitional properties; either their direct object is optional (Patient is usually implicit or generalized) or they do not form mediopassives (several transitively aligned experiential verbs and a couple exceptions). A few verbs like ‘look’ or ‘wait’ alternate between intransitive and transitive patterns with shifts in meaning. On the other hand, verbs that do not take direct objects, i.e. intransitive verbs, typically do not form a mediopassive but do form morphological causatives. However, this distribution of valency derivations between the two major morphosyntactic groups is not a strict rule, and availability of passivization and causativization also varies between individual verbs (see Table 2 in the appendix). Armenian verbs further split into several major morphosyntactic groups: verbs taking no arguments at all; intransitive verbs taking only one argument; dative intransitive verbs that add a second obligatory argument in the dative (notably, not only Blume’s (1998) interaction verbs but also verbs of contact, reflecting the connection of the dative case to the Goal); transitive verbs; dative transitive verbs (causative counterparts to dative intransitive verbs); and ablative verbs (verbs indicating the locus of impact). Other valencies are mostly formed by extending the primary valency of the verb by various slots including Place, Goal, Source, Beneficiary, Instrument or Cause, to introduce additional participants, more often optional but sometimes also obligatory (such as dative Beneficiary/Recipient with ‘give’, postpositional Place/Goal with ‘put’).

Valency classes in Eastern Armenian

519

Both extensions and unmarked valency alternations play an important role in a further, more fine-grained classification of Eastern Armenian verbs, one that goes beyond transitivity. Thus, the contentive-locative (alias locative) alternation, typical of many Standard Average European languages, allows verbs of filling to alternate between (DO=container; INS=filler) and (DO=filler; LOC=container) patterns, depending on which wins the patient slot. A less obvious alternation is the proprietive one which involves removal verbs – the same participant is construed either as a Source (ablative marking) or a Possessor (genitive marking within the Patient noun phrase). This alternation is actually typical of many languages for human Sources (take someone’s book ~ take a book from someone); Eastern Armenian is unusual in that it expands this alternation to some inanimate Possessors and to verbs that designate removing a superficial layer of an object (e.g., ‘wipe’) or a part from the whole (e.g., ‘pluck’). As these examples show, unmarked alternations are motivated by the availability of different construals of the participants. Extensions can be viewed as tendencies to introduce specific additional participants and thus, through valency effects, also reflect semantic differences between verb classes. Motion verbs are often extended by spatial noun phrases, caused motion verbs may be re-interpreted as transfer verbs extending their valency with a human participant, Source or Goal depending on the type of the primary motion verb. By taking into account transitivity and morphosyntactic class of the verb, available unmarked alternations, extensions, argument marking and the semantics of the verb, we arrive at a classification of Eastern Armenian verbs into the following valency classes: weather (e.g., ‘rain’), motion and change of posture (‘jump’, ‘sit down’) and caused motion (‘put’) and the extension of the latter into property transfer (‘bring’), internal state and physiological processes (‘ache’, ‘cough’), sound production (‘laugh’), contact (‘touch’), interaction verbs (‘help’), two classes of experiential verbs, one inconsistently, the other consistently transitive (‘want’ and ‘love’, respectively), caused change of state (‘burn’), instrumental change of state (‘cut’), filling (‘pour’), removal (‘peel’), creation (‘build’) and speech (‘say’). The relations among the classes, their connection to various extensions and to morphosyntactic properties connected to transitivity are shown in Figure 3. The classification of the verbs provided in this paper is an approximation based exclusively on semantics and morphosyntax but disregarding purely syntactic properties of the argument noun phrases (such as reflexivization control) and relative frequencies of the alternating constructions (e.g., one uncoded alternation vs. another). Our main purpose was to discover behind individual verbal valencies some common features that group them together on semantic grounds. Hopefully, this contributes to a cross-linguistic comparison of how similar verbal meanings are grouped into valency classes, which is the ultimate objective of the present volume.

semitransitive

transitive

spatial extension: apɾel ‘live’, gal ‘come’, gnal ‘go’, nstel ‘sit’

motion and posture: baɾʒɾanal ‘climb’, vazel ‘run’, cʰatkel ‘jump’, ənknel ‘fall’, gal ‘come’, gnal ‘go’, meknel ‘leave’, nstel ‘sit’, apɾel ‘live’

caused motion: tal ‘give’, uʁaɾkel ‘send’, beɾel ‘bring’, dnel ‘put’, gloɾel ‘roll’, tanel ‘carry’, hɾel ‘push’, goʁanal ‘steal’, veɾcʰnel ‘take’

creation verbs: anel ‘do’, karucʰel ‘build’, šinel ‘build’, saɾkʰel ‘make’

caused (change of) state: ajɾel ‘burn (tr)’, cackel ‘cover’, cecel ‘beat’, gɾkel ‘hug’, hagcʰnel ‘dress’, jepʰel ‘cook’, ktɾel ‘cut’, loʁacʰnel ‘wash’, patrel ‘tear’, pʰoɾel ‘dig’, pokel ‘tear’, spanel ‘kill’, ǯaɾdel ‘break’, χoɾtakel ‘sink’

target extension: cicaʁel ‘laugh’

havanel ‘like’, jergel ‘sing’, tʰaɾtʰel ‘blink’, tesnel ‘see’, utel ‘eat’, uzel ‘want’

sound production: cicaʁel ‘laugh’, hazal ‘cough’

cause extended valency: cʰavel ‘ache’, mahanal ‘die’, mernel ‘die’ vaχenal ‘fear’

dual intransitive

experiential (tr aligned): čanačʰel ‘know’, lsel ‘hear’, siɾel ‘love’, sovoɾel ‘learn’

(change of) state: jeral ‘boil’, mahanal ‘die’, mernel ‘die’

weather: anʒɾevel ‘rain’

experiential verbs: mɾsel ‘feel cold’, sarel ‘feel cold’, sovac linel (sovel) ‘be hungry’, cʰavel ‘ache’, vaχenal ‘fear’

intransitive

zero

Appendix 1

(motion) + Goal dnel ‘put’, tanel ‘carry’, hɾel ‘push’

(filling verbs) + Goal baɾʒel ‘load’, lcʰnel ‘pour’

spatial extension: (transfer verbs) + Source goʁanal ‘steal’, veɾcʰnel ‘take’

(caused motion verbs) + Beneficiary: tal ‘give’, uʁaɾkel ‘send’, beɾel ‘bring’, anel ‘do’, karucʰel ‘build’, šinel ‘build’, saɾkʰel ‘make’

benefactive extension: (creation verbs) + Beneficiary anel ‘do’, karucʰel ‘build’, šinel ‘build’, saɾkʰel ‘make’

(filling verbs) + Instrument baɾʒel ‘load’, lcʰnel ‘pour’

instrumental extension: (caused change of state) + Instrument cecel ‘beat’, ktɾel ‘cut’, pʰoɾel ‘dig’

hypertransitive

520 Michael Daniel and Victoria Khurshudian

Dative contact: haɾvacel ‘hit’, kpčʰel ‘touch’

interaction: handipel ‘meet’, hetevel ‘follow’, najel ‘look at’, ognel ‘help’

comitative: amusnanal ‘marry’, χosel ‘talk’

Ablative

Dative (transitive)

removal: makʰɾel ‘peel wipe’

locus of contact: kapel ‘tie’

contact verbs: netel ‘throw (at)’

speech: asel ‘say’, patmel ‘tell’, bacʰakančʰel ‘shout’, goral ‘shout’

Valency classes in Eastern Armenian

521

SCREAM

FEEL PAIN

FEEL COLD

FEEL COLD

DIE

DIE

BE SAD

58

59

60

60

61

61

63

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]



















+

+

+

+

+

m

m

+



m



+































































































3 Estimation of frequencies could not be based on corpus data for uncoded alternation. Therefore it is based on the judgment of the authors.

tχuɾ e

mernel

mahanal

sarel

mɾsel

cʰavel

goral

cʰatkel































JUMP

1-nom V.subj[1]

m





52

vazel







RUN

1-nom V.subj[1]







49

hazal





COUGH

1-nom V.subj[1]



Proprie- Reciptive rocal

47

sapʰɾvel (med)

m

Object Omission

SHAVE

V

uncoded 3

Conten- Intertive nal Object

14

anʒɾevel

CausativeMediopassive

RAIN

Causative

69

coded

Mediopassive

Meaning label

#

Coding frame schema

Michael Daniel and Victoria Khurshudian

Appendix 2

>Verb form

522

BE HUNGRY

ROLL

SINK

BURN

BE DRY

BE A HUNTER

BOIL

BE ILL

CRY

SAY

THINK

TALK

EAT

HUG

SEE

SMELL

FRIGHTEN

LIKE

LIKE

LIKE

64

65

66

67

68

70

80

82

83

22

10

18

 1

 2

 4

 5

 7

 8

 8

 8

siɾel

dur gal

havanel

vaχecʰnel (caus)

hototel

tesnel

gɾkel

utel

χosel

mtacel

asel

lacʰel

hivand e

jeral

voɾsoɾd e

čʰoɾ e

ajɾvel (med)

χoɾtakvel (med)

gloɾvel (med)

sovac e

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2+het 3+masin V.subj[1]

1-nom 2+masin V.subj[1]

1-nom UTT2 (3-dat) V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

+



m



m

+

+

+

+

+

+



















m



m









+

+





m



+







































m











































































































m

















































































+























Valency classes in Eastern Armenian

523

Meaning label

KNOW

SEARCH FOR

WASH

WASH

DRESS

KILL

TAKE

BLINK

SING

DIG

HEAR

COOK

WANT

SEND

CARRY

THROW

#

 9

11

12

12

13

26

31

46

53

73

78

79

87

37

38

39

netel

tanel

uʁaɾkel

uzel

jepʰel

lsel

pʰoɾel

jeɾgel

tʰaɾtʰel

arnel

spanel

hagcʰnel (caus)

lval

loʁacʰnel (caus)

pʰntɾel

čanačʰel

>Verb form

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

1-nom 2-nomdat V.subj[1]

Coding frame schema

+



+

m

+

+

+

+

+



+



+



+

+

coded

Mediopassive











m



m















m

Causative

































CausativeMediopassive

































uncoded















m

m















Conten- Intertive nal Object









m





+

















Object Omission

































































Proprie- Reciptive rocal

524 Michael Daniel and Victoria Khurshudian

TIE

PUT

PUT

POUR

FILL

LOAD

PUSH

BRING

NAME

TELL

SHOW

GIVE

TEACH

40

41

41

42

44

45

74

75

23

21

35

36

77

sovoɾecʰnel (caus)

tal

cʰujcʰ tal

patmel

anvanel

beɾel

hɾel

baɾʒel

lcʰnel (caus)

lcʰnel (caus)

daɾsel

dnel

kapel

1-nom 2-nomdat 3-dat V.subj[1]

1-nom 2-nomdat 3-dat V.subj[1]

1-nom 2-nomdat 3-dat V.subj[1]

1-nom 2-nomdat 3-dat V.subj[1]

1-nom 2-nomdat 3-nom V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]

1-nom 2-nomdat LOC3 V.subj[1]







+

+



+

+





+



+



































































+

m

m















































































































Valency classes in Eastern Armenian

525

Meaning label

ASK FOR

BUILD

TAKE

TEAR

TEAR

PEEL

PEEL

PEEL

HIDE

WIPE

STEAL

#

19

24

31

32

32

33

33

33

34

72

76

goʁanal

makʰɾel

tʰakʰcʰnel (caus)

klpel

makʰɾel

kʰeɾel

pokel

patrel

veɾcʰnel (caus?)

karucʰel

χndɾel

>Verb form

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

Coding frame schema



+



+

+

+

+

+



+

+

coded

Mediopassive

m





















Causative























CausativeMediopassive























uncoded























Conten- Intertive nal Object























Object Omission



m



+

m



+

+





























Proprie- Reciptive rocal

526 Michael Daniel and Victoria Khurshudian

MAKE

MAKE

MAKE

GET

BREAK

KILL

KILL

BEAT

HIT

CUT

COVER

GRIND

SHOUT AT

85

85

85

86

25

26

26

27

28

30

43

71

20

gorgoral

aʁal

cackel

ktɾel

haɾvacel

cecel

moɾtʰel

spanel

ǯaɾdel

stanal

šinel

steʁcel

saɾkʰel

+

+

+



+

+

+

+



+

+

+

1-nom 2-gen+vra V.subj[1] –

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat (3-instr) V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

1-nom 2-nomdat 3-abl V.subj[1]

m





m















































































































































































Valency classes in Eastern Armenian

527

Meaning label

SMELL

LOOK AT

HELP

FOLLOW

MEET

SHOUT AT

TOUCH

TOUCH

FEAR

LEAVE

CLIMB

SIT

SIT DOWN

GO

LIVE

#

 5

 3

15

16

17

20

29

29

 6

55

48

50

51

54

56

apɾel

gnal

nstel

nstac e

baɾʒɾanal

meknel

vaχenal

dipčʰel

kpčʰel

bacʰakančʰel

handipel

hetevel

ognel

najel

hot arnel

>Verb form

1-nom (LOC2) V.subj[1]

1-nom (LOC2) V.subj[1]

1-nom (LOC2) V.subj[1]

1-nom (LOC2) V.subj[1]

1-nom (LOC2) V.subj[1]

1-nom 2-abl V.subj[1]

1-nom 2-abl V.subj[1]

1-nom 2-dat 3-instr V.subj[1]

1-nom 2-dat 3-instr V.subj[1]

1-nom 2-dat V.subj[1]

1-nom 2-dat V.subj[1]

1-nom 2-dat V.subj[1]

1-nom 2-dat V.subj[1]

1-nom 2-dat V.subj[1]

1-nom 2-gen V.subj[1]

Coding frame schema

+

m

















m

+



+



coded

Mediopassive

+



+



+



+

+





+









Causative





m

























CausativeMediopassive































uncoded































Conten- Intertive nal Object































Object Omission



















































+





+



Proprie- Reciptive rocal

528 Michael Daniel and Victoria Khurshudian

FALL

LAUGH

PLAY

84

57

62

χaʁal

cicaʁel

ənknel

hajtnvel (med)

1-nom (2-instr) V.subj[1]

1-nom (2+vra) V.subj[1]

1-nom (LOC2) V.subj[1]

1-nom (LOC2) V.subj[1]

+







+

+













Legend: + = alternation occurs regularly; − = alternation never occurs; m = alternation occurs marginally

APPEAR

81









































Valency classes in Eastern Armenian

529

JUMP

SCREAM

FEEL PAIN

FEEL COLD

52

58

59

60

intr

body part as subject, experien- intr cer as its possessor

intr

intr

intr

2916

9126

11058

3172

20799

1661















57

482

2

3

135

5













Caus: ‘cause to feel cold’

4 This table shows the statistics of occurrences of the primary verb and its coded valency alternations in Eastern Armenian National Corpus (www.eanc.net) in columns 6 through 9. Column 6 shows total occurrences of the verb in the corpus, also including occurrences of derived alternations shown in columns 7 through 9; for those forms that significantly overlap with other lexical items, as well as for periphrastic constructions (such as adjective with a copula), only an estimate is provided (marked with ~). Shaded cases indicate that the form is morphologically impossible (valency alternations from a copular construction, double causatives, double mediopassives and medio-causatives where the causative does not exist). The fourth column shows the source of derivation for non-primary verbs and provides comments on unexpected role-to-argument correspondences. Column 5 shows valency class (bracketed for those verbs that are derived alternations themselves; extended valencies such as locations, instruments and sources are not shown). The last column comments on semantics of the valency alternation (when necessary).

mɾsel

cʰavel

goral

cʰatkel

vazel

intr

1173

body part as causee

RUN

49

hazal

[intr]



COUGH

Med from sapʰɾel ‘shave’ (tr)

47

sapʰɾvel



SHAVE

2

24734



14

anʒɾevel

MedioComment causative

RAIN

Causative

69

Mediopassive

Total occurrences

Valency class

Meaning label

#

Comments on derivation and semantics

Michael Daniel and Victoria Khurshudian

Appendix 3

Verb form

530

FEEL COLD

DIE

DIE

BE SAD

BE HUNGRY

ROLL

SINK

BURN

BE DRY

BE A HUNTER

BOIL

BE ILL

CRY

SAY

THINK

60

61

61

63

64

65

66

67

68

70

80

82

83

22

10

mtacel

asel

lacʰel

hivand e

jeral

voɾsoɾd e

čʰoɾ e

ajɾvel

χoɾtakvel

gloɾvel

sovac e

tχuɾ e

mernel

mahanal

sarel

Med from ajɾel ‘burn’ (tr)

Med from χoɾtakel ‘sink’ (tr)

Med from gloɾel ‘roll’ (tr)

Res Ptcp from the literary and rare sovel ‘be hungry’

from mah ‘death’

intr +masin ‘about’

tr +Dat

intr

cop

intr

cop

cop

[intr]

[intr]

[intr]

cop

cop

intr

intr

intr

74696

592935

2698

~7000

3493

1

~3000

6529

1422

3424

~2000

~9000

32534

9078

5575

1044

13053















2

1236

569

79

1595



12

21



641

Med: most frequently as a Res Ptcp ‘the planned’

Med: demoted speaker (reportative contexts)

Caus: indirect causation

Caus: indirect causation

Caus: ‘freeze’

Valency classes in Eastern Armenian

531

Meaning label

TALK

EAT

HUG

SEE

SMELL

FRIGHTEN

LIKE

LIKE

LIKE

KNOW

SEARCH FOR

WASH

WASH

#

18

1

2

4

5

7

8

8

8

9

11

12

12

lval

loʁacʰnel

pʰntɾel

čanačʰel

siɾel

dur gal

havanel

vaχecʰnel

hototel

tesnel

gɾkel

utel

χosel

Verb form

Caus from loʁanal ‘wash’ (intr)

‘love’

compl. verb based on ‘come’

Caus from vaχenal ‘fear’ (V)

Comments on derivation and semantics

tr

[tr]

tr

tr

tr

tr

tr

[tr]

tr

tr

tr

tr

intr +het ‘with’ +masin ‘about’

Valency class

4134

648

13874

46262

64136

4309

3899

3005

895

217783

11845

35531

108095

Total occurrences

2189



322

10215

2240



6



1

2904

286

216

3377

Mediopassive





20

6



6







195

568

Causative











Med: reflexive

Med: ‘be wanted’, as of a criminal or worker

Med: ‘become known, famous’

Med: passive

Med: ‘meet each other’

Med: reciprocal

Med: demoted speaker (reportative contexts)

MedioComment causative

532 Michael Daniel and Victoria Khurshudian

DRESS

KILL

TAKE

BLINK

SING

DIG

HEAR

COOK

WANT

SEND

CARRY

THROW

TIE

PUT

PUT

POUR

FILL

LOAD

13

26

31

46

53

73

78

79

87

37

38

39

40

41

41

42

44

45

baɾʒel

lcʰnel

lcʰnel

daɾsel

dnel

kapel

netel

tanel

uʁaɾkel

uzel

jepʰel

lsel

pʰoɾel

jeɾgel

tʰaɾtʰel

arnel

spanel

vhagcʰnel

cf. 42

Caus from li ‘full’ (Adj)?

‘pile up’

Caus from hagnel ‘put on (clothes)’

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

[tr]

2976

16314

16314

1420

88165

62864

20538

112285

39108

132482

3787

101678

5367

32959

1118

~100000

31186

2381

202





473



34809

7505



6192

51

551

17157

815

502

25



8828

























39



10











Med: passive

Med: anticausative

Med: ‘can be heard’ of a sound

passive

Med: demoted speaker (reportative contexts)

Valency classes in Eastern Armenian

533

Meaning label

PUSH

BRING

NAME

TELL

SHOW

GIVE

TEACH

ASK FOR

BUILD

TAKE

TEAR

TEAR

#

74

75

23

21

35

36

77

19

24

31

32

32

pokel

patrel

veɾcʰnel

karucʰel

χndɾel

sovoɾecʰnel

tal

cʰujcʰ tal

patmel

anvanel

beɾel

hɾel

Verb form

‘pluck’

Caus from veɾcʰ ‘up’ (Adv)?

Caus from sovoɾel ‘learn’

from anun ‘name’ (N)

Comments on derivation and semantics

tr

tr

tr

tr

tr

[tr] +dat

tr +dat

tr +dat

tr +dat

tr +nom

tr

tr

Valency class

8993

3471

50688

32245

53910

5984

413908

~30000

43880

17573

113991

7399

Total occurrences

3769

1170



20017

586







548

1607



98

Mediopassive





















Causative

Med: anticausative or passive

Med: anticausative or passive

Med: passive

Med: demoted speaker (‘(we) ask that’)

Med: demoted speaker (reportative contexts)

Med: ‘be called’

Med: ‘push oneself away from something’

MedioComment causative

534 Michael Daniel and Victoria Khurshudian

PEEL

PEEL

PEEL

HIDE

WIPE

STEAL

MAKE

MAKE

MAKE

GET

BREAK

KILL

KILL

BEAT

HIT

33

33

33

34

72

76

85

85

85

86

25

26

26

27

28

haɾvacel

cecel

moɾtʰel

spanel

ǯaɾdel

stanal

šinel

steʁcel

saɾkʰel

goʁanal

makʰɾel

tʰakʰcʰnel

klpel

makʰɾel

kʰeɾel

also ‘repair’

from goʁ ‘thief’

primary meaning ‘clean’ (cf. 33)

Caus from tʰakʰ(un) ‘secret’ (Adj)

primary meaning ‘clean’ (cf. 33)

(as by rubbing)

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

tr

6682

12560

3801

31186

8767

105726

11503

78683

10810

5107

11022

10001

125

11022

984



1224

590

8828

2678



1817

38916

659



2203



4

2203

88



















9











Med: passive

Med: passive

Med: passive

Med: anticausative or passive

Med: passive

Med: passive

Med: passive

Med: potential passive

Med: potential passive

Med: potential passive

Med: potential passive

Valency classes in Eastern Armenian

535

Meaning label

CUT

COVER

GRIND

SHOUT AT

SMELL

LOOK AT

HELP

FOLLOW

MEET

SHOUT AT

TOUCH

TOUCH

FEAR

LEAVE

#

30

43

71

20

 5

 3

15

16

17

20

29

29

 6

55

meknel

vaχenal

dipčʰel

kpčʰel

bacʰakančʰel

handipel

hetevel

ognel

najel

hot arnel

gorgoral

aʁal

cackel

ktɾel

Verb form

from vaχ ‘fear’ (N)

primary meaning ‘become glued (intr) to smth’, coll. also ‘touch smth’

lit. ‘take smell (of smth.)’

primary meaning ‘shout’

Comments on derivation and semantics

abl

abl

dat

dat

dat

dat

dat

dat

dat

[intr] +Gen

intr +vɾa ‘on’

tr

tr

tr

Valency class

30068

30420

5620

~7000

6483

41738

28738

27174

146331

~400

992

15034

18903

30189

Total occurrences











3

+



166





89

7696

8160

Mediopassive



2953

399





110









1





5

Causative











‘touch with smth’

causative only from the primary meaning

Med: ‘can be found’, ‘occurs’

Med: irreg. hetevecʰvel or hetevvel

Med: ‘look good’ etc.

Med: passive

Med: anticausative

Med: passive or potential passive

MedioComment causative

536 Michael Daniel and Victoria Khurshudian

CLIMB

SIT

SIT DOWN

GO

LIVE

APPEAR

FALL

LAUGH

PLAY

48

50

51

54

56

81

84

57

62

χaʁal

cicaʁel

ənknel

hajtnvel

apɾel

gnal

nstel

nstac e

baɾʒɾanal

Med from hajtnel ‘announce’

Res Ptcp from ‘sit down’ (see 51)

intr

intr +vɾa ‘on’

intr

[intr]

intr

intr

intr

cop

primary meaning ‘go/move up’ intr

34647

23495

89371

25718

82418

219872

80317

~15000

65870

320





207

11





1432

250



400



3300

31718







72



Caus: ‘involve in a game (e.g., children)’, ‘feign (a sentiment)’

Caus: ‘make sit’, ‘send to jail’ Medio-causative (marginal): ‘force smth into place’

‘raise’

Valency classes in Eastern Armenian

537

538

Michael Daniel and Victoria Khurshudian

Abbreviations ADJVZ AOR COND CVB.CONNEG CVB.IPFV CVB.PFV MED PROH PTCP.RES PTCP.SBJ SBJV

adjectivizer aorist conditional (widely used as future) converb used with the negative auxiliary imperfective converb perfective converb (mostly perfect) mediopassive prohibitive (preverbal particle) resultative participle subject participle subjunctive (subordinate mood)

References Arkhangelskiy, Timofey & Oleg Belyaev. 2011. A Comparison of Eastern Armenian and Iron Ossetic spatial systems. In Vittorio S. Tomelleri, Manana Topadze & Anna Lukianowicz (eds.), Languages and Cultures in the Caucasus, 285–299. München–Berlin: Verlag Otto Sagner. Arkhangelskiy, Timofey & Xenia Semionova (in preparation). Locative constructions in Modern Eastern Armenian: Structuring spatial and temporal domain. Arkhipov, Alexandre. 2009. Comitative as a cross-linguistically valid category. In Alexandre Arkhipov & Patience Epps (eds.), New Challenges in Typology: Transcending the Borders and Refining the Distinctions, 223–246. Berlin–New York: Mouton de Gruyter. Blume, Kerstin. 1998. A Contrastive analysis of interaction verbs with dative complements. Linguistics 36(2). 253–280. Comrie, Bernard. 1984. Some formal properties of focus in Modern Eastern Armenian. Annual of Armenian Linguistics 5. 1–21. Dixon, R. M. W. & Alexandra Yu. Aikhenvald. 2000. Introduction. In R. M. W. Dixon & Alexandra Yu. Aikhenvald (eds.), Changing Valency. Case Studies in Transitivity, 1–29. Cambridge: CUP. Donabédian, Anaïd. 1997. Neutralisation de la diathèse des participes en -ac de l’arménien moderne occidental. Studi italiani di linguistica teorica ed applicata (Nuova serie (Padova)) 26(2). 327–339. Dum-Tragut, Jasmine. 2010. A Grammar of Modern Eastern Armenian. Amsterdam: Benjamins. EANC. Eastern Armenian National Corpus (www.eanc.net) Haspelmath, Martin. 1993. More on the typology of inchoative/causative verb alternations. In Bernard Comrie & Maria Polinsky (eds.), Causatives and Transitivity (Studies in Language Companion Series 23), 87–120. Amsterdam: Benjamins. Haspelmath, Martin. 2001. Non-canonical marking of core arguments in European languages. In Alexandra Y. Aikhenvald, R. M. W. Dixon & Masayuki Onishi (eds.), Non-Canonical Marking of Subjects and Objects (Typological Studies in Language 46), 53–83. Amsterdam: Benjamins. Khanina, Alesya & Andrej Shluinsky (submitted to Linguistics). A Rare type of benefactive construction: Evidence from Enets. Kortlandt, Frederik. 1999. The Armenian causative. Annual of Armenian Linguistics 20(1999– 2000). 47–49. Kozinceva, Natalia. 1974. Zalogi v armjanskom jazyke. In Aleksandr Xolodovič (ed.), Tipologija passivnyx konstrukcij. Diatezy i zalogi, 73–90. Leningrad: Nauka.

Valency classes in Eastern Armenian

539

Kozinceva, Natalia. 1981. Refleksivnye glagoly v armjanskom jazyke. In Victor Xrakovskij (ed.), Zalogovye konstrukcii v raznostrukturnyx jazykax, 81–98. Leningrad: Nauka. Kulikov, Leonid. 2001. Causatives. In Armin Burkhardt, Hugo Steger & Herbert Ernst Wiegand (eds.), Language Typology and Language Universals, 886–899. HSK 20.2. Berlin–New York: Walter de Gruyter. Maslova, Elena & Giuliano Bernini. 2006. Sentence topics in the languages of Europe and beyond. In Giuliano Bernini & Marcia L. Schwarz (eds.), Pragmatic Organization of Discourse in the Languages of Europe 20(8), 67–120. Berlin–New York: Mouton de Gruyter. Nichols, Johanna, David A. Peterson & Jonathan Barnes. 2004. Transitivizing and detransitivizing languages. Linguistic Typology 8. 149–211. Payne, Doris L. & Immanuel Barshi. 1999. External Possession. Amsterdam: Benjamins. Shibatani, Masayoshi. 1996. Applicatives and benefactives. A Cognitive account. In Masayoshi Shibatani & Sandra Thompson (eds.), Grammatical Constructions. Their Form and Meaning, 157–195. Oxford: Oxford University Press.

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

15 Valency and valency classes in Bezhta 1 Introduction Bezhta belongs to the Tsezic subgroup of the Nakh-Daghestanian (Northeast Caucasian, East Caucasian) language family and is spoken in the Republic of Daghestan in the Russian Federation. The most common morpho-syntactic features of Bezhta are as follows. Bezhta is a verb-final language, though with no rigid order of the major clause constituents. It is a dependent-marking ergative language. Bezhta has an elaborate spatial case system, with the possibility of combining orientation and direction suffixes, e.g., the sequence super-abl means ‘from on (top of)’. It also has gender agreement (4 genders in the Bezhta Proper dialect, to which we restrict ourselves here), gender being employed to indicate cross-referencing of arguments on the verb. The agreement is always with the Absolutive argument. In general only – though not all – vowel-initial verbs have a prefixal slot for agreement, plus a small number of verbs with internal vowel change. Throughout, it should be borne in mind that agreement will only be manifested if the verb in question has an agreement slot. The article is structured as follows: Bezhta basic argument structure types are discussed in Section 2. Then valency classes are treated in Section 3. In Sections 4 and 5 we discuss valency alternations, coded and uncoded respectively.

2 Basic argument structure types in Bezhta Bezhta has intransitive, unergative, transitive, and affective verbs. Intransitive verbs have the single S argument in the Absolutive case. The verb agrees with S. Examples of intransitive verbs are y-iyāl ‘to cry’, y-ogic’al ‘to jump’, y-e nƛ’al ‘to go’, kok’al ‘to ache’, qoqal ‘to be dry’, y-aqal ‘to happen’, y-uɣal ‘to die’, ɣayƛal ‘to boil’. (Here and below, verbs are cited in the Infinitive in -al, before which a short stem-final vowel drops. Verbs with a prefixal agreement slot are marked with initial y-.) (1) biƛo y‑ek’e‑š house(iv) iv‑burn‑prs ‘The house burns.’ In Bezhta unergative verbs constitute a small group of onomatopoetic verbs, e.g., ɣäʔƛäl ‘to caw’, hicƛal ‘to sneeze’, öhƛäl ‘to cough’, hik’ƛal ‘to hiccup’, hahƛal

542

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

‘to yawn’. The single SA argument is in the Ergative.1 Another unergative verb not belonging to this semantic type is č’än-äɣäl ‘to run’, for which see further § 6.1. The verb shows no agreement. Note that we restrict the term “unergative” to this formally definable class of verbs; verbs that we classify as “intransitive” may have more agent-like or more patient-like single arguments (and this further distinction will be relevant in the discussion of the Potential Alternation in § 5.3). (2) öždi lalaƛo‑yo boy.obl(erg) shout‑prs ‘The boy shouts.’ Canonical transitive verbs have the A argument in the Ergative case and the P argument in the Absolutive case. Verb agreement is with the P argument. Examples of transitive verbs are y-owal ‘to make, build’, y-ü nqal ‘to eat’, nizāl ‘to wash’. (3) öždi bäbä m‑üq‑iyo boy.obl(erg) bread(iii) iii‑eat‑pst ‘The boy ate bread.’ Affective verb are verbs of perception, emotion and cognition. The experiencer is in the Lative, and the stimulus in the Absolutive. Verb agreement is with the stimulus. Bezhta has 11 affective verbs, namely y-egāl ‘to see’, tuqal ‘to hear’, č’aɬal ‘to feel’, y-at’al ‘to like, love, want’, y-iqal ‘to get’, y-iq’al ‘to know’, zoɣal ‘to find’, čoqal ‘to hear’, bidži y-aqal ‘to understand’, kezi y-aqal ‘to meet’ – these two are light verb constructions – and yak’ƛ’a y-ec-al ‘to remember’ (lit. ‘to be on heart’). (4) kibba‑l quy tuq‑iyo girl.obl‑lat noise hear‑pst ‘The girl heard the noise.’ At least intransitive, unergative, and transitive clauses can be extended by further oblique NPs/PPs. In some cases these are arguments (e.g., with case forms governed by the predicate, ex. 5), in others adjuncts (e.g., with case forms determined purely semantically, ex. 6). Extended intransitive verbs have an agent-like argument in the Absolutive case and another argument in a spatial case (ex. 5, 6). A number of extended unergative verbs have an agent-like argument in the Ergative and another argument in a spatial case (ex. 7). Extended transitive verbs have an agent argument in the Ergative case, a patient in the Absolutive, and another

1 Diachronically these onomatopoetic verbs seem to have arisen through the incorporation of an onomatopoetic element like hah ‘yawning’ into the verb iƛal, which in Bezhta means ‘to call’ but has the more general meaning ‘to say’ in some other Tsezic languages.

Valency and valency classes in Bezhta

543

argument (recipient, goal, or other) in some other grammatical or spatial case (ex. 8, 9). (5)

öžö c’oy‑qa hič’e‑š boy fire.obl‑poss fear‑prs ‘The boy fears the fire.’

(6)

öžö xöx‑ƛ’ä ƛ’odo 0‑̸ e nƛ’e‑yo boy(i) tree.obl‑super up i‑go‑pst ‘The boy climbed up the tree.’

(7)

öždi dī‑ƛ’a lalaƛo‑yo boy.obl(erg) me.obl‑super shout‑pst ‘The boy shouted at me.’

(8)

öždi kibba‑l xabar m‑ē‑yo boy.obl(erg) girl.obl‑lat story(iii) iii‑let_out‑pst ‘The boy told a story to the girl.’

(9)

öždi c’it’a‑d bäbä b‑üč’‑iyo boy.obl(erg) knife.obl‑ins bread(iii) iii‑cut‑pst ‘The boy cut the bread with the knife.’

3 Valency classes (main verb types found in the database) Bezhta seems to lack zero‑valency predicates, i.e. all verbs take at least one argument in the Absolutive case or in the Ergative case (unergative verbs). (10) wodo g‑s rain(iv) come‑prs ‘It is raining.’ (lit. ‘Rain comes.’) Bezhta is a pro-drop language, where retrievable arguments can be easily left out. Weather expressions are usually introduced with no overt arguments, as in (11), where the verb might seem to lack an Absolutive argument. Though the Absolutive argument is not overtly expressed the predicate (here adjectival) still shows agreement in gender IV (ex. 11a). The noun mex ‘time, day’ or the noun wodo ‘day, rain’, both of gender IV, can be understood (cf. ex. 11b). (11) a. že hoƛoʔ y-äč’č’ö gey now here iv-cold be.prs ‘Now it is cold here.’

544

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

b. že hoƛoʔ y-äč’č’ö mex/wodo gey now here iv-cold day(iv)/day(iv) be.prs ‘Now the day is cold here.’ Bezhta makes extensive use of light verb constructions. The element accompanying the light verb is often (though not always) an argument, giving the Bezhta verb one argument more than its English equivalent. The light verb always agrees with a P argument of this type. (13) öždi aƛa‑ʔ ömrö b‑ō‑s boy.obl(erg) village.obl‑in life(iii) iii‑do‑prs ‘The boy lives in the village.’ (lit. ‘does life’) (14) öždi kibba‑l komak b‑ō‑yo boy.obl(erg) girl.obl‑lat help(iii) iii‑do‑pst ‘The boy helped the girl.’ (lit. ‘does help’) Where possible, Bezhta seems to prefer a possessor within an NP rather than an extra NP, giving rise to examples where the Bezhta valency is one less than that of English. However, the latter construction is available, for instance if it is necessary to specify that the possessor is temporary, as in (15b). (The difference between permanent and temporary possession is also found with transfer verbs, using the difference between Lative and Poss-essive; see § 4.1.) (15) a. öždi kibba‑s t’ek y‑u nco‑yo boy.obl(erg) girl.obl‑gen1 book(iv) iv‑steal‑pst ‘The boy stole the book from the girl.’ [i.e. the girl’s book] b. öždi kibba‑qa‑s t’ek y‑u nco‑yo boy.obl(erg) girl.obl‑poss‑abl book(iv) iv‑steal‑pst ‘The boy stole the book from the girl.’ [i.e. a book that was with the girl, but not belonging to her] For verbs of covering, filling, loading, tying, etc. examined from the database, Bezhta treats the moving object as P and expresses the goal of the movement as a location (NB: using Essive, not Lative or Directional). (16) öždi hudo m‑oso‑yo mašinal‑ƛ’a boy.obl(erg) firewood(iii) iii‑collapse‑pst car.obl‑super ‘The boy loaded the car with the firewood (lit. loaded the firewood on the car).’

Valency and valency classes in Bezhta

545

The use of the genitive to encode verbal arguments is not very common in Bezhta, but is found with the two verbs: y-oc’lal ‘to fill’,2 and y-iq’lašal ‘to talk’. (17) kibba ɬī‑s wedra b‑oc’‑il‑lo girl.obl(erg) water.obl‑gen1 bucket(iii) iii‑fill‑caus‑pst ‘The girl filled the bucket with the water.’ Verbs of contact typically take the moving object as P and the goal of the movement in the bare Lative, though an alternative with the goal as P and the instrument in the Instrumental is sometimes also possible, but only with the verb y-äƛ’elal ‘to hit’. (18) öždi kibba‑l kō y‑e nxe‑l‑lo boy.obl(erg) girl.obl‑lat hand(iv) iv‑touch‑caus‑pst ‘The boy touched the girl with his hand (lit. touched his hand to the girl).’ (19) a. öždi kibba‑l k’obala b‑äƛ’el‑lo boy.obl(erg) girl.obl‑lat stick(iii) iii‑hit‑pst ‘The boy hit the girl with the stick (lit. hit the stick to the girl).’ b. öždi kid k’obalali‑d y‑äƛ’el‑lo boy.obl(erg) girl(ii) stick.obl‑ins ii‑hit‑pst ‘The boy hit the girl with the stick.’ All the verbs of transfer take an indirective construction (theme (T) encoded like the P of a transitive construction, recipient (R) in an oblique case), but with variation in the oblique case used for R depending on such semantic factors as permanent versus temporary transfer; for permanent transfer, the bare Lative is usual (ex. 20a); for temporary transfer, there is some lexical variation among spatial cases, both Essive and Lative (ex. 20c, 20b). (20) a. öždi kibba‑l diʔi niƛ‑iyo boy.obl(erg) girl.obl‑lat flower give‑pst ‘The boy gave flowers to the girl.’ [permanently] b. öždi kibba‑qa t’ek niƛ‑iyo boy.obl(erg) girl.obl‑poss book give‑pst ‘The boy gave the book to the girl.’ [temporarily]

2 The same is also found in Akhvakh (Creissels, ms.).

546

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

c. öždi kibba‑qa‑l t’ek niƛ‑iyo boy.obl(erg) girl.obl‑poss‑lat book give‑pst ‘The boy gave the book to the girl.’ [so that she would read it, i.e. with some purpose, but not for permanent possession] (21) a. öždi kibba‑l diʔi y‑e ne‑yo boy.obl(erg) girl.obl‑lat flower(iv) iv‑send‑pst ‘The boy sent flowers to the girl.’ [permanently] b. öždi kibba‑doy diʔi y‑e ne‑yo boy.obl(erg) girl.obl‑apud flower(iv) iv‑send‑pst ‘The boy sent flowers to the girl.’ [temporarily] Verbs of speech take the P argument in the Absolutive and the addressee can be marked with the Poss-essive, bare Lative, or Super-essive. But very often verbs of speech occur in sentential complement constructions where the complement clause encoding the reported utterance is treated as an Absolutive argument (ex. 22). The most common speech verbs are nisal ‘to say’, nisolal ‘to ask’ (which is formally the causative of the verb nisal ‘say’) and harzi y-owal ‘ask, beg’, which take the addressee in the Poss-essive. The addressee is in the bare Lative with the verb maɬci y-owal ‘to teach’ and the expression for ‘to tell a story’, which is not an actual speech verb but a combination of verb y-e nyal ‘to send’ and the noun xabar ‘story’, i.e. xabar m-eyal. Occasionally the addressee can be marked with the Superessive as with the verb lalaƛal ‘to shout’. (22) iyo-qa nisol-ca holco, nā-d gey diye abo-ƛo mother-poss ask-prs he.obl(erg) where-q be.prs 1sg.gen1 father-quot ‘He asks his mother: “Where is my dad?”’ (23) öždi kibba‑qa bit’arab žo niso‑yo boy.obl(erg) girl.obl‑poss right thing say‑pst ‘The boy said the right thing to the girl.’ (24) öždi kibba‑l xabar m‑ē‑yo boy.obl(erg) girl.obl‑lat story(iii) iii‑let.out‑pst ‘The boy told a story to the girl.’ (25) öždi ʔäräb mic kibba‑l maɬciō‑s boy.obl(erg) Arabic language(iii) girl.obl‑lat teach‑prs ‘The boy teaches the girl Arabic.’

Valency and valency classes in Bezhta

547

(26) iyo kibba‑ƛ’a lalaƛo‑yo mother.erg girl.obl‑super shout‑pst ‘The mother shouted at the girl.’ Bezhta has an accidental construction which expresses a potential or accidental situation. The most agent‑like argument appears in the Poss-essive case, the P in the Absolutive; agreement is with the P. The basic verb can be intransitive or transitive. See the detailed discussion in sections 4.4, 5.2, and 5.3. (27) a. ɣarapink’a y‑iše‑yo vase(iv) iv‑break‑pst ‘The vase broke.’ b. dī‑qa ɣarapink’a y‑iše‑yo me.obl‑poss vase(iv) iv‑break‑pst ‘I broke the vase accidentally.’ / ‘I could break the vase.’ (28) a. öždi hini‑s zoƛ’o b‑üč’‑iyo boy.obl(erg) self.obl‑gen1 finger(iii) iii‑cut‑pst ‘The boy cut his finger.’ b. öždi‑qa hini‑s zoƛ’o b‑üč’‑iyo boy.obl‑poss self.obl‑gen1 finger(iii) iii‑cut‑pst ‘The boy cut his finger accidentally.’ / ‘The boy could cut his finger.’

4 Case alternations (uncoded alternations) 4.1 The Recipient/Goal/Location Alternation The following alternations concern the encoding of the recipient/goal/location argument. These alternations are limited to a small group of verbs: ‘to bring’, ‘to carry’, ‘to send’, ‘to throw’, and ‘to give’. This alternation marks the differentiation between permanent and temporary transfer of possession, which is widespread in Nakh-Daghestanian languages (see Daniel at el. 2010). Depending on the verb the alternation is always between bare Lative for permanent transfer of possession and some other spatial case, either Poss-essive, Apud-essive, or Poss-lative for temporary transfer.

4.1.1 The Recipient/Goal/Location Alternation 1 with POSS The Lative-Poss-essive Alternation is found regularly with the verb niƛal ‘to give’. The recipient is marked either with the Lative (for permanent transfer of posses-

548

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

sion) or with the Poss-essive (for temporary transfer of possession). Additionally, the recipient can be marked with the Poss-lative for non-permanent transfer, but this version occurs only very occasionally. (29) a. öždi kibba‑l diʔi niƛ‑iyo boy.obl(erg) girl.obl‑lat flower give‑pst ‘The boy gave flowers to the girl.’ (permanent) b. öždi kibba‑qa t’ek niƛ‑iyo boy.obl(erg) girl.obl‑poss book give‑pst ‘The boy gave the book to the girl.’ (temporary) c. öždi kibba‑qa‑l t’ek niƛ‑iyo boy.obl(erg) girl.obl‑poss‑lat book give‑pst ‘The boy gave the book to the girl.’ (so that she would read it, i.e. with some purpose, but not for permanent possession) Another verb that undergoes this alternation is verb y-iƛ’al ‘to throw’, which marks its goal argument with the Poss-essive. (30) ‘throw to someone (so that they would catch)’ öždi kibba‑qa burti b‑iƛ’e‑yo boy.obl(erg) girl.obl‑poss ball(iii) iii‑throw‑pst ‘The boy threw a ball to the girl.’

4.1.2 The Recipient/Goal/Location Alternation 2 with APUD The Apud-essive-Lative Alternation is found with the two verbs, y-aq’al ‘to bring’ and y-e nyal ‘to send’. The recipient is either marked with the Lative (for permanent transfer) or with the Apud-essive (for temporary transfer). (31) a. öždi kibba‑l žamɣo b‑aq’o‑yo boy.obl(erg) girl.obl‑lat dock(iii) iii‑bring‑pst ‘The boy brought dock leaves to the girl.’ b. öždi kibba‑doy žamɣo b‑aq’o‑yo boy.obl(erg) girl.obl‑apud dock(iii) iii‑bring‑pst ‘The boy brought dock leaves to the girl.’ (32) a. öždi kibba‑l diʔi y‑e ne‑yo boy.obl(erg) girl.obl‑lat flower(iv) iv‑send‑pst ‘The boy sent flowers to the girl.’

Valency and valency classes in Bezhta

549

b. öždi kibba‑doy diʔi y‑e ne‑yo boy.obl(erg) girl.obl‑apud flower(iv) iv‑send‑pst ‘The boy sent flowers to the girl.’ The Apud-essive-Lative Alternation is found with one verb y-iƛ’al ‘to throw’. The goal argument can be marked either with the Apud-essive or the Lative or the Poss-essive. Each encoding has a slightly different meaning. (33) a. ‘throw to someone (in their direction)’ öždi kibba‑doy burti b‑iƛ’e‑yo boy.obl(erg) girl.obl‑apud ball(iii) iii‑throw‑pst ‘The boy threw a ball to the girl.’ b. ‘throw to someone (in order to hit them)’ öždi kibba‑l burti b‑iƛ’e‑yo boy.obl(erg) girl.obl‑lat ball(iii) iii‑throw‑pst ‘The boy threw a ball to the girl.’ An inanimate location argument is marked with the appropriate Essive. (34) a. ‘throw onto the table’ öždi istoliya‑ƛ’a q’alam b‑iƛ’e‑yo boy.obl(erg) table.obl‑super pencil(iii) iii‑throw‑pst ‘The boy threw the pencil onto the table.’ b. ‘to throw into the bucket’ öždi wedrala-ʔ q’alam b‑iƛ’e‑yo boy.obl(erg) bucket.obl-in pencil(iii) iii‑throw‑pst ‘The boy threw the pencil into the bucket.’

4.2 The Ambitransitive Alternation There is only one verb that has the Ambitransitive Alternation. This is the S=P labile verb helal ‘to cook’. (35) a. k’atu hele-yo potato cook-pst ‘The potato has cooked.’ b. kibba k’atu hele-yo girl.obl(erg) potato cook-pst ‘The girl cooked the potato.’

550

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

4.3 The Instrument Alternation This alternation is present in one verb y-äƛ’elal ‘to hit’. This alternation concerns the encoding of the instrument argument, which can be marked either with the Absolutive or with the Instrumental case. (36) a. öždi kibba-l k’obala b-äƛ’el-ca boy.obl(erg) girl.obl-lat stick(iii) iii-hit-prs ‘The boy hits the girl with the stick.’ b. öždi kid k’obala-li-d y-äƛ’el-ca boy.obl(erg) girl(ii) stick-obl-ins ii-hit-prs ‘The boy hits the girl with the stick. ’

4.4 Accidental/Potential Alternation (unmarked) This alternation has two subtypes: Accidental/Potential Intransitive and Accidental/Potential Transitive Alternations.

4.4.1 Accidental/Potential Intransitive Alternation This alternation is found with 10 verbs in the database. The Accidental/Potential Intransitive Alternation is an unmarked alternation. The Accidental/Potential expresses an accidental or potential situation. The most agent‑like argument appears in the Poss‑essive case, the P in the Absolutive; agreement is with the P. The basic verbs are mostly intransitives, in fact semantically patientive intransitives. The Accidental/Potential Intransitive Alternation constructions with patientive intransitives require a new argument in the Poss-essive, while the Absolutive S argument remains unchanged. This alternation increases the valency of the intransitive verb by one. The 10 patientive intransitive verbs in the database undergoing this alternation are: ɣayƛal ‘to boil’, helal ‘to cook’ – this one actually ambitransitive – qoqal ‘to dry’, y-ä č’ɬä l ‘to be cold’, y-ek’al ‘to burn’, y-ekal ‘to fall’, y-uɣal ‘to die’, y-o nq’al ‘to come, to appear’, y-isal ‘break’, pacpaƛal ‘to blink’. (37) Intransitive verb a. kid y-uɣo-yo girl(ii) ii-die-pst ‘The girl died.’ b. öždi-qa kid y-uɣo-yo boy.obl-poss girl(ii) ii-die-pst ‘The boy could kill the girl. / The boy accidentally killed the girl.’

Valency and valency classes in Bezhta

551

(In (37b), the corresponding derived transitive y-uɣo-l-lo, with the causative suffix -l, is impossible.) (38) Intransitive verb a. ɬi ɣayƛo-yo water boil-pst ‘The water boils.’ b. kibba-qa ɬi ɣayƛo-yo girl.obl-poss water boil-pst ‘The girl could boil the water. / The girl accidentally boiled the water.’

4.4.2 Accidental/Potential Transitive Alternation This alternation is found with 3 verbs in the database, namely y-ü č’äl ‘to cut’, haʔal ‘to grind’, y-išelal ‘to break’. The Accidental/Potential Transitive Alternation is an unmarked alternation. The Accidental/Potential expresses an accidental or potential situation. The most agent‑like argument is marked with the Poss‑essive case, the P is in the Absolutive; agreement is with the P. The number of arguments of a transitive verb does not change in the Accidental/Potential Transitive Alternation, the accidental/potential agent stands in the Poss-essive case and the P remains in the Absolutive case. (39) Transitive verb a. öždi hini-s zoƛ’o b-üč’-iyo boy.obl(erg) self.obl-gen1 finger(iii) iii-cut-pst ‘The boy cut his finger.’ b. öždi-qa hini-s zoƛ’o b-üč’-iyo boy.obl-poss self.obl-gen1 finger(iii) iii-cut-pst ‘The boy could cut his finger. / The boy cut his finger accidentally.’

5 Verb coded alternations 5.1 The Antipassive Alternation 5.1.1 The Antipassive 1 Alternation The antipassive is a marked (coded) alternation: it is formed with the antipassive suffixes/infixes -la/ā, -da/ā, -ya. The antipassive can change the number of arguments.

552

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

An antipassive can be formed from intransitive, unergative, transitive, and affective verbs. The general meaning of the antipassive is durative (which with the semelfactive verbs like ‘jump’, ‘cough’ implies iterativity). As will be seen in the treatment of individual argument structure types below, the Bezhta antipassive sometimes changes the valency of the verb (with unergatives and transitives) and sometimes does not (with intransitives). One might therefore argue that the construction overall should be called “durative” rather than “antipassive”. For historical-comparative purposes, we prefer to retain the term “antipassive”, but note that with intransitive verbs this does not involve a valency alternation. The 44 verbs that undergo this alternation include 32 transitives: y-oq’olal ‘to cover’, y-o nsal ‘to load’, y-i ncal ‘to tie’, git’al ‘to pour’, glal ‘to put’, y-iƛ’al ‘to throw’, y-iƛ’al ‘to kill’, y-aq’al ‘to bring’, y-ežal ‘to carry’, y-e nyal ‘to send’, niƛal ‘to give’, y-u ncal ‘to steal’, y-u ncal ‘to hide’, y-äxelal ‘to tear’, y-a ncolal ‘to wipe’, yüč’äl ‘to cut’, y-e nxelal ‘to touch’, y-äƛ’elal ‘to hit/beat’, xabar m-eyal ‘to tell’, yü nqäl ‘to eat’, keč’ iƛal ‘to sing’, y-okčal ‘to take’, kusuʔal ‘to shave’, cā n glal ‘to name’, haʔal ‘to grind’, ɬič elal ‘to dress’, y-äxä l ‘to dig’, a nɬo y-oc’al ‘to hug’, zistqlal ‘to push’, č’an y-aɣal ‘to run’, nizāl ‘to wash’, y-owal ‘to make’; 8 intransitives: y-ogic’al ‘to jump’, ɣayƛal ‘to boil’, y-ek’al ‘to burn’, k’äyƛäl ‘to roll’, y-e nƛ’äl ‘to go, to leave’, k’ok’al ‘feel pain’, ɬowal ‘to laugh’, pacpaƛal ‘to blink’; 2 unergatives: lalaƛal ‘to shout/scream’, öhƛäl ‘to cough’; and 2 affective verbs: y-egāl ‘to see’, y-iq’al ‘to know’ . The antipassive from intransitive verbs does not change the general case frame; there is still a single argument in the Absolutive case. (40) ‘to jump’ a. öždä ä b‑ogic’‑iyo boy.pl hpl‑jump‑pst ‘The boys jumped once.’ b. öždä ä b‑ogic‑ca boy.pl hpl‑jump‑prs ‘The boys jump many times.’ When the antipassive is formed from unergative verbs, the Ergative argument shows up in the Absolutive. (41) ‘to cough’ a. öždi öhƛö‑yö boy.obl(erg) cough‑pst ‘The boy coughed (once).’ b. öžö öhdä ‑yö boy cough.antip‑pst ‘The boy was coughing.’

Valency and valency classes in Bezhta

553

(42) ‘to shout’ a. öždi lalaƛo‑s boy.obl(erg) shout‑prs ‘The boy shouts.’ b. öžö lala-da‑s boy shout-antip‑prs ‘The boy is shouting.’ When the antipassive is formed from transitive verbs the agentive Ergative argument of the source verb is marked with the Absolutive case and its Absolutive patient appears in the Instrumental. (43) ‘to eat’ a. öždi bäbä m-üq-čä boy.obl(erg) bread(iii) iii-eat-prs ‘The boy eats the bread.’ b. öžö bäbälä-d 0-̸ ü nq-dä -š boy(i) bread.obl-ins i-eat-antip-prs ‘The boy is busy eating the bread.’ Just as with the P argument of the transitive construction, the T argument in the antipassive construction from a ditransitive verb shifts to the Instrumental. (44) a. öždi t’ek kibba-l niƛ-iyo boy.obl(erg) book girl.obl-lat give-pst ‘The boy gave the book to the girl.’ b. öžö kibba-l t’ek-la-d niƛ-da-s boy girl.obl-lat book-pl.obl-ins give-antip-prs ‘The boy is giving books to the girl.’

5.1.2 The Antipassive 2 Alternation The antipassive 2 has reflexive meaning. This is only found in the one verb nizāl ‘to wash’.3 The A of the basic verb appears as S of the antipassive; the P of the basic verb, which in the antipassive of this verb is coreferential with the A, cannot

3 Similar behavior of this verb is also found in other closely related Tsezic languages, Tsez, Hinuq (Forker 2013) and Hunzib (van den Berg 1995: 110); Khwarshi, however, has a labile verb usana meaning both ‘wash (something)’ and ‘wash (oneself)’ (Khalilova 2009: 336).

554

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

be expressed overtly (45b). Note that this verb also has a semantically regular antipassive (45c). (45) ‘wash’ a. kibba ɬic’o nizā‑yo girl.obl(erg) clothes wash‑pst ‘The girl washed the clothes.’ b. kid niza‑lā-yo girl wash‑antip-pst ‘The girl washed herself.’ c. kid ɬic’o-li-d niza‑lā-yo girl clothes-obl-instr wash‑antip-pst ‘The girl was busy washing clothes.’

5.2 The Accidental/Potential Alternation 2 (marked) Bezhta has an Accidental/Potential Alternation 2, although this is not found with any verbs in the database. It is found, for instance, with the transitive verb y-aɣal ‘to remove’. The accidental/potential construction of this transitive verb is formed with the suffix -yc’; the Absolutive argument is left unchanged, and the Ergative agent is changed to the Poss-essive case. (46) ‘remove’ a. öždi k’atula-s beš b-aɣo-yo boy.obl(erg) potato.obl-gen1 skin(iii) iii-remove-pst ‘The boy peeled the skin off the potato.’ b. öždi-qa k’atula-s beš b-aɣo-yc’-iyo boy.obl-poss potato.obl-gen1 skin(iii) iii-remove-pot-pst ‘The boy could peel the skin off the potatoes. / The boy accidentally peeled the skin off the potatoes.’

5.3 The Potential Alternation The Potential Alternation (which is distinct from the Accidental/Potential Alternations) is a marked alternation (with the suffix -yɬ after a vowel, -iyɬ after a consonant); it is by far the most complex of the alternations in terms of differences across argument structures, as detailed in the rest of this section. The argument of the potential construction to which ability is assigned (i.e. the X of ‘X can Verb’, hereafter the potential agent) nearly always stands in the Poss-essive case, irrespective of the valency of the basic or potential verb.

Valency and valency classes in Bezhta

555

The Potential Alternation can be formed from 48 verbs including 34 transitive verbs, 1 affective verb, 11 intransitive verbs, and 2 unergative verbs. Transitive verbs are y-o nsal ‘to load’, y-i ncal ‘to tie’, git’al ‘to pour’, glal ‘to put’, y-iƛ’al ‘to throw’, y-aq’al ‘to bring’, y-ezal ‘to carry’, y-e nyal ‘to send’, niƛal ‘to give’, y-u ncal ‘to steal’, y-u ncal ‘to hide’, y-üč’al ‘to cut’, y-äƛ’elal ‘to hit/beat’, yiƛ’al ‘to kill’, nisal ‘to say’, xabar m-eyal ‘to tell’, y-ü nqal ‘to eat’, nizāl ‘to wash’, komak b-owal ‘to help’, keč’ iƛal ‘to sing’, ömrö b-owal ‘to live’, y-owal ‘to make’, y-okčal ‘to take’, kusuʔal ‘to shave’, č’än y-aɣal ‘to run’, maɬci y-owal ‘to teach’, cā n glal ‘to name’, haʔal ‘to grind’, ɬič’alal ‘to dress’, y-äxä l ‘to dig’, xal b-owal ‘to search for’, harzi y-owal ‘to ask’, a nɬo y-oc’al ‘to hug’, y-ä xelä l ‘to tear’. Note that seven of these verbs are transitive by virtue of having a light verb that takes a noun as its P, namely keč’ iƛal (lit. ‘to call song’), xabar m-eyal (lit. ‘to send information’), komak b-owal (lit. ‘to do help’), ömrö b-owal (lit. ‘to do life’), č’än y-aɣal (lit. ‘to remove running’), cā n glal (lit. ‘to put name’), a nɬo y-oc’al ‘to hug’ (lit. ‘to ? embrace’). It will be noted that nearly all of these 34 verbs are basic in the sense that they are not derived from a corresponding morphologically simpler intransitive verb; cf. the behavior of transitive and intransitive verbs under the Accidental/ Potential Alternation (§ 4.4). However, two verbs, y-äƛ’elal ‘to hit/beat’ and y-ä xelä l ‘to tear’ are derived transitives, and form their Potential from the corresponding intransitive, i.e. y-äƛ’eyɬal ‘to hit/beat’ and y-äxeyɬäl, respectively. The number of arguments of a transitive verb does not change in the Potential Alternation, the potential agent stands in the Poss-essive case and the P remains in the Absolutive case. (47) hini-la yakʼi-ʔ-is unti-urɣel holco-qa gisa self.obl-gen2 heart.obl-in-abl disease-sorrow(iii) he.obl-poss out b-aɣo-yɬ-aʔa-s iii-take-pot-neg-prs ‘ he was not able to discuss (lit. take out) his sorrow.’ (Iqla 3.128) The set of verbs undergoing the Potential Alternation includes 2 unergative verbs: öhƛäl ‘to cough’ and lalaƛal ‘to shout/scream’. In the Potential, the number of arguments does not change, but the potential SA appears in the Poss-essive. (48) öždi-qa öhƛö-yɬ-iyo boy.obl-poss cough-pot-pst ‘The boy could cough.’ About 11 intransitive verbs undergo the Potential Alternation. Among these intransitives, there is one intransitive verb, semantically agentive, namely goc’oqal ‘to look’, which remains with one core argument in the Potential, this single core argument appearing in the Poss-essive corresponding to the Absolutive of the basic verb, as in (49).

556

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

(49) a. öžö kibba-qa goc’oq-iyo boy girl.obl-poss look-pst ‘The boy looked at the girl.’ b. öždi-qa kibba-qa goc’oq-iyɬ-iyo boy.obl-poss girl.obl-poss look-pot-pst ‘The boy could look at the girl.’ Intransitive verbs that undergo the Potential Alternation are müɣäƛ y-e nƛ’al ‘to follow’, y-e nƛ’al ‘to go’, q’ey y-ecal ‘to sit (down)’, k’isāl ‘to play’, y-iyaal ‘to cry’, urɣezi y-aqal ‘to think’, y-iq’lašal ‘to talk’, y-ogic’al ‘to jump’, urɣeliƛ’a y-aqal ‘be sad’, yuɣal ‘to die’. They behave in the same way except that their potential agent quite exceptionally stands in the Absolutive, as in (50). Note that two of these involve the same lexical item y-e nyal ‘to go’. Also note that two verbs y-iq’lašal and y-ogic’al may also follow the Potential Transitivizing Alternation. (50) öžö biƛo-ɣa 0-̸ e nƛ’e-yɬ-iyo boy.abs house-at i-go-pot-pst ‘The boy could go home.’ (51) öžö mašinal-ƛ’a 0-̸ ogic’-iyɬ-iyo boy.abs car.obl-super i-jump-pot-pst ‘The boy could jump on the car.’ The Potential Alternation is found with one affective verb in the database, ‘smell’. The Experiencer stands in the Poss-essive and the Stimulus in the Absolutive, i.e. the number of arguments remains the same but the case of the Experiencer shifts. (52) kibba-qa k’ima-li-s mäh b-iq’e-ył-iyo girl.obl-poss cheese-obl-gen1 smell(iii).abs iii-know-pot-pst ‘The girl could smell the cheese.’ The verb y-egāl ‘see’ forms a Potential y-ega-yɬ-al, but this is only used in the idiom mot’o b-ega-yɬ-al, literally ‘to be able to show face’, with the meaning ‘to no longer be angry’; and in the sense ‘to be able to see something small’; the fact that it can mean ‘be able to show’ rather than ‘be able to see’, i.e. a three-place predicate rather than a two-place predicate, presumably parallels the fact that the potential of a patientive intransitive verb is a two-place predicate, in each case an increase of one in the valency (see § 5.4).

5.4 The Potential Transitivizing Alternation The Potential Transitivizing Alternation is a coded alternation (with the suffix -yɬ after a vowel, -iyɬ after a consonant). The basic verb is intransitive. The argument

Valency and valency classes in Bezhta

557

of the potential construction to which ability is assigned (i.e. the X of ‘X can Verb’) stands in the Poss-essive case. This alternation increases the valency of the intransitive verb by one. This alternation is found with 12 verbs: k’ok’al ‘to ache’, qoqal ‘to be dry’, y-ek’al ‘to burn’, y-ekal ‘to fall, to drop’, ɣanq’izi y-aqal ‘to sink’, ɣayƛal ‘to boil’, k’äyƛäl ‘to roll’, nukal ‘be hungry’, y-o nq’al ‘to come’, pacpaƛal ‘to blink’, y-ogic’al ‘to jump’, y-iq’lašal ‘to talk’, and 1 ambitransitive verb helal ‘to cook’. Note that the verb ‘to jump’ also undergoes the Potential Alternation. (53) a. ɬi ɣayƛo-s water boil-prs ‘The water boils.’ b. kibba-qa ɬi ɣayƛo-yɬ-iyo girl.obl-poss water boil-pot-pst ‘The girl could boil the water.’ (54) a. öžö mašinal-ƛ’a-s 0-̸ ogic’-iyo boy.abs car.obl-super-abl i-jump-pst ‘The boy jumped off the car.’ b. öždi-qa kid mašinal-ƛ’a-s y-ogic’-iyɬ-iyo boy.obl-poss girl(ii) car.obl-super-abl ii-jump-pot-pst ‘The boy could make the girl jump off the car.’

5.5 The Causative Alternations The Causative Alternations are a set of marked alternations that are valency increasing mechanisms. There are four types of Causative Alternations, and at least one of these types is found for each verb in the database. The four types are the Causative with -k’, Causative with yowal, Causative with -l/-ll, Causative with golal.

5.5.1 The Causative Alternation with -k’ The Causative Alternation with -k’ makes use of the transitive suffix -k’ which relates transitive verbs to intransitive inchoative verbs with the suffix -ɬ. In Bezhta such inchoative-causative verb pairs are a distinct class of verbs. The inchoativecausative verbs are verbs derived from adjectives and adverbs, expressing a change of state. The inchoative verbs are derived with the suffix -ɬ and the causative verbs are derived with the suffix -k’. In the database only one such verb pair is found, as in (55).

558

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

(55) a. öžö 0-̸ ač’-ɬ-is boy(i) i-be_cold-inch-prs ‘The boy feels cold.’ b. kibba öžö 0-̸ äč’-k’-iyo girl.obl(erg) boy(i) i-be_cold-tr-pst ‘The girl made the boy feel cold.’

5.5.2 The Causative Alternation with yowal Bezhta has a distinct class of light verbs which consist of two parts, the lexical word and the auxiliary verbs y-aqal ‘to happen’ and y-owal ‘to do’. When the auxiliary verb in the light verb construction is the verb y-aqal ‘to happen’ then the construction is intransitive or affective, and when the auxiliary verb is the verb yowal ‘to do’ then it is transitive. Many intransitive and affective compound verbs formed with the auxiliary verb y-aqal ‘to happen’ derive transitives by changing the auxiliary verb to y-owal ‘to do’; kezi y-aqal ‘to meet (intransitive)’ (affective in Bezhta), kezi y-owal ‘to introduce (to each other) (transitive)’, and urɣezi y-aqal ‘to think’, urɣezi y-owal ‘to cause to think’. Note that the inverse derivation does not hold, i.e. there are many transitive verbs using the light verb y-owal that do not have a corresponding intransitive with y-aqal. (56) a. öždi-l kid keziaq-iyo boy.obl-lat girl(ii) meet-pst ‘The boy met the girl.’ b. kibba öždä keziō-yo girl.obl(erg) boy.pl meet-pst ‘The girl introduced the boys.’ ̸ aq-iyo (57) a. öžö kibba-ƛ’a-s urɣezi boy(i) girl.obl-super-abl think-pst ‘The boy thinks about the girl.’ b. öždi iyo kibba-ƛ’a urɣeziō-yo boy.obl(erg) mother(ii) girl.obl-super think-pst ‘The boy made the mother think about the girl.’

5.5.3 The Causative Alternation with -l/-ll The causative suffix -l/-ll derives transitive verbs from intransitive or affective verbs, and ditransitive verbs from transitive verbs; in the latter, the causee appears

Valency and valency classes in Bezhta

559

in the Instrumental case. The basis of the choice between -l and -ll (which has a variant -ill after consonants) requires further work. A geminate is sometimes the causative of a causative (see below), but often clearly a simple causative. A few verbs allow alternatives, e.g., y-owal ‘to do’ has both y-oy-l-al and y-oy-ill-al. Causatives of intransitive verbs introduce a new argument, an agent in the Ergative case, and the former Absolutive subject of the intransitive clause becomes the Absolutive patient of the transitive clause. (58) a. sidda-ƛ’a maɣo k’ok’o-s one.obl-super body ache-prs ‘(My) body aches from one side .’ b. untila k’ok’o-ll-o maɣo disease.obl(erg) ache-caus-pst body ‘The disease made the body ache.’) Causatives from transitive verbs derive a ditransitive construction with the causer in the Ergative case, the causee in the Instrumental case, and the patient in the Absolutive. (59) a. öždi ɬi č’aƛe-š boy.obl(erg) water pour-prs ‘The boy pours the water.’ b. kibba öždi-d ɬi č’aƛe-ll-is girl.obl(erg) boy.obl-ins water pour-caus-prs ‘The girl makes the boy pour the water.’ When the causative is from an unergative verb, the newly introduced agent argument, the causer, is in the Ergative and the causee in the Instrumental case. (60) a. öždi kibba-ƛ’a lalaƛo-yo boy.obl(erg) girl.obl-super shout-pst ‘The boy shouted at the girl.’ b. öždi ist’i-d kibba-ƛ’a lalaƛo-ll-iyo boy.obl(erg) brother.obl-ins girl.obl-super shout-caus-pst ‘The boy made the brother shout at the girl.’ A causative can be derived from the P-labile verb helal ‘to boil’, but only with the transitive meaning, and never from the same verb with intransitive meaning. This causative regularly derives a ditransitive construction from a monotransitive one, with the causer in the Ergative, the causee in the Instrumental and the patient in the Absolutive.

560

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

(61) a. k’atu hele-yo potato boil-pst ‘The potato has boiled.’ b. kibba k’atu hele-yo girl.obl(erg) potato boil-pst ‘The girl boiled the potato.’ c. kibba öždi-d k’atu hele-ll-iyo girl.obl(erg) boy.obl-ins potato boil-caus-pst ‘The girl made the boy boil the potato.’ When the causative is based on affective verbs, the construction becomes ditransitive with an agentive argument in the Ergative case, an Absolutive theme, and a Lative experiencer. (62) a. hogco-l raɬad b-egā-yo he.obl-lat sea(iii) iii-see-pst ‘He saw the sea.’ b. hogco kibba-l raɬad b-ega-l-lo he.obl(erg) girl.obl-lat sea(iii) iii-see-caus-pst ‘He showed the sea to the girl.’ The causative with -(i)l/-(i)ll can have idiosyncratic meaning when used with some affective verbs, such as y-at’al ‘to love’ and y-iq’al ‘to know’. (63) a. kibbal iyo y-at’-ca daughter.lat mother(ii) ii-love-prs ‘The daughter loves the mother.’ b. kibba iyo y-at’-il-ca daughter.erg mother(ii) ii-love-caus-prs ‘The daughter cuddles with the mother.’ (64) a. öždil dars y-iq’e-š boy.lat lesson(iv) iv-know-prs ‘The boy knows the lesson.’ b. öždi dars y-iq’e-l-ca boy.erg lesson(iv) iv-know-caus-prs ‘The boy learns the lesson.’ Causatives can be formed from almost all verbs in the database (and in the language). The only exceptions seem to be inchoatives with the suffix -ɬ, which

Valency and valency classes in Bezhta

561

only allow the corresponding transitive with the suffix -k’, and not a sequence of inchoative plus causative suffixes. However, light verb constructions with the intransitive light verb y-aqal can form causatives, in addition to transitive equivalents using the transitive light verb y-owal. Causatives can be formed from transitive verbs with the suffix -k’ (leading to a suffix sequence of transitive plus causative), from transitive constructions with the light verb y-owal, and also from causatives, i.e. verbs that already have one causative suffix can add a further causative suffix with the meaning “causative of causative” (double causative). At least some double causatives are perfectly natural, e.g., from intransitive ɣayƛal ‘to boil’ one can form transitive ɣayƛo-l-al, and from this one can in turn form ɣayƛo-l-l-al ‘cause [someone] to boil [something], cause [someone] to cause [something] to boil’.

5.5.4 The Causative Alternation with gulal Another alternation is Causative with gulal. A small group of intransitive verbs form transitives analytically with the verb gulal with gender variants golal, gilal ‘put’. These verbs are hic’al ‘to be afraid’, y-iyāl ‘to cry’, k’isal ‘to play’, ɬowal ‘to laugh’, y-u nqal ‘to eat’, nukal ‘to be hungry’, öhdǟl ‘to cough’. (65) a. kid łowa-s. girl laugh-prs ‘The girl laughs.’ b. öždi kid łowa-gil-ca boy.erg girl(ii) laugh-make(ii)-prs ‘The boy made the girl laugh.’

6 Others 6.1 The Object Incorporation Alternation This alternation is only found with the one verb ‘to run’. This alternation reduces the valency of this verb by one. The verb ‘to run’ is transitive with two core arguments, agent and patient (66a). Additionally, the verb ‘to run’ can undergo contraction of patient and the verb resulting in a one-argument construction where the agent is still marked with the Ergative (66b), i.e. the result is unergative. (66) a. öždi yiƛ’a č’än y-aɣo-s boy.obl(erg) fast run(iv) iv-take-prs ‘The boy runs fast.’

562

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

b. öždi yiƛ’a č’änäɣö‑š boy.obl(erg) fast run‑prs ‘The boy runs fast.’

7 Conclusion In Bezhta most of the alternations (case-coded and verb-coded) have a semantic effect beyond mere valency change. For example, the Recipient Alternation relates to permanent versus temporary transfer of possession to the recipient. The Accidental/Potential Intransitive and Transitive Alternations refer to the situation where the agent-like argument can be an involuntary actor or a potential actor in an event. The Potential Alternation and the Potential Transitivizing Alternation can only express a potential agent-like argument, not an involuntary actor. The Antipassive Alternation is a valency-changing derivation only when used with transitive and unergative verbs; since intransitives can take antipassive suffixes, and since the Bezhta antipassive always indicates iterativity, it is more properly considered an iterative rather than a canonical antipassive, i.e. it is basically a device for changing semantics, with only incidental valency-changing properties. Only the Causative Alternations and a few minor alternations (the Ambitransitive, Instrumental, and Object Incorporation Alternations) have no semantic effect beyond valency change. It is difficult in Bezhta to make generalizations relating to semantic classes of verbs. Most alternations are either highly restricted, sometimes to a single verb in the database, while others, such as the Causative Alternation, are so productive that they do not subgroup verbs in any significant way. Only the Accidental/Potential Intransitive Alternation, the Antipassive 1 Alternation, and the Potential Alternation and the Potential Transitivizing Alternation apply to large but restricted sets of verbs, but in none of these cases does the relevant set of verbs seem easily definable in semantic terms. It is thus difficult to make generalizations across classes of verbs.

Appendix: Summary of Bezhta valency classes Verbs are cited in their stem form; a leading hyphen or angled brackets indicate an agreement slot. There are differences between the presentation here and in the on-line database, reflecting our view of how valency and valency alternations operate internal to Bezhta here, versus the cross-linguistic perspective imposed in the on-line version.

play

feel pain

appear

feel cold

be dry

be hungry

be sad

blink

boil

burn

cry

die

fall/drop

go

jump

62

59

81

60

68

64

63

46

80

67

83

61

84

54

52

-ogic’-

S[Abs] jump

S[Abs] go

-e ƛ’e-

S[Abs] fall

n

S[Abs] die

S[Abs] cry

S[Abs] burn

S[Abs] boil

-ek-

-uɣo-

-iya-

-ek’e-

ɣayƛ-

S[Abs] blink

S[Abs] be sad

urɣeliƛ’a gey5

pacpaƛo-

S[Abs] be hungry

nuko-

S[Abs] be dry

S[Abs] be cold

qoqo-

S[Abs] appear

-äč’ɬ-4

S[Abs] ache

S[Abs] play

Case frame

-o q’-

n

k’ok’o-

k’isa-

Verb form

4 Variant -äƛƛ-, presumably reflecting assimilation. 5 lit. ‘to be on sorrow’

Meaning label

#

Minor Alternations





+

+



+

+

+





+

+

+





Acc/ Pot Intr































Acc/ Pot Trans

+

+



+



+

+

+











+

+

Antipassive 1

+

+



+

+





+











+

Pot

+

+





+

+

+

+

+



+

+

-

Pot Trans

Valency and valency classes in Bezhta

563

cough

shout at

run

laugh

follow

leave

look at

think

fear

47

20

49

57

16

55

 3

10

 6

Exp[Abs] fear X[Poss]

A[Abs] thinks about X[Super-abl]

A[Abs] look at P[Poss]

A[Abs] leave L[Abl]

A[Abs] follow P[Apud]

S[Abs] laugh at X[Super]

S[Erg] run

S[Erg] shout

S[Erg] cough

S[Abs] sit

S[Abs] sink

S[Abs] roll

Case frame

Minor Alternations























+

Acc/ Pot Intr

























Acc/ Pot Trans







+

+

+



+

+







Antipassive 1

+

+

+

+





+

+

+





Pot



-



-













+

+

Pot Trans

6 Bezhta does not distinguish lexically between stative sit and inchoative sit down. 7 lit. ‘to be down’ 8 lit. ‘to go after’ 9 This is the same as the verb for ‘go’, and is included separately only because of the distinct valency frame, although the literal translation is simply ‘X goes from Y’, with Y in one of the combinations of an orientational case with the Ablative. The text does not include the translation of ‘leave’ as a separate item.

hič’e-

urɣezi -aq-

goc’oq-

-e

n

ƛ’e-9

müɣäƛ -e

ɬowa-

č’änäɣö-

lalaƛo-

öhƛö-

ƛ’e-8

q’ey -eče-7

sit (down)6

51

n

ɣanq’izi -aq-

sink

66

k’äyλö-

roll

65

Verb form

Meaning label

#

564 Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

get

hear

know

like/ want 10

meet

see

smell

live

help

hug

run

86

78

 9

 8

17

 4

 5

56

15

 2

49

Exp[Lat] smell Stim[Gen] A[Erg] live in Loc[In] A[Erg] help P[Lat] A[Erg] hug P[Lat] A[Erg] run (a run[Abs])

mäh b-iq’e-12

ömrö b-ō-13

komak b-ō-14

y-oc’-15

y-aɣo-16

č’än

a ɬo

n

Exp[Lat] see Stim[Abs]

Exp[Lat] meet X[Abs]

Exp[Lat] like Stim[Abs]

Exp[Lat] know Stim[Abs]

Exp[Lat] hear Stim[Abs]

Exp[Lat] get T[Abs]

A[Abs] talks to X[Com] about Y[Gen]

-egā-

kezi -aq-

-at’-

-iq’e-

tuq-

-iq-

-iq’laše-

















































+

+







+





+







+

+

+

+













–11 +











+











+

10 We have combined the items like and want into a single entry since they correspond to a single lexical item in Bezhta, with no difference in valency frame. 11 Only in the idiom mot’o b-egayɬal, literally ‘to be able to show face’ (where b- is the gender III prefix indexing the gender III noun mot’o ‘face’ in the verb), in the meaning ‘to no longer be angry’, and in the sense ‘to be able to see something small’. 12 lit. ‘to know [Stim’s] smell’ 13 lit. ‘to do life’ 14 lit. ‘to do help’ 15 lit. ‘to V hug’; The identity of the verb -oc’- in the translation of ‘hug’ is unclear; it is apparently not the intransitive verb -oc’ ‘to fill’. 16 lit. ‘to remove run’

talk

18

Valency and valency classes in Bezhta

565

sing

cook

dig

dress

eat

frighten

wash

search for

shave

push

ask for

say

53

79

73

13

 1

7

12

11

14

74

19

22

niso-

hardizi -ō-

zistuql-

kusuʔ-

xal

b-ō-19

nizā-

A[Erg] say P[Abs] to Y[Poss]

A[Erg] ask X [Abs] from Y[Poss]

A[Erg] push P[Abs]

A[Erg] shave

P[Abs] 20

A[Erg] search for P[Abs]

A[Erg] wash P[Abs]

A[Erg] frighten P[Abs]

A[Erg] eat P[Abs]

-ü nq-

hič’egol-

A[Erg] dress P[Abs]

A[Erg] dig P[Abs]

ɬič’el-

-ä xä

A[Erg] cook P[Abs]

A[Erg] sing (a song [Abs])

keč’ iƛe-17

hel-

Case frame

Verb form

Minor Alternations





















+



Acc/ Pot Intr

























Acc/ Pot Trans





+

+



+

+

+

+

+

+

18

Antipassive 1

+

+

+

+

+

+

+

+

+

+

Pot

























Pot Trans

17 lit. ‘to call song’ 18 and Antipassive 2 19 Although this appears to consist of a gender iii noun xal (not attested independently) as P of the verb y-owal ‘to do’, the construction nonetheless takes a P in the Absolutive; the gender prefix b- on the verb is, however, fixed and does not agree with this lexical P. 20 The verb requires a P such as bab ‘beard’.

Meaning label

#

566 Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

grind

cut

kill

bring

carry

give

make

send

show

tell

71

30

26

75

38

36

85

37

35

21

m-ē-21

xabar

A[Erg] tell P[Abs] to Y[Lat]

A[Erg] show T[Abs] to R[Lat]

A[Erg] send T[Abs] to R[Lat]

-e ne-

-egal-

A[Erg] make P[Abs] out of X[Lat]

A[Erg] give T[Abs] to R[Lat]

A[Erg] carry T[Abs] to R [Lat]

A[Erg] bring T[Abs] to R[Lat]

A[Erg] kill P[Abs] with Instr[Ins]

A[Erg] cut P[Abs] with Instr[Ins]

A[Erg] grind P[Abs] with Instr[Ins]

A[Erg] break P [Abs] with Instr[Ins]

-ō-

niƛ-

-eze-

-aq’-

-iƛ’e-

-üč’-

haʔ-

-išel-

21 lit. ‘to send information’

break

25

A[Erg] send T[Abs] to R[Apud] (§ 4.1)

A[Erg] give T[Abs] to R[Poss] ((§ 4.1) A[Erg] give T[Abs] to R[Poss.lat] (§ 4.1)

A[Erg] carry T[Abs] to R [Apud] (§ 4.1)







































+

+

+

+

+

+

+

+

+

+

+

+

+

+

+



+

+

+

+

+

+

+

+

























Valency and valency classes in Bezhta

567

fill

tear

hide

load

put

pour

tie

throw

wipe

take

44

32

34

45

41

42

40

39

72

31

A[Erg] tear T[Abs] from X[Gen] A[Erg] hide T[Abs] from X[Abl] A[Erg] load P[Abs] on Loc[Super]

-äxel-23

-u nco-

-o nso-

A[Erg] take P[Abs] from X[Poss-abl]

A[Erg] wipe T[Abs] off X[Loc]

-a ncol-

-okč-

A[Erg] throw T[Abs] at Loc[Apud]

A[Erg] tie P[Abs] to Loc[Inter]

-i nco-

-iƛ’e-

A[Erg] pour P[Abs] in Loc[Inter]

git’-

A[Erg] put P[Abs] on Loc[Super]

A[Erg] fill P[Abs] with X[Gen]

-oc’il-22

gl-

Case frame

Verb form

22 caus of -oc’- ‘to become full’ 23 caus of -äxe- ‘to be torn’ 24 The Potential is formed from the corresponding intransitive stem.

Meaning label

#

A[Erg] throw T[Abs] at Loc[Lat] (§ 4.1) A[Erg] throw T[Abs] at Loc[Poss] (§ 4.1)

Minor Alternations





















Acc/ Pot Intr





















Acc/ Pot Trans

+

+

+

+

+

+

+

+

+

+

Antipassive 1

+



+

+

+

+

+

+

+24



Pot





















Pot Trans

568 Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

teach

touch

28

77

29

A[Erg] touch P[Lat] with X[Abs]

A[Erg] teach R[Lat] T [Abs]





– –









+



+

+

+



+

+27

+













25 lit. ‘to put name’ 26 We have combined the items hit and beat into a single entry since they correspond to a single lexical item in Bezhta, with no difference in valency frame. 27 The Potential is formed from the corresponding intransitive stem.

Legend: + occurs regularly; m occurs marginally; − occurs never

-e xe-

n

maɬci -ō-

-äƛ’el-

hit/beat 26



A[Erg] hit P[Lat] with Instr[Abs] A[Erg] hit P[Abs] with Instr[Ins] – (§ 4.3)

A[Erg] name X[Super] Y[Abs]

cā n gl-25

name

23

А[Erg] cover P[Super] with X[Abs]

-oq’ol-

cover

43

Valency and valency classes in Bezhta

569

570

Bernard Comrie, Madzhid Khalilov and Zaira Khalilova

Abbreviations apud at ess Exp hpl in inch inter lat poss pot Stim super

Apud-essive locative ‘at’ Essive (location) Experiencer-like argument Human plural locative ‘in’ Inchoative Inter-essive Lative (motion to) Poss-essive Potential Stimulus-like argument locative ‘on’

References Creissels, Denis. ms. Valency properties of Northern Akhvakh verbs. http://www.deniscreissels.fr/ public/Creissels-valency_classes_project_Akhvakh.pdf. (Last consulted 2014-11-16) Daniel, Michael, Zaira Khalilova & Zarina Molochieva. 2010. Ditransitive constructions in East Caucasian: A family overview. In Andrej Malchukov, Martin Haspelmath & Bernard Comrie (eds.), Studies in Ditransitive Constructions: A Comparative Handbook, 277–315. Berlin: De Gruyter Mouton. Forker, Diana. 2013. A grammar of Hinuq. Berlin: De Gruyter Mouton. Khalilova, Zaira. 2009. A Grammar of Khwarshi. Utrecht: LOT. van den Berg, Helma. 1995. A Grammar of Hunzib: With Texts and Lexicon. Munich: Lincom Europa. Xalilov, Madžid Š. 1995. Bežtinsko-russkij slovar´. Makhachkala: IJaLI DNC RAN.

Andrej Malchukov and Igor V. Nedjalkov

16 Valency classes in Even (North Tungusic) in a comparative Tungusic perspective 1 Introduction1 The present chapter provides an overview of valency patterns and valency alternations in Even, a North Tungusic language, from a comparative Tungusic perspective. The paper is structured as follows. Section 2 provides general information on the morphosyntax of Even relevant to issues of valency classification. Section 3 discusses diagnostics for distinguishing between arguments and adjuncts, which is crucial for identifying valency frames. Section 4 describes valency patterns in Even in some detail. Section 5 addresses uncoded argument alternations (“case alternations”), while section 6 describes verb-coded alternations (“voice alternations”). Section 7 discusses availability of other case/voice alternations familiar from other languages. Section 8 provides a concluding discussion of valency classes in Even. The discussion of Even valency in the above-mentioned sections draws heavily on the monographic descriptions of Even syntax in (Malchukov 1999) and (Malchukov 2008). The last section (section 9) provides a comparative-contrastive perspective on the issue of valency classification in Even, and it addresses the question as to what extent variation in case inventories and in the function/distribution of valency-changing categories has an impact on valency classification across Tungusic languages.

2 Basics of Even morphosyntax 2.1 General characteristics of Even As a North Tungusic language, Even2 displays many of the typical properties of Altaic languages (see Malchukov 1995 for a brief presentation, and Benzing 1955a;

1 We are very grateful to Bernard Comrie and Martin Haspelmath for comments on an earlier version of this draft. We are indebted to our colleagues who provided native judgments on their languages, in particular, to Dr. Sverchkova (Evenki) and Dr. Zaksor (Nanai). The usual disclaimers apply. 2 The Even data, unless otherwise indicated, comes from Andrej Malchukov’s fieldnotes during his fieldwork in the village of Tompo (Yakut/Sakha Republic). The Tompo dialect belongs to the group of mid-western dialects, but it is fairly representative of other dialects with respect to issues of valency classification.

572

Andrej Malchukov and Igor V. Nedjalkov

Novikova 1960; and Malchukov 2008 for more detail). In particular, it is an agglutinating language making use of suffixation. Morphophonological variation in suffixes is due to vowel harmony and assimilation processes. Even is a nominative-accusative language with the basic word order SOV. Word order is fairly free, so it will be disregarded in the following discussion of valency patterns. Even shows dependent marking at the clause level (i.e. it uses cases/postpositions to code arguments), or double marking in the case of subjects (which are additionally cross-referenced through person/number agreement suffixes), but it shows head marking at the NP level. That is, in a possessive construction, the head is marked with possessive agreement, while a genitive is lacking. This is generally true for nouns (cf. bejd’u-n [man house-3sg.poss] ‘the man’s house’), while for pronouns a special possessive form is used; thus, the construction is double-marked (cf. hin d’u-s [your house-2sg.poss] ‘your house’). The possessive agreement endings include personal-possessive endings and reflexive-possessive endings. The latter are used when the possessor is coreferential with the subject.

2.2 Basic clause structure The basic intransitive and transitive constructions in Even are exemplified in (1) and (2) below: (1) Oron-0̸ köke-n. reindeer-nom die-nf.3sg ‘The reindeer died.’ (2) Etiken-0̸ oro-m ma-n. old_man-nom reindeer-acc kill-nf.3sg ‘The old man killed the reindeer.’3 As illustrated above, the clause structure is straightforwardly accusative with respect to both flagging and indexing. The subject (S/A argument) is in the nominative, while the direct object (P) is in the accusative (for other forms of the direct object see § 4.4.1). Since the NOM case is zero marked in what follows, it is omitted from the glosses. Agreement is possible only with the (nominative) subject; there is no object agreement in Even.

3 In Even a distinction is made between oron ‘domestic reindeer’ and bujun ‘wild reindeer’. The difference is ignored in the glosses and translations of the examples.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

573

2.3 Case-markers Argument coding is accomplished primarily through case markers. There are 14 distinct case-markers in Even (Malchukov 1995). Apart from the unmarked nominative (NOM), the following 10 markers will be of interest here: 4 – Accusative (ACC): -v/-u/-m – Dative (DAT): -du/-tu – Comitative (COM): -n’un – Designative (DES): -gA-/-kA- 5 – Instrumental (INST): -č/-d’i/-n’ – Locative (LOC): -lA/-dulA – Prolative (PROL): -li/-duli – Allative (ALL): -tki/-tAki – Ablative (ABL): -duk/-tuk – Elative (ELAT): -gič/-kič Unlike the other case-markers, DES occurs only in the “possessive declension”, that is, only when followed by a possessive person suffix: cf. d’u-ga-s [house-DESyour] ‘the house for you’; see 5.2 below for a discussion of the peculiarities of designative case marking.

2.4 Voice and valency-changing markers Even has the following valency-changing categories (see Malchukov 1995 for an overview; and Novikova 1980; Robbek 1984; Burykin 1984; and Malchukov 1999, 2008 for details): causative in -vkAn‑ (cf. ič- ‘see’ → ič-uken- ‘show’), adversative passive in -v-/-m- (ma- ‘kill’ → ma-v- ‘be killed’), mediopassive in -b-/-p- (aŋa- ‘open (tr.)’ → aŋa-b- ‘open (intr.)’, reciprocal in -mAt- (it- ‘see’ → it-met- ‘see each other’), sociative in -ldA- (bak- ‘find’ → baka-lda- ‘meet’). The latter may be used as a reciprocal marker with some verbs (otherwise it does not change valency) expressing a ‘together’ meaning. Another marker which can be used to reduce valency is the resultative in -t-/-č-, which, when derived from transitives is usually used as a stative passive (cf. aŋa- ‘open’ → aŋa-t- ‘be open’). The adversative passive, implying an adverse effect on the subject of the passive form, is unusual in that it can

4 Other case-markers are: ‘directive-locative’ in -kla, ‘directive-prolative’ in -kli, and equative in -g(a)čin. 5 Here and elsewhere, allomorphy in suffixes conditioned by vowel harmony is represented by capital letters, a convention common in Turkic studies. Otherwise orthographic conventions largely follow (Malchukov 1995), except for the fact that vowel length is not indicated. Orthographic conventions for other Tungusic languages follow those in Malchukov and Nedjalkov (2010), except for the cases when examples come from secondary sources.

574

Andrej Malchukov and Igor V. Nedjalkov

both reduce and increase verbal valency (see 6.2 below for further discussion of the adversative passives).

3 Identifying valency frames Even lacks an anaphoric ‘do so-construction’, so the verb-anaphoric test which is the most straightforward diagnostic for distinguishing between arguments and adjuncts cannot be applied. Some Even dialects, however, display a construction reminiscent of the ‘do-so construction’ as found in English and other languages. This construction involves ‘ordinal pro-verbs’ that are based on numerals and have the meaning ‘do for nth time’ (e.g., gia-v- ‘do for the second time’ from gia ‘another; the second’; ili-v- ‘do for the third time’ from ili(ten) ‘the third’, etc). These constructions can also be used to distinguish between complements and adjuncts (Malchukov 2008: 39). (3) a. Etiken buju-n. Öme-m buju-m ma-n. old_man hunt_go-nf.3sg one-acc reindeer-acc killed-nf.3sg ‘The old man went hunting. (He) killed a (wild)reindeer.’ b. Timinni buju-rid’i giav-ra-n ma-n. next_day hunt_go-cvb again_do-nf-3sg kill-nf.3sg ‘Next day (he) went hunting (and) again killed (a reindeer).’ Note that in the context of (3a), substitution of the direct object bujum in (3a) for another direct object (as in (3c)) results in ungrammaticality or contradiction: (3) c. #Timinni buju-rid’i, giav-ra-n toki-v ma-n. next_day hunt_go-cvb again_do-nf-3sg elk-acc kill-nf.3sg ‘Next day (he) went hunting (and) again killed an elk.’ However, this test is not available for those dialects where ordinal pro-verbs are non-productive. Thus for other dialects we must rely on other tests. As argued in Malchukov (1995, 2008), for Even the relativization test is a fairly reliable diagnostic for distinguishing between arguments and adjuncts. Only arguments can be relativized by the primary participial strategy, a prenominal gapping strategy in terms of Keenan and Comrie (1977). This test applies equally to direct and oblique objects, for example, and to objects of bivalent intransitives as well as to the indirect object of trivalent transitives. The following examples illustrate relativization of arguments of the trivalent verb bö- ‘give’ from the basic clause in (4a). Note that object relativization, unlike subject relativization, involves marking of the partici-

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

575

ple by possessive markers in agreement with the embedded subject (the gap in the relative clause is indicated by the underscore): (4) a. Etiken hurken-du oro-m bö-n. old man boy-dat reindeer-acc give-nf:3sg ‘The old man gave the reindeer to the boy.’ b. [_ hurken-du oro-m bö-če] etiken. boy-dat reindeer-acc give-perf part old man ‘The old man who gave the reindeer to the boy.’ c. [etiken hurken-du _ bö-če-n] oron. old man boy-dat give-perf part:3sg reindeer ‘The reindeer that the old man gave to the boy.’ d. [etiken _ oro-m bö-če-n] hurken. old man reindeer-acc give-perf part:3sg boy ‘The boy to whom the old man gave the reindeer.’ Relativization of adjuncts is generally impossible with the primary strategy (some dialects have developed, under Russian influence, finite relative clauses making use of relative/interrogative pronouns, but these are irrelevant for the present discussion). The relativization test identifies the status of the benefactive/recipient NP as an argument in the case of bö- ‘give’ but as an adjunct in the case of ga- ‘take’. (5) a. Etiken hurken-duk oro-m ga-d-ni. old man boy-abl reindeer-acc take-nf-3sg ‘The old man took the reindeer from the boy.’ b. [etiken _ oro-m ga-ča-n] hurken. old man reindeer-acc take-perf part:3sg boy ‘The boy from whom the old man took the reindeer.’ Note that oro-m ga-ča-n hurken in (5b) cannot be interpreted as ‘the boy for whom I have taken the reindeer’ but only as ‘the boy from whom I have taken the reindeer’. In this case, it would correspond to a relativized Source argument (in the ablative case) which belongs to the valency frame but crucially not to the beneficiary adjunct. Another test which can be used to delimit arguments is reciprocal formation (see § 6.5 below), which is also sensitive to the argument/adjunct distinction. The reciprocal test is more restrictive though as it can be used only for constructions with animate arguments. Yet, to the extent that the reciprocal test and relativization test are both applicable, the results of these tests converge on the same NPs

576

Andrej Malchukov and Igor V. Nedjalkov

(arguments). In what follows, we will determine the number of verbal arguments on the basis of the relativization test since it has the broadest applicability.

4 Valency classes in Even: an overview 4.1 Introductory As explained above, the valency classification of Even verbs is established on the basis of the relativization test: only arguments, but not adjuncts, can be relativized by the primary gapping strategy. This means that avalent verbs do not allow relativization at all; monovalent verbs allow relativization of subjects only; bivalent verbs – both transitive and intransitive with oblique objects – allow relativization of both subjects and objects, while trivalent verbs allow for relativization of all three arguments (see above). Only arguments identified in this way are regarded as part of the valency frame in the following sections (and are represented in the coding frame in the summary table in the appendix). As coding frames are manifested mainly in case-marking (flagging), and (subject-)indexing is less important, case frames rather than full coding frames are given below. The list of verbs cited to illustrate individual valency patterns is not exhaustive and is biased towards the verbs on the Leipzig Verb List introduced in the Database Manual. For more details on syntactic classes of verbs, see Malchukov (1999, 2008).

4.2 Avalent verbs (coding frame: < >) This class includes meteoverbs, some of which do not combine with a subject at all and others which may optionally take a cognate subject; bav- ‘dawn’ is representative of the first group (Ba-v-ra-n ‘It dawned’), while udan- ‘rain’ is representative of the second. (6) (Udan) udn-a-n. rain rain-nf-3sg ‘It rains.’ The verb in (6) counts as avalent since it can only take a cognate (non-referential) subject and thus counts as a variety of impersonal construction (Malchukov & Ogawa 2011). On this account, avalent predicates (in a cognate subject construction) seem to be more widespread in Even when compared to other languages. Cf. N’ölten n’ölten-ni ‘The sun shines’ which would be expressed as a simple intransitive in most languages.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

577

4.3 Monovalent verbs (coding frame: ) Monovalent verbs constitute a fairly heterogeneous group, including verbs with both animate/agentive and inanimate/non-agentive subjects: evi- ‘play’ is representative of the first class, while hegdem- ‘tear (intr)’ is representative of the second group. (7) Kuŋa evi-n. child play-nf.3sg ‘The child plays.’ (8) Mavut hegdem-re-n. lasso tear-nf-3sg ‘The lasso tore/got torn.’ For different kinds of subjects, the case is invariably the nominative; the only exception are predicates like ačča ‘be absent’, which also allow designative case marking on subjects (see § 5.1 below on the nominative/designative alternation).

4.4 Bivalent verbs 4.4.1 The transitive pattern Transitive verbs may occur in three different types of transitive structures, depending on the form of the object. The most general form of direct object marking is the accusative case (see (2) above). Accusative marking is generally obligatory. Thus, Even, like other North Tungusic languages, does not show the kind of “Differential Object Marking” (DOM) as familiar with from Turkic languages (and also found in East and South Tungusic languages; see § 9.3 below) where the accusative can be dropped if an object is indefinite/non-specific. However, Even shows another kind of alternation, where the accusative is replaced by the possessive-reflexive form or the designative case marker (see § 5.2, § 5.3 for discussion of accusative-reflexive and accusative-designative alternations). The accusative is replaced by the reflexive possessive marker if the subject is the possessor of the object: (9) Etiken or-mi ma-n. old_man reindeer-refl.poss.sg kill-nf.3sg ‘The old man killed his (own) reindeer.’ Another form of direct object marking is through designative case (always in combination with the possessive markers).

578

Andrej Malchukov and Igor V. Nedjalkov

(10) Etiken hin or-ŋa-s emu-n. old_man your reindeer-des-2sg. poss bring-nf.3sg ‘The old man brought the reindeer for himself.’ While the accusative pattern is equally applicable to all verbs (see below for a qualification), ACC~REFL and ACC~DES can roughly partition the verbs into subclasses of transitive verbs with “affected” vs. “effected” (i.e. created) objects. (See § 5.2 and § 5.3 below for further discussion of accusative-reflexive and accusativedesignative alternations.) The transitive pattern is not restricted to events with affected Patients; thus, many verbs of perception (gele(t)- ‘look for’) and social interaction (bel- ‘help’) as well as verbs of knowledge (ha- ‘know’) and some emotion verbs (ajav- ‘like; love’) also display this pattern. (11) Asi kuŋa-v gelet-te-n. Woman child-acc look_for-nf-3sg ‘The woman looked for the child.’ (12) Asatkan hurke-m ajav-ra-n. girl boy-acc like-nf.3sg ‘The girl liked the boy.’ Yet, the classes of verbs of perception and emotion also include verbs taking oblique objects or verbs alternating between a transitive and intransitive pattern, possibly with a concomitant variation in meaning (see § 4.4.2.2 on it- ‘see; look at’).

4.4.2 Bivalent intransitives 4.4.2.1 The case frame This case frame is found with a closed class of verbs mostly referring to non-volitional (often inadvertent) events involving a second animate argument like dabda‘submit oneself, lose to’, huptu- ‘fall behind’, and tat- ‘get used to’: (13) Oron hunŋi-di tat-ta-n. reindeer master-dat.refl get_used-nf-3sg ‘The reindeer got used to its master.’ It is worth noting that the Agent in the adversative passive constructions, if expressed, also takes the dative case (see 6.2).

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

579

4.4.2.2 The case frame This case frame is found with a small class of verbs including some verbs of intentional visual perception like it- ‘look at’ and köjet- ‘look at’. (14) Bej hurken-teki it-te-n. man boy-all see-nf-3sg ‘The man looked at the boy.’ Note that it- can also be used for resultative (but not necessarily intentional) perception (‘see’); in that case, it takes the transitive frame: (15) Bej hurke-m it-te-n. man boy-acc see-nf-3sg ‘The man saw the boy.’

4.4.2.3 The case frame A few emotion verbs, including ŋel- ‘fear’, ol- ‘fear’, girbe- ‘be shy (of)’, and n’umari- ‘be shy (of)’ belong to this group: (16) Kuŋa nakata-č ŋel-re-n. child bear-instr fear-nf-3sg ‘The child was afraid of the bear.’ Some other emotion verbs (like as- ‘be angry at’, tiku- ‘be angry’, nab- ‘be upset (about)’, and öreŋči- ‘rejoice at’) may also take the stimulus in the instrumental case; however, they allow an alternative marking of the stimulus by the allative case (see 5.4. on the instrumental-allative alternation). Thus, verbs of emotion are heterogeneous syntactically: some take the instrumental pattern, others alternate between instrumental and allative, while still others like ajav- ‘like’ are transitive. Note that verbs of cognition like ha- ‘know’ and d’oŋči- ‘remember’ as well as (resultative) perception verbs like it- ‘see’ and dolda- ‘hear’ also take a transitive pattern.

4.4.2.4 Verbs with spatial case frames Verbs of motion and (change) of location fall into several classes which are diagnosed by the availability of relativization for a particular argument. Thus, em‘come; go to’ is goal-oriented, while hör- ‘go; leave’ is source-oriented as shown by results of relativization tests: em-če-v d’u [come-part-1sg.poss house] can only be interpreted as ‘the house to which I came’ (cf. (19) below), while hör-če-v d’u [gopart-1sg.poss house] can only be interpreted as ‘the house from which I went’ (cf.

580

Andrej Malchukov and Igor V. Nedjalkov

(20) below). Thus, relativization allows us to distinguish three main groups of change of location verbs: (i) goal-oriented; (ii) source-oriented; and (iii) non-oriented motion verbs which take a route or place argument.

4.4.2.4.1 The case frame The verbs of static location including bi- ‘be’, mevd’en- ‘live somewhere; be settled’, and teget- ‘sit’ belong to this group; they take an argument referring to the place of location. (17) Hurken bilek-le bi-s-ni. boy village-loc be-nf-3sg ‘The boy lived in the village.’ Some of these verbs also allow the use of dative along with the locative for the place of location, but this alternation will not be considered here.6

4.4.2.4.2 The case frame Verbs of non-oriented motion (i.e. which are neither source-oriented nor goal-oriented), including tut- ‘run’, take the object in a prolative case, which refers to the route of motion. (18) Kuŋa avlan-duli tut-te-n. child plain-prol run-nf-3sg ‘The child runs on/across the plain.’ The prolative case can be replaced by the locative -lA if the general location rather than the route of (unidirectional) motion is intended.

4.4.2.4.3 The case frame This case frame is found with a fairly large group of verbs which take a goal argument in the allative case (such as em- come’, i- ‘enter’, and muču- ‘return’). (19) Bej d’u-tki em-re-n. man house-all come-nf-3sg ‘The man came to the house.’

6 The locative ~ dative alternation with some Locations and Goals depends on topological properties of the argument and discourse factors (see Malchukov 1995: 14, Malchukov 2008: 73) and will not be discussed here since it is not relevant for issues of valency classification.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

581

These verbs allow for an alternation of allative with locative to indicate actual attainment of goal, as illustrated in 5.4.1.

4.4.2.4.4 The case frame This case frame is found with hör- ‘go; leave’ and a few other verbs (n’ö- ‘go out’, n’ön- ‘run away; flee’) which take a Source-of-motion argument in the ablative case: (20) Bej d’u-duk hör-re-n. man house-abl go-nf-3s ‘The man went from the house.’ The ablative case alternates with the elative case when the direction rather than actual source of motion is implied (see § 5.4.2 on this alternation).

4.4.2.5 Verbs with sentential actants Verbs of cognition and perception, like ha- ‘know’ and it- ‘see’, can also take a sentential argument (as in (21b)), alternating with the nominal argument (as in (21a)). The dependent clause takes the participial form and is marked with the accusative case. The subordinate subject is unmarked if nominal, but is expressed by a possessive pronoun if pronominal; in both cases, it is cross-referenced by possessive agreement on the participle. (21) a. Bej-u ha-ra-m. man-acc know-nf-1sg ‘I know/saw the man.’ b. Bej em-če-ve-n ha-ra-m. man come-perf.part-acc-3sg.poss know-nf-1sg ‘I know/saw that the man returned.’ We will not have much to say about this pattern as this article does not focus on sentential complements, and, furthermore, these verbs can also combine with nominal arguments (as in (21a)).

4.5 Trivalent verbs Most trivalent verbs can be seen as causative counterparts of bivalent verbs; some of these also include (non-productive) causative markers. This is especially true of caused motion verbs. Apart from the productive causative form, they frequently

582

Andrej Malchukov and Igor V. Nedjalkov

make use of the non-productive (lexical) causative in -u- (cf. em- ‘come’ > em-u ‘bring’, muču- ‘return (intr.)’ > muču-v- ‘return (tr.)’, hör- ‘leave’ > hör-u ‘lead away’, etc.). Derived verbs preserve their valency pattern so that the first two verbs are goal-oriented, while the third one is source-oriented (as diagnosed by the relativization test).

4.5.1 The case frame This class of verbs involves canonical ditransitives like bö- ‘give’, ič-uken- ‘show’ but also includes some semantically similar verbs such as ulit- ‘feed’: (22) Hurken kniga-v hupkučimŋe-du bö-n. boy book-acc teacher-dat give-nf.3sg ‘The boy gave the book to the teacher.’ (23) Hurken kniga-v hupkučimŋe-du ič-uken-ni. boy book-acc teacher-dat see-caus-nf.3sg ‘The boy showed the book to the teacher.’ Note that the latter verb is a morphological causative; as explained in 5.1, morphological causatives of transitives regularly follow this pattern. This class of verbs is not formally differentiated from the class of verbs with a benefactive adjunct in the dative case, although dative arguments and adjuncts show a distinct syntactic behavior (the latter are inaccessible to relativization). Note, however, that addressees take the allative rather than the dative case. Caused motion verbs based on goal-oriented intransitives also follow this pattern as long as the goal is animate. If the goal is inanimate, however, they follow the locative pattern, frequently in alternation with the allative one (see 5.5.5 below)

4.5.2 The case frame To this group belong the verbs of impact taking an instrumental object, including iet- ‘hit’, mine- ‘cut’, and also hon- ‘chop’ and nemke- ‘shoot with’. With other verbs like čelge- ‘break’ or ma- ‘kill’, the Instrument is an adjunct, as revealed by inaccessibility to relativization. (24) Hurken kileb-u hirka-n’ mine-n. boy bread-acc knife-instr cut-nf-3sg ‘The boy cut the bread with the knife.’

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

583

A different group of verbs formally following the same pattern are verbs like gerbet‘call’ (with this verb, however, it is less clear that the instrumental NP is a complement; it is obligatory, yet the usual tests do not apply as it is non-referential). (25) Bej hurke-m nuŋu-č gerbet-te-n. man boy-acc fool-instr call-nf-3sg ‘The man called the boy a fool.’

4.5.3 The case frame This case frame is found with verbs of deprivation such as ga- ‘take’ and d’ormi‘steal’: (26) Hurken kniga-v hupkučimŋe-duk d’ormi-n. boy book-acc teacher-abl steal-nf.3sg ‘The boy stole the book from the teacher.’ Verbs of removal such as atal- ‘tear’ formally belong to this class also. These verbs might be regarded as a causative counterpart to intransitive source-oriented motion verbs (and some of them are the lexical causatives of corresponding intransitives: cf. hör- ‘leave’ → hör-u- ‘bring away’). Most verbs of removal (such as tin- ‘let go; release’ and ilbe- ‘drive away (a reindeer herd)’, but not atal- ‘tear’), show a partially distinctive behavior from verbs of dispossession in that they also allow an elative pattern, like their intransitive counterparts.

4.5.4 The case frame Verbs like emen- ‘leave’ and bak- ‘find’ belong to this class and take the oblique argument in locative (with some objects, also the dative case): (27) Mut asa-l-bu d’u-la emen-i-tan. we woman-pl-acc house-loc leave-pst-3sg ‘We left the women at home.’ A different group of verbs displaying a similar case frame are goal-oriented caused motion verbs with the Goal in the locative case. While most of these verbs take a locative frame in alternation with the allative frame (see 5.5.1 below), a few verbs (intrinsically denoting a resultative motion, such as ne- ‘put’ and in-u- ‘load’) take only the locative pattern:

584

Andrej Malchukov and Igor V. Nedjalkov

(28) Bej hut-i turki-le in-u-n. man child-refl sledge-loc load-caus-nf.3sg ‘The man put the child on the sledge.’

4.5.5 The case frame This case frame is found with Goal-oriented verbs of caused motion (such as emu‘bring’ and ŋal- ‘carry’): (29) Hurken kniga-v škola-tki emu-n. boy book-acc school-all bring-nf-3sg ‘The boy brought the book to school.’ Like Goal-oriented intransitive verbs, caused motion verbs also allow locative in alternation with allative on the Goal if the Goal is attained (see 5.5.1). The only verbs that only admit the allative case are verbs of speech, including gön- ‘tell; say’ and ukčen- ‘tell’: (30) Bej hurken-teki ad’it-u gön-ni. man boy-all truth-acc say-nf.3sg ‘The man told the truth to the boy.’ (31) Bej hurken-teki ukčenek-u ukčen-ni. man boy-all story-acc tell-nf.3sg ‘The man told the story to the boy.’ Verbs of sound omission (such as irkan- ‘shout) also take an allative addressee, but with these verbs, the message is normally not expressed, so they do not count as trivalent. As in other languages, the distinction between transitive and ditransitive is partly blurred for verbs of speech and related semantic types.

4.6 Non-verbal predication Non-verbal predication is based on the copula verb bi- ‘be’ (which may be absent with 3rd person subjects) and o- ‘become’: (32) Bi bujusemŋe bi-se-m. I hunter be-nf-1sg ‘I am a hunter.’

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

585

4.7 Concluding remarks As is clear from this presentation, syntactic classification through coding frames only partially reflects the semantic classification as many of the patterns are semantically heterogeneous. Nonetheless, many verbs are able to take alternative frames, which helps to disambiguate valency patterns and arrive at more coherent semantic classes. Those (uncoded) alternations that are more productive will be introduced in the next section.

5 Uncoded alternations (“Case alternations”) Case alternations play a limited role in Even, which is expected given that Even has a rich array of both valency-increasing and valency-reducing operations, especially in the domain of subject alternations (causative-anticausative alternation, etc). In this section we will briefly discuss, in this order, subject alternations (NOM/ DES alternation), followed by (direct) object alternations (ACC/REFL alternation and ACC-DES alternation), an oblique object alternation (instrumental-allative alternation), spatial object coding alternations, and object-rearranging alternations (possessor raising alternation and dative designative alternation).

5.1 The nominative-designative alternation This alternation is primarily found with the negative existential predicate ačča and can be regarded as the only variety of ‘differential subject marking’ in Even: (33) D’u-0-̸ v ačča. house-nom-1sg absent ‘I have no house.’ (34) D’u-ga-ku ačča. house-des-1sg absent ‘I have no house; I am homeless.’ (Lit. ‘There’s no house for me.’) Since the designative case is also found on objects in addition to intransitive subjects, Kazama (2012), in his overview of the designative case in Tungusic languages, qualifies this pattern as ergative.7 It can also be characterized as “unaccu-

7 “The designative case not only encodes objects ‘which are newly achieved’ but also intransitive subjects ‘which are newly brought into existence’. That is, the designative case encoding follows ergative pattern.” (Kazama 2012: 142).

586

Andrej Malchukov and Igor V. Nedjalkov

sative” given that it applies to only some intransitive verbs (generally non-agentive, but primarily existence verbs). In some dialects, it can also be extended to other verbs in a presentative construction: (35) Kuma-0̸ hie-n. seal-nom appear-nf.3sg ‘A seal appeared.’ (36) Kuma-ŋ-ga-k hie-n. seal-al.poss-des-1sg appear-nf.3sg ‘A seal appeared for me. (that is, to my benefit)’

5.2 The accusative-reflexive alternation As noted in § 4.4.1 above, the accusative-reflexive alternation is syntactically conditioned: the reflexive possessive form routinely replaces the accusative case when the subject is coreferential with the possessor of the direct object: (37) Etiken oro-m ma-n. old_man reindeer-acc kill-nf.3sg ‘The old man killed the reindeer.’ (38) Etiken (men) or-mi ma-n. old_man (self’s) reindeer-refl.poss.sg kill-nf.3sg ‘The old man killed his (own) reindeer.’ The alternation is regular with two qualifications. First, with verbs requiring a body part as a direct object (like nimru- ‘close (eyes); blink’ and nöŋke- ‘nod’), marking with a reflexive-possessive form is virtually the only possibility. (39) Kuŋa iasal-i nimru-n. child eyes-refl.poss close-nf.3sg ‘The child blinked/closed his eyes.’ Since very few verbs (only nimru- ‘close (eyes); blink’ on the Leipzig Verb List) show this pattern, it has been assimilated to the NOM-ACC pattern in this paper (see also Footnote 5 above). We will return to discussion of verbs with an “internal object” in 6.4 as they show a distinctive behavior in the derivation of resultatives (see also the resultative marker in the examples above). Second, the reflexive form is infelicitous on the direct objects of verbs of creation; in that function, the designative case should be used instead (compare (44) below with (40), the latter being of questionable grammaticality):

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

587

(40) ??Bej d’u -j o-ča. man.nom house-refl.poss make-perf ‘The man built a house (for himself).’ Thus, accusative-reflexive and accusative-designative alternations can partition the transitive verbs into finer subclasses, roughly into transitive verbs with “affected” vs. “effected” (i.e. created) objects.

5.3 The accusative-designative alternation The designative case replaces the accusative case when the possessor of the direct object is the beneficiary of the action. The designative case obligatory combines with the possessive suffixes. If the possessor/beneficiary is further coreferential to the subject, reflexive-possessive forms are used (as in (44) below); otherwise corresponding personal possessive forms are used: (41) Etiken oro-m ma-n. old_man reindeer-acc kill-nf.3sg ‘The old man killed a reindeer.’ (42) Etiken [hin or-ŋa-s] ma-n. old_man your reindeer-des-2sg.poss kill-nf.3sg ‘The old man killed a reindeer for you.’ The accusative-designative alternation as described in this section is valency-preserving: although the designative construction introduces a beneficiary, the beneficiary is expressed as a possessor, so the overall number of arguments remains unchanged. However, with ditransitive verbs, the designative alternations can arguably reduce verbal valency. For this reason we will treat the designative alternation as a separate alternation (the ‘dative-designative alternation’ in 5.7), even though formally we are dealing with the same alternation. The accusative-designative alternation applies regularly to transitive verbs. Unlike the reflexive form, it is perfectly felicitous with verbs of creation. In fact, the designative form is more common with these verbs: (43) Bej d’u-v o-ča. man house-acc make-perf ‘The man built a house.’ (44) Bej d’u-ga-j o-ča. man house-des-refl.poss make-perf ‘The man built a house for himself.’

588

Andrej Malchukov and Igor V. Nedjalkov

With some verbs, this alternation is used to disambiguate verbal meanings; thus, duru- ‘burn’ is used with an accusative or reflexive form when combined with the patient (cf. mo-v duru-n ‘wood burn’) but with the designative case when combined with the result argument (cf. tog-ga-j duru-n [fire-des-self burned] ‘he made a fire’).

5.4 The oblique object alternation With a small group of bivalent intransitives referring to emotion (e.g., as- ‘be angry at’, tiku- ‘be angry’, nab- ‘be upset (about)’, öreŋči- ‘rejoice at’), the stimulus argument can be marked with the allative or instrumental case: (45) Bej min-teki~ mine-č as-sa-n. man i-all ~ i-instr get_angry-nf-3sg ‘The man got angry with me/at me.’ This alternation might reflect different conceptualizations of the stimulus argument – as the source of emotion (as is typical with emotion verbs; see (16) above), or as a goal of an emotional attitude.

5.5 Spatial case alternations These are found with motion verbs, which may take two different forms for encoding Goals and two forms for encoding Sources.

5.5.1 Goal alternation ( ~ ) For Goal marking, either Allative (-tki) or Locative (-la) is used, the former marking the direction, the latter an achieved Goal: (46) D’u-tki em-re-n. house-all come-nf-3sg ‘(S/He) came to/towards the house.’ (47) D’u-la em-re-n. house-loc come-nf-3sg ‘(S/He) came (in)to the house.’ The same difference carries over to verbs of caused motion (like em-u- ‘bring’), which may also take two different forms of Goal arguments under the same conditions:

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

589

(48) Hurken kniga-v škola-la/škola-tki emu-n. boy book-acc school-loc/school-all bring-nf.3sg ‘The boy brought the book to school.’ Importantly it is not found with ditransitives (transfer of possession like bö- ‘give’), which always select for the dative Recipient.

5.5.2 Source alternation ( ~ ) Similarly to the Goal alternation, there are two different ways of coding the source of motion: the elative case in -gič vs. the ablative in -duk. Elative is used instead of ablative where the direction rather than the actual source is implied (mirroring a similar contrast with Goals): (49) D’u-gič hör-re-n. house-elat go-nf-3sg ‘(S/He) came from the house.’ (50) D’u-duk hör-re-n. house-abl go-nf-3sg ‘(S/He) came out of the house.’ Under similar conditions, the Source alternation is found with Source-oriented transitive verbs such as tin- ‘let go; release’ and ilbe- ‘drive away (a reindeer herd)’, mentioned in § 4.5.3 above. As noted above, exempted from this alternation are verbs of dispossession (like ga- ‘take’), which only allow for the ablative form.

5.6 Possessor Raising ‘Possessor Raising’ (or ‘Possessor Ascension’) is a term used for an alternation where a possessor can be expressed NP-externally as a clausal constituent with an oblique marker (cf. Mary hit Joe’s arm / Mary hit Joe on the arm). A similar alternation is possible in Even with verbs of obtaining (including ga- ‘take’ but also d’av‘catch/snatch’) and verbs of impact (like iet- ‘hit’, na- ‘hit the target’, etc.). Notably, these two groups of verbs select for different cases of the affected body part – dative for verbs of obtaining, and locative for verbs of impact: (51) a. [Bej ŋal-du-n] ga-d-ni. man (poss) hand-dat-3sg took.nf.3sg ‘(S/He) took the man by the hand.’

590

Andrej Malchukov and Igor V. Nedjalkov

b. [Bej-u] [ŋal-du-n] ga-d-ni. man-acc hand-dat-3sg took.nf.3sg ‘(S/He) took the man by the hand.’ (52) a. [Bej dil-la-n] iet-te-n. man.poss head-loc-3sg hit.nf.3sg ‘(S/He) hit the man on the head.’ b. [Bej-u] [dil-la-n] iet-te-n. man-acc head-loc-3sg hit.nf.3sg ‘(S/He) hit the man on the head.’

5.7 The dative-designative alternation This alternation is, in a sense, the opposite of possessor raising as it involves “indirect object lowering” (in the terminology of Croft 1985; cf. Creissels 1979) rather than “possessor raising”. Formally, the dative-designative alternation is a variety of the accusative-designative alternation (discussed in 5.3 above), yet it is different in that it arguably performs a valency-decreasing function with ditransitive verbs. The dative-designative alternation involves ditransitive verbs with a recipient or benefactive argument. The regular way of expressing a recipient or beneficiary is through the dative case, but there is also an option of expressing it as a possessor of the object. In the latter case the object (the possessed) takes the designative case. (53) a. Hurken [hin-du] [turki-v] emu-n. boy you-dat sledge-acc brought.nf.3sg ‘The boy brought a sledge to/for you.’ b. Hurken [hin turki-ga-s] emu-n. boy your sledge-des-2sg.poss brought nf.3sg ‘The boy brought a sledge to/for you.’ The two constructions differ in their distribution insofar as the designative construction is used more restrictedly – it can be used to encode Beneficiaries in a transitive clause but not in an intransitive one (*sing for him).

6 Voice alternations and valency change (verb-coded alternations) As briefly discussed in § 2.4, the main valency-changing markers (verbal suffixes) are as follows: causative, adversative passive, mediopassive, reciprocal, and result-

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

591

ative. It is not always straightforward to characterize valency-changing markers in terms of valency-increasing vs. valency-decreasing. Some voice-markers, like causative, are indeed primarily valency-increasing, and some others, like mediopassive and reciprocal, are predominantly valency-decreasing. However, the adversative passive may either increase or decrease verbal valency, while the resultative may either decrease valency or leave it unaffected. The presentation here generally proceeds from valency-increasing to valency-decreasing voice alternations.

6.1 Causative The causative formation (the marker -vkAn- and its allomorphs) is equally possible for intransitive and transitives. In the former case, the verb becomes expectedly transitive, the causee taking the accusative case (cf. miltere-mken- ‘fill’ from milteren- ‘be full’): (54) Bej turki-v mo-č miltere-mken-ni. man sledge-acc wood-instr fill-caus-nf.3sg ‘The man filled the sledge with wood.’ The same is true of bivalent intransitives (such as ŋel- ‘fear’ with the instrumental pattern), which yield a transitive pattern when causativized (cf. ŋel-uken‘frighten’). With causatives of intransitives, case marking does not depend on the variety of causative involved, e.g., factitive/coercive (make-causatives), or permissive (letcausatives). Things are different with causatives of transitives. When derived from transitives, the causee usually appears in the dative case. (55) Hurken kniga-v hupkučimŋe-du ič-uken-ni. boy book-acc teacher-dat see-caus-nf.3sg ‘The boy showed the book to the teacher.’ However, the causee may take the accusative instead of the dative case if a factitive/coercive meaning is implied (see Malchukov 1995; (56b) originally from Novikova 1980): (56) a. Eve-sel Kad’d’ak-tu mine-v kol-ukan. Even-pl Kad’d’ak-dat wine-acc drink-caus.nf.3pl ‘Evens let/made Kad’d’ak drink the wine.’ b. Eve-sel Kad’d’ak-u mine-v kol-ukan. Even-pl Kad’d’ak-acc wine-acc drink-caus.nf.3pl ‘Evens made Kad’d’ak drink the wine.’

592

Andrej Malchukov and Igor V. Nedjalkov

On the whole, the causative formation makes a strict partition between transitive and intransitive verbs; it is less sensitive to the semantic subclasses of intransitives and transitives.

6.2 Adversative passive Adversative passive is peculiar syntactically in that it can fulfill both a valencyincreasing and a valency-decreasing function. The common denominator of the adversative form in both uses is that the subject of the adversative form (which may, but need not correspond to the P argument of the underlying verb) is adversely affected. Its use in a valency-decreasing function is illustrated below; the basic P surfaces as the subject of the adversative form, while the basic A (the subject in (57a)) is expressed by an adjunct in the dative case (in (57b)): (57) a. Nugde etike-m ma-n. bear old_man-acc kill-nf.3sg ‘The old man killed the bear.’ b. Etiken nugde-du ma-v-ra-n. old_man bear-dat kill-ad-nf-3sg ‘The old man was killed by the bear.’ This construction looks like a canonical passive, yet it cannot be used with verbs implying a positive effect on the P (like bel- ‘help’, etc.). Further, the same construction need not reduce the verbal valency, as in the following example, where the subject of the adversative form is the possessor of P rather than P itself: (58) Etiken nugde-du gia-j ma-v-ra-n. old_man bear-dat friend-refl.poss.sg kill-ad-nf-3sg ‘The bear killed the old man’s friend. (the old man was negatively affected)’ In the following examples, the subject of the adversative form (as in (59b)) is identical to the possessor of the intransitive subject (as in 59a)): (59) a. Huličan bödele-n ene-l-re-n. fox feet-3sg.poss hurt-inch-nf-3sg ‘The fox’s paws began to hurt.’ b. Huličan bödel-i ene-le-v-re-n. fox feet-refl.poss.sg hurt-inch-ad-nf-3sg ‘The fox’s paws began to hurt; it was negatively affected.’

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

593

It might even increase the argument structure, as is most clear with atmospheric verbs which are otherwise avalent syntactically (take a cognate subject); cf. (6) above and (60): (60) Etiken uda-la-v-ra-n. old_man rain-inch-ad-nf-3sg ‘The old man is caught by the rain.’ Thus, the adversative formation cuts across the classes of transitive vs. intransitive verbs. It is more sensitive to the animacy of an argument and its ability to be construed as adversely affected than to transitivity per se. (More discussion of adversative passives in Even can be found in Malchukov 1993).

6.3 Mediopassive The main functions of the “mediopassive” marker -b-/-p- are: a) anticausative; b) agentless passive (“quasipassive” in Geniušiene 1987’s terminology); c) stative passive; and d) “middle” (alias ‘potential passive’ or ‘facilitative passive’); cf. aŋa-bta-n [open-med-nf-3sg] ‘It opened/was opened/is open/can be opened’. With a few verbs (of negative impact) it can also perform a reflexive function (cf. ma- ‘kill’ → ma-b- ‘be killed’ but also ‘get hurt’). The choice of interpretations is dependent on the aspectual properties and transitivity and, hence, also indirectly on valency classes. The general rules can be summarized as follows: – perfective accomplishments in the mediopassive form have an anticausative or an agentless passive interpretation: (61) Urke aŋa-p-ta-n. door open-med-nf-3sg ‘The door opened/was opened/is open.’ –

activities, are preferentially interpreted as potential/middle forms (the agentless passive interpretation is less likely here):

(62) Ulre e-s-ni mine-p-te. meat neg-nf-3sg cut-med-conneg ‘The meat does not cut (easily).’ –

for intransitives, the potential interpretation is the only possibility:

(63) Ele (ajič) nulge-p-te-n. here well nomadize-med-nf-3sg ‘It is good (convenient) to nomadize here.’

594

Andrej Malchukov and Igor V. Nedjalkov

Note that when derived from intransitives (as in (63)), the mediopassive construction is impersonal. Impersonal uses of mediopassives derived from transitive verbs are also possible but occur rarely. So the ‘personal’ passive with the Patient in the nominative case (as in (62)) is definitely preferred to the impersonal one with the Patient in the accusative. Mediopassive differs syntactically from the adversative passive insofar as the agent cannot be expressed. Also semantically an agent can be absent (resulting in the anticausative meaning) unless the verbal meaning suggests otherwise.

6.4 Resultative The term resultative is used here for a verbal form referring to a resultant state of a verbal action (Nedjalkov 1988). In Even, the resultative form in -t-/-č- when applied to intransitives refers to a state of the underlying subject (‘S-resultative’). (64) a. Bej (tör-le) teg-re-n. man (ground-loc) sit_down-nf-3sg ‘The man sat down (on the ground).’ b. Bej (tör-le) teget-t-te-n. man (ground-loc) sit_down-res-nf-3sg ‘The man sits (on the ground).’ In this case, verbal valency does not change. When applied to transitives, the resultative form most usually pertains to the state of the underlying object (P-resultative or stative passive). (65) a. Bej učiki-j (hiakita-du/la) ön’e-n. man reindeer-refl.poss.sg tree-dat/loc tie-nf.3sg ‘I tied the reindeer to the tree’ b. Učiki-u ( hiakita-du/la) ön’e-t-te-n. reindeer-1sg.poss tree-dat/loc tie-res-nf-3sg ‘My reindeer is tied to the tree.’ This is in accordance with the general tendency of resultatives to pattern “ergatively”, taking an S/P subject, as observed by Nedjalkov (1988) and others. However, in contrast to Evenki (Nedjalkov 1992) and most other Tungusic languages, resultatives built on transitives usually also allow a transitive A-diathesis with a meaning ‘keep P in a state V’: (65) c. Bej učiki-j (hiakita-du/la) ön’e-t-te-n. man reindeer-refl.poss.sg (tree-dat/loc) tie-res-nf-3sg ‘The man keeps his reindeer tied to the tree.’

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

595

Malchukov (2008) suggests that a preference for A-resultatives, which is more pronounced in Even (as compared to Evenki, see 9.4.3), correlates with functional transitivity features. In the present context, what is most relevant is that A-resultatives predominate with semi-transitive verbs like d’on- ‘recall’ (cf. doŋ-či- ‘remember’) and verbs with an ‘internal object’ related to a body part, where the change of the state or pose of the A is at issue (cf. nimru-t- ‘hold (eyes) shut’ in (39) above and (39′) below). (39′) Kuŋa iasal-i nimru-t-te-n. child eyes-refl.poss close-res-nf-3sg ‘The child blinks/holds his eyes shut.’ Resultatives are semiproductive: they apply only to telic verbs, primarily verbs implying a change of spatial configuration (for intransitive resultatives, this is mostly a change of pose). Beyond this group, the form in -č-/-t- may also be found, but it will have a different – durative/progressive or multiplicative – interpretation (see Malchukov 2008: 103).

6.5 Reciprocal The verbal reciprocal form in -mAt- can be used to indicate cross-coreferentiality between two animate arguments. The arguments may correspond to the subject and the direct object (cf. ma- ‘kill’ → ma-mat- ‘kill each other’) in a ‘direct reciprocal’ construction: (66) a. Hurken nö-j čor- ča. boy brother-refl hit-part ‘The boy hit (with his fist) the (younger) brother.’ b. Akanur čor-ma-ča-l-ča-l. brothers hit-recp-inch-part-pl ‘The brothers fought (with fists).’ However, the subject of the reciprocal construction may also correspond to an animate oblique object of an intransitive verb (‘oblique reciprocals’): (67) a. Bej nakata-č ŋel-re-n. Man bear-instr fear-nf-3sg ‘The man is afraid of the bear.’ b. Bej-il ŋel-met-te. man-pl fear-recp-nf.3pl ‘The men are afraid of each other.’

596

Andrej Malchukov and Igor V. Nedjalkov

Further, the subject can correspond to an animate indirect object of a trivalent verb (‘indirect reciprocals’, in Nedjalkov’s 2007 terms), as is common with semantic ditransitives like bö- ‘give’ (Malchukov 2007; (68b) originally from Novikova 1980: 136): (68) a. Oroč tara-v oro-m nimek-tu borit-ti-n. even that-acc reindeer-acc neighbor-dat give.a.share-past-3sg ‘The Even gave that reindeer to the neighbor as his share.’ b. Oroči-l tara-v oro-r-bu mer do-li even-pl that-acc reindeer-pl-acc among _themselves borit-mat-ti-tan. give.a.share-recp-past-3pl ‘Evens divided these reindeer among themselves.’ Importantly, only arguments, not adjuncts, can be coded by verbal reciprocals (Malchukov 2007). Thus, with verbs of transfer of possession (ditransitives like bö‘give’), the reciprocal form can be used to cross-reference the Recipient/Beneficiary (cf. bö-met- ‘give to each other’). With verbs of dispossession (like ga- ‘take’ and d’örmi- ‘steal’), by contrast, this form indicates cross-coreferentiality of the agent with a malefactive source argument, rather than with a benefactive adjunct (gamat- ‘take from each other’, but not ‘take to/for each other’): (69) a. Etiken nimek-tuk-i olra-v ga-vat-ta-n. old_man neighbor-abl-refl fish-acc take-iter-nf-3sg ‘The old man usually takes fish from his neighbor.’ b. Nimeke-l olra-v ga-mač-čot-ta. neighbor-pl fish-acc take-recp-iter-nf.3pl ‘The neighbors usually take fish from each other.’ (not: ‘for each other.’) Thus, reciprocal formation shows the same argument sensitivity as observed above for relativization (see section 3 above). On the other hand, reciprocal formation can be used to diagnose valency status more restrictively as it involves only animate arguments. Apart from animacy, it seems that the semantic role is also important; thus, the verbal reciprocal cannot be used to indicate cross-coreferentiality of the agent with the goal, even if the goal is animate (Malchukov 2007). Compare the ungrammaticality of reciprocal forms from motion verbs, which are impossible even with animate Goals (cf. *em-met-te [come-recp-nf.3pl] intended: ‘come to each other’). On the other hand, with some motion verbs, this marker may have an aspectual multiplicative interpretation (cf. melu-met- ‘to jump (over sth) repeatedly; jump repeatedly’; cf. melu-n ‘to jump (over sth/out of sth)’).

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

597

There is another variety of reciprocal derivation which does not involve valency decrease: with ‘possessive reciprocals’ the Agent is coreferential with a possessor of an object rather than with the object itself. This type of reciprocal formation does not affect verbal valency but is also sensitive to the argument status of the head NP (adjunct NPs are outside the scope of reciprocal markers). See Malchukov 2007 for an extensive discussion of reciprocal constructions in Even.

6.6 Sociative Apart from the reciprocal voice in -mAt-, Even also features a “sociative voice” in -ldA-, which refers to a joint action of several subjects. Normally, sociative derivation does not affect verbal valency (cf. höre-lde-r [go-soc-nf.3pl] ‘go/leave together’). However, there is a small group of verbs mostly including ‘symmetrical predicates’ and similar verbs with which the sociative form can be used in the reciprocal function: (70) a. Bej bej-u kusi-n. man man-acc fight-nf.3sg ‘The man fights with another man.’ b. Bej-il kusi-lde-r. man-pl fight-soc-nf.pl ‘The men fight.’ These forms are, however, unproductive and quite often lexicalized (involving a change in meaning): cf. bak- ‘find’ → baka-lda- ‘meet (each other)’, it- ‘see’ → ičilde- ‘to meet’; gön- ‘to say, tell’ → gö-lde- ‘to come to agreement’. These forms usually allow for two versions (as in Russian), where symmetrical actants may be either in the subject position (and control verbal agreement), or else the second argument is expressed as a comitative object. In addition, a few forms allow the second argument to be expressed as the direct object (cf. bejil baka-lda-r ‘The men met (each other)’ ~ bi beju baka-lda-ram ‘I met the man (acc.)’).

7 Some other alternations This section deals with the availability of certain other case/voice alternations as familiar from other languages. In particular, it addresses the “inchoative-causative alternation”, where the directionality of alternation (valency-increasing or decreasing) is not clear, and it also comments on other kinds of object-demoting or objectrearranging alternations.

598

Andrej Malchukov and Igor V. Nedjalkov

7.1 The “inchoative-causative” alternation As noted above, the inchoative-causative alternation (The man broke the stick/The stick broke) is predominantly coded through the use of the mediopassive voice (emen- ‘leave sth’ → eme-p- ‘stay; be left’). For events which are less likely to be spontaneous, causative derivation is possible (cf. miltere-mken- ‘to fill’ vs. milteren‘get full’), etc. In addition, there are a number of non-productive (lexicalized) causative/anticausative alternations which might involve the following: a) a nonproductive causative marker (e.g., -u-: tet- ‘dress (intr).’ → tet-u- ‘put sth on sb’; un‘melt (intr.) → un-ŋi- ‘melt (tr.)); b) a non-productive anticausative (e.g., in -rga-; cf. kapta- ‘flatten (tr.)’ → kapta-rga- ’get flat’); c) equipollent alternations where both transitive and intransitive forms are marked (cf. egd’e-lte- ‘enlarge’ ~ egd’elbe- ‘get larger’). Especially interesting among the latter is a four-way alternation found with “destruction verbs” (frequently of ideophonic origin), specifying different ways of destruction (breaking, splitting, tearing, etc.): (71) a. čelgel- ‘break (of a hand, leg, etc)’ b. čelge-l- ‘break (pfv.tr.)’ → c. čelge-r- ‘break/get broken (pfv.intr.)’ d. čelge-k- ‘break (ipfv.tr.)’ → e. čelge-m- ‘break/be broken (ipfv.intr.)’

(72) a. tekel- ‘tear’: b. teke-l- ‘tear (pfv.tr.)’ → d. teke-k- ‘tear (ipfv.tr.)’ →

c. teke-r- ‘tear/get torn (pfv.intr.)’ e. teke-m- ‘tear/be torn (ipfv.intr.)’

As (71)–(72) show, many verbs which would be labile in other languages (cf. Haspelmath 1993) form equipollent pairs in Even; in addition, some such verbs with a general transitive-bias (like aŋa- ‘open (tr.)’ and öŋke- ‘spill’) form anticausative pairs through the use of mediopassive forms (cf. aŋa- ‘open (tr.)’ → aŋa-b- ‘open (intr.)’), while verbs with a general intransitive bias use a causative form (either a productive morphological causative or an unproductive lexical causative such as un- ‘melt (intr.)’ → un-ŋi- ‘melt (tr.)’).

7.2 Object-demoting/deleting alternations Generally, Even does not have a productive object-deleting alternation of the conative type (cf. the “conative alternation” in English of the type ‘hit the table’ ~ ‘hit at the table’) or the antipassive or incorporating type. Nonetheless, Even, like many other languages, features a number of object-deleting alternations restricted to certain verbs: a) “reflexive deletion”; b) “indefinite object deletion”; c) and “cognateobject construction”. (The terminology in this section partly follows Levin 1993.)

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

599

7.2.1 The “reflexive deletion” alternation This transformation, known as reflexive deletion (The man washed himself ~ The man washed), is found with a few verbs, mostly “natural reflexives”/grooming verbs such as av- ‘wash’ and tet- ‘dress’. Elsewhere, the pronominal reflexive form meni ‘oneself’ must be used. As noted above, reflexivity is not rendered through the use of a verbal form: mediopassive forms may only have this meaning with a few verbs of negative impact; cf. ma- ‘kill’ → ma-b- ‘be killed; also: hurt (oneself)’ and mine- ‘cut’ → mine-p- ‘be cut’ but also ‘inadvertently cut oneself’.8

7.2.2 The “object deletion” alternation The “object deletion” alternation (of the type: He ate the bread ~ He ate) is available for verbs having a cognate object: d’epten (d’eplev) ‘eat (food)’, (teti-v) tet-ten ‘put on (clothes)’, (ike-v) iken ‘sing a song’, (teleŋ-u) teleŋren ‘tell a story’, or (ukčenek-u) ukčenni ‘tell a story’ (in (31)).

7.3 Object-rearranging alternations Object-rearranging alternations are not productive in Even either; thus, one does not have any equivalent to the dative alternation (Mary gave the book to John ~ Mary gave John the book) or the benefactive alternation (Mary made some soup for John ~ Mary made John some soup). Normally, recipients and beneficiaries are expressed by the dative; alternatively, they can be expressed in the designative construction (see 5.7). We also find no direct correspondence for the “locative alternation” (John loaded the truck with hay / John loaded the hay on the truck) either; rather we find the lexical differentiation of verbs. Thus ‘load’ is rendered by either caused motion verbs (tev-, ne- ‘put’; also morphological causatives in-u- from in‘carry on one’s back’, see (28) above), which take a locative pattern, or, less frequently, ‘load’ is rendered by ‘fill’ verbs like miltere-mken- ‘fill’, which take the instrumental pattern (see (54) above). The latter pattern is also found with das‘cover’ as well as with many denominal verbs (meaning ‘ACT with X’) with an instrument (X) “incorporated”, such as gid-la- ‘hit with a spear’, čor-da- ‘hit with a fist’, hulda-la- ‘cover with a blanket’, etc. The instrument is usually covert, but, if overt, it must be expressed through the noun phrase in the instrumental case.

8 See Malchukov (1993, 1995), where it is argued that the valency-decreasing adversative construction (as in (57)) is derived from the valency-increasing one (as in (58)) through reflexive object deletion.

600

Andrej Malchukov and Igor V. Nedjalkov

7.4 Final remarks As is clear from the previous discussion, Even features subject-deleting alternations (through the use of mediopassive, etc.) rather than object-deleting or objectrearranging voice alternations. It also lacks the “subject-substitution” or “subjectinversion” alternations which we are familiar with from English. Thus, Even lacks the “instrumental subject alternation” illustrated by Levin (1993) with English examples such as John hit the fence with a stick ~ The stick hit the fence. Generally, constructions with inanimate transitive subjects that are Instruments rather than Agents (like *Hirkamu ajič ulrev minen ‘My knife cuts meat well.’) are ungrammatical in Even (Malchukov 2008: 66). Nor does Even feature “subject inversion” constructions, as attested in English (There comes a man) and also in some Bantu languages with “locative inversion”.

8 Valency classes in Even: concluding discussion Argument coding in Even is manifested primarily through case marking (indexing is confined to subject agreement and is therefore of little use for differentiating valency types in this language). Case marking in Even is mostly semantically transparent, with a few idiosyncrasies among “semitransitive” verbs, mostly psychverbs, some of which take the object (stimulus) in the instrumental or allative case. Verbs of pursuit (such as gelet- ‘search for’) are transitive. There is a distinction between ditransitives and caused motion verbs: only the latter display a spatial alternation (as found with intransitive motion verbs). Verbs of speech differ from ditransitive verbs, aligning themselves partially with verbs of caused motion (both select the allative case). Even makes limited use of case alternations, having recourse to voice alternations instead. Thus, subject alternations are coded through a mediopassive marker, which has different uses with different verb classes (and also depends on transitivity and aspect). There are no object alternations (like a ‘load’-alternation), either uncoded (as in English) or verb-coded (i.e. there are no applicative or directional markers). Rather, we are dealing here with a lexical split – the use of different verbs for different uses of ‘load’. The different uses are rendered either by caused motion verbs (taking a locative pattern) or through ‘affect’ verbs with the patient as the direct object. The latter verbs take the instrumental pattern, except for the verbs which have the instrument “incorporated” (rendered through the stem of a nominal verb; cf. hulra-la ‘cover with a blanket’ from hulra ‘blanket’, etc). Reciprocal forms are used fairly indiscriminately across verb classes insofar as the reciprocal form can mark cross-coreferentiality not only between verbal arguments but also between the subject and the possessor of the object. Nonetheless, the syntactic construction is different for different classes (e.g., for transitive and intransitive

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

601

verbs). Also, adversative passives may apply to both transitives and intransitives (provided that the derived subject is affected). Yet there are again structural differences between these constructions. Finally, resultative markers are semiproductive; they mostly occur with verbs denoting a change in location or state. With intransitives, resultative forms describe the state of the intransitive subject (S-resultatives). When built on transitives, resultative forms may be either P-resultatives (stative passive) or A-resultatives (the latter with less canonical transitives). Thus, although there are certain unusual types of verb-clustering due to the language-particular semantics of individual constructions used as diagnostics here (e.g., the semantics of the adversative passive construction), there are also some general tendencies discernable (as manifested in the similar behavior of certain verb classes cross-linguistically): – the mediopassive voice in -b- has an anticausative meaning, with accomplishments allowing for a spontaneous interpretation; elsewhere it has the resultative, agentless passive or middle-potential meaning (not unlike the reflexive -sja forms in Russian); – “natural reflexives” (grooming verbs like av- ‘wash’) allow for reflexive deletion (as in English); – “natural reciprocals” can be singled out by the possibility of using a sociative form in the reciprocal function (cf. baka-lda- find-soc- ‘meet’); – resultative forms are semi-productive and are found among subclasses of verbs of change of state/location (“concrete resultatives”); they may also be found with some less canonical transitive verbs, but in the latter case they have the function of A-resultatives rather than P-resultatives. Thus, one arrives at a finer classification once one takes syntactic and semantic properties of particular constructions into account. This is especially conspicuous in the case of the transitive construction, which has a fairly broad extension, but fewer canonical transitives will show a deviant behavior with respect to voice alternations or have alternative case patterns (case alternations). By way of exemplification, I represent extensions of several constructions over some verb types on the semantic map for the transitive-intransitive domain sketched in Malchukov (2005). This map includes only a few verb meanings from the Leipzig List and, furthermore, largely disregards the domain of three argument verbs (but see Malchukov et al. 2010 for a semantic map of the three argument domain centering on ditransitives). It includes canonical transitives (like ‘kill’ and ‘break’) as well as verbs of pursuit (‘follow’, ‘search for’) and interaction (‘help’, ‘succumb to’), “middles” (Kemmer 1993) covering the domain of “natural reflexives” (‘wash’) and “natural reciprocals” (‘fight’), affected agent verbs (‘eat’, ‘find’), verbs of perception (‘see’), cognition (‘know’), and emotion (‘fear’, ‘love’) and spontaneous verbs (‘burst’, ‘sink’) among others (see Malchukov 2005; cf. Tsunoda 1981; Kemmer 1993). On this map I indicate the transitive and intransitive coding

602

Andrej Malchukov and Igor V. Nedjalkov

frames (an overlap represents either an alternation or a lexical split). In addition, the domain of application of certain syntactic properties such as the P-resultative construction is shown. The map below reveals a (fairly large) transitive domain across the verb types, which extends quite far beyond the domain of canonical transitives (effected action verbs, in terms of Tsunoda 1981). Yet, for pursuit and interaction verbs, the dative object construction is attested for some verbs (huptu‘fall behind’, dabda- ‘lose to’), and with emotion verbs, the instrumental object construction is the dominant construction (see § 4.4.2.3). Middle verbs often alternate with a transitive pattern, either through a ‘reflexive-deleting alternation’ (grooming verbs), or a comitative alternation (“natural reciprocals” like baka-ldafind-soc- ‘meet’ in § 6.6). On the other hand, canonical transitives can be singled out by featuring the P-resultative diathesis (although it is not found with all the verbs).

contact

pursuit

motion

interaction Canonical Transitives middle

affected Agent

Legend:

perception

cognition

spontaneous

emotion

I n t r a n s i t i v e

sensation

transitive domain dative object construction instrumental object construction intransitive monovalent construction P-resultative construction

Fig. 1: Extensions of several Even constructions on the semantic map for the transitive-intransitive domain.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

603

9 Even valency classes in a comparative perspective 9.1 Introductory In this section we will briefly discuss Even valency classes from a comparativecontrastive perspective highlighting similarities and differences with respect to other Tungusic languages. Tungusic languages include about a dozen languages belonging to one of three branches (see Malchukov & Nedjalkov 2010 for a brief summary as well as Benzing (1955b) for more detail). For contrastive purposes, we will mostly focus on “major” Tungusic languages taking Evenki (Nedjalkov 1997; cf. Konstantinova 1964; Bulatova & Grenoble 1999) as representative for the North Tungusic branch, Nanai (Avrorin 1961) as representative for East Tungusic, and Manchu (Gorelova 2002; Avrorin 2000), as representative for South Tungusic; but occasional mention will also be made of other “smaller” Tungusic languages. The outline of exposition generally follows the outline in the Even part (§ 3–8): first we consider coding frames in a contrastive perspective in § 9.2; subsequent sections deal with uncoded alternations (case alternations) in § 9.3, and coded alternations (voice alternations) in § 9.4. The concluding section summarizes findings from this comparative study. What makes this study interesting is that different branches of Tungusic, in spite of uncontroversial genealogical relatedness, reveal considerable structural variation. While North Tungusic languages have typical agglutinating characteristics featuring rich morphology, South Tungusic languages show reduced morphology and increased isolating traits (presumably under the influence of Sinitic languages), and East Tungusic languages show an intermediate position in this respect. The question to be addressed is what consequences of these structural differences condition valency differentiation of the verbal lexicon.

9.2 Coding frames With respect to coding frames, we expect considerable differences in Tungusic languages as the inventory of case markers differs greatly across languages. Thus, while North Tungusic languages feature 12–14 cases (14 in Even, 12 in Evenki)9, in East Tungusic, the system is reduced (to 8–9 cases in Nanai dialects), and it is still more reduced in Manchu, which features only 5 cases (nominative, accusative, genitive, dative-locative, and ablative). (Note that in all Tungusic languages, nominative is unmarked.) The genitive case is absent in most Tungusic languages, ex-

9 Unless otherwise indicated, the Polygus dialect is taken as representative for Evenki and the Naikhin dialect for Nanai since the standard (“literary”) varieties are based on these dialects.

604

Andrej Malchukov and Igor V. Nedjalkov

cept for Manchu, since the possessive relations are head-marked. The following table (based on Sunik 1962: 160–161; and Benzing 1955b) summarizes case inventories in four major Tungusic languages, taking into consideration only cases directly pertaining to our study: Tab. 1: Case inventories of Tungusic languages. Even

Evenki

Nanai

Manchu

NOM (nominative)









ACC (accusative)

-v/-u/-m/-bu

-vA

-vA/-bA

-be

ACC 2 (designative, indefinite, accusative)

-gA

-jA

-gA



DAT (dative)

-du

-du

-du

-de

LOC (locative)

-lA

-lA

-lA



ALL (allative)

-tki

-tki

-či

– -či/-ci

ABL (ablative)

-duk

-duk

–10

Instr (instrumental)

-d’i

-d’i

-d’i



PROL (prolative)

-li

-li

(-li)



ELAT (elative)

-gič

-git

-diAdi



The differences between individual Tungusic languages can be summarized as follows: – Evenki has a case system comparable to Even, except for the fact that instead of the designative case we find indefinite accusative, which has a somewhat different distribution (see below); – The case system in Nanai is reduced: it distinguishes between two varieties of directional cases (cf. locative vs. allative) but does not distinguish between two cases related to Source (in the Naikhin dialect, only cognates of the elative are found) and also lacks a prolative (in the Naikhin dialect; some other dialects preserve the prolative case though); – Manchu further lacks a distinction between allative and locative. Moreover, it lacks a special instrumental case – in the latter functions, either a dative or the genitive is used. The origin of these discrepancies in the case systems of Tungusic languages is not totally clear. In one possible scenario, North Tungusic languages have grammati-

10 Ablative is absent in the Naikhin dialect but found in some other dialects.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

605

calized additional cases after the split from Proto-Tungusic. In another scenario, these discrepancies result from a reduction of morphology in Manchu (presumably under the areal influence of Chinese; Doerfer 1978: 7). In line with the second scenario, Benzing (1955b) points out possible cognates to productive case suffixes in North Tungusic which are fossilized in Manchu. Whatever the correct scenario (and it might well be that different positions are justified with respect to different casemarkers), here we will rather focus on the consequences of these differences for the issues of valency classification. As noted above, while North Tungusic languages show an expanded system featuring a variety of spatial cases, in Nanai other oblique cases beyond locative, elative/ablative, dative, and (for some dialects) prolative are lost; and in Manchu instrumental is also lacking, ablative is residual, and Location and Goal functions are taken over by the dative case. This has obvious consequences for valency classification. Thus, while the North Tungusic Even and Evenki distinguish between verbs taking a Source argument (like hör- ‘leave, go’), Goal-argument (em- ‘come’), static location (teget- ‘sit’), and route (ŋen- ‘walk’), the other Tungusic languages neutralize this distinction. This is most obvious for Manchu, which neutralizes the latter three groups (both take a locative argument in the dative-locative case), but this neutralization is also found in Nanai. In particular, Nanai uses the locative case in -lA much more freely than North Tungusic language, for different types of spatial arguments including Location, but also Goal, Source, and Route (Avrorin 1959: 182). Admittedly, this neutralization is only partial in Nanai as it also has dedicated cases (allative, ablative, and, in some dialects, also prolative) for the respective functions, which alternate with the locative. Thus, reduction in the case inventories in East Tungusic and, in particular, in South Tungusic expectedly leads to under-differentiation of verb classes. Indeed, one could argue that the dative case in Manchu approximates the function of the oblique case, given that the use of ablative and genitive for verbal arguments is highly restricted. Under these conditions, locative verbs taking the Location or Goal argument feature the same NOM-DAT frame as psych-verbs (like gele- ‘be afraid’) or interaction verbs (e.g., gečule- ‘make friends with’) taking the dative case for the stimulus (Gorelova 2002: 167).11 Similarly, for trivalent verbs, the absence of the instrumental case, which is replaced by dative, results in the same NOM-ACC-DAT pattern being used both for ditransitives (with Recipient in the dative case) and verbs of impact (with Instrument in the dative case). Yet, it should be noted that this neutralization of argument marking in East Tungusic and South Tungusic is partially resolved through alternations (see below). Thus, the reduction of case

11 On the other hand, some (bivalent) verbs taking an oblique argument are assimilated to transitives in Manchu, even though it is difficult to show this for cognates. Thus, both basu- ‘laugh at’ and sebjele- ‘rejoice’ take the Stimulus argument in the accusative case (Gorelova 2002: 167).

606

Andrej Malchukov and Igor V. Nedjalkov

paradigms as one moves from North Tungusic to East Tungusic and further to South Tungusic is not paralleled by a similar dramatic reduction in valency classes.

9.3 Uncoded alternations Above we considered two main types of uncoded alternations in Even: 1) Spatial alternations, including Goal, Source, and Location alternations (§ 5.4 above); 2) Object alternations, including Accusative-Designative alternations (see § 5.2) and Accusative-Reflexive alternations (§ 5.3). Similar alternations are found in other Tungusic languages, albeit to different degrees. Thus, the spatial alternations are also found in Evenki, although the markers are distinct (Nedjalkov 1997: 170–171). This distinction shows limited verb-sensitivity though, at least on the Leipzig Verb List. Other languages lack such an alternation since they have one spatial case for the Goal argument and one locative case for the Source argument. In addition the object alternations are not of equal importance for valency classification. Thus, an accusative-reflexive alternation is found in all Tungusic languages which have a special possessive-reflexive form (i.e. all languages except for Manchu), under similar conditions as in Even. That is, a reflexive form replaces the accusative if the possessor of the object is coreferential with the subject. Apart from the small set of verbs which only take a body part as their object (cf. Even nimru- ‘close (eyes); blink’ in (39)), this alternation is regular with transitive verbs and does not yield any subgrouping. More interesting is the accusative-designative alternation, which shows some variation across languages (see Kazama 2012 for a recent overview of designative case marking in Tungusic). Interestingly, Nanai, though belonging to the Eastern group, produces a better match to Even, as far as the form and the function of designative case are concerned, than the more closely related Evenki (see Table 1 above). In Nanai, the designative case in -gA- is obviously cognate to the Even form and is used only in conjunction with possessive suffixes. It is also similar to Even in that it is preferentially used with “effective” verbs with a result argument rather than “affective” verbs with a Patient argument (in Kazama’s (2012) count aŋgo- ‘make’ is the third most frequent verb taking the designative object in Nanai). Moreover, again as in Even (see § 5.1 above), the designative in Nanai can appear on some intransitive subjects (Kazama 2012). Manchu does not have a designative case; less expectedly, we do not find a designative case in -gA- in Evenki. Instead, Evenki features an “indefinite accusative” in -jA, which shows functional similarities to the designative case but is not restricted to possessive forms and can generally appear with indefinite/non-specific objects. The alternation between definite and indefinite ACC in Evenki is not verb-sensitive though; rather, transitive verbs may take either (definite) ACC -vA or indefinite ACC -jA depending on the properties of the object.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

607

(73) Evenki Beje muu-ve (tyge-du) uŋku-re-n. man water-acc (bowel-dat) pour-nf-3sg ‘The man poured the water (into the glass).’ (74) Muu-je uŋku-kel! water-acc2 pour-imper.2sg ‘Pour (some) water!’ In this respect, the alternation between the two accusative cases in Evenki is not unlike more conventional kinds of object alternation, as found in Manchu and Nanai, where the indefinite/non-specific object is left unmarked (formally ACC~NOM alternation; see Avrorin 2000: 79–85 on Manchu). This latter alternation, known under the label of Differential Object Marking (see e.g., Malchukov & de Swart 2009), is also common in Turkic and Mongolian languages (see Pakendorf 2007: 149–152 for discussion of this areal feature). This kind of differential object marking is not verb-sensitive but rather depends on the features of arguments. A more interesting alternation found in Manchu is an alternative marking of Instruments either by the dative case or the genitive case: (75) Manchu a. Beje-i gala-i gaj-su. your.own-gen hand-gen take-imp ‘Take with your own hand(s).’

(Avrorin 2000: 79)

b. Beje-i gala-de ǯafa-ha-bi. your.own-gen hand-dat take-past-aux ‘(He) took (it) with his own hand(s).’ The function of this alternation is not totally clear,12 yet it is relevant for verb classification insofar as it can be used to differentiate Instruments from other arguments in the dative case (Recipients, Goals, etc). Thus, the neutralization of verb classes conditioned by the absence of the instrumental case (a lack of distinct case frames for “giving” vs. “hitting” verbs) is disambiguated through the possibility of dative ~ genitive alternations with Instruments (but not with Recipients). There are some other uncoded alternations in Tungusic languages, such as alternations with ditransitive verbs discussed in Malchukov & Nedjalkov (2010). Thus, Evenki shows remarkable variation in the marking of recipients and addressees of verbs of speech, which may take the allative, dative, or accusative case. Yet,

12 Avrorin (2000: 79) suggests that genitive is used in an instrumental function when the action is not completed or is in the future, whereas dative is used if the action is completed.

608

Andrej Malchukov and Igor V. Nedjalkov

these alternations are not relevant for verb classification. In some cases, this variation may be due to a different conceptualization of the third argument (use of dative for Recipients and allative for Goals), and in some cases (use of dative with Addressees), case variation is arguably due to Russian influence.

9.4 Coded alternations Finally, let us consider verb-coded alternations, addressing in particular four voice alternations: a) Passive; b) Causative; c) Resultative; d) Reciprocal. We discuss the first alternation in some detail since it also provides an opportunity to raise some methodological issues relevant for valency classification. The passive alternation is of particular interest here because the cognates of the passive marker *-bu- show considerable functional differences across languages, which raises a number of general questions with respect to valency classification. The other three voice alternations will be considered in brief.

9.4.1 Passive The passive in *-bu- is found in all Tungusic languages but shows significant differences in function and distribution. As noted above for Even (in 6.2), the passive form in -v-/-m- is primarily used as an adversative passive, although it is also found in the causative function with few intransitives. It is also used as a conventional (valency-decreasing personal) passive, of the type illustrated in (57b) above, yet this function seems to be derivative from the first one. In particular, the conventional (valency-decreasing) passive as in (57b) is found with the same groups of verbs (primarily, verbs of adverse effect on the animate object, like ma- ‘kill’) as the valency-increasing adversative passive in (58). In Evenki, the situation is more complex. The suffix is used as a conventional passive with no adversative connotation (in (76a)), as an impersonal passive (in (76b)), and as a causative (in (76c)), anticausative (in (76d)), or adversative passive (in (76e)): (76) Evenki a. Er d’u tar beje-du oo-v-ča. this house that man-dat make-pass-past ‘This house is built by that man.’ b. Tar amut-tu ollomo-či-v-d’aŋa. that lake-dat fish-prf-pass-mod.part ‘It is possible to catch fish in that lake.’ c. Asi hute-vi suru-v-re-n. woman child-refl.poss.sg go.away-caus-nf-3sg ‘The woman led her child away.’

(Nedjalkov 2013)

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

609

d. D’av sukča-v-ra-n. boat break-pass-nf-3sg ‘The boat broke/ got broken (by itself).’ e. Bi udun-mu-0-̸ m. I rain-pass-nf-1sg ‘I was caught by the rain (and got wet).’ In contrast to Even, in Evenki, conventional passive seems to be the basic function of the -v-/-mu- marker. The adversative function is less characteristic but may be found, for example, with meteo-verbs (see (76e)). The causative function is predominantly found with intransitive motion verbs (like suru- ‘go’ in (76c)) but is more productive than in Even. In Even, the suffix -u-, historically cognate but synchronically distinct from the passive marker, is used in that function (cf. Evenki suru-v- and Even hör-u- ‘bring’). Similarly, Even does not use the passive -v-/-mmarker in the anticausative function but uses the mediopassive -b- instead (as in (61) above). The latter suffix may also be historically related to the passive marker (Cincius 1947: 180)13 but cannot be identified with the passive synchronically, while in Evenki, the two markers can be considered as allomorphs (cf. Evenki: ula- ‘make wet’ → ula-v-/ula-p- ‘become wet’). The same mediopassive marker is used in Even in impersonal constructions of the type exemplified above for Evenki (cf. (76b) and (63) above). Thus, a more restricted use of the Even passive when compared to Evenki is due to the fact that the passive marker split into three distinct markers (passive, anticausative, and mediopassive) in Even (these markers cannot be regarded as allomorphs of the same marker). In Nanai, the picture is different insofar as the passive marker is used exclusively in the impersonal passive construction: (77) Nanai Inda-sal-du dalom-ba bu-vu-ri. dog-pl-dat food-acc give-pass-pres.part ‘It is time (necessary) to give food to the dogs.’

(L. Zaksor; p.c)

For North Tungusic, such uses are also found (cf. (76b) from Evenki), but they are mostly restricted to particular (participial) forms (moreover, as mentioned above, for Even the mediopassive rather than the adversative passive is used). Finally, in Manchu, the use of passive is again different in that it is productively used in the causative function when occurring with intransitives and with a passive function when derived from transitives:14 13 Benzing (1955b: 1070), however, considers *-bu- and *-p- suffixes as historically distinct. 14 Nedjalkov (1993) performed a count on the use of the passive forms as recorded in Zakharov’s (1875) Manchu-Russian dictionary and found the following correlation between the transitivity of the base verb and the function of the -bu- marker: with intransitive verbs, it has predominantly

610

Andrej Malchukov and Igor V. Nedjalkov

(78) Manchu a. Bata i-mbe va-ha. enemy he-acc kill-past ‘The enemy killed him.’

(Nedjalkov 1991: 5)

b. Bata-be va-bu-ha. enemy-acc kill-caus-past ‘(He) made (somebody) kill the enemy.’ c. Bata-de va-bu-ha. enemy-dat kill-pass-past ‘(He) was killed by the enemy.’ While the passive use is unremarkable here and finds parallels in other Tungusic languages, the use of the passive marker in the causative sense with transitives is exceptional in other Tungusic languages. As mentioned below, other Tungusic languages use a complex marker -bu-kan- of a more recent origin in a regular causative function. Diachronically, these differences in the use of the causative/passive marker can be explained if one assumes along with much of the literature (see e.g., Zakharov 1875; Nedjalkov 1992; Malchukov 1993; Robbeets 2007; Li & Whaley 2012) that the original function of this marker was a causative one which developed into a passive function through an intermediate reflexive permissive stage. As repeatedly noted in the literature, causative suffixes in the Altaic languages (Turkic, Mongolic, and Tungusic) may perform a passive function (Kormushin 1976; Nedjalkov 1991); a similar pattern has been reported for the neighboring South East Asian languages (Yap & Iwasaki 1998). The causative function is still best attested in Manchu but has also left its mark in the adversative passive uses in other languages. In this scenario, the development of a passive function is a later development, accelerated through the rise of the new causative marker -bu-kan-, presumably the old causative-passive strengthened by the emphatic -kan- marker (Nedjalkov 1993; Li & Whaley 2012). The general path of development of the Tungusic passive can be represented as in Figure 2. The passive marker in Even can be taken to represent stage II of this development (adversative passive), given that it is restricted to verbs of adverse effect. As illustrated in (76), the Evenki passive performs all functions but is mostly found in function III (conventional passive). In Nanai, the passive marker is mostly found in the impersonal function (stage IVa). It is also reported in the anticausative func-

causative uses (86.1 %); only 4,6 % of verbs have a passive function, and 9,3 % have both a causative and a passive function. With transitive verbs, the pattern is different: 45.2 % of verbs allow only for passive uses, 24.8 % only causative uses, and the remaining 30 % both passive and causative uses.

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

611

Impersonal (passive)(IVa) Causative (I)

Adversative (passive) (II)

Passive (III) Anticausative (IVb)

Fig. 2: Development of the passive marker in Tungusic languages.

tion (stage IVb), which Avrorin (1961: 40–42) somewhat misleadingly calls passive, but such derivations are few and lexicalized. Finally, in Manchu, the passive marker is attested for stage I (causative) and stage III (passive) as well as for the intermediate adversative stage II (e.g., for passives from intransitives). As noted above, in the majority of Tungusic languages, the causative function has been renewed by the complex marker *-bu-kan-, of a more recent origin. Now, what are the consequences of such discrepancies in the functions of the passive marker for valency classification in Tungusic? If one takes a morphological stance and studies the distribution of the common Tungusic passive marker *-buacross Tungusic languages, the differences will seem enormous and are derivative from its function. Thus, in Evenki, the passive is mostly restricted to transitive verbs (apart from a few meteo-verbs, where it appears in an adversative function, and a few intransitive verbs of motion, where it appears with a causative function). In Even it is less restricted insofar as it is productively used with an adversative function in the case of intransitives. However, it is more restricted in another respect – it is found only in constructions with animate patients depicted as adversely affected (so verbs taking inanimate objects like anŋa- ‘open’ do not productively have this form). In Nanai, on the other hand, the form can be used without any restriction since it is used for impersonal passives with transitives and intransitives alike. Finally, in Manchu the form is also less restricted but for another reason – it is still productively used in the causative function when used with intransitives. Thus, differences across Tungusic appear to be pervasive. Nonetheless, if we take syntactic characteristics of the passive construction into account as well (i.e. the coding frame of the derived construction), then the differences are less dramatic. Thus, if we focus on the use of the conventional passive (with P in the nominative and A in the dative case), then this alternation will apply largely to transitives. Exceptionally, a few intransitives like udan-/udun‘rain’ and em-/eme- ‘come’ will be found in this frame in Even and Evenki (Nedjalkov 1993). For Nanai, which features only impersonal passives, this construction is not available at all and cannot be used as a diagnostic for verb classification. If one takes into further account a relation between the basic and derived frames (i.e. “diathesis”), the degree of convergence will increase accordingly since then the exceptional case of meteo-verbs would also be treated distinctly from transitives, reflecting differences in the basic frames. There may be residual differences, such

612

Andrej Malchukov and Igor V. Nedjalkov

as reduced acceptability of passives with verbs taking inanimate objects, as observed for Even, but generally the differences are less dramatic. As expected, the more syntactic information is taken into consideration, the more semantically coherent groupings emerge across languages, even though certain differences remain.

9.4.2 Causative The causative formation through the use of the common Tungusic marker *-bukan- shows fewer discrepancies across languages: when derived from intransitives, the causative yields a transitive frame; when derived from transitives, it normally yields a ditransitive frame with the causee in the dative (see (56a) above). Some languages (Even, Evenki, Nanai) also allow for a double object construction under similar conditions as Even (cf. (56b)), that is, in cases when the factitive-coercive causation is intended (see Malchukov & Nedjalkov 2010 for exemplification). On the whole, causative formation as a valency diagnostic gives a clear distinction between intransitives and transitives in Tungusic; “extended intransitives” (psychverbs with oblique objects) invariably pattern with intransitives in that respect. Further discussion of causatives in Tungusic is provided in Malchukov and Nedjalkov (2010).

9.4.3 Resultative Resultative forms with a cognate marker -ča- are found in North and East Tungusic (the Manchu parallels are less certain; cf. Avrorin 2000: 162–166)). In all Tungusic languages, the resultative form shows ergative characteristics insofar as it takes an S-argument when derived from intransitives and the Patient-argument when derived from transitives (cf. Evenki: doo- ‘land; sit down (of a bird)’ → doo-ča- ‘sit (of a bird)’, and loko- ‘hang (tr.)’ → loku-ča- ‘be hanging’, corresponding to Even examples (64b, 65b)). In contrast to Evenki, Even also features some A-oriented intransitives (as illustrated in (65c) above), and the same is true of Nanai (cf. Nanai: daila‘take a pipe into the mouth’ → daila-ča-‘hold a pipe in the mouth’). The use of the resultative form as a diagnostic for valency classification would again yield different results depending on whether we focus on morphological distribution of a particular form or take the syntactic properties of the derived construction into account. A classification on purely morphological grounds will yield a classification which cross-cuts the transitive/intransitive dichotomy as the resultative forms will be found both with (telic) intransitive verbs (S-resulatives) and (telic) transitive verbs (P-resultatives). If, however, one focuses on P-resultative as a diagnostic frame, the same subgroups emerge across languages: the resultative forms are pref-

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

613

erentially derived from transitive verbs designating a (caused) change of state (e.g., destruction, etc.) or of location as applied to the Patient (see Nedjalkov 1997 on resultatives in Evenki).

9.4.4 Reciprocal The reciprocal formation (cognates of common Tungusic *-ma-či- (Benzing 1955b: 122 (1070); Sunik 1962: 123) is interesting insofar as this formation is not solely dependent on transitivity or (numerical) valency but is rather sensitive to a number of roles which are invariably associated with animate arguments (such as agents and recipients as opposed to, say goals, which may be animate or inanimate). This has been shown to be true of Even, where apart from ‘direct reciprocals’ (with a subject coreferential to an animate direct object; cf. (66b) above), we find ‘indirect reciprocals’ (with a subject coreferential to an animate indirect object; cf. (68b)) as well as ‘oblique reciprocals’ (with a subject coreferential to an animate oblique object of an intransitive; cf. (67b)). In other languages, the formation of indirect and oblique reciprocals is more restricted though. They are most productive in Even (Malchukov 2007), less so in Evenki (cf. gogo-mot- ‘bark at each other’; Nedjalkov & Nedjalkov 2007: 1604), and least so in Udihe (cf. teluŋu-masi- ‘talk to each other’; Nikolaeva 2007: 941). In Nanai, the reciprocal marker is found almost exclusively with transitives (Avrorin 1961: 42–43), while in Manchu the reciprocal formation is further lexicalized and very restricted, even with transitives (Nedjalkov & Nedjalkov 2007).15 Thus, as we move from North Tungusic, to East Tungusic and further to South Tungusic, the reciprocal formation becomes more restricted. What makes reciprocal formation interesting for issues of valency classification is that the impact of animacy cannot be exclusively attributed to the animacy of the noun phrase in a particular construction. Rather, as noted above, reciprocal formation depends on the availability of semantic roles associated with animate arguments (such as Agent and Recipient) in the argument structure of a particular verb. Thus, in all Tungusic languages, verbs of motion (‘go to’, etc) fail to undergo this alternation, even in cases where the Goal is animate (see § 6.5 above). Similar restrictions are found for trivalent verbs: thus, indirect reciprocals are found with ditransitives taking Recipient and Addressee arguments (cf. Even: bö-met- ‘give to each other’, gö-met- ‘tell to each other’) but not with caused motion events (cf. Even: *n’odumat- ‘throw to each other’; see § 6.5 above). Thus, this restriction is verb-sensitive

15 Direct comparison with Manchu is problematic since the basic reciprocal marker -ndu- in Manchu does not correspond to the North Tungusic form -mat/-mač- but is related to the North Tungusic sociative in -lda (discussed in 6.6 above; see Nedjalkov & Nedjalkov 2007 for further discussion).

614

Andrej Malchukov and Igor V. Nedjalkov

insofar as it does not exclusively depend on the properties of the object but rather is rooted in the argument structure of individual verbs and verb classes.

9.5 Conclusions The main question we addressed in this section is what consequences the structural differences in inventories of case markers and valency-changing morphology have for valency classifications in individual Tungusic languages. Our general conclusion on coding frames is that reduction of case paradigms leads expectedly to neutralization of coding frames, but this is partially compensated for by alternations targeting individual semantic roles. Concerning the effect of discrepancies in valency-changing morphology, the answer would be different depending on how far one takes into account syntactic properties of the cognate morphemes in addition to morphological information. Thus, the more consistently syntactic information is taken into account, the more coherent semantic classes emerge. This is expected given the hypothesis that syntactic classification is a reflection of semantic classification, as argued by Apresjan (1969), Levin (1993), and subsequent work.

Appendix. Valency classes and alternations in Even The appendix includes a tabular summary of the coding patterns of alternations, extracted from the database. Not all alternations are represented, only more regular ones. In some cases only a particular syntactic type of alternation is taken into account. Thus, for an adversative passive, a valency-decreasing type (as illustrated in (57b)) is taken into account as it shows more restrictions. Similarly, for mediopassive, the anticausative function (as in (61)) is taken as diagnostic, while potential-impersonal uses (as in (63)) are ignored as they can be found with both transitives and intransitives. For resultative forms, P-resultatives (as in (65b)) and Sresultatives (as in (64b)) are taken into account. For the reciprocal formation, only direct reciprocals are considered as diagnostic (cf. (66b)). The database includes a few more alternations. The footnotes comment on the internal composition of individual verbs, or mention synonymous verbs.

kuni-

en-

iŋem-

GO

LAUGH

SCREAM

FEEL PAIN16

FEEL COLD17

DIE

PLAY

BE SAD

 54

 57

 58

 59

 60

 61

 62

 63

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]







m











Coding frame Adversative schema Passive











m

m





Direct Reciprocal



m















Mediopassive



















Object deletion













m



+

Resultative

16 Alternatively this meaning can be rendered by denominal verbs in -m- derived from the body parts: dila-m- ‘have headache’ (cf. dil ‘head’), böde-m- ‘hurt (of a leg)’ (cf. bödel ‘leg’),etc. 17 Also can be rendered by beji- (tr.) ‘frost-bite’; cf. Bi bödel-i beji-re-m ‘I had my legs frost-bitten’.

nabu-t

evi-

kö’ke-

inin-

hör- (1)

hiemken-

COUGH

 47

Verb form

Meaning label

#

Valency classes in Even (North Tungusic) in a comparative Tungusic perspective

615

d’em-

ileŋ bi-

udn-

BE HUNGRY18

ROLL

SINK

BURN

BE DRY

RAIN19

BE A HUNTER

APPEAR

FALL

 64

 65

 66

 67

 68

 69

 70

 81

 84

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]

1-nom V.subj[1]



m



m











Coding frame Adversative schema Passive



















Direct Reciprocal

18 The verb d’em- is historically a desiderative form